[SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e #8782

aelovikov-intel · 2023-03-27T15:42:24Z

This PR was implemented using the following steps:

In intel/llvm: remove current limited subset of end-to-end tests in sycl/test-e2e/ directory
Create a fresh clone of intel/llvm-test-suite and filter it using https://github.com/newren/git-filter-repo with the following command:

git-filter-repo --path SYCL  --message-callback 'return re.sub(b"\\(#([0-9]+)\\)", b"(intel/llvm-test-suite#\\1)", message)'

In the filtered repo, perform

$ git mv SYCL sycl/test-e2e && git commit -m "Move SYCL -> sycl/test-e2e"

Return back to intel/llvm, add filtered repo as a remote and merge it in. It resulted in two conflicts in CMakeLists.txt and lit.site.cfg.py.in. The resolution is to keep intel/llvm version with an additional change of uncommenting previously skipped add_subdirectory(External) and add_subdirectory(ExtraTests) in CMakeLists.txt.
Update sycl/test-e2e/README.md

…llvm-test-suite#1359) Signed-off-by: Konstantin S Bobrovsky <konstantin.s.bobrovsky@intel.com>

* add tests for a bunch of fp16 utils * add check for platform without fp16 support Signed-off-by: jinge90 <ge.jin@intel.com>

While extending the original test did some refactoring: 1. Reduced types it tests 2. Test only 1 and 2 dimensions

…test-suite#1362)

…vm-test-suite#1386)

…est-suite#1396) CUDA and HIP BE do not support intel::reqd_sub_group_size, so it's not possible to control sub-group size which is needed to get the size of temporary memory for sort algorithm.

…est-suite#1387) intel#7401 fixes host-side minimal subnormal half values, which in turn fixes generate_ref_conv_data in value_conv.hpp. This causes ctor_broadcast_fp_extra to unexpectedly pass. This commit enables ctor_broadcast_fp_extra. Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

…llvm-test-suite#1314) Related compiler changes: intel#6989 and intel#7419

[SYCL] Fix type Signed-off-by: Rauf, Rana <rana.rauf@intel.com>

…ite#1358)

…lvm-test-suite#1403) Complementary to intel#7478.

…lvm-test-suite#1400) Copy-pasted from reduction_nd_N_vars.cpp and adjusted to use sycl::range.

…llvm-test-suite#1408)

Also added a try catch block to print possible "synchronous" exceptions.

…llvm-test-suite#1409)

…ns-env.cpp (intel/llvm-test-suite#1412) While the investigation is in progress.

…suite#1373) That way we can bypass strategy selection and test any of them with the same type/operation.

…uite#1401) ctor_broadcast_fp_extra uses fp16 and fp64 but does not check for the requirement before running the kernels on the device. This commit adds the requirement check to ctor_broadcast. Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

…ite#1404) Two simple tests to check that code using the kernel fusion extension API compiles correctly. The tests currently do not yet execute the compiled application, as the necessary functionality will only be added to the implementation in a later PR. Spec: intel#7098 Implementation: intel#7416 Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

…ad.cpp (intel/llvm-test-suite#1418)

…ntal status. (intel/llvm-test-suite#1129) Tests changes for intel#6524 Signed-off-by: Rajiv Deodhar <rajiv.deodhar@intel.com> Co-authored-by: JackAKirk <jack.kirk@codeplay.com>

…l/llvm-test-suite#1398) * [ESIMD] Separate ieee sqrt/div functions from ext_math.cpp test This patch: - separates the ieee_sqrt() and ieee_div() test cases to a new test, which is not started on Gen12 because it doesn't support those operations even for fp32. - enables ext_math.cpp on Windows. Signed-off-by: Vyacheslav N Klochkov <vyacheslav.n.klochkov@intel.com>

…vm-test-suite#1422) intel#6524 moved bfloat16 out of the experimental namespace. This commit removes the last remaining uses of the experimental namespace in bfloat16 for ESIMD and matrix tests. Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

…on (intel/llvm-test-suite#1365) This test is expected to throw an exception so we need to disable the ZE_DEBUG test. Signed-off-by: Byoungro So byoungro.so@intel.com

…est-suite#1423) * [SYCL][CUDA] Add SM version check to bfloat16 CUDA test bfloat16 requires SM80 on the CUDA backend. This commit changes CUDA tests to do an early exit if that requirement is not met. * Remove CUDA specific test and try run SM80 Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

…ty (intel/llvm-test-suite#1414)

* [SYCL] Fix CUDA tests using bfloat16 * Add missing using in element_wise_wi_marray_legacy Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

…suite#1424) See intel#7330

This commit should fix the is_compatible test for non-SPIRV targets. Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com> Co-authored-by: Vyacheslav Klochkov <vyacheslav.n.klochkov@intel.com>

…m-test-suite#1663)

…ite#1671)

…st-suite#1695) Revert "[SYCL] Refactor invalid use of local accessor (intel/llvm-test-suite#1646)" This reverts commit c257c0b.

