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

Add summary tables for v1.00.00 tput/mad results for CHEP24 #1021

Closed
wants to merge 25 commits into from

Conversation

valassi
Copy link
Member

@valassi valassi commented Oct 12, 2024

Hi @oliviermattelaer this PR only adds summary tables for v1.00.00 tput/mad results for CHEP24.

It does not touch CODEGEN. Hopefully not controversial. Can you have a look please?

Thanks
Andrea

PS Examples with fptype=m in ggttggg

CUDA on itscrd90

===========================================================================================================
|            | mad                        | mad               | mad               | sa/brdg   | sa/full   |
-----------------------------------------------------------------------------------------------------------
| ggttggg    | [sec] tot = mad + MEs      | [TOT/sec]         | [MEs/sec]         | [MEs/sec] | [MEs/sec] |
===========================================================================================================
| nevt/grid  |                            |                   |                   |      8192 |      8192 |
| nevt total |                      81920 |             81920 |             81920 |  256*32*1 |  256*32*1 |
-----------------------------------------------------------------------------------------------------------
| FORTRAN    | 1003.89 =   4.43 +  999.46 |  8.16e+01 (= 1.0) |  8.20e+01 (= 1.0) |       --- |       --- |
| CPP/none   | 1239.64 =   4.53 + 1235.11 |  6.61e+01 (x 0.8) |  6.63e+01 (x 0.8) |  7.86e+01 |  7.89e+01 |
| CPP/sse4   |  618.78 =   4.43 +  614.35 |  1.32e+02 (x 1.6) |  1.33e+02 (x 1.6) |  1.60e+02 |  1.60e+02 |
| CPP/avx2   |  270.59 =   4.38 +  266.21 |  3.03e+02 (x 3.7) |  3.08e+02 (x 3.8) |  3.73e+02 |  3.71e+02 |
| CPP/512y   |  238.98 =   4.34 +  234.64 |  3.43e+02 (x 4.2) |  3.49e+02 (x 4.3) |  4.31e+02 |  4.30e+02 |
| CPP/512z   |  245.96 =   4.35 +  241.61 |  3.33e+02 (x 4.1) |  3.39e+02 (x 4.1) |  3.68e+02 |  3.69e+02 |
| CUDA/8192  |   14.32 =   5.69 +    8.63 |  5.72e+03 (x70.1) |  9.49e+03 (x115.) |  9.39e+03 |  1.08e+04 |
===========================================================================================================
| nevt/grid  |                                                                    |     16384 |     16384 |
| nevt total |                                                                    |  512*32*1 |  512*32*1 |
--------------                                                                    -------------------------
| CUDA/max   |                                                                    |  1.11e+04 |  1.16e+04 |
|            |                                                                    |           |   (x141.) |
==============                                                                    =========================

AVX512z on itgold91

===========================================================================================================
|            | mad                        | mad               | mad               | sa/brdg   | sa/full   |
-----------------------------------------------------------------------------------------------------------
| ggttggg    | [sec] tot = mad + MEs      | [TOT/sec]         | [MEs/sec]         | [MEs/sec] | [MEs/sec] |
===========================================================================================================
| nevt/grid  |                            |                   |                   |      8192 |      8192 |
| nevt total |                      81920 |             81920 |             81920 |  256*32*1 |  256*32*1 |
-----------------------------------------------------------------------------------------------------------
| FORTRAN    |  854.16 =   2.80 +  851.36 |  9.59e+01 (= 1.0) |  9.62e+01 (= 1.0) |       --- |       --- |
| CPP/none   |  983.59 =   2.96 +  980.63 |  8.33e+01 (x 0.9) |  8.35e+01 (x 0.9) |  9.85e+01 |  9.85e+01 |
| CPP/sse4   |  491.53 =   2.87 +  488.66 |  1.67e+02 (x 1.7) |  1.68e+02 (x 1.7) |  1.98e+02 |  1.98e+02 |
| CPP/avx2   |  227.77 =   2.83 +  224.94 |  3.60e+02 (x 3.8) |  3.64e+02 (x 3.8) |  4.40e+02 |  4.40e+02 |
| CPP/512y   |  202.04 =   2.83 +  199.22 |  4.05e+02 (x 4.2) |  4.11e+02 (x 4.3) |  5.09e+02 |  5.08e+02 |
| CPP/512z   |  116.60 =   2.82 +  113.78 |  7.03e+02 (x 7.3) |  7.20e+02 (x 7.5) |  8.79e+02 |  8.78e+02 |
===========================================================================================================

HIP on LUMI (gg_ttgg only, building gg_ttggg fails)

===========================================================================================================
|            | mad                        | mad               | mad               | sa/brdg   | sa/full   |
-----------------------------------------------------------------------------------------------------------
| ggttgg     | [sec] tot = mad + MEs      | [TOT/sec]         | [MEs/sec]         | [MEs/sec] | [MEs/sec] |
===========================================================================================================
| nevt/grid  |                            |                   |                   |      8192 |      8192 |
| nevt total |                      81920 |             81920 |             81920 |  256*32*1 |  256*32*1 |
-----------------------------------------------------------------------------------------------------------
| FORTRAN    |   26.54 =   1.39 +   25.15 |  3.09e+03 (= 1.0) |  3.26e+03 (= 1.0) |       --- |       --- |
| CPP/none   |   33.21 =   1.38 +   31.83 |  2.47e+03 (x 0.8) |  2.57e+03 (x 0.8) |  2.68e+03 |  2.68e+03 |
| CPP/sse4   |   16.67 =   1.38 +   15.29 |  4.91e+03 (x 1.6) |  5.36e+03 (x 1.6) |  5.55e+03 |  5.57e+03 |
| CPP/avx2   |    8.25 =   1.40 +    6.85 |  9.93e+03 (x 3.2) |  1.20e+04 (x 3.7) |  1.24e+04 |  1.23e+04 |
| CPP/512y   |    0.00 =   0.00 +    0.00 |  0.00e+00 (x 0.0) |           (x 0.0) |  0.00e+00 |  0.00e+00 |
| CPP/512z   |    0.00 =   0.00 +    0.00 |  0.00e+00 (x 0.0) |           (x 0.0) |  0.00e+00 |  0.00e+00 |
| HIP/8192   |    2.84 =   1.77 +    1.07 |  2.88e+04 (x 9.3) |  7.65e+04 (x23.5) |  7.50e+04 |  5.01e+04 |
===========================================================================================================
| nevt/grid  |                                                                    |    524288 |    524288 |
| nevt total |                                                                    |  16k*32*1 |  16k*32*1 |
--------------                                                                    -------------------------
| HIP/max    |                                                                    |  1.80e+05 |  6.82e+04 |
|            |                                                                    |           |   (x20.9) |
==============                                                                    =========================

… fix this so that only CUDA or HIP is printed out
…ut - still need to exclude GPU tags in CPU-only tests
…ee separate chep24 tables for rd90/gold/lumi
@valassi valassi self-assigned this Oct 12, 2024
@valassi
Copy link
Member Author

valassi commented Oct 12, 2024

Some tests are failing because the github CI has upgraded ubuntu-latest from 22.04 to 24.04. I downgraded it back (only for codegen) to 22.04 in PR #1022

image

@valassi
Copy link
Member Author

valassi commented Oct 12, 2024

Closing as discussed with @oliviermattelaer in #1024. Will move this to my fork after CHEP.

@valassi valassi closed this Oct 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant