Enable dpnp build on AMD GPU #2302

vlad-perevezentsev · 2025-02-10T15:24:54Z

This PR updates СMakeLists files and build_locally.py to enable building dpnp for AMD targets.

To build dpnp on AMD:

python scripts/build_locally.py --target-hip=gfx90a

To find the architecture, use

rocminfo | grep 'Name: *gfx.*'

Have you provided a meaningful PR description?
Have you added a test, reproducer or referred to issue with a reproducer?
Have you tested your changes locally for CPU and GPU devices?
Have you made sure that new changes do not introduce compiler warnings?
Have you checked performance impact of proposed changes?
If this PR is a work in progress, are you filing the PR as a draft?

github-actions · 2025-02-10T16:12:02Z

Array API standard conformance tests for dpnp=0.18.0dev1=py312he4f9c94_23 ran successfully.
Passed: 1222
Failed: 0
Skipped: 9

github-actions · 2025-02-10T17:15:04Z

View rendered docs @ https://intelpython.github.io/dpnp/pull/2302/index.html

dpnp/backend/extensions/indexing/CMakeLists.txt

antonwolfy · 2025-02-17T19:10:02Z

scripts/build_locally.py

+        if not arch:
+            raise ValueError("--arch is required when --target=hip")
+        cmake_args += [
+            "-DDPNP_TARGET_HIP=ON",


For what do we need to define two variables? Can it be combined in a single one, like in dpctl: -DDPNP_TARGET_HIP={arch}?

Additionally, --target=cuda is current dpnp approach, but:

dpctl and dpnp should consider supporting targeting specific CUDA architectures

--target=hip means that there is no way to build simultaneously for HIP and CUDA (which is very, very much an edge case, but should be considered)

For these reasons, I think it is most sensible to move away from --target= universal approach to --target-cuda= and --target-hip= or something to that effect

@ndgrigorian it is a great suggestion.
I have added support for --target-hip and I am going to add --target-cuda instead of --target in the next PR.
Thanks

scripts/build_locally.py

coveralls · 2025-03-18T11:28:02Z

coverage: 72.271%. remained the same
when pulling 5e2cc3d on enable_amd_build
into 2966ae6 on master.

antonwolfy · 2025-03-31T13:26:17Z

doc/quick_start_guide.rst

+
+.. code-block:: bash
+
+    python scripts/build_locally.py --target-hip=gfx90a


In general it might be unclear what gfx90a means here. It'd be great to clarify.

antonwolfy · 2025-03-31T13:34:50Z

scripts/build_locally.py

@@ -104,6 +105,15 @@ def run(
        # Always builds using oneMKL interfaces for the cuda target
        onemkl_interfaces = True

+    if target_hip is not None:
+        if target_hip == "default":


We need a special handling for python scripts/build_locally.py --target-hip=.
Now it is equal to python scripts/build_locally.py, which was not intended, I guess.

The same comment is applicable to python scripts/build_locally.py --target=.

antonwolfy · 2025-03-31T13:36:51Z

scripts/build_locally.py

@@ -104,6 +105,15 @@ def run(
        # Always builds using oneMKL interfaces for the cuda target
        onemkl_interfaces = True

+    if target_hip is not None:
+        if target_hip == "default":


It is a bit unclear what the use case assumed here? Is it about python scripts/build_locally.py --target-hip="default" only?
Then I believe the error message below needs to be rephrase a bit to something like No default HIP architecture is supported. It must be specified explicitly.

I have changed the logic here by removing the check for default

antonwolfy · 2025-03-31T13:41:21Z

CMakeLists.txt

@@ -75,27 +75,64 @@ option(DPNP_USE_ONEMKL_INTERFACES
    "Build DPNP with oneMKL Interfaces"
    OFF
 )
+set(HIP_TARGETS "" CACHE STRING "HIP architecture for target")


I assume there is no support for multiple values:

Suggested change

set(HIP_TARGETS "" CACHE STRING "HIP architecture for target")

set(HIP_TARGET "" CACHE STRING "HIP architecture for target")

At some point, it was clear in docs that only one architecture was supported at a time, but now it isn't as clear and should be tested

Also, there is new information in the extension guide

The compiler driver also offers alias targets for each target+architecture pair to make the command line shorter and easier to understand for humans. Thanks to the aliases, the -Xsycl-target-backend flags no longer need to be specified.

It shows that the command

icpx -fsycl -fsycl-targets=spir64_gen,amdgcn-amd-amdhsa,nvptx64-nvidia-cuda \ -Xsycl-target-backend=spir64_gen '-device pvc' \ -Xsycl-target-backend=amdgcn-amd-amdhsa --offload-arch=gfx1030 \ -Xsycl-target-backend=nvptx64-nvidia-cuda --offload-arch=sm_80 \ -o sycl-app sycl-app.cpp

is equivalent to

icpx -fsycl -fsycl-targets=intel_gpu_pvc,amd_gpu_gfx1030,nvidia_gpu_sm_80 \ -o sycl-app sycl-app.cpp

so maybe both dpctl and dpnp can simplify by removing the need for -Xsycl-target-backend=amdgcn-amd-amdhsa --offload-arch=[X] completely

list of aliases:
https://intel.github.io/llvm/UsersManual.html

Aliases list seems to claim only one alias is supported at a time. So probably only one architecture at once is possible? That would be my guess

antonwolfy · 2025-03-31T13:47:49Z

CMakeLists.txt

-   set(_dpnp_sycl_targets ${DPNP_SYCL_TARGETS})
+    set(_dpnp_sycl_targets ${DPNP_SYCL_TARGETS})
+
+    if (NOT "x${HIP_TARGETS}" STREQUAL "x")


