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.

Sign up for GitHub

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

Simplify and rewrite reference packing kernels. #610

Merged
merged 17 commits into from
Apr 7, 2022
Merged

Simplify and rewrite reference packing kernels. #610

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
0ca0313
Add compile-time constants (macros) for the MR and NR block sizes.
devinamatthews Jan 31, 2022
88dd4d6
PoC implementation of reference gemm kernel.
devinamatthews Feb 5, 2022
ac920eb
Simplify cntx_t structure.
devinamatthews Feb 6, 2022
ccf99db
Missed `bli_kernel_defs_zen3.h` somehow.
devinamatthews Feb 6, 2022
eebe527
Fix two ARM issues.
devinamatthews Feb 7, 2022
87b76c1
Merge remote-tracking branch 'origin/master' into mr_nr_macros_phase1
devinamatthews Feb 7, 2022
84af20d
Re-write reference packing kernels to make use of MR/NR values when a…
devinamatthews Feb 9, 2022
68ffb78
Fix missing `_ROW_PREF`s.
devinamatthews Feb 11, 2022
9f9700b
Merge cosmetic changes and fix packm_1er kernel.
devinamatthews Feb 12, 2022
0b86e10
New packing kernels completely working.
devinamatthews Feb 16, 2022
9baa417
Merge branch 'master' into mr_nr_macros
devinamatthews Feb 16, 2022
034d88c
Fix one last bug.
devinamatthews Feb 16, 2022
97d124a
Fix performance bug.
devinamatthews Mar 15, 2022
99e4c80
Attempt to resolve some conflicts before merge.
fgvanzee Apr 5, 2022
08aea6a
Merge branch 'master' into mr_nr_macros
fgvanzee Apr 5, 2022
9533519
Fixed some stale bli_cntx_get_l1v_ker_dt() calls.
fgvanzee Apr 5, 2022
2998bce
Renamed BLIS_IS_KERNEL; whitespace changes.
fgvanzee Apr 6, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Fix missing _ROW_PREFs.
  • Loading branch information
@devinamatthews
devinamatthews committed Feb 11, 2022
commit 68ffb78a94ac43d54da70a01d92f3725c13bc1d9
4 changes: 2 additions & 2 deletions config/cortexa57/bli_cntx_init_cortexa57.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,8 +61,8 @@ void bli_cntx_init_cortexa57( cntx_t* cntx )
cntx,

// level-3
BLIS_GEMM_UKR, BLIS_FLOAT, FALSE,
BLIS_GEMM_UKR, BLIS_DOUBLE, FALSE,
BLIS_GEMM_UKR_ROW_PREF, BLIS_FLOAT, FALSE,
BLIS_GEMM_UKR_ROW_PREF, BLIS_DOUBLE, FALSE,

-1
);
Expand Down
60 changes: 30 additions & 30 deletions config/zen/bli_cntx_init_zen.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,39 +198,39 @@ void bli_cntx_init_zen( cntx_t* cntx )
BLIS_GEMMTRSM_U_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,

// gemmsup
BLIS_GEMMSUP_RRR_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RRC_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RCR_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RCC_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CRR_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CRC_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CCR_UKR, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CCC_UKR, BLIS_DOUBLE, TRUE,

BLIS_GEMMSUP_RRR_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RRC_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RCR_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RCC_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CRR_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CRC_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CCR_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CCC_UKR, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RRR_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RRC_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RCR_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_RCC_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CRR_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CRC_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CCR_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,
BLIS_GEMMSUP_CCC_UKR_ROW_PREF, BLIS_DOUBLE, TRUE,

BLIS_GEMMSUP_RRR_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RRC_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RCR_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_RCC_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CRR_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CRC_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CCR_UKR_ROW_PREF, BLIS_FLOAT, TRUE,
BLIS_GEMMSUP_CCC_UKR_ROW_PREF, BLIS_FLOAT, TRUE,

#if 0
// NOTE: This set of kernels is likely broken and therefore disabled.
BLIS_GEMMSUP_RRR_UKR, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_RCR_UKR, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CRR_UKR, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_RCC_UKR, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CCR_UKR, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CCC_UKR, BLIS_SCOMPLEX, TRUE,

BLIS_GEMMSUP_RRR_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_RCR_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CRR_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_RCC_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CCR_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CCC_UKR, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_RRR_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_RCR_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CRR_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_RCC_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CCR_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,
BLIS_GEMMSUP_CCC_UKR_ROW_PREF, BLIS_SCOMPLEX, TRUE,

BLIS_GEMMSUP_RRR_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_RCR_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CRR_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_RCC_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CCR_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
BLIS_GEMMSUP_CCC_UKR_ROW_PREF, BLIS_DCOMPLEX, TRUE,
#endif

-1
Expand Down