Add 2.8 release notes worksheet for Intel GPU (XPU). #12

etaf · 2025-07-08T01:13:26Z

No description provided.

facebook-github-bot · 2025-07-08T01:13:32Z

Thank you for your pull request.

We require contributors to sign our Contributor License Agreement, and yours needs attention.

You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

2.8.0/done/result_xpu.md

EikanWang · 2025-07-08T01:43:34Z

2.8.0/done/result_xpu.md

+- Support int4 WOQ GEMM on Intel GPU ([#137566](https://github.com/pytorch/pytorch/pull/137566))
+- Support third_party SYCL kernels with CPP Extension API ([#132945](https://github.com/pytorch/pytorch/pull/132945))
+- Support safe softmax for SDPA on Intel GPU ([#151999](https://github.com/pytorch/pytorch/pull/151999))
+- SupportGQA and different head_dim of value for SDPA on Intel GPU([#150992](https://github.com/pytorch/pytorch/pull/150992))


It should be an improvement rather than a new feature.

EikanWang · 2025-07-08T01:44:03Z

2.8.0/done/result_xpu.md

+- Support Intel distributed backend (XCCL) in PyTorch ([#141856](https://github.com/pytorch/pytorch/pull/141856))
+- Support int4 WOQ GEMM on Intel GPU ([#137566](https://github.com/pytorch/pytorch/pull/137566))
+- Support third_party SYCL kernels with CPP Extension API ([#132945](https://github.com/pytorch/pytorch/pull/132945))
+- Support safe softmax for SDPA on Intel GPU ([#151999](https://github.com/pytorch/pytorch/pull/151999))


It should be an improvement rather than a new feature.

EikanWang · 2025-07-08T01:45:08Z

2.8.0/done/result_xpu.md

+- Improve scalar tensor case handling in addmm, baddmm on Intel GPU ([#153051](https://github.com/pytorch/pytorch/pull/153051))
+- Support f32 intermediate dtype, headdim <= 576, and f32 causal mask for SDPA ([#152091](https://github.com/pytorch/pytorch/pull/152091))
+- Add Intel GPU device support for nested_layer_norm ([#148593](https://github.com/pytorch/pytorch/pull/148593))
+- Refine oneDNN context manager API on Intel GPU ([#147349](https://github.com/pytorch/pytorch/pull/147349))


Let's remove it.

2.8.0/done/result_xpu.md

EikanWang · 2025-07-08T01:57:23Z

2.8.0/done/result_xpu.md

+
+### bug fixes
+- Fix matmul accuracy when offset > 0 ([#154495](https://github.com/pytorch/pytorch/pull/154495))
+- Correct torch.xpu.is_bf16_supported to return False if no Intel GPU detected ([#152317](https://github.com/pytorch/pytorch/pull/152317))


Suggested change

- Correct torch.xpu.is_bf16_supported to return False if no Intel GPU detected ([#152317](https://github.com/pytorch/pytorch/pull/152317))

- Fix the issue that `torch.xpu.is_bf16_supported` always returns `True` even if Intel GPU is not available ([#152317](https://github.com/pytorch/pytorch/pull/152317))

2.8.0/done/result_xpu.md

EikanWang · 2025-07-08T02:00:30Z

2.8.0/done/result_xpu.md

+- Add Intel GPU device support for nested_layer_norm ([#148593](https://github.com/pytorch/pytorch/pull/148593))
+- Refine oneDNN context manager API on Intel GPU ([#147349](https://github.com/pytorch/pytorch/pull/147349))
+- Improve error handling and reporting in Intel GPU CMake files ([#149353](https://github.com/pytorch/pytorch/pull/149353))
+- Support multi_arch_kernel_binary option in AOTInductor for Intel GPU ([#154514](https://github.com/pytorch/pytorch/pull/154514))


Pls. consolidate AOTI-related improvements into a single bullet.

EikanWang · 2025-07-08T02:01:05Z

2.8.0/done/result_xpu.md

+- Support Intel GPU profiler toggle functionality ([#155135](https://github.com/pytorch/pytorch/pull/155135))
+- Support distributed memory tracker integration for Intel GPU ([#150703](https://github.com/pytorch/pytorch/pull/150703))
+- Improve scalar tensor case handling in addmm, baddmm on Intel GPU ([#153051](https://github.com/pytorch/pytorch/pull/153051))
+- Support f32 intermediate dtype, headdim <= 576, and f32 causal mask for SDPA ([#152091](https://github.com/pytorch/pytorch/pull/152091))


Pls. consolidate SDPA-related improvement into a single bullet.

