Skip to content

Add ASIMD Small GEMM kernels #4963

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

Draft
wants to merge 1 commit into
base: develop
Choose a base branch
from
Draft
Changes from all commits
Commits
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
18 changes: 18 additions & 0 deletions kernel/arm64/KERNEL.ARMV8
Original file line number Diff line number Diff line change
@@ -143,6 +143,15 @@ endif
SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX)
SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX)

SGEMM_SMALL_K_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_B0_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_B0_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_B0_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_TN = sgemm_small_kernel_tn_asimd.c
SGEMM_SMALL_K_B0_TN = sgemm_small_kernel_tn_asimd.c

DGEMMKERNEL = dgemm_kernel_$(DGEMM_UNROLL_M)x$(DGEMM_UNROLL_N).S
DTRMMKERNEL = dtrmm_kernel_$(DGEMM_UNROLL_M)x$(DGEMM_UNROLL_N).S

@@ -171,6 +180,15 @@ endif
DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX)
DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX)

DGEMM_SMALL_K_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_B0_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_B0_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_B0_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_TN = dgemm_small_kernel_tn_asimd.c
DGEMM_SMALL_K_B0_TN = dgemm_small_kernel_tn_asimd.c

CGEMMKERNEL = cgemm_kernel_$(CGEMM_UNROLL_M)x$(CGEMM_UNROLL_N).S
CTRMMKERNEL = ctrmm_kernel_$(CGEMM_UNROLL_M)x$(CGEMM_UNROLL_N).S
ifneq ($(CGEMM_UNROLL_M), $(CGEMM_UNROLL_N))
50 changes: 34 additions & 16 deletions kernel/arm64/KERNEL.ARMV8SVE
Original file line number Diff line number Diff line change
@@ -132,14 +132,14 @@ SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX)
SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX)

SGEMM_SMALL_M_PERMIT = gemm_small_kernel_permit_sve.c
SGEMM_SMALL_K_NT = sgemm_small_kernel_nt_sve.c
SGEMM_SMALL_K_B0_NT = sgemm_small_kernel_nt_sve.c
SGEMM_SMALL_K_NN = sgemm_small_kernel_nn_sve.c
SGEMM_SMALL_K_B0_NN = sgemm_small_kernel_nn_sve.c
SGEMM_SMALL_K_TT = sgemm_small_kernel_tt_sve.c
SGEMM_SMALL_K_B0_TT = sgemm_small_kernel_tt_sve.c
SGEMM_SMALL_K_TN = sgemm_small_kernel_tn_sve.c
SGEMM_SMALL_K_B0_TN = sgemm_small_kernel_tn_sve.c
# SGEMM_SMALL_K_NT = sgemm_small_kernel_nt_sve.c
# SGEMM_SMALL_K_B0_NT = sgemm_small_kernel_nt_sve.c
# SGEMM_SMALL_K_NN = sgemm_small_kernel_nn_sve.c
# SGEMM_SMALL_K_B0_NN = sgemm_small_kernel_nn_sve.c
# SGEMM_SMALL_K_TT = sgemm_small_kernel_tt_sve.c
# SGEMM_SMALL_K_B0_TT = sgemm_small_kernel_tt_sve.c
# SGEMM_SMALL_K_TN = sgemm_small_kernel_tn_sve.c
# SGEMM_SMALL_K_B0_TN = sgemm_small_kernel_tn_sve.c

STRMMUNCOPY_M = trmm_uncopy_sve_v1.c
STRMMLNCOPY_M = trmm_lncopy_sve_v1.c
@@ -163,14 +163,14 @@ DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX)
DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX)

DGEMM_SMALL_M_PERMIT = gemm_small_kernel_permit_sve.c
DGEMM_SMALL_K_NT = dgemm_small_kernel_nt_sve.c
DGEMM_SMALL_K_B0_NT = dgemm_small_kernel_nt_sve.c
DGEMM_SMALL_K_NN = dgemm_small_kernel_nn_sve.c
DGEMM_SMALL_K_B0_NN = dgemm_small_kernel_nn_sve.c
DGEMM_SMALL_K_TT = dgemm_small_kernel_tt_sve.c
DGEMM_SMALL_K_B0_TT = dgemm_small_kernel_tt_sve.c
DGEMM_SMALL_K_TN = dgemm_small_kernel_tn_sve.c
DGEMM_SMALL_K_B0_TN = dgemm_small_kernel_tn_sve.c
# DGEMM_SMALL_K_NT = dgemm_small_kernel_nt_sve.c
# DGEMM_SMALL_K_B0_NT = dgemm_small_kernel_nt_sve.c
# DGEMM_SMALL_K_NN = dgemm_small_kernel_nn_sve.c
# DGEMM_SMALL_K_B0_NN = dgemm_small_kernel_nn_sve.c
# DGEMM_SMALL_K_TT = dgemm_small_kernel_tt_sve.c
# DGEMM_SMALL_K_B0_TT = dgemm_small_kernel_tt_sve.c
# DGEMM_SMALL_K_TN = dgemm_small_kernel_tn_sve.c
# DGEMM_SMALL_K_B0_TN = dgemm_small_kernel_tn_sve.c

DTRMMUNCOPY_M = trmm_uncopy_sve_v1.c
DTRMMLNCOPY_M = trmm_lncopy_sve_v1.c
@@ -227,3 +227,21 @@ ZHEMMUTCOPY_M = zhemm_utcopy_sve.c

ZSYMMUCOPY_M = zsymm_ucopy_sve.c
ZSYMMLCOPY_M = zsymm_lcopy_sve.c

SGEMM_SMALL_K_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_B0_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_B0_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_B0_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_TN = sgemm_small_kernel_tn_asimd.c
SGEMM_SMALL_K_B0_TN = sgemm_small_kernel_tn_asimd.c

DGEMM_SMALL_K_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_B0_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_B0_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_B0_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_TN = dgemm_small_kernel_tn_asimd.c
DGEMM_SMALL_K_B0_TN = dgemm_small_kernel_tn_asimd.c
18 changes: 18 additions & 0 deletions kernel/arm64/KERNEL.NEOVERSEV2
Original file line number Diff line number Diff line change
@@ -1 +1,19 @@
include $(KERNELDIR)/KERNEL.ARMV8SVE

SGEMM_SMALL_K_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_B0_NT = sgemm_small_kernel_nt_asimd.c
SGEMM_SMALL_K_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_B0_NN = sgemm_small_kernel_nn_asimd.c
SGEMM_SMALL_K_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_B0_TT = sgemm_small_kernel_tt_asimd.c
SGEMM_SMALL_K_TN = sgemm_small_kernel_tn_asimd.c
SGEMM_SMALL_K_B0_TN = sgemm_small_kernel_tn_asimd.c

DGEMM_SMALL_K_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_B0_NT = dgemm_small_kernel_nt_asimd.c
DGEMM_SMALL_K_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_B0_NN = dgemm_small_kernel_nn_asimd.c
DGEMM_SMALL_K_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_B0_TT = dgemm_small_kernel_tt_asimd.c
DGEMM_SMALL_K_TN = dgemm_small_kernel_tn_asimd.c
DGEMM_SMALL_K_B0_TN = dgemm_small_kernel_tn_asimd.c
Loading