[SYCL][USM] Improve USM allocator test and fix improper behavior. by jbrodman · Pull Request #1538 · intel/llvm

jbrodman · 2020-04-16T19:10:51Z

Signed-off-by: James Brodman james.brodman@intel.com

Signed-off-by: James Brodman <james.brodman@intel.com>

sycl/include/CL/sycl/usm/usm_allocator.hpp

bd4 · 2020-04-17T14:07:45Z

sycl/test/usm/allocator_vector.cpp

+    vec.resize(N);
+
+    int *res = &vec[0];
+    int *vals = &vec[0];


why not int *res = vec.data()?

No good reason.

bd4 · 2020-04-17T14:14:07Z

sycl/test/usm/allocator_vector.cpp

+
+    auto e0 = q.submit([=](handler &h) {
+      h.single_task<class baz_init>([=]() {
+        res[0] = 0;


Confused by this, res and vals point to the same device vector data, so res[0] = 0 is redundant to the first loop iteration below where vals[i=0] = 0. This also means the res[0] == vals[0] is added to itself an extra time, which works because is happens to be 0, but again confusing.

I see i=0 is skipped in the second loop so no double add. Anyway I know it's just a test, but having res and vals point at the same memory is still confusing to me, takes a minute to verify that it's not a problem in this case.

The computation is sort of irrelevant.

sycl/test/usm/allocator_vector.cpp

bd4 · 2020-04-17T16:02:49Z

Testing with inteloneapi v 2021.1-beta05, with usm_allocator.hpp header change manually applied. The test program works for me when SYCL_DEVICE_TYPE=HOST, but fails (nonzero exit code) with CPU and GPU. Added a debug print when there is a mismatch, only the usm::alloc::device case is failing, with value 0 instead of 28.

Note that I am also seeing this behavior in my own usm test programs, with usm_allocator and sycl::malloc_device: waiting on the event (or the queue) is not sufficient to ensure the kernel is complete and it's safe to copy memory back. For example: https://github.com/bd4/dpcpp-test/blob/master/usmmem.cxx#L65

Perhaps there is something I am missing from latest git commits? I will try to build from source and see if I can still reproduce.

bd4 · 2020-04-17T16:09:42Z

Perhaps there is something I am missing from latest git commits? I will try to build from source and see if I can still reproduce.

Looks like this was already fixed in #1132, sorry for the noise. Looks like USM is still very much in active development and anyone wanting to use it would be advised to track develop.

s-kanaev · 2020-04-20T16:29:55Z

@jbrodman What was wrong with previous version of the test?

jbrodman · 2020-04-20T16:34:52Z

@jbrodman What was wrong with previous version of the test?

It was only testing the allocator with host allocations, not device/shared allocations - hence a method on device allocations wasn't working as intended.

s-kanaev · 2020-04-20T17:04:18Z

sycl/test/usm/allocator_vector.cpp

  auto ctxt = q.get_context();

-  if (!dev.get_info<info::device::usm_host_allocations>()) 
+  if (!(dev.get_info<info::device::usm_host_allocations>() &&


In this test you expect that a single device/environment/runtime will provide you with all three types of allocations. Dunno if it's possible that, say, CPU device/runtime provide you with either two out of three allocation facilities. Anyway, I suggest splitting the cases into three distinct ones. One way I see how this can be achieved is to if test blocks with appropriate dev.get_info<info::device::usm_xxx_allocations>().

s-kanaev · 2020-04-28T08:40:05Z

@jbrodman ping.

jbrodman · 2020-04-28T14:18:44Z

Sorry, what exactly is the request again?

s-kanaev · 2020-04-28T16:23:35Z

This one: #discussion_r411543369

Signed-off-by: James Brodman <james.brodman@intel.com>

s-kanaev

LGTM

bader · 2020-04-29T09:15:25Z

@jbrodman, @bd4, @Ruyk, I marked some comments as resolved, but there are a couple of them which might require attention. Are you okay to merge or you expect additional fixes?

bader · 2020-05-01T11:39:39Z

@jbrodman, @bd4, @Ruyk, I marked some comments as resolved, but there are a couple of them which might require attention. Are you okay to merge or you expect additional fixes?

Ping.

bd4 · 2020-05-01T13:17:25Z

@bader looks good to me. I ran the tests with some added debug prints against beta 05. With SYCL_DEVICE_TYPE=GPU, the usm::alloc::device case fails, other two pass. For HOST and CPU, all pass.

In beta06, which appears to have #1132 but not this, the tests almost pass but the feature not supported exception is thrown on device vector destructor. If I comment out the throw at /opt/intel/inteloneapi/compiler/2021.1-beta06/linux/include/sycl/CL/sycl/usm/usm_allocator.hpp line 88 and rebuild, then the test runs successfully.

