PortPilot
Benchmark · pre-generated, always available← back to leaderboard
Migration Readiness Report

NVIDIA/cuda-python

https://github.com/NVIDIA/cuda-python ↗
Manual blockers present

With 843 manual blockers against only 42 items that work as-is, this repo is very far from ROCm readiness — the vast majority of CUDA Python bindings, driver/runtime API wrappers, and kernel-launch paths will require deliberate porting effort, not just search-and-replace. The 34 mechanical items (likely header renames, enum substitutions, and API signature tweaks) are low-hanging fruit but represent a tiny fraction of the work. The first step should be triaging the 843 blockers by subsystem — separating driver-API bindings (which may map to HIP's driver API) from runtime/launcher code (which needs deeper restructuring) — to identify whether a shim/compatibility layer is feasible or whether a from-scratch HIP Python binding is more practical. Until that triage is done, treat the 6/100 score as a hard stop, not a starting point.

42Works as-is
34Mechanical change
843Manual blocker
Findings
919
Python files
297
Files scanned
859
Custom CUDA kernels
detected

Detected but out of scope (not analyzed): C++, C/C++ header

Findings by file

919 findings · 223 files
benchmarks/cuda_bindings/benchmarks/bench_ctx_device.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver import is the cuda-python low-level driver API, which has no direct ROCm equivalent. Migration requires switching to hip-python (hip.bindings.driver) or the HIP C API via ctypes, and reviewing all downstream driver calls for API differences.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_ctx_device.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_ctx_device.py
@@ -9,1 +9,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: replace with hip-python low-level driver API
+from hip.bindings import driver as hip
benchmarks/cuda_bindings/benchmarks/bench_enum.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:7

The cuda.bindings.driver module is part of cuda-python's low-level driver bindings with no drop-in ROCm equivalent. Migration requires either the hip-python project (HIP Python bindings) or direct ctypes/Cython bindings to HIP/ROCm runtime APIs; the API surface differs significantly so call-sites must be reviewed individually.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_enum.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_enum.py
@@ -4,7 +4,8 @@
 # Advisory only: cuda-python driver bindings have no direct ROCm equivalent.
 # Replace with hip-python or custom HIP/ROCm bindings after auditing call-sites.
-from cuda.bindings import driver as cuda
+# from cuda.bindings import driver as cuda  # CUDA-only
+# TODO: migrate to hip-python or ctypes-based HIP runtime bindings
benchmarks/cuda_bindings/benchmarks/bench_event.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver import is the cuda-python low-level driver API, which has no direct drop-in ROCm equivalent. The closest analog is hipPython (hip.bindings.driver), but API names and signatures differ significantly, so each downstream call must be reviewed and adapted individually.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_event.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_event.py
@@ -9,1 +9,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: migrate to hipPython; verify all downstream driver API calls
+from hip.bindings import driver as hip
benchmarks/cuda_bindings/benchmarks/bench_launch.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:10

The cuda.bindings.driver module provides Python bindings to the CUDA low-level driver API (context, module, stream, memory management). There is no drop-in ROCm equivalent Python package with the same API surface; migration requires either using ctypes to bind against libamdhip64.so directly, or porting to a higher-level framework (e.g., PyTorch HIP, Numba ROCm). All downstream driver-API calls in this benchmark will need manual re-mapping to HIP runtime/driver equivalents.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_launch.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_launch.py
@@ -10,1 +10,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: no direct ROCm Python binding equivalent for cuda-python driver API.
+# Consider ctypes bindings to libamdhip64.so or port to a HIP-aware framework.
+# from rocm_bindings import driver as hip  # placeholder — manual port required
benchmarks/cuda_bindings/benchmarks/bench_memory.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:10

The cuda.bindings.driver module is the cuda-python low-level driver API binding, which has no direct ROCm equivalent in the same package form. For AMD Instinct migration, this should be replaced with hipPython (the hip Python bindings) or the HIP runtime API, keeping in mind that API names and semantics differ (e.g., cuDeviceGet → hipDeviceGet).

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_memory.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_memory.py
@@ -10,1 +10,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: replace with hipPython low-level driver bindings for ROCm
+# Requires: pip install hip-python (package availability may vary)
+# from hip.bindings import driver as hip  # placeholder; verify exact import path
benchmarks/cuda_bindings/benchmarks/bench_module.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver module is part of cuda-python and provides low-level CUDA driver API access with no drop-in ROCm equivalent. Migration requires either using amdsmi (for system management / device queries) or ctypes-based bindings to the HIP runtime, depending on which driver calls are actually used downstream. This import alone is advisory — the real migration effort depends on the specific driver API calls invoked after this import.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_module.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_module.py
@@ -9,1 +9,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: no direct ROCm equivalent for cuda-python low-level driver API.
+# Consider `amdsmi` for device queries or ctypes HIP runtime bindings.
+# from amdsmi import amdsmi as cuda  # placeholder — requires downstream API adaptation
benchmarks/cuda_bindings/benchmarks/bench_nvrtc.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver import is the low-level CUDA driver API from the cuda-python package, which has no direct ROCm equivalent in the same package. On AMD Instinct GPUs, the advisory migration target is the hip-python package (e.g., from hip import hip), but the API surface differs significantly and all downstream driver calls must be individually translated.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_nvrtc.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_nvrtc.py
@@ -6,7 +6,8 @@
 
-# CUDA low-level driver API import; requires cuda-python package
-from cuda.bindings import driver as cuda
+# Advisory: migrate to hip-python for ROCm; API surface differs from CUDA driver
+# from hip import hip as cuda  # uncomment after translating downstream calls
C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:10

The cuda.bindings.nvrtc module is part of cuda-python's low-level driver bindings and has no direct ROCm equivalent in the same package ecosystem. For ROCm, NVRTC runtime compilation should be replaced with HIP's hiprtc API, typically accessed via amdsmi/pyhip or by calling hiprtc through ctypes/Cython bindings, or by pre-compiling HIP kernels offline with hipcc.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_nvrtc.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_nvrtc.py
@@ -10,1 +10,1 @@
-from cuda.bindings import nvrtc
+# Advisory: replace with ROCm hiprtc binding, e.g. `from pyhip import hiprtc` or ctypes-based hiprtc wrapper
benchmarks/cuda_bindings/benchmarks/bench_pointer_attributes.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver module is the cuda-python low-level driver API with no direct ROCm equivalent. Pointer attribute queries (cudaPointerGetAttributes) map conceptually to HIP's hipPointerGetAttributes, but no mature Python bindings package provides the same surface; this benchmark will require substantial rework or a custom ctypes wrapper around the HIP runtime. Migration impact is high — the entire driver-level API surface used in this file must be replaced.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- benchmarks/cuda_bindings/benchmarks/bench_pointer_attributes.py
+++ benchmarks/cuda_bindings/benchmarks/bench_pointer_attributes.py
@@ -9,1 +9,4 @@
-from cuda.bindings import driver as cuda
+# Advisory: cuda-python has no direct ROCm equivalent.
+# Consider hip-python (https://github.com/ROCm/hip-python) or a ctypes
+# wrapper around the HIP runtime for hipPointerGetAttributes.
+# from hip import hip as cuda  # placeholder — API surface differs
benchmarks/cuda_bindings/benchmarks/bench_stream.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver import is the low-level CUDA driver API from cuda-python, which has no direct drop-in equivalent in ROCm. Migration requires switching to hip-python's hip.bindings.driver module or using PyTorch/hipify utilities, and all downstream driver API calls (e.g., cuDeviceGet, cuMemAlloc) must be translated to their HIP equivalents (hipDeviceGet, hipMalloc).

