Skip to content

second exp baseline#350

Merged
superlopuh merged 29 commits intomainfrom
szerdick/exp_baseline_manual
Mar 24, 2026
Merged

second exp baseline#350
superlopuh merged 29 commits intomainfrom
szerdick/exp_baseline_manual

Conversation

@szerdick
Copy link
Copy Markdown
Contributor

@szerdick szerdick commented Mar 4, 2026

No description provided.

@szerdick szerdick marked this pull request as draft March 4, 2026 14:01
@szerdick szerdick requested a review from superlopuh March 4, 2026 14:01
@github-actions
Copy link
Copy Markdown

github-actions Bot commented Mar 4, 2026

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,825,752,248,,
ddot 128xf64,924,924,,177,546
dense 8x8xf64,5551,3843,,2693,2673
exp 64xf16,6432,,,,
exp 64xf32,6046,,,,
exp 64xf64,6886,,,,
fill 4x4xf64,35,35,39,,
fill 4x8xf32,43,43,47,,
matmul 4x16x8xf64,2880,2567,668,,
matmul_transb 4x16x16xf32,3667,4885,,762,755
matmul_transb 4x16x16xf64,,4981,1244,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,711,396,224,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,982,728,230,,
relu 4x4xf64,114,83,50,,
relu 4x8xf32,408,149,,51,46
saxpy 64xf32,597,597,,,100
sum 4x4xf64,157,104,45,,
sum 4x8xf32,307,198,51,,44
sum 8x8xf16,981,,59,,

@github-actions
Copy link
Copy Markdown

github-actions Bot commented Mar 4, 2026

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,0.18,0.19,0.72,,
ddot 128xf64,0.14,0.14,,0.77,0.24
dense 8x8xf64,0.12,0.17,,0.26,0.26
exp 64xf16,0.14,,,,
exp 64xf32,0.13,,,,
exp 64xf64,0.12,,,,
fill 4x4xf64,0.03,0.03,0.46,,
fill 4x8xf32,0.02,0.02,0.38,,
matmul 4x16x8xf64,0.18,0.20,0.87,,
matmul_transb 4x16x16xf32,0.19,0.21,,0.84,0.89
matmul_transb 4x16x16xf64,,0.21,0.93,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,0.32,0.37,0.79,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,0.20,0.20,0.77,,
relu 4x4xf64,0.16,0.22,0.36,,
relu 4x8xf32,0.24,0.23,,0.37,0.41
saxpy 64xf32,0.11,0.11,,,0.65
sum 4x4xf64,0.11,0.16,0.38,,
sum 4x8xf32,0.11,0.17,0.33,,0.39
sum 8x8xf16,0.26,,0.29,,

@szerdick
Copy link
Copy Markdown
Contributor Author

szerdick commented Mar 4, 2026

This does not work yet, but working on it.

@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,825,752,248,,
ddot 128xf64,924,924,,177,546
dense 8x8xf64,5551,3843,,2693,2673
exp 64xf16,19582,,,,3735
exp 64xf32,20738,,,,3735
exp 64xf64,28528,,,,3371
exp_experiment 64xf64,,,,,8287
fill 4x4xf64,35,35,39,,
fill 4x8xf32,43,43,47,,
matmul 4x16x8xf64,2880,2567,668,,
matmul_transb 4x16x16xf32,3667,4885,,762,755
matmul_transb 4x16x16xf64,,4981,1244,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,711,396,224,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,982,728,230,,
relu 4x4xf64,114,83,50,,
relu 4x8xf32,408,149,,51,46
saxpy 64xf32,597,597,,,100
sum 4x4xf64,157,104,45,,
sum 4x8xf32,307,198,51,,44
sum 8x8xf16,981,,59,,

@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,0.18,0.19,0.72,,
ddot 128xf64,0.14,0.14,,0.77,0.24
dense 8x8xf64,0.12,0.17,,0.26,0.26
exp 64xf16,0.05,,,,0.21
exp 64xf32,0.04,,,,0.21
exp 64xf64,0.03,,,,0.19
exp_experiment 64xf64,,,,,0.08
fill 4x4xf64,0.03,0.03,0.46,,
fill 4x8xf32,0.02,0.02,0.38,,
matmul 4x16x8xf64,0.18,0.20,0.87,,
matmul_transb 4x16x16xf32,0.19,0.21,,0.84,0.89
matmul_transb 4x16x16xf64,,0.21,0.93,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,0.32,0.37,0.79,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,0.20,0.20,0.77,,
relu 4x4xf64,0.16,0.22,0.36,,
relu 4x8xf32,0.24,0.23,,0.37,0.41
saxpy 64xf32,0.11,0.11,,,0.65
sum 4x4xf64,0.11,0.16,0.38,,
sum 4x8xf32,0.11,0.17,0.33,,0.39
sum 8x8xf16,0.26,,0.29,,

