Skip to content

llamafile : improve sgemm.cpp #6796

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

Merged
merged 4 commits into from
Apr 22, 2024
Merged

llamafile : improve sgemm.cpp #6796

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3e4fc41
llamafile : improve sgemm.cpp
jart Apr 20, 2024
9d4d14c
Address review comments
jart Apr 21, 2024
6b220dc
Help clang produce fma instructions
jart Apr 21, 2024
d22488c
Address review comments
jart Apr 22, 2024
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
4 changes: 0 additions & 4 deletions Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -384,10 +384,6 @@ ifdef LLAMA_OPENBLAS
MK_LDFLAGS += $(shell pkg-config --libs openblas)
endif # LLAMA_OPENBLAS

# TODO: temporary disable until MoE is fixed
# https://github.com/ggerganov/llama.cpp/pull/6716
LLAMA_NO_LLAMAFILE := 1

ifndef LLAMA_NO_LLAMAFILE
MK_CPPFLAGS += -DGGML_USE_LLAMAFILE
OBJS += sgemm.o
Expand Down
4 changes: 2 additions & 2 deletions ggml.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10825,7 +10825,7 @@ static void ggml_compute_forward_mul_mat(
#endif

#if GGML_USE_LLAMAFILE
if (nb10 == ggml_type_size(src1->type)) {
if (src1_cont) {
for (int64_t i13 = 0; i13 < ne13; i13++)
for (int64_t i12 = 0; i12 < ne12; i12++)
if (!llamafile_sgemm(ne01, ne11, ne00/ggml_blck_size(src0->type),
Expand Down Expand Up @@ -10878,7 +10878,7 @@ UseGgmlGemm1:;
const size_t row_size = ggml_row_size(vec_dot_type, ne10);

#if GGML_USE_LLAMAFILE
if (nb10 == ggml_type_size(src1->type) || src1->type != vec_dot_type) {
if (src1_cont) {
for (int64_t i13 = 0; i13 < ne13; i13++)
for (int64_t i12 = 0; i12 < ne12; i12++)
if (!llamafile_sgemm(ne01, ne11, ne00/ggml_blck_size(src0->type),
Expand Down
Loading