Co-authored-by: Eikan Wang <[email protected]>

EikanWang · 2025-07-08T03:36:28Z

2.8.0/done/result_xpu.md

+- Add memory reporting to Memory Profiler for Intel GPU ([#152842](https://github.com/pytorch/pytorch/pull/152842))
+- Support Intel GPU profiler toggle functionality ([#155135](https://github.com/pytorch/pytorch/pull/155135))
+- Support distributed memory tracker integration for Intel GPU ([#150703](https://github.com/pytorch/pytorch/pull/150703))
+- Improve scalar tensor case handling in addmm, baddmm on Intel GPU ([#153051](https://github.com/pytorch/pytorch/pull/153051))


Let's move this bullet to performance section. Because it reduces the oneDNN integration overhead.

riverliuintel

Only one minor comments and looks good.

riverliuintel · 2025-07-08T05:13:36Z

2.8.0/done/result_xpu.md

+- Add Intel GPU device support for nested_layer_norm ([#148593](https://github.com/pytorch/pytorch/pull/148593))
+- Refine oneDNN context manager API on Intel GPU ([#147349](https://github.com/pytorch/pytorch/pull/147349))
+- Improve error handling and reporting in Intel GPU CMake files ([#149353](https://github.com/pytorch/pytorch/pull/149353))
+- Support multi_arch_kernel_binary option in AOTInductor for Intel GPU ([#154514](https://github.com/pytorch/pytorch/pull/154514))


if code, use single quotes 'multi_arch_kernel_binary'

riverliuintel · 2025-07-08T23:12:48Z

@jbschlosser please help to comment and merge it. Thanks.

Add release notes worksheet for Intel GPU (XPU).

39b6b7d

etaf force-pushed the intel_gpu branch from 1390abb to 39b6b7d Compare July 8, 2025 01:18

EikanWang reviewed Jul 8, 2025

View reviewed changes

Update 2.8.0/done/result_xpu.md

d49187a

Co-authored-by: Eikan Wang <[email protected]>

facebook-github-bot added the CLA Signed This label is managed by the Meta Open Source bot. label Jul 8, 2025

etaf and others added 4 commits July 8, 2025 10:36

Update 2.8.0/done/result_xpu.md

4bb15c3

Co-authored-by: Eikan Wang <[email protected]>

Update 2.8.0/done/result_xpu.md

f9bd104

Co-authored-by: Eikan Wang <[email protected]>

Apply suggestions from code review

cc7187c

Co-authored-by: Eikan Wang <[email protected]>

Apply suggestions from code review

d47e82b

etaf force-pushed the intel_gpu branch from b770fd7 to d47e82b Compare July 8, 2025 03:05

etaf requested a review from EikanWang July 8, 2025 03:06

EikanWang approved these changes Jul 8, 2025

View reviewed changes

riverliuintel approved these changes Jul 8, 2025

View reviewed changes

Update

fdda7b1

etaf marked this pull request as ready for review July 8, 2025 05:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add 2.8 release notes worksheet for Intel GPU (XPU). #12

Add 2.8 release notes worksheet for Intel GPU (XPU). #12

etaf commented Jul 8, 2025

Uh oh!

facebook-github-bot commented Jul 8, 2025

Uh oh!

Uh oh!

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

EikanWang Jul 8, 2025

Uh oh!

riverliuintel left a comment

Uh oh!

riverliuintel Jul 8, 2025

Uh oh!

riverliuintel commented Jul 8, 2025

Uh oh!

Uh oh!

	- Correct torch.xpu.is_bf16_supported to return False if no Intel GPU detected ([#152317](https://github.com/pytorch/pytorch/pull/152317))
	- Fix the issue that `torch.xpu.is_bf16_supported` always returns `True` even if Intel GPU is not available ([#152317](https://github.com/pytorch/pytorch/pull/152317))

Add 2.8 release notes worksheet for Intel GPU (XPU). #12

Are you sure you want to change the base?

Add 2.8 release notes worksheet for Intel GPU (XPU). #12

Conversation

etaf commented Jul 8, 2025

Uh oh!

facebook-github-bot commented Jul 8, 2025

Process

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

riverliuintel left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

riverliuintel commented Jul 8, 2025

Uh oh!

Uh oh!