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_stream.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_stream.py
@@ -9,1 +9,1 @@
-from cuda.bindings import driver as cuda
+from hip.bindings import driver as hip  # Advisory: replace all cu* driver calls with hip* equivalents
benchmarks/cuda_bindings/benchmarks/bench_tensormap.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:9

The cuda.bindings.driver module is part of cuda-python and provides low-level CUDA Driver API bindings with no direct ROCm equivalent. Migration requires replacing this with ROCm's HIP Python bindings or amdsmi for device management, and all downstream driver API calls (e.g., cuDeviceGet, cuCtxCreate) must be individually remapped to HIP equivalents.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/benchmarks/bench_tensormap.py
+++ b/benchmarks/cuda_bindings/benchmarks/bench_tensormap.py
@@ -6,7 +6,8 @@
 # ADVISORY: cuda.bindings.driver has no direct ROCm equivalent.
 # Replace with HIP Python bindings or amdsmi as appropriate.
 # All downstream cu* calls must be remapped to hip* equivalents.
-from cuda.bindings import driver as cuda
+# from cuda.bindings import driver as cuda
+import amdsmi as cuda  # advisory placeholder; API surface differs significantly
benchmarks/cuda_bindings/runner/runtime.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:7

The cuda.bindings.driver module is part of cuda-python, a low-level CUDA driver API binding with no direct ROCm equivalent. Migration requires replacing this with a ROCm-compatible Python binding (e.g., amdsmi, rocm_smi, or HIP Python bindings) and rewriting all downstream driver API calls, as the function signatures and object types differ substantially.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/runner/runtime.py
+++ b/benchmarks/cuda_bindings/runner/runtime.py
@@ -7,1 +7,1 @@
-from cuda.bindings import driver as cuda
+# Advisory: no 1:1 ROCm equivalent for cuda-python driver bindings.
+# Consider `import amdsmi` or HIP Python bindings; downstream calls must be rewritten.
+from amdsmi import amdsmi as cuda  # placeholder — verify API coverage before use
C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:8

The cuda.bindings.nvrtc module provides Python bindings to NVIDIA's NVRTC runtime compilation library; on ROCm the equivalent is HIP RTC (hiprtc). There is no drop-in Python binding package with the same API surface, so this import and all downstream NVRTC calls must be replaced with a HIP RTC-based alternative (e.g., via hip Python bindings, ctypes wrappers, or a framework like CuPy-on-ROCm/Triton).

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_bindings/runner/runtime.py
+++ b/benchmarks/cuda_bindings/runner/runtime.py
@@ -8,1 +8,1 @@
-from cuda.bindings import nvrtc
+# Advisory: replace with a HIP RTC Python binding (e.g., hiprtc via ctypes or CuPy-on-ROCm); no direct drop-in exists
+import hiprtc  # placeholder — verify availability and API parity before use
benchmarks/cuda_core/benchmarks/bench_ctx_device.py· 1
C
cuda-python low-level driver API
from cuda.core import Device
:9

The cuda.core module is part of cuda-python's low-level driver API and has no direct ROCm equivalent. For device enumeration/selection on AMD Instinct GPUs, use amdsmi (AMD System Management Interface Python bindings) or HIP runtime APIs via hip/pyhip packages. This is advisory — the replacement depends on what Device methods are actually called downstream.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_core/benchmarks/bench_ctx_device.py
+++ b/benchmarks/cuda_core/benchmarks/bench_ctx_device.py
@@ -9,1 +9,1 @@
-from cuda.core import Device
+# Advisory: replace with ROCm device API, e.g.:
+# import amdsmi  # amdsmi.amdsmi_initialize(); amdsmi.amdsmi_get_processor_handles()
+# or use HIP runtime Python bindings if available
+from amdsmi import amdsmi_initialize, amdsmi_get_processor_handles  # advisory placeholder
benchmarks/cuda_core/benchmarks/bench_launch.py· 1
C
cuda-python low-level driver API
from cuda.core import LaunchConfig, launch
:9

The cuda.core module is NVIDIA's experimental low-level Python API with no direct ROCm equivalent; LaunchConfig and launch must be replaced with a ROCm-compatible kernel launch path. For benchmark code, the most practical migration is to use PyTorch's stream/launch facilities or hipPython's hip module, but the exact replacement depends on how kernels are compiled and launched downstream.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/benchmarks/cuda_core/benchmarks/bench_launch.py
+++ b/benchmarks/cuda_core/benchmarks/bench_launch.py
@@ -9,1 +9,1 @@
-from cuda.core import LaunchConfig, launch
+# ADVISORY: cuda.core has no ROCm equivalent; replace with a ROCm launch path
+# (e.g., hipPython `hip` module or PyTorch stream-based launch). Pending migration.
+# from cuda.core import LaunchConfig, launch
benchmarks/cuda_core/runtime.py· 1
C
cuda-python low-level driver API
from cuda.core import Device, Program, ProgramOptions
:14

The cuda.core high-level API from cuda-python has no direct ROCm equivalent; migration requires either hip-python (which exposes a lower-level hip module) or manual porting to PyHIP/ctypes. Device enumeration, program compilation, and program options semantics differ between CUDA and HIP, so each usage site must be reviewed individually.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- benchmarks/cuda_core/runtime.py
+++ benchmarks/cuda_core/runtime.py
@@ -14,1 +14,1 @@
-from cuda.core import Device, Program, ProgramOptions
+# Advisory: cuda.core has no ROCm equivalent. Consider hip-python:
+# from hip import hip  # low-level HIP runtime API
+# Device/Program/ProgramOptions must be reimplemented manually
+# (e.g., hipGetDeviceCount/hipDeviceGet for Device,
+#  hiprtcCreateProgram/hiprtcCompileProgram for Program).
conftest.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:9

The cuda.pathfinder module is part of the cuda-python package and has no ROCm equivalent; it will fail to import on AMD Instinct systems. The path lookup logic should be replaced with a ROCm-specific mechanism, typically referencing the ROCM_PATH environment variable or falling back to /opt/rocm.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- conftest.py
+++ conftest.py
@@ -9,1 +9,1 @@
-from cuda.pathfinder import get_cuda_path_or_home
+import os
+def get_rocm_path_or_home():
+    return os.environ.get("ROCM_PATH", "/opt/rocm")
cuda_bindings/build_hooks.py· 12
C
cuda-python low-level driver API
import cuda.pathfinder
:43

