Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

benchdnn matmul with --wtag=acbd correctness test failure on CPU #2180

Closed
shu1chen opened this issue Oct 23, 2024 · 1 comment
Closed

benchdnn matmul with --wtag=acbd correctness test failure on CPU #2180

shu1chen opened this issue Oct 23, 2024 · 1 comment
Labels
bug A confirmed library bug platform:cpu-x64 Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64

Comments

@shu1chen
Copy link
Contributor

Summary

When working on #2175, it's found that there is also a bug on x86_64. After the commit 3877097 on main branch, benchdnn correctness check fails after adding --wtag=acbd. It's good on the previous commit 6bfb6a8.

Version

onednn_verbose,v1,info,oneDNN v3.7.0 (commit 3877097)
onednn_verbose,v1,info,cpu,runtime:OpenMP,nthr:64
onednn_verbose,v1,info,cpu,isa:Intel AVX-512 with float16, Intel DL Boost and bfloat16 support and Intel AMX with bfloat16 and 8-bit integer support

Environment

  • CPU: Intel (R) Xeon (R) CPU Max 9462
  • OS version: Ubuntu 22.04
  • Compiler version: gcc 11.4.0

Steps to reproduce

$ ./tests/benchdnn/benchdnn --engine=cpu --matmul --wtag=acbd --dtag=abcd 1x1x2x3:3x1x3x3

Observed behavior

[   6][DST][1:0:0:0] exp_f32:        7186 exp:        7186 got:        -nan diff:     nan rdiff:     nan
[   7][DST][1:0:0:1] exp_f32:       -5487 exp:       -5487 got:        -nan diff:     nan rdiff:     nan
[   8][DST][1:0:0:2] exp_f32:       -5549 exp:       -5549 got:        -nan diff:     nan rdiff:     nan
[   9][DST][1:0:1:0] exp_f32:       -3131 exp:       -3131 got:        -nan diff:     nan rdiff:     nan
[  10][DST][1:0:1:1] exp_f32:        2511 exp:        2511 got:        -nan diff:     nan rdiff:     nan
[  11][DST][1:0:1:2] exp_f32:        -240 exp:        -240 got:        -nan diff:     nan rdiff:     nan
[  12][DST][2:0:0:0] exp_f32:       -1727 exp:       -1727 got:        -nan diff:     nan rdiff:     nan
[  13][DST][2:0:0:1] exp_f32:       -9256 exp:       -9256 got:        -nan diff:     nan rdiff:     nan
[  14][DST][2:0:0:2] exp_f32:        2826 exp:        2826 got:        -nan diff:     nan rdiff:     nan
[  15][DST][2:0:1:0] exp_f32:        1391 exp:        1391 got:        -nan diff:     nan rdiff:     nan
[COMPARE_STATS][DST]: trh=0 err_max_diff:     nan err_max_rdiff:     nan all_max_diff:     nan all_max_rdiff:     nan
0:FAILED (errors:12 total:18) __REPRO: --matmul --wtag=acbd --dtag=abcd 1x1x2x3:3x1x3x3
tests:1 passed:0 skipped:0 mistrusted:0 unimplemented:0 invalid_arguments:0 failed:1 listed:0
total: 0.01s; fill: 0.01s (48%); compute_ref: 0.00s (24%); compare: 0.00s (7%);

Expected behavior

Test pass. It's better to add a test for this use case.

@shu1chen shu1chen added bug A confirmed library bug platform:cpu-x64 Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64 labels Oct 23, 2024
@dzarukin
Copy link
Contributor

Must be fixed by 6be35fd4.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug A confirmed library bug platform:cpu-x64 Intel64/AMD64 processors. Codeowner: @oneapi-src/onednn-cpu-x64
Projects
None yet
Development

No branches or pull requests

2 participants