Why is that applicable only to HIP target? What is the use case? Should it be supported for CUDA target also?

antonwolfy · 2025-03-31T13:49:50Z

CMakeLists.txt

+
+    if (NOT "x${HIP_TARGETS}" STREQUAL "x")
+        set(_dpnp_amd_targets ${HIP_TARGETS})
+        set(_use_onemkl_interfaces_hip ON)


Do we need here something similar to above?

set(_dpnp_sycl_targets "amdgcn-amd-amdhsa,${_dpnp_sycl_targets}")

I think if we set DPNP_SYCL_TARGETS via --cmake_opts we expect them to be the right target e.g. amdgcn-amd-amdhsa or nvptx64-nvidia-cuda

vlad-perevezentsev added 3 commits February 10, 2025 06:53

Enable CMake options to build dpnp on AMD

72bc4d4

Add build_locally args for AMD build

5f11917

Remove unused lines

c07e0a7

vlad-perevezentsev self-assigned this Feb 10, 2025

vlad-perevezentsev requested review from antonwolfy, AlexanderKalistratov and vtavana as code owners February 10, 2025 15:24

vlad-perevezentsev added 3 commits February 11, 2025 04:58

Remove ROCM_PATH logic

323bbb4

Support amd build for indexing extension

e111ce1

Merge master into enable_amd_build

ccc7b72

antonwolfy reviewed Feb 17, 2025

View reviewed changes

dpnp/backend/extensions/indexing/CMakeLists.txt Show resolved Hide resolved

antonwolfy reviewed Feb 17, 2025

View reviewed changes

scripts/build_locally.py Outdated Show resolved Hide resolved

antonwolfy added this to the 0.18.0 release milestone Feb 26, 2025

vlad-perevezentsev added 7 commits March 14, 2025 04:22

Merge master into enable_amd_build

efbab02

Support amd build for window extension

310cd82

Set HIP specific flags for MKL

c3adf4e

pdate logic to use --target-hip

574ea90

Merge master into enable_amd_build

d6c5925

Add docs for dpnp build on AMD

5bca529

Remove unnecessary HIP_TARGETS validation in CMake

b858ae2

A small docs update

273113e

antonwolfy reviewed Mar 31, 2025

View reviewed changes

vlad-perevezentsev added 4 commits April 16, 2025 03:31

Improve validation of --target and --target-hip

c4da3ef

Clarify --target-hip usage in doc

e6c280e

Update SYCL target selection logic in CMakeLists

b27a8a1

Merge master into enable_amd_build

2238372

Avoid false HIP error when building for default target

5e2cc3d

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Enable dpnp build on AMD GPU #2302

Enable dpnp build on AMD GPU #2302

vlad-perevezentsev commented Feb 10, 2025 •

edited

Loading

github-actions bot commented Feb 10, 2025 •

edited

Loading

github-actions bot commented Feb 10, 2025

antonwolfy Feb 17, 2025 •

edited

Loading

ndgrigorian Feb 21, 2025 •

edited

Loading

vlad-perevezentsev Mar 18, 2025

coveralls commented Mar 18, 2025 •

edited

Loading

antonwolfy Mar 31, 2025

vlad-perevezentsev Apr 16, 2025

antonwolfy Mar 31, 2025

antonwolfy Mar 31, 2025

antonwolfy Mar 31, 2025

vlad-perevezentsev Apr 16, 2025

antonwolfy Mar 31, 2025

ndgrigorian Apr 10, 2025 •

edited

Loading

ndgrigorian Apr 10, 2025

antonwolfy Mar 31, 2025

vlad-perevezentsev Apr 16, 2025

antonwolfy Mar 31, 2025

vlad-perevezentsev Apr 16, 2025


		.. code-block:: bash

		python scripts/build_locally.py --target-hip=gfx90a

	set(HIP_TARGETS "" CACHE STRING "HIP architecture for target")
	set(HIP_TARGET "" CACHE STRING "HIP architecture for target")

Enable dpnp build on AMD GPU #2302

Are you sure you want to change the base?

Enable dpnp build on AMD GPU #2302

Conversation

vlad-perevezentsev commented Feb 10, 2025 • edited Loading

github-actions bot commented Feb 10, 2025 • edited Loading

github-actions bot commented Feb 10, 2025

antonwolfy Feb 17, 2025 • edited Loading

Choose a reason for hiding this comment

ndgrigorian Feb 21, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coveralls commented Mar 18, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ndgrigorian Apr 10, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vlad-perevezentsev commented Feb 10, 2025 •

edited

Loading

github-actions bot commented Feb 10, 2025 •

edited

Loading

antonwolfy Feb 17, 2025 •

edited

Loading

ndgrigorian Feb 21, 2025 •

edited

Loading

coveralls commented Mar 18, 2025 •

edited

Loading

ndgrigorian Apr 10, 2025 •

edited

Loading