This is a build-time dependency on the cuda-python pathfinder utility, which locates CUDA SDK components and has no ROCm equivalent. During migration, this import and any logic relying on it must be replaced with ROCm-specific path discovery (e.g., via ROCM_PATH env var or amdsmi/rocminfo), or the build hook must be forked/removed if it is not relevant to the ROCm target.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- a/cuda_bindings/build_hooks.py
+++ b/cuda_bindings/build_hooks.py
@@ -40,7 +40,11 @@
 # Advisory: cuda.pathfinder is CUDA-specific; replace with ROCm path discovery.
-import cuda.pathfinder
+try:
+    import cuda.pathfinder
+except ImportError:
+    # ROCm migration: fall back to ROCM_PATH or system default
+    cuda = None
A
Device string "cuda"
if exc.name not in ("cuda", "cuda.pathfinder"):
:45

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
import cuda
:48

This import pulls in the cuda-python low-level driver API bindings, which have no direct ROCm equivalent. Migration requires replacing the cuda-python package with hip-python (AMD's HIP Python bindings) or an equivalent ROCm Python wrapper, and the surrounding build_hooks infrastructure would need corresponding adjustments.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
--- cuda_bindings/build_hooks.py
+++ cuda_bindings/build_hooks.py
@@ -48,1 +48,1 @@
-import cuda
+import hip  # Advisory: replace cuda-python with hip-python (ROCm equivalent)
A
Device string "cuda"
sp_cuda = os.path.join(p, "cuda")
:53

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
import cuda.pathfinder
:62

cuda.pathfinder is a CUDA-specific utility for locating the CUDA SDK/toolkit paths at build time; it has no ROCm equivalent. For ROCm migration, path discovery should rely on the ROCM_PATH environment variable or /opt/rocm as a fallback, or use hipconfig --rocmpath.

RecommendRewrite driver-API calls against the HIP runtime.

Suggested change · advisory
-import cuda.pathfinder
+# Advisory: cuda.pathfinder has no ROCm equivalent.
+# Use ROCM_PATH env var or /opt/rocm fallback for SDK path discovery:
+# import os
+# rocm_path = os.environ.get('ROCM_PATH', '/opt/rocm')
A
Device string "cuda"
path = os.path.join("cuda", "bindings", "_internal")
:283

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda"),
:370

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda", "bindings"),
:371

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda", "bindings", "_bindings"),
:372

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda", "bindings", "_internal"),
:373

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda", "bindings", "_lib"),
:374

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
os.path.join("cuda", "bindings", "utils"),
:375

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_bindings/cuda/bindings/__init__.py· 2
C
cuda-python low-level driver API
from cuda.bindings import utils
:4

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._version import __version__
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/_example_helpers/common.py· 4
C
cuda-python low-level driver API
from cuda import pathfinder
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/_example_helpers/helper_cuda.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:4

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/_test_helpers/arch_check.py· 2
C
cuda-python low-level driver API
from cuda.bindings import nvml
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.utils import FunctionNotFoundError as NvmlSymbolNotFoundError
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/_test_helpers/mempool.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver, runtime
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.nvml as nvml
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/utils/_nvvm_utils.py· 2
C
cuda-python low-level driver API
from cuda.bindings import nvvm
:54

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.nvvm import _inspect_function_pointer
:60

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/cuda/bindings/utils/_version_check.py· 1
C
cuda-python low-level driver API
from cuda.bindings import driver
:39

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/clock_nvrtc.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, find_cuda_device, requirement_not_met
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/simple_cubemap_texture.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, find_cuda_device, requirement_not_met
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/simple_p2p.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, requirement_not_met
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/simple_zero_copy.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/system_wide_atomics.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, find_cuda_device, requirement_not_met
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/vector_add_drv.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, find_cuda_device_drv, requirement_not_met
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/0_Introduction/vector_add_mmap.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, find_cuda_device_drv, requirement_not_met
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/2_Concepts_and_Techniques/stream_ordered_allocation.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/3_CUDA_Features/global_to_shmem_async_copy.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/3_CUDA_Features/simple_cuda_graphs.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import (
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/4_CUDA_Libraries/conjugate_gradient_multi_block_cg.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/4_CUDA_Libraries/nvidia_smi.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/extra/iso_fd_modelling.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._example_helpers import KernelHelper, check_cuda_errors, requirement_not_met
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/examples/extra/jit_program.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver as cuda
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/conftest.py· 1
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/cython/build_tests.py· 2
C
cuda-python low-level driver API
import cuda.bindings
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
if not (root / "cuda" / "bindings").is_dir():
:28

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_bindings/tests/nvml/__init__.py· 1
C
cuda-python low-level driver API
from cuda.bindings._test_helpers.arch_check import hardware_supports_nvml
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/conftest.py· 2
C
cuda-python low-level driver API
from cuda.bindings import nvml
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.arch_check import unsupported_before # noqa: F401
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_compute_mode.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_cuda.py· 2
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvml
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_device.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_gpu.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_init.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_nvlink.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_page_retirement.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_pci.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_pynvml.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/test_unit.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/nvml/util.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_basics.py· 1
C
cuda-python low-level driver API
from cuda.bindings._internal import _fast_enum
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_cuda.py· 5
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.runtime as cudart
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.mempool import xfail_if_mempool_oom
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.driver import _inspect_function_pointer
:871

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_cudart.py· 5
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.runtime as cudart
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda import pathfinder
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import runtime
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.mempool import xfail_if_mempool_oom
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_cufile.py· 1
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_examples.py· 1
C
cuda-python low-level driver API
from cuda.bindings._test_helpers.pep723 import has_package_requirements_or_skip
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_graphics_apis.py· 1
C
cuda-python low-level driver API
from cuda.bindings import runtime as cudart
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_interoperability.py· 3
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.runtime as cudart
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.mempool import xfail_if_mempool_oom
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_kernelParams.py· 3
C
cuda-python low-level driver API
import cuda.bindings.driver as cuda
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.nvrtc as nvrtc
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.runtime as cudart
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_nvfatbin.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvfatbin, nvrtc
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_nvjitlink.py· 2
C
cuda-python low-level driver API
from cuda.bindings import nvjitlink, nvrtc
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal import nvjitlink as inner_nvjitlink
:68

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_nvrtc.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_nvvm.py· 2
C
cuda-python low-level driver API
from cuda.bindings import nvvm
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.utils import FunctionNotFoundError
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_utils.py· 5
C
cuda-python low-level driver API
from cuda.bindings import driver, runtime
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.utils import get_c_compiler
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.utils import (
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal.nvvm import _inspect_function_pointer
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import DynamicLibNotFoundError
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/tests/test_version_check.py· 2
C
cuda-python low-level driver API
from cuda.bindings import driver
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.utils import _version_check, warn_if_cuda_major_version_mismatch
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/build_hooks.py· 8
C
cuda-python low-level driver API
import cuda.pathfinder
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
if exc.name not in ("cuda", "cuda.pathfinder"):
:40

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
import cuda
:43

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
sp_cuda = os.path.join(p, "cuda")
:48

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
import cuda.pathfinder
:57

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings
:135

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
root_path = os.path.sep.join(["cuda", "core", ""])
:155

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
import cuda.bindings
:245

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/__init__.py· 19
C
cuda-python low-level driver API
from cuda.core._version import __version__
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda import bindings
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import checkpoint, system, utils
:71

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._context import Context, ContextOptions
:72

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device import Device
:73

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device_resources import (
:74

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._event import Event, EventOptions
:81

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._graphics import GraphicsResource
:82