Comment thread docker/Dockerfile Outdated
Comment thread Snakefile Outdated
Comment thread Snakefile Outdated
Comment thread truth.txt Outdated
Comment thread results/pivoted.fast.csv Outdated
Comment thread results/pivoted.fast.csv Outdated
Comment thread kernels/exp_micro/main.c
Comment thread kernels/exp/64xf64/snrt.c Outdated
Comment thread kernels/exp_macro/main.c Outdated
Comment thread kernels/exp_macro/main.c Outdated
Comment thread kernels/exp_macro/optimized.h
Comment thread Snakefile Outdated
Comment thread plots-mia-thesis/output/exp_plots_2.pdf Outdated
Comment thread plots-mia-thesis/output/exp_macro_plots.pdf
@szerdick szerdick marked this pull request as ready for review March 18, 2026 09:56
@szerdick
Copy link
Copy Markdown
Contributor Author

I think this is done as well

@superlopuh
Copy link
Copy Markdown
Collaborator

CI is failing, does make clean && make fast work for you locally?

Comment on lines +2 to +3
matmul_transb,1x12x40xf32,snitch_stream,416,2957,2954,0.0,0,0,322,0,0.7740384615384616,0.8846153846153846,364,0,0,0.875,0,0,1.7009345794392525,1.7009345794392523,1,0,214,0.7062706270627063,89,0,0,0.21394230769230768,0,2542,2961,1.0889423076923077,2544
matmul_transb,1x16x40xf32,snitch_stream,497,3119,3116,0.0,0,0,402,0,0.8088531187122736,0.9054054054054054,444,0,0,0.8933601609657947,0,0,2.074766355140187,2.074766355140187,1,0,214,0.7039473684210527,90,0,0,0.18108651911468812,0,2623,3123,1.0744466800804828,2625
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this is still a surprising diff, can you re-run all again on this branch? I don't see why it should be different

Comment thread results/pivoted.all.csv
@@ -1,52 +1,60 @@
kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 12x12xf64,7373,6726,2449,,
conv2d_d1_s1_3x3 12x12xf64,7373,6726,2482,,
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

similarly here, this is unexpected

@szerdick
Copy link
Copy Markdown
Contributor Author

This Is very weird to me. Locally make clean && make fast works without failure.

@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,825,752,248,,
ddot 128xf64,924,924,,177,546
dense 8x8xf64,5551,3843,,2693,2673
exp_macro 128xf64,,,,,2855
exp_macro 64xf64,,,,,1897
exp_micro 16xf16,1636,,,,
exp_micro 16xf32,1542,,,,
exp_micro 16xf64,1748,,,,
exp_micro 32xf16,3236,,,,
exp_micro 32xf32,3047,,,,
exp_micro 32xf64,3467,,,,
exp_micro 48xf16,4832,,,,
exp_micro 48xf32,4551,,,,
exp_micro 48xf64,5189,,,,
exp_micro 64xf16,6432,,,,
exp_micro 64xf32,6046,,,,
exp_micro 64xf64,6886,,,,
fill 4x4xf64,35,35,39,,
fill 4x8xf32,43,43,47,,
matmul 4x16x8xf64,2880,2567,668,,
matmul_transb 4x16x16xf32,3667,4885,,762,755
matmul_transb 4x16x16xf64,,4981,1244,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,711,396,224,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,982,728,230,,
relu 4x4xf64,114,83,50,,
relu 4x8xf32,408,149,,51,46
saxpy 64xf32,597,597,,,100
sum 4x4xf64,157,104,45,,
sum 4x8xf32,307,198,51,,44
sum 8x8xf16,981,,59,,

