Skip to content

ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) #1179

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 10 commits into from
Apr 25, 2023
Merged

ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) #1179

Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
f83c321
ggml : add Q8_0 quantization format (rename the old one to Q8_1)
ggerganov Apr 25, 2023
79cfdf5
tests : fix test-quantize-fns
ggerganov Apr 25, 2023
d8bf720
ggml : finalize Q8_0 implementation
ggerganov Apr 25, 2023
6496b79
ggml : use q4_0_q8_0 and q4_2_q8_0
ggerganov Apr 25, 2023
88618ab
ggml : fix Q8_0 dot product bug (ARM)
ggerganov Apr 25, 2023
6e0f0b6
ggml : Q8_0 unroll x2
ggerganov Apr 25, 2023
46fc696
ggml : fix bug - using wrong block type
ggerganov Apr 25, 2023
91bfa51
ggml : extend quantize_fns_t with "vec_dot_type"
ggerganov Apr 25, 2023
4ddb983
ggml : fix Q8_0 to use 255 values out of 256
ggerganov Apr 25, 2023
e8c3731
ggml : fix assert using wrong QK4_2 instead of QK4_3
ggerganov Apr 25, 2023
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
ggml : fix Q8_0 dot product bug (ARM)
  • Loading branch information
@ggerganov
ggerganov committed Apr 25, 2023
commit 88618ab7f5935cb96ce487dc72262a5ad625531f
4 changes: 0 additions & 4 deletions ggml.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3094,10 +3094,6 @@ static void ggml_vec_dot_q8_0_q8_0(const int n, float * restrict s, const void *
sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_0, v1_0),
vdotq_s32(vdupq_n_s32(0), v0_1, v1_1))), x0->d*y0->d);

sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_0, v1_1),
vdotq_s32(vdupq_n_s32(0), v0_1, v1_0))), x0->d*y0->d);
#else
const int16x8_t p0l = vmull_s8(vget_low_s8 (v0_0), vget_low_s8 (v1_0));
const int16x8_t p0h = vmull_s8(vget_high_s8(v0_0), vget_high_s8(v1_0));
Expand Down