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._host import Host
:83

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launch_config import LaunchConfig
:84

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launcher import launch
:85

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._linker import Linker, LinkerOptions
:86

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory import (
:87

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import Kernel, ObjectCode
:102

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import Program, ProgramOptions
:103

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import (
:104

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._tensor_map import TensorMapDescriptor, TensorMapDescriptorOptions
:110

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core.graph
:120

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core.texture
:121

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_memory/_legacy.py· 10
C
cuda-python low-level driver API
from cuda.core._stream import Stream
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import DevicePointerType
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._buffer import Buffer, MemoryResource
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:52

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:76

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:112

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:124

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_memory/_managed_buffer.py· 10
C
cuda-python low-level driver API
from cuda.core._device import Device
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._host import Host
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._buffer import Buffer
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_memory_ops import (
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version, driver_version
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._buffer import MemoryResource
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_memory/_managed_location.py· 5
C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version, driver_version
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device import Device
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._host import Host
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device import Device
:62

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._host import Host
:63

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_memory/_virtual_memory_resource.py· 10
C
cuda-python low-level driver API
from cuda.core._stream import Stream
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device import Device
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._buffer import Buffer, MemoryResource
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:517

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept
:589

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_utils/driver_cu_result_explanations.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.enum_explanations_helpers import get_best_available_explanations
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.driver_cu_result_explanations_frozen import _FALLBACK_EXPLANATIONS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/_utils/runtime_cuda_error_explanations.py· 3
C
cuda-python low-level driver API
from cuda.bindings import runtime
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.enum_explanations_helpers import get_best_available_explanations
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.runtime_cuda_error_explanations_frozen import _FALLBACK_EXPLANATIONS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/checkpoint.py· 5
C
cuda-python low-level driver API
from cuda.bindings import driver as _driver
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import handle_return as _handle_cuda_return
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version as _binding_version
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import driver_version as _driver_version
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import ProcessStateType as _ProcessStateType
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/system/__init__.py· 1
C
cuda-python low-level driver API
from cuda.core.system import typing
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/system/exceptions.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvml
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/system/typing.py· 3
C
cuda-python low-level driver API
from cuda.core._utils.pycompat import StrEnum
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvml as _nvml
:324

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._internal._fast_enum import FastEnum as _FastEnum
:327

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/texture/__init__.py· 4
C
cuda-python low-level driver API
from cuda.core.texture._array import OpaqueArray, OpaqueArrayOptions
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture._mipmapped_array import MipmappedArray, MipmappedArrayOptions
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture._surface import SurfaceObject
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture._texture import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/typing.py· 3
C
cuda-python low-level driver API
from cuda.core._context import DeviceResourcesType
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import IsStreamType
:31

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver
:32

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/utils/__init__.py· 3
C
cuda-python low-level driver API
from cuda.core._memory._managed_memory_ops import (
:5

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memoryview import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache import (
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/utils/_program_cache/_abc.py· 1
C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/utils/_program_cache/_file_stream.py· 1
C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/utils/_program_cache/_in_memory.py· 1
C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/cuda/core/utils/_program_cache/_keys.py· 8
C
cuda-python low-level driver API
from cuda.core._program import ProgramOptions
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import (
:33

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
# Order is preserved on purpose: ptxas accepts ordering-sensitive flag
:91

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
cuda-python low-level driver API
from cuda.bindings import nvjitlink as _nvjitlink
:200

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:222

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._linker import _decide_nvjitlink_or_driver
:488

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/docs/source/conf.py· 1
C
cuda-python low-level driver API
from cuda.core._system import System
:179

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/cuda_graphs.py· 2
B
CuPy dependency
import cupy as cp
:20

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/gl_interop_fluid.py· 3
C
cuda-python low-level driver API
from cuda.core import (
:102

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture import (
:110

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:115

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/gl_interop_mipmap_lod.py· 3
C
cuda-python low-level driver API
from cuda.core import (
:95

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture import (
:103

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:108

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/gl_interop_plasma.py· 1
C
cuda-python low-level driver API
from cuda.core import (
:67

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/graph_update.py· 1
C
cuda-python low-level driver API
from cuda.core import (
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/jit_lto_fractal.py· 2
B
CuPy dependency
import cupy as cp
:22

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Linker, LinkerOptions, Program, ProgramOptions, launch
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/memory_ops.py· 2
B
CuPy dependency
import cupy as cp
:19

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import (
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/memory_pool_resources.py· 1
C
cuda-python low-level driver API
from cuda.core import (
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/pytorch_example.py· 8
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
torch.cuda API usage
pt_stream = torch.cuda.current_stream()
:51

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
Device string "cuda"
a = torch.tensor([10.0], dtype=dtype, device="cuda")
:73

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
x = torch.rand(size, dtype=dtype, device="cuda")
:74

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
y = torch.rand(size, dtype=dtype, device="cuda")
:75

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
a = torch.tensor([42.0], dtype=dtype, device="cuda")
:97

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
x = torch.rand(size, dtype=dtype, device="cuda")
:98

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
y = torch.rand(size, dtype=dtype, device="cuda")
:99

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_core/examples/saxpy.py· 3
C
cuda-python low-level driver API
from cuda import pathfinder
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

B
CuPy dependency
import cupy as cp
:24

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/show_device_properties.py· 1
C
cuda-python low-level driver API
from cuda.core import Device, system
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/simple_multi_gpu_example.py· 2
B
CuPy dependency
import cupy as cp
:18

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch, system
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/strided_memory_view_constructors.py· 3
B
CuPy dependency
import cupy as cp
:18

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/strided_memory_view_cpu.py· 1
C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView, args_viewable_as_strided_memory
:29

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/strided_memory_view_gpu.py· 3
B
CuPy dependency
import cupy as cp
:20

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView, args_viewable_as_strided_memory
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/texture_sample.py· 3
C
cuda-python low-level driver API
from cuda.core import (
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture import (
:32

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:37

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/thread_block_cluster.py· 2
C
cuda-python low-level driver API
from cuda.core import (
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/tma_tensor_map.py· 4
B
CuPy dependency
import cupy as cp
:32

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import (
:35

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:42

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:43

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/examples/vector_add.py· 2
B
CuPy dependency
import cupy as cp
:16

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, Program, ProgramOptions, launch
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/conftest.py· 7
C
cuda-python low-level driver API
import cuda.core
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import (
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError, handle_return
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.mempool import xfail_if_mempool_oom
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.nvml as nvml
:42

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/cython/build_tests.py· 2
C
cuda-python low-level driver API
import cuda.bindings
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
if not (root / "cuda" / "bindings").is_dir():
:28

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_core/tests/example_tests/test_basic_examples.py· 3
C
cuda-python low-level driver API
from cuda.core import Device, ManagedMemoryResource, system
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _can_load_generated_ptx
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.pep723 import has_package_requirements_or_skip
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_device_launch.py· 2
C
cuda-python low-level driver API
from cuda.core import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphCompleteOptions
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_builder_conditional.py· 2
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, LegacyPinnedMemoryResource, launch
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_builder.py· 2
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, LegacyPinnedMemoryResource, launch
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder, GraphDefinition
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_definition_errors.py· 6
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import (
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:193