…te#1679) Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com>

…-test-suite#1685) This patch updates the E2E tests to account for the changes made to aspects in [Section 4.6.4.3](https://registry.khronos.org/SYCL/specs/sycl-2020/html/sycl-2020.html#sec:device-aspects) of the specification: - The concept of a "host" device no longer exists in SYCL 2020 - any `host` should now be a `cpu`, so the aspect has been deprecated accordingly. - `usm_restricted_shared_allocations` has been removed. - `usm_system_allocator` has been renamed `usm_system_allocations`. - `int64_*` aspects were replaced with `atomic64`, so no longer should be tested.

We're doing the final move and it makes sense to make a "clean" merge (except that CMake/LIT config stuff that had some customizations).

…t-suite#1687)

Signed-off-by: Vyacheslav N Klochkov <vyacheslav.n.klochkov@intel.com>

kbobrovs and others added 30 commits November 14, 2022 11:01

[ESIMD] Allow small epsilon in FP division result comparison. (intel/…

1c29bbb

…llvm-test-suite#1359) Signed-off-by: Konstantin S Bobrovsky <konstantin.s.bobrovsky@intel.com>

Add tests for trivial fp16 util functions. (intel/llvm-test-suite#1250)

9706152

* add tests for a bunch of fp16 utils * add check for platform without fp16 support Signed-off-by: jinge90 <ge.jin@intel.com>

[SYCL] Add test for sorting over sub-group (intel/llvm-test-suite#1380)

e451c08

While extending the original test did some refactoring: 1. Reduced types it tests 2. Test only 1 and 2 dimensions

[SYCL] Add end-to-end tests for queue shortcut functions (intel/llvm-…

4165920

…test-suite#1362)

[SYCL] Add test for memory clock rate and bus width queries (intel/ll…

277942d

…vm-test-suite#1386)

[SYCL] Disable sort over sub-group for CUDA and HIP BEs (intel/llvm-t…

a8f030e

…est-suite#1396) CUDA and HIP BE do not support intel::reqd_sub_group_size, so it's not possible to control sub-group size which is needed to get the size of temporary memory for sort algorithm.

[SYCL] Test exception throwing for unsupported kernel feature (intel/…

204ea3e

…llvm-test-suite#1314) Related compiler changes: intel#6989 and intel#7419

[SYCL] Fix typo (intel/llvm-test-suite#1399)

93b4cf1

[SYCL] Fix type Signed-off-by: Rauf, Rana <rana.rauf@intel.com>

[SYCL] Add a new query test that uses the new API (intel/llvm-test-su…

f12037b

…ite#1358)

[SYCL] Add a test with reduction kernel accepting sycl::item (intel/l…

3436cc3

…lvm-test-suite#1403) Complementary to intel#7478.

[SYCL][Reduction] Add a test for range+multiple reds version (intel/l…

b9f7014

…lvm-test-suite#1400) Copy-pasted from reduction_nd_N_vars.cpp and adjusted to use sycl::range.

[SYCL] Remove CL namespace from level_zero_sub_sub_device.cpp (intel/…

6da932a

…llvm-test-suite#1408)

[LIT][SYCL] Avoid fp64 support requirement (intel/llvm-test-suite#1410)

b61a984

Also added a try catch block to print possible "synchronous" exceptions.

[SYCL][Level Zero][USM] Add a test for read-only shared alloc (intel/…

2ed479d

…llvm-test-suite#1409)

[SYCL] Temporarily disable KernelAndProgram/kernel-bundle-merge-optio…

f0e0fba

…ns-env.cpp (intel/llvm-test-suite#1412) While the investigation is in progress.

[SYCL] Add a test targeting internal reduction APIs (intel/llvm-test-…

d7b225b

…suite#1373) That way we can bypass strategy selection and test any of them with the same type/operation.

