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

Jump to bottom

Fix compile failure and further optimize by using vec_msum & vec_sum4s on ppc64le #849

Merged
merged 4 commits into from
Jun 16, 2024
Merged

Fix compile failure and further optimize by using vec_msum & vec_sum4s on ppc64le #849

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
81ab568
fix compile issues introduced by loongarch_asx
penghongbo Jun 4, 2024
f6e002a
restore quant changes to merge
penghongbo Jun 4, 2024
0ba346e
fix compile issues introduced by loongarch_asx
penghongbo Jun 4, 2024
9672484
further optimize by using vec_msum & vec_sum4s on ppc64le
penghongbo Jun 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
Prev Previous commit
Next Next commit
fix compile issues introduced by loongarch_asx
  • Loading branch information
@penghongbo
penghongbo committed Jun 4, 2024
commit 0ba346e1415d8234006b9b064d6e32c999523b07
4 changes: 4 additions & 0 deletions src/ggml-quants.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1078,6 +1078,7 @@ void quantize_row_q8_0(const float * restrict x, void * restrict vy, int64_t k)
}
vec_xst(vec_pack(vec_pack(vi[0], vi[1]), vec_pack(vi[2], vi[3])), 0, &y[i].qs[0]);
vec_xst(vec_pack(vec_pack(vi[4], vi[5]), vec_pack(vi[6], vi[7])), 16, &y[i].qs[0]);
}

#elif defined(__loongarch_asx)
for (int i = 0; i < nb; i++) {
Expand Down Expand Up @@ -1437,6 +1438,7 @@ void quantize_row_q8_1(const float * restrict x, void * restrict vy, int64_t k)
accv = vec_add(accv, vec_sld(accv, accv, 4));
accv = vec_add(accv, vec_sld(accv, accv, 8));
y[i].s = GGML_FP32_TO_FP16(d * vec_extract(accv, 0));
}

#elif defined(__loongarch_asx)
for (int i = 0; i < nb; i++) {
Expand Down Expand Up @@ -5984,6 +5986,8 @@ void ggml_vec_dot_q2_K_q8_K(int n, float * restrict s, size_t bs, const void * r
vector signed int vsumi6 = vec_splats((int32_t)0);
vector signed int vsumi7 = vec_splats((int32_t)0);

const uint8_t * restrict q2 = x[i].qs;
const int8_t * restrict q8 = y[i].qs;

for (int j = 0; j < QK_K/128; ++j) {
__builtin_prefetch(q2, 0, 1);
Expand Down