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:222

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:252

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_definition_integration.py· 3
C
cuda-python low-level driver API
from cuda.core import Device, EventOptions, LaunchConfig, Program, ProgramOptions
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphDefinition
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_definition_lifetime.py· 15
C
cuda-python low-level driver API
from cuda.core import Device, DeviceMemoryResource, EventOptions, Kernel, LaunchConfig
:65

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import (
:66

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:330

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:475

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:564

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:601

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:637

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:664

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:689

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:718

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils._weak_handles import weak_handle
:753

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:780

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:816

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:877

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:906

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_definition_mutation.py· 3
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, LegacyPinnedMemoryResource
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphDefinition, KernelNode, MemsetNode
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_definition.py· 11
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import (
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import GraphConditionalType, GraphMemoryType
:36

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version, driver_version
:53

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:62

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph._graph_node import _node_registry
:733

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:801

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:1055

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:1196

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:1231

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver as drv
:1265

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_memory_resource.py· 4
C
cuda-python low-level driver API
from cuda.core import (
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphCompleteOptions
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._graph_memory_resource import GraphMemoryResourceAttributes
:295

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_graph_update.py· 3
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, LegacyPinnedMemoryResource, launch
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphDefinition
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/graph/test_options.py· 2
C
cuda-python low-level driver API
from cuda.core import Device, LaunchConfig, launch
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphBuilder, GraphCompleteOptions, GraphDebugPrintOptions
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/__init__.py· 3
C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import handle_return
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver # type: ignore
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/buffers.py· 2
C
cuda-python low-level driver API
from cuda.core import Buffer, Device, MemoryResource
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver, handle_return
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/graph_kernels.py· 3
C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device, Program, ProgramOptions
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import NVRTCError, handle_return
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/latch.py· 1
C
cuda-python low-level driver API
from cuda.core import (
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/misc.py· 1
C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/helpers/nanosleep_kernel.py· 1
C
cuda-python low-level driver API
from cuda.core import (
:4

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_errors.py· 4
C
cuda-python low-level driver API
from cuda.core import Buffer, Device, DeviceMemoryResource, DeviceMemoryResourceOptions
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory import IPCBufferDescriptor
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._ipc import IPCAllocationHandle
:48

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_event_ipc.py· 1
C
cuda-python low-level driver API
from cuda.core import Device, EventOptions
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_ipc_duplicate_import.py· 1
C
cuda-python low-level driver API
from cuda.core import Buffer, Device
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_memory_ipc.py· 1
C
cuda-python low-level driver API
from cuda.core import Buffer, DeviceMemoryResource
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_peer_access.py· 2
C
cuda-python low-level driver API
from cuda.core import Device, DeviceMemoryResource, DeviceMemoryResourceOptions
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_send_buffers.py· 1
C
cuda-python low-level driver API
from cuda.core import Device, DeviceMemoryResource, DeviceMemoryResourceOptions
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_serialize.py· 1
C
cuda-python low-level driver API
from cuda.core import Buffer, Device, DeviceMemoryResource, PinnedMemoryResource
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory_ipc/test_workerpool.py· 1
C
cuda-python low-level driver API
from cuda.core import Buffer, Device, DeviceMemoryResource, DeviceMemoryResourceOptions
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/memory/test_managed_ops.py· 21
C
cuda-python low-level driver API
from cuda.bindings import driver
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device, Host, ManagedBuffer
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_buffer import _get_int_attr
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:154

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:162

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:168

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:169

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:178

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:179

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:187

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:193

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:198

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_location import _coerce_location
:205

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import prefetch_batch
:215

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import prefetch_batch
:230

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import discard_batch, prefetch_batch
:255

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import discard_prefetch_batch, prefetch_batch
:274

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import utils
:303

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import utils
:317

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Buffer
:351

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:391

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/system/conftest.py· 3
C
cuda-python low-level driver API
from cuda.core import system
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.arch_check import hardware_supports_nvml
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.arch_check import unsupported_before as nvml_unsupported_before
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/system/test_nvml_context.py· 3
C
cuda-python low-level driver API
from cuda.core.system import _nvml_context
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import _nvml_context
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import _nvml_context
:60

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/system/test_system_device.py· 6
C
cuda-python low-level driver API
from cuda.core import system
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import typing
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvml
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.nvml import DeviceArch
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import _device
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device as CudaDevice
:56

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/system/test_system_events.py· 2
C
cuda-python low-level driver API
from cuda.core import system
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import typing
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/system/test_system_system.py· 3
C
cuda-python low-level driver API
from cuda.bindings import driver
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import system
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import handle_return
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_build_hooks.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder import get_cuda_path_or_home
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_checkpoint.py· 2
C
cuda-python low-level driver API
from cuda.core import Device, checkpoint
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_context.py· 2
C
cuda-python low-level driver API
import cuda.core
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_cuda_utils.py· 7
C
cuda-python low-level driver API
from cuda.bindings import driver, runtime
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils import cuda_utils
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.clear_error_support import assert_type_str_or_bytes_like, raise_code_path_meant_to_be_unreachable
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.enum_explanations_helpers import _binding_version_has_usable_enum_docstrings
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:195

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvrtc
:203

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_device.py· 8
C
cuda-python low-level driver API
import cuda.core
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver, runtime
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import ComputeCapability, handle_return
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version, driver_version
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import _system
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings._test_helpers.arch_check import hardware_supports_nvml
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import Device as SystemDevice
:35

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_enum_coverage.py· 7
C
cuda-python low-level driver API
import cuda.core
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core.typing
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import system
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core.system.typing as system_typing
:125

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvml
:126

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.system import _device, _system_events
:127

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_event.py· 6
C
cuda-python low-level driver API
import cuda.core
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import (
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._event as _event_module
:240

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._event as _event_module
:249

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._event as _event_module
:267

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._event as _event_module
:278

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_graphics.py· 6
C
cuda-python low-level driver API
from cuda.core import (
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._graphics import _parse_register_flags
:160

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._graphics import _parse_register_flags
:165

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._graphics import _parse_register_flags
:171

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._graphics import _parse_register_flags
:177

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_green_context.py· 5
C
cuda-python low-level driver API
from cuda.core import (
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import WorkqueueSharingScopeType
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device, system
:240

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import StreamOptions
:403

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_helpers.py· 1
C
cuda-python low-level driver API
from cuda.core import Device
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_host.py· 1
C
cuda-python low-level driver API
from cuda.core import Host
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_launcher.py· 19
B
CuPy dependency
import cupy as cp
:11

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
from cuda.core import (
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._legacy import _SynchronousMemoryResource
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import ObjectCodeFormatType, SourceCodeType
:29

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launch_config import _to_native_launch_config
:100

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launch_config import _to_native_launch_config
:130

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _launch_config as _lc_mod
:147

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:162

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _launch_config as _lc_mod
:163

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launch_config import _to_native_launch_config
:164

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _launch_config as _lc_mod
:190

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _launch_config as _lc_mod
:207

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:236

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._launch_config import _to_native_launch_config
:237

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._kernel_arg_handler import ParamHolder
:515

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._kernel_arg_handler import ParamHolder
:523

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._kernel_arg_handler import ParamHolder
:594