[SYCL] Lock mutex only for critical region in interop-level-zero-thre…

07733a8

…ad.cpp (intel/llvm-test-suite#1418)

[SYCL] Test corrections after moving bfloat16 support out of experime…

74100da

…ntal status. (intel/llvm-test-suite#1129) Tests changes for intel#6524 Signed-off-by: Rajiv Deodhar <rajiv.deodhar@intel.com> Co-authored-by: JackAKirk <jack.kirk@codeplay.com>

[SYCL] disable the ZE_DEBUG test that is expected to throw an excepti…

77f9d52

…on (intel/llvm-test-suite#1365) This test is expected to throw an exception so we need to disable the ZE_DEBUG test. Signed-off-by: Byoungro So byoungro.so@intel.com

[SYCL] Add support for sycl::ext::oneapi::property::queue::use_priori…

ea83f27

…ty (intel/llvm-test-suite#1414)

[SYCL] Fix CUDA tests using bfloat16 (intel/llvm-test-suite#1421)

162c033

* [SYCL] Fix CUDA tests using bfloat16 * Add missing using in element_wise_wi_marray_legacy Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

[SYCL] Disable SYCL/Basic/barrier_order.cpp on CUDA (intel/llvm-test-…

c7d10e7

…suite#1424) See intel#7330

[SYCL] Test sycl::Is_compatible() (intel/llvm-test-suite#1411)

1eed057

[SYCL] Add target to is_compatible test (intel/llvm-test-suite#1428)

18fd4fe

This commit should fix the is_compatible test for non-SPIRV targets. Signed-off-by: Larsen, Steffen <steffen.larsen@intel.com>

sarnex and others added 7 commits March 24, 2023 11:57

[SYCL][ESIMD] Add large GRF test (intel/llvm-test-suite#1686)

44692bd

Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com> Co-authored-by: Vyacheslav Klochkov <vyacheslav.n.klochkov@intel.com>

[SYCL][CUDA] Re-enable subgroup shuffle_fp64 test for CUDA (intel/llv…

f7a7b69

…m-test-suite#1663)

[SYCL][ESIMD] Add a runtime test for BFN function (intel/llvm-test-su…

7e9ed9e

…ite#1671)

Revert "[SYCL] Refactor invalid use of local accessor" (intel/llvm-te…

9d296b0

…st-suite#1695) Revert "[SYCL] Refactor invalid use of local accessor (intel/llvm-test-suite#1646)" This reverts commit c257c0b.

[SYCL][InvokeSIMD] Delete ret size mismatch test (intel/llvm-test-sui…

9ef2050

…te#1679) Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com>

Remove exerimental subset of the end-to-end tests

5569b32

We're doing the final move and it makes sense to make a "clean" merge (except that CMake/LIT config stuff that had some customizations).

aelovikov-intel requested review from a team as code owners March 27, 2023 15:42

aelovikov-intel requested a review from cperkinsintel March 27, 2023 15:42

aelovikov-intel added the disable-lint Skip linter check step and proceed with build jobs label Mar 27, 2023

aelovikov-intel closed this Mar 27, 2023

aelovikov-intel reopened this Mar 27, 2023

aelovikov-intel temporarily deployed to aws March 27, 2023 18:59 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to aws March 27, 2023 20:24 — with GitHub Actions Inactive

[SYCL] Add feature test for gpu_cache_config property (intel/llvm-tes…

1fcd268

…t-suite#1687)

aelovikov-intel requested a review from bader March 27, 2023 20:58

v-klochkov and others added 2 commits March 27, 2023 17:03

[ESIMD] Add tests for private/stack memory (intel/llvm-test-suite#1682)

c1c4394

Signed-off-by: Vyacheslav N Klochkov <vyacheslav.n.klochkov@intel.com>

Move SYCL -> sycl/test-e2e

b0bdf35

aelovikov-intel force-pushed the final-test-move branch from c0989a1 to 280498d Compare March 27, 2023 22:20

aelovikov-intel added 2 commits March 27, 2023 15:28

Merge intel/llvm-test-suite in-tree

f33f8b2

Updates to README.md

9fd0ee6

aelovikov-intel force-pushed the final-test-move branch from 280498d to 9fd0ee6 Compare March 27, 2023 22:29

bader changed the title ~~Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e~~ [SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e Mar 27, 2023

bader merged commit ee211b6 into intel:sycl Mar 27, 2023

aelovikov-intel temporarily deployed to aws March 27, 2023 23:25 — with GitHub Actions Inactive

aelovikov-intel temporarily deployed to aws March 28, 2023 00:48 — with GitHub Actions Inactive

EwanC mentioned this pull request Mar 29, 2023

Expand on graphs llvm-test-suite tests reble/llvm#33

Closed

aelovikov-intel deleted the final-test-move branch April 7, 2023 15:57

aelovikov-intel mentioned this pull request Aug 21, 2025

[SYCL][NATIVECPU] Port ock compiler internals into sycl native cpu #19841

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e #8782

[SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e #8782

Uh oh!

aelovikov-intel commented Mar 27, 2023 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

43 participants

[SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e #8782

[SYCL] Move intel/llvm-test-suite SYCL tests to in-tree sycl/test-e2e #8782

Uh oh!

Conversation

aelovikov-intel commented Mar 27, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

43 participants

aelovikov-intel commented Mar 27, 2023 •

edited

Loading