Using my source build llvm install with this fix applied (and the already merged queue fix from #1132), all pass.

bader · 2020-05-01T15:43:38Z

@bd4, thanks!

…_docs * origin/sycl: (6482 commits) [SYCL][NFC] Clean formatting in Markdown documents (intel#1635) [SYCL][Doc] Remove obsolete parens from README (intel#1637) [SYCL] Fix failing ABI tests when LLVM_LIBDIR_SUFFIX is set (intel#1605) [SYCL] Fix warnings in libdevice (intel#1630) [SYCL][CUDA] Triage and clean LIT (intel#1620) [SYCL][NFC] Fix GCC 8 compilation warnings (intel#1631) [SYCL] Minor fixes in LowerWGScope [SYCL] PI: correct default interoperability plugin selection [SYCL] Add faster reduction implementations using atomic or/and intel::reduce() (intel#1615) [SYCL] Add sycl-ls utility for listing devices discovered/selected by SYCL RT (intel#1575) [SYCL] Fix getDeviceFromHandler declarations (intel#1626) [SPIR-V] Correct/improve declaration of SPIR-V builtins (intel#1519) [SYCL][USM] Improve USM allocator test and fix improper behavior. (intel#1538) [SYCL] Fix failing ABI LITs (intel#1622) [SYCL] Add support for MSVC internal math functions in device library (intel#1441) [SYCL] Add runtime library versioning (intel#1604) [SYCL] Check weak symbols in ABI dumps (intel#1609) [NFC][SYCL] Improve kernel metadata test (intel#1610) Revert "[SYCL] XFAIL LIT test due to duplicate diagnostic" (intel#1460) [SYCL] Move the reduction command group funcs out of handler.hpp (intel#1602) ...

… (#154163) This reverts commit 7c53c61. This patch hit an issue in hip test. revert and will reopen later

Improve USM allocator test and fix improper behavior.

de73790

Signed-off-by: James Brodman <james.brodman@intel.com>

jbrodman requested a review from bader April 16, 2020 19:10

jbrodman requested a review from a team as a code owner April 16, 2020 19:10

jbrodman requested a review from s-kanaev April 16, 2020 19:10

Ruyk reviewed Apr 17, 2020

View reviewed changes

sycl/include/CL/sycl/usm/usm_allocator.hpp Show resolved Hide resolved

bd4 reviewed Apr 17, 2020

View reviewed changes

sycl/test/usm/allocator_vector.cpp Show resolved Hide resolved

s-kanaev reviewed Apr 20, 2020

View reviewed changes

Gate different parts of test appropriately

6475531

Signed-off-by: James Brodman <james.brodman@intel.com>

jbrodman requested a review from s-kanaev April 28, 2020 22:33

s-kanaev approved these changes Apr 29, 2020

View reviewed changes

bader merged commit bb8fce9 into intel:sycl May 1, 2020

MartinWehking mentioned this pull request May 8, 2024

AMD CUDA Event Reuse #13697

Closed

MartinWehking pushed a commit to MartinWehking/llvm that referenced this pull request May 13, 2024

CI testing for intel#1538

fe654a0

iclsrc pushed a commit that referenced this pull request Aug 18, 2025

Revert "[AMDGPU][True16][CodeGen] use vgpr16 for zext patterns (#1538……

d49aab1

… (#154163) This reverts commit 7c53c61. This patch hit an issue in hip test. revert and will reopen later

Conversation

jbrodman commented Apr 16, 2020

Uh oh!

Uh oh!

bd4 Apr 17, 2020

Choose a reason for hiding this comment

Uh oh!

jbrodman Apr 17, 2020

Choose a reason for hiding this comment

Uh oh!

bd4 Apr 17, 2020

Choose a reason for hiding this comment

Uh oh!

bd4 Apr 17, 2020

Choose a reason for hiding this comment

Uh oh!

jbrodman Apr 17, 2020

Choose a reason for hiding this comment

Uh oh!

Uh oh!

bd4 commented Apr 17, 2020

Uh oh!

bd4 commented Apr 17, 2020

Uh oh!

s-kanaev commented Apr 20, 2020

Uh oh!

jbrodman commented Apr 20, 2020

Uh oh!

s-kanaev Apr 20, 2020

Choose a reason for hiding this comment

Uh oh!

s-kanaev commented Apr 28, 2020

Uh oh!

jbrodman commented Apr 28, 2020

Uh oh!

s-kanaev commented Apr 28, 2020

Uh oh!

s-kanaev left a comment

Choose a reason for hiding this comment

Uh oh!

bader commented Apr 29, 2020

Uh oh!

bader commented May 1, 2020

Uh oh!

bd4 commented May 1, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bader commented May 1, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

bd4 commented May 1, 2020 •

edited

Loading