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._kernel_arg_handler import ParamHolder
:608

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_linker.py· 6
C
cuda-python low-level driver API
from cuda.core import Device, Linker, LinkerOptions, Program, ProgramOptions, _linker
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _can_load_generated_ptx
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvjitlink
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:323

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_managed_memory_warning.py· 4
C
cuda-python low-level driver API
import cuda.bindings
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device, ManagedMemoryResource, ManagedMemoryResourceOptions
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._managed_memory_resource import reset_concurrent_access_warning
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_memory_peer_access_utils.py· 1
C
cuda-python low-level driver API
from cuda.core._memory._peer_access_utils import PeerAccessPlan, plan_peer_access_update
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_memory_peer_access.py· 4
C
cuda-python low-level driver API
from cuda.core import Device, DeviceMemoryResource, DeviceMemoryResourceOptions, system
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory import _peer_access_utils
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._peer_access_utils import PeerAccessibleBySetProxy
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_memory.py· 16
C
cuda-python low-level driver API
from cuda.bindings import driver
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import (
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import (
:41

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._dlpack import DLDeviceType
:44

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory import IPCBufferDescriptor
:45

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError, handle_return
:46

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:47

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:55

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import Stream_accept, default_stream
:530

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core
:1750

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core.utils
:1751

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core
:1763

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._legacy import _SynchronousMemoryResource
:1788

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._legacy import _SynchronousMemoryResource
:1821

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._device_memory_resource import DMR_mempool_get_access
:1869

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._device_memory_resource import DMR_mempool_get_access
:1878

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_module.py· 9
C
cuda-python low-level driver API
import cuda.core
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device, Kernel, Linker, LinkerOptions, ObjectCode, Program, ProgramOptions
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _can_load_generated_ptx
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError, driver, handle_return
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.version import binding_version, driver_version
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import find_nvidia_binary_utility
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvfatbin
:44

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import nvfatbin
:144

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
ptx_file = tmp_path / "test.ptx"
:280

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

cuda_core/tests/test_multiprocessing_warning.py· 5
C
cuda-python low-level driver API
from cuda.core import DeviceMemoryResource, DeviceMemoryResourceOptions, EventOptions
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._event import _reduce_event
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._device_memory_resource import _deep_reduce_device_memory_resource
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._ipc import _reduce_allocation_handle
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import check_multiprocessing_start_method, reset_fork_warning
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_object_protocols.py· 3
C
cuda-python low-level driver API
from cuda.core import (
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _can_load_generated_ptx
:31

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.graph import GraphDefinition
:32

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_optional_dependency_imports.py· 1
C
cuda-python low-level driver API
from cuda.core import _linker, _program
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_program_cache_multiprocess.py· 9
C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:54

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:80

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:101

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:130

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:139

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:158

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_program_cache.py· 121
C
cuda-python low-level driver API
from cuda.core import _linker
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import ProgramCacheResource
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import ProgramCacheResource
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import ProgramCacheResource
:41

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import ProgramCacheResource
:79

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import ProgramOptions
:113

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import make_program_cache_key
:124

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:203

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:386

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:398

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:420

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:444

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:454

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import make_program_cache_key
:482

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _program
:545

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _program
:569

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:579

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:614

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:624

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:637

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:666

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:694

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:715

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:744

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:772

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:814

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._version as _version_mod
:1032

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:1045

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:1070

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
pytest.param("_linker_backend_and_version", "ptx", ".ptx", id="linker"),
:1085

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:1095

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _linker
:1114

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:1115

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
k_ok = _make_key(code=".ptx", code_type="ptx")
:1121

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
k_broken1 = _make_key(code=".ptx", code_type="ptx")
:1123

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
k_broken2 = _make_key(code=".ptx", code_type="ptx")
:1124

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:1131

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1142

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1155

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1164

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1180

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1198

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1208

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1218

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1232

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1250

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1281

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1304

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1322

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache._file_stream import _prune_if_stat_unchanged
:1323

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1348

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache._file_stream import _touch_atime
:1349

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1380

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1394

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1402

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1412

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:1422

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1423

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:1444

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1445

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1465

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:1484

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1485

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1509

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache
:1526

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache._file_stream import _default_cache_dir
:1527

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1553

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1571

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1590

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:1618

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1649

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1685

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1722

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1766

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1791

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1805

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1814

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1836

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1861

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1876

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1901

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1924

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:1962

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache import _file_stream
:1963

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2020

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2029

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2045

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2059

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2071

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache
:2080

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:2094

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils._program_cache._file_stream import _touch_atime
:2095

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:2117

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, _program_cache
:2150

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import ProgramOptions
:2204

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache, make_program_cache_key
:2205

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Program, ProgramOptions
:2234

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import Kernel, ObjectCode
:2235

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import FileStreamProgramCache, make_program_cache_key
:2236

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2278

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2288

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2297

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2305

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:2317

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2318

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2333

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2343

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2355

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2363

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2374

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2384

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2395

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2403

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2411

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2425

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2439

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2446

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2459

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2475

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2484

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import InMemoryProgramCache
:2500

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_program_compile_cache.py· 5
C
cuda-python low-level driver API
from cuda.core import Program, ProgramOptions
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _program as _program_module
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import ObjectCode
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import (
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import _program_cache as _pc
:341

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_program.py· 27
C
cuda-python low-level driver API
from cuda.core import _linker
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._device import Device
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._module import Kernel, ObjectCode
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import Program, ProgramOptions
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import CUDAError, handle_return
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import CompilerBackendType, PCHStatusType
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import nvrtc
:39

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.bindings.utils as utils
:76

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.utils import check_nvvm_compiler_options
:90

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.utils import check_nvvm_compiler_options
:99

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:112

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
%0 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x()
:129

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
%1 = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
:130

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
%2 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
:132

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
declare i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() nounwind readnone
:140

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
declare i32 @llvm.nvvm.read.ptx.sreg.ntid.x() nounwind readnone
:141

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() nounwind readnone
:142

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import NVRTCError
:386

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:420

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings.nvvm import nvvmError
:471

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:562

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._program import _get_nvvm_module
:599

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
Inline PTX assembly
%tid = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
:612

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
Inline PTX assembly
declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() nounwind readnone
:623

Inline PTX is NVIDIA ISA and cannot run on AMD. The block must be hand-rewritten in HIP/GCN or replaced with a portable path.

RecommendRewrite the PTX block in HIP or a portable high-level equivalent.

C
cuda-python low-level driver API
import cuda.pathfinder
:898

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import _program
:899

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_rlcompleter_patch.py· 1
C
cuda-python low-level driver API
from cuda.core import Device
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_stream.py· 4
C
cuda-python low-level driver API
from cuda.core import Device, Stream, StreamOptions
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._event import Event
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import LEGACY_DEFAULT_STREAM, PER_THREAD_DEFAULT_STREAM, Stream_accept
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.cuda_utils import driver
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_strided_layout.py· 1
C
cuda-python low-level driver API
from cuda.core._layout import _StridedLayout
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_tensor_map.py· 6
C
cuda-python low-level driver API
from cuda.core import (
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._dlpack import DLDeviceType
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._tensor_map import (
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._tensor_map import _NUMPY_DTYPE_TO_TMA
:346

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._tensor_map import _NUMPY_DTYPE_TO_TMA
:354

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_texture_surface.py· 6
C
cuda-python low-level driver API
import cuda.core
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture import (
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.texture._texture import _normalize_address_modes
:238

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.bindings import driver
:406

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_typing_imports.py· 4
C
cuda-python low-level driver API
from cuda.core.typing import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memory._buffer import DevicePointerType as _DevicePointerT
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._stream import IsStreamType as _IsStreamT
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.typing import (
:25

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_utils_dlpack.py· 2
C
cuda-python low-level driver API
from cuda.core._dlpack import DLDeviceType
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_utils_enum_explanations_helpers.py· 4
C
cuda-python low-level driver API
from cuda.core._utils import enum_explanations_helpers
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._utils.enum_explanations_helpers import (
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._utils.driver_cu_result_explanations as driver_explanations
:150

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
import cuda.core._utils.runtime_cuda_error_explanations as runtime_explanations
:162

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_core/tests/test_utils.py· 21
B
CuPy dependency
import cupy as cp
:10

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
import cuda.core
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core import Device
:31

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._dlpack import DLDeviceType
:32

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._layout import _StridedLayout
:33

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core.utils import StridedMemoryView, args_viewable_as_strided_memory
:34

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
pytest.param(torch.arange(12, dtype=torch.float32, device="cuda"), True, id="torch-1d"),
:208

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
torch.arange(24, dtype=torch.float32, device="cuda").reshape(2, 3, 4),
:210

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
pytest.param(torch.tensor(42.0, dtype=torch.float32, device="cuda"), False, id="torch-scalar"),
:214

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
pytest.param(torch.empty(0, dtype=torch.float32, device="cuda"), False, id="torch-empty"),
:215

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
torch.arange(12, dtype=torch.float32, device="cuda").reshape(3, 4).t(),
:217

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
torch.arange(100, dtype=torch.int64, device="cuda")[10:20],
:222

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
torch.arange(60, dtype=torch.float32, device="cuda").reshape(6, 10)[1:4, 2:7],
:227

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
a = torch.tensor([1, 2, 3, 4, 5, 6], dtype=torch.bfloat16, device="cuda").reshape(2, 3)[slices]
:765

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
from cuda.core._memoryview import view_as_cai
:950

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memoryview import view_as_cai
:959

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memoryview import view_as_cai
:980

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memoryview import view_as_array_interface
:989

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.core._memoryview import view_as_array_interface
:999

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
a = torch.tensor([1, 0, 1], dtype=torch_dtype, device="cuda")
:1035

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
from cuda.core._memoryview import _StridedMemoryViewProxy
:1066

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/__init__.py· 24
C
cuda-python low-level driver API
from cuda.pathfinder._binaries.find_nvidia_binary_utility import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._binaries.supported_nvidia_binaries import SUPPORTED_BINARIES as _SUPPORTED_BINARIES
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import (
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError as DynamicLibNotFoundError
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import (
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import LoadedDL as LoadedDL
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import load_nvidia_dynamic_lib as load_nvidia_dynamic_lib
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.supported_nvidia_libs import (
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.find_nvidia_headers import LocatedHeaderDir as LocatedHeaderDir
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.find_nvidia_headers import find_nvidia_header_directory as find_nvidia_header_directory
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.find_nvidia_headers import (
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.supported_nvidia_headers import SUPPORTED_HEADERS_CTK as _SUPPORTED_HEADERS_CTK
:31

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:32

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:35

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:41

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:44

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:47

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:50

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:53

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:56

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:59

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home as get_cuda_path_or_home
:62

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._version import __version__ # isort: skip
:64

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_binaries/find_nvidia_binary_utility.py· 4
C
cuda-python low-level driver API
from cuda.pathfinder._binaries import supported_nvidia_binaries
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_all_sitepackages
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/dynamic_lib_subprocess.py· 8
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError, LoadedDL
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.platform_loader import LOADER
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import (
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import load_nvidia_dynamic_lib
:55

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import _load_lib_no_cache
:56

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.supported_nvidia_libs import (
:57

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:61

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/lib_descriptor.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.descriptor_catalog import (
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/load_dl_common.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/load_dl_linux.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import LoadedDL
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/load_dl_windows.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import LoadedDL
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/load_nvidia_dynamic_lib.py· 7
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import (
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.platform_loader import LOADER
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import (
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import (
:29

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:37

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:40

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/platform_loader.py· 5
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import LoadedDL
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_dl_windows as _impl
:35

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_dl_linux as _impl
:37

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/search_platform.py· 4
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:19

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.supported_nvidia_libs import is_suppressed_dll_file
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_all_sitepackages
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/search_steps.py· 4
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LibDescriptor
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError
:29

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_platform import PLATFORM, SearchPlatform
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:31

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/supported_nvidia_libs.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.descriptor_catalog import DESCRIPTOR_CATALOG
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_headers/find_nvidia_headers.py· 6
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import derive_ctk_root
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.header_descriptor import (
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_all_sitepackages
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.header_descriptor import HeaderDescriptor
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_headers/header_descriptor.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._headers.header_descriptor_catalog import (
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_headers/supported_nvidia_headers.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._headers.header_descriptor_catalog import HEADER_DESCRIPTOR_CATALOG
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_optional_cuda_import.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_static_libs/find_bitcode_lib.py· 3
C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_all_sitepackages
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_static_libs/find_static_lib.py· 3
C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_all_sitepackages
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/cuda/pathfinder/_utils/driver_info.py· 3
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
loaded_cuda = _load_nvidia_dynamic_lib("cuda")
:65

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_pathfinder/tests/child_load_nvidia_dynamic_lib_helper.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/conftest.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_nvidia_dynamic_lib as load_nvidia_dynamic_lib_module
:39

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_ctk_root_discovery.py· 8
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_nvidia_dynamic_lib as load_mod
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import search_steps as steps_mod
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError, LoadedDL
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import (
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import (
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:33

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_descriptor_catalog.py· 1
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.descriptor_catalog import DESCRIPTOR_CATALOG, DescriptorSpec
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_driver_lib_loading.py· 8
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS
:20

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError, LoadedDL
:21

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import STATUS_NOT_FOUND, parse_dynamic_lib_subprocess_payload
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils import driver_info
:28

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS, quote_for_shell
:29

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
_CUDA_DESC = LIB_DESCRIPTORS["cuda"]
:37

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import LOADER
:58

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_find_bitcode_lib.py· 3
C
cuda-python low-level driver API
import cuda.pathfinder._static_libs.find_bitcode_lib as find_bitcode_lib_module
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_bitcode_lib import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_find_nvidia_binaries.py· 7
C
cuda-python low-level driver API
from cuda.pathfinder import find_nvidia_binary_utility
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._binaries import find_nvidia_binary_utility as binary_finder_module
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._binaries.find_nvidia_binary_utility import UnsupportedBinaryError
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._binaries.supported_nvidia_binaries import (
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
cuda_home = os.path.join(os.sep, "cuda")
:46

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
cuda_home = os.path.join(os.sep, "cuda")
:80

