Skip to content

(Prototype) q1_0 nrc = 2 and diabolic tiles branches #21

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

Draft
pl752 wants to merge 9 commits into
base: master
Choose a base branch
from
Draft

(Prototype) q1_0 nrc = 2 and diabolic tiles branches #21

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
195593b
Implemented optimized q1_0 dot for x86 and generic
pl752 Apr 7, 2026
e29cd48
Removed redundant helper definition
pl752 Apr 7, 2026
4d0a787
Initial AVX2 nrc2 experiment
pl752 Apr 9, 2026
7ee5400
Experiment with rectangular tile
pl752 Apr 9, 2026
50d5d39
Implemented prototype nrc2 branches for SSSE3 and AVX
pl752 Apr 9, 2026
e5ce6d4
Hooked up special q1_0 4x2 dot kernel for AVX2/512
pl752 Apr 9, 2026
34f547d
Implemented 4x4 dot fastpath
pl752 Apr 9, 2026
bec8051
Added 2x1 kernel for slightly faster q1_0 decoding
pl752 Apr 9, 2026
a88ee90
Completed x86 SIMD extremities demo experiments
pl752 Apr 9, 2026
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
1 change: 0 additions & 1 deletion ggml/src/ggml-cpu/arch-fallback.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,7 +83,6 @@
#elif defined(__x86_64__) || defined(__i386__) || defined(_M_IX86) || defined(_M_X64)
// quants.c
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
// repack.cpp
#define ggml_quantize_mat_q8_0_4x4_generic ggml_quantize_mat_q8_0_4x4
#define ggml_quantize_mat_q8_K_4x4_generic ggml_quantize_mat_q8_K_4x4
Expand Down
Loading