@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,0.18,0.19,0.72,,
ddot 128xf64,0.14,0.14,,0.77,0.24
dense 8x8xf64,0.12,0.17,,0.26,0.26
exp_macro 128xf64,,,,,0.45
exp_macro 64xf64,,,,,0.34
exp_micro 16xf16,0.14,,,,
exp_micro 16xf32,0.13,,,,
exp_micro 16xf64,0.12,,,,
exp_micro 32xf16,0.14,,,,
exp_micro 32xf32,0.13,,,,
exp_micro 32xf64,0.12,,,,
exp_micro 48xf16,0.14,,,,
exp_micro 48xf32,0.13,,,,
exp_micro 48xf64,0.12,,,,
exp_micro 64xf16,0.14,,,,
exp_micro 64xf32,0.13,,,,
exp_micro 64xf64,0.12,,,,
fill 4x4xf64,0.03,0.03,0.46,,
fill 4x8xf32,0.02,0.02,0.38,,
matmul 4x16x8xf64,0.18,0.20,0.87,,
matmul_transb 4x16x16xf32,0.19,0.21,,0.84,0.89
matmul_transb 4x16x16xf64,,0.21,0.93,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,0.32,0.37,0.79,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,0.20,0.20,0.77,,
relu 4x4xf64,0.16,0.22,0.36,,
relu 4x8xf32,0.24,0.23,,0.37,0.41
saxpy 64xf32,0.11,0.11,,,0.65
sum 4x4xf64,0.11,0.16,0.38,,
sum 4x8xf32,0.11,0.17,0.33,,0.39
sum 8x8xf16,0.26,,0.29,,

@superlopuh superlopuh merged commit efc5737 into main Mar 24, 2026
1 check passed
@superlopuh superlopuh deleted the szerdick/exp_baseline_manual branch March 24, 2026 17:46
@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,825,752,248,,
ddot 128xf64,924,924,,177,546
dense 8x8xf64,5551,3843,,2693,2673
exp_macro 128xf64,,,,,2855
exp_macro 64xf64,,,,,1897
exp_micro 16xf16,1636,,,,
exp_micro 16xf32,1542,,,,
exp_micro 16xf64,1748,,,,
exp_micro 32xf16,3236,,,,
exp_micro 32xf32,3047,,,,
exp_micro 32xf64,3467,,,,
exp_micro 48xf16,4832,,,,
exp_micro 48xf32,4551,,,,
exp_micro 48xf64,5189,,,,
exp_micro 64xf16,6432,,,,
exp_micro 64xf32,6046,,,,
exp_micro 64xf64,6886,,,,
fill 4x4xf64,35,35,39,,
fill 4x8xf32,43,43,47,,
matmul 4x16x8xf64,2880,2567,668,,
matmul_transb 4x16x16xf32,3667,4885,,762,755
matmul_transb 4x16x16xf64,,4981,1244,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,711,396,224,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,982,728,230,,
relu 4x4xf64,114,83,50,,
relu 4x8xf32,408,149,,51,46
saxpy 64xf32,597,597,,,100
sum 4x4xf64,157,104,45,,
sum 4x8xf32,307,198,51,,44
sum 8x8xf16,981,,59,,

@github-actions
Copy link
Copy Markdown

kernels,baseline,linalg,linalg_xdsl,snitch_stream,snrt
conv2d_d1_s1_3x3 4x4xf64,0.18,0.19,0.72,,
ddot 128xf64,0.14,0.14,,0.77,0.24
dense 8x8xf64,0.12,0.17,,0.26,0.26
exp_macro 128xf64,,,,,0.45
exp_macro 64xf64,,,,,0.34
exp_micro 16xf16,0.14,,,,
exp_micro 16xf32,0.13,,,,
exp_micro 16xf64,0.12,,,,
exp_micro 32xf16,0.14,,,,
exp_micro 32xf32,0.13,,,,
exp_micro 32xf64,0.12,,,,
exp_micro 48xf16,0.14,,,,
exp_micro 48xf32,0.13,,,,
exp_micro 48xf64,0.12,,,,
exp_micro 64xf16,0.14,,,,
exp_micro 64xf32,0.13,,,,
exp_micro 64xf64,0.12,,,,
fill 4x4xf64,0.03,0.03,0.46,,
fill 4x8xf32,0.02,0.02,0.38,,
matmul 4x16x8xf64,0.18,0.20,0.87,,
matmul_transb 4x16x16xf32,0.19,0.21,,0.84,0.89
matmul_transb 4x16x16xf64,,0.21,0.93,,
pooling_nchw_max_d1_s2_3x3 4x4xf64,0.32,0.37,0.79,,
pooling_nchw_sum_d1_s2_3x3 4x4xf64,0.20,0.20,0.77,,
relu 4x4xf64,0.16,0.22,0.36,,
relu 4x8xf32,0.24,0.23,,0.37,0.41
saxpy 64xf32,0.11,0.11,,,0.65
sum 4x4xf64,0.11,0.16,0.38,,
sum 4x8xf32,0.11,0.17,0.33,,0.39
sum 8x8xf16,0.26,,0.29,,

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.

2 participants