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
cuda_home = os.path.join(os.sep, "cuda")
:138

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_pathfinder/tests/test_find_nvidia_headers.py· 6
C
cuda-python low-level driver API
import cuda.pathfinder._headers.find_nvidia_headers as find_nvidia_headers_module
:24

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import LocatedHeaderDir, find_nvidia_header_directory, locate_nvidia_header_directory
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:27

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._headers.supported_nvidia_headers import (
:30

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:38

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:39

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_find_static_lib.py· 4
C
cuda-python low-level driver API
import cuda.pathfinder._static_libs.find_static_lib as find_static_lib_module
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._static_libs.find_static_lib import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import get_cuda_path_or_home
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import quote_for_shell
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_lib_descriptor.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.supported_nvidia_libs import (
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_load_nvidia_dynamic_lib_subprocess.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import dynamic_lib_subprocess as subprocess_mod
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import (
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_load_nvidia_dynamic_lib_using_mocker.py· 8
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_nvidia_dynamic_lib as load_mod
:6

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import search_steps as steps_mod
:7

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError, LoadedDL
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_nvidia_dynamic_lib import (
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import EARLY_FIND_STEPS
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import find_in_conda
:74

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import find_in_conda
:153

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py· 5
C
cuda-python low-level driver API
from cuda.pathfinder import DynamicLibNotAvailableError, DynamicLibUnknownError, load_nvidia_dynamic_lib
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import load_nvidia_dynamic_lib as load_nvidia_dynamic_lib_module
:16

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs import supported_nvidia_libs
:17

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.subprocess_protocol import (
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.platform_aware import IS_WINDOWS, quote_for_shell
:22

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_optional_cuda_import.py· 3
C
cuda-python low-level driver API
import cuda.pathfinder._optional_cuda_import as optional_import_mod
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder import DynamicLibNotFoundError
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._optional_cuda_import import _optional_cuda_import
:10

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_search_steps.py· 4
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.lib_descriptor import LIB_DESCRIPTORS, LibDescriptor
:12

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import DynamicLibNotFoundError
:13

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_platform import LinuxSearchPlatform, WindowsSearchPlatform
:14

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.search_steps import (
:15

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_utils_driver_info.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.load_dl_common import LoadedDL
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils import driver_info
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_pathfinder/tests/test_utils_env_vars.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._utils.env_vars import (
:11

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
d = tmp_path / "cuda"
:79

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_pathfinder/tests/test_utils_find_sub_dirs.py· 4
C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import (
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_cached
:140

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_cached
:196

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

C
cuda-python low-level driver API
from cuda.pathfinder._utils.find_sub_dirs import find_sub_dirs_cached
:251

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_python_test_helpers/cuda_python_test_helpers/nvvm_bitcode.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvvm
:9

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

toolshed/_catalog_writer.py· 2
C
cuda-python low-level driver API
from cuda.pathfinder._dynamic_libs.descriptor_catalog import ( # noqa: E402
:26

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

A
Device string "cuda"
CATALOG_PATH = _PATHFINDER_ROOT / "cuda" / "pathfinder" / "_dynamic_libs" / "descriptor_catalog.py"
:31

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

toolshed/build_static_bitcode_input.py· 1
C
cuda-python low-level driver API
from cuda.bindings import nvvm
:23

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

toolshed/dump_cutile_b64.py· 2
B
CuPy dependency
import cupy
:16

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

C
cuda-python low-level driver API
import cuda.tile as ct
:18

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

toolshed/run_cuda_pathfinder.py· 1
C
cuda-python low-level driver API
from cuda import pathfinder
:8

cuda-python (from cuda import cudart/cuda) wraps the CUDA driver directly. There is no drop-in ROCm binding; the driver calls must be rewritten against HIP.

RecommendRewrite driver-API calls against the HIP runtime.

cuda_bindings/pyproject.toml· 1
B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cudla==13.*; platform_system == 'Linux' and platform_machine == 'aarch64'",
:43

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

cuda_core/pyproject.toml· 2
B
CuPy dependency
test-cu12 = [ {include-group = "ml-dtypes" }, {include-group = "test" }, "cupy-cuda12x; python_version < '3.14'", "cuda-toolkit[cudart]==12.*"]
:76

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

B
CuPy dependency
test-cu13 = [ {include-group = "ml-dtypes" }, {include-group = "test" }, "cupy-cuda13x; python_version < '3.14'", "cuda-toolkit[cudart]==13.*"]
:77

CuPy targets CUDA directly. AMD ships a ROCm build of CuPy, so this is a mechanical swap rather than a rewrite.

RecommendInstall the ROCm CuPy build (cupy-rocm-6-0) instead of cupy-cudaXXx.

cuda_core/tests/test_binaries/saxpy.cu· 1
B
Custom CUDA kernels (.cu/.cuh)
CUDA kernel source
file

Hand-written .cu/.cuh kernels compile only with nvcc. AMD's HIPify tools translate the vast majority of CUDA C++ to HIP automatically.

RecommendRun hipify-perl over the .cu/.cuh sources and build with hipcc.

cuda_pathfinder/pyproject.toml· 18
B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cublasmp-cu12; sys_platform != 'win32'",
:25

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cudss-cu12",
:26

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cufftmp-cu12; sys_platform != 'win32'",
:27

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cusolvermp-cu12; sys_platform != 'win32'",
:28

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cusparselt-cu12",
:29

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-libmathdx-cu12",
:30

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-nccl-cu12; sys_platform != 'win32'",
:31

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-nvshmem-cu12<3.7; sys_platform != 'win32'",
:32

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cublasmp-cu13; sys_platform != 'win32'",
:38

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cudla; platform_system == 'Linux' and platform_machine == 'aarch64'",
:39

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cudss-cu13",
:40

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cufftmp-cu13; sys_platform != 'win32'",
:41

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cusolvermp-cu13; sys_platform != 'win32'",
:42

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cusparselt-cu13",
:43

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-libmathdx-cu13",
:44

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-nccl-cu13; sys_platform != 'win32'",
:45

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-nvshmem-cu13<3.7; sys_platform != 'win32'",
:46

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

B
Pinned nvidia-* / +cuXXX wheel
"nvidia-cutlass; python_version < '3.15'",
:50

A CUDA-pinned wheel (nvidia-* runtime package or a +cuXXX local version) forces the CUDA toolchain. Repin to the ROCm index.

RecommendRemove nvidia-* pins and install from the ROCm PyTorch index (--index-url .../whl/rocm6.1).

benchmarks/cuda_bindings/tests/test_runner.py· 2
A
Device string "cuda"
cuda_module = types.ModuleType("cuda")
:88

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
monkeypatch.setitem(sys.modules, "cuda", cuda_module)
:95

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

ci/tools/merge_cuda_core_wheels.py· 1
A
Device string "cuda"
base_dir = Path("cuda") / "core"
:135

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_core/setup.py· 1
A
Device string "cuda"
_AOTI_SHIM_DEF_FILE = _ROOT_DIR / "cuda" / "core" / "_include" / "aoti_shim.def"
:16

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

cuda_pathfinder/cuda/pathfinder/_dynamic_libs/descriptor_catalog.py· 1
A
Device string "cuda"
name="cuda",
:400

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.