Skip to content

Q1_0 repack kernels for Arm NEON+DP #34

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
khosravipasha merged 7 commits into from
Jun 8, 2026
+762 −0
Merged

Q1_0 repack kernels for Arm NEON+DP #34

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
33055fc
Implemented ARM NEON DP q1 4x4 repack
pl752 May 17, 2026
b162fdc
Hoisted out scaling by b_d in gemm
pl752 May 17, 2026
5e677e6
Added 4x8 NEON I8MM repack kernels
pl752 May 18, 2026
00289b0
Cleanup for q1 arm repack
pl752 May 18, 2026
e4c4c4a
Added missing aliases for arch fallback
pl752 May 20, 2026
8ac75b2
Corrected unused var statements
pl752 May 20, 2026
c2145be
Extended table guard condition to account for i8mm w/o dp build
pl752 Jun 4, 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
28 changes: 28 additions & 0 deletions ggml/src/ggml-cpu/arch-fallback.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_0_8x8_q8_0_generic ggml_gemm_q4_0_8x8_q8_0
Expand All @@ -70,6 +72,8 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__aarch64__) || defined(__arm__) || defined(_M_ARM) || defined(_M_ARM64)
// repack.cpp
#define ggml_quantize_mat_q8_K_4x4_generic ggml_quantize_mat_q8_K_4x4
Expand Down Expand Up @@ -97,6 +101,8 @@
#define ggml_gemv_mxfp4_4x4_q8_0_generic ggml_gemv_mxfp4_4x4_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_K_8x4_q8_K_generic ggml_gemm_q4_K_8x4_q8_K
Expand All @@ -108,6 +114,8 @@
#define ggml_gemm_mxfp4_4x4_q8_0_generic ggml_gemm_mxfp4_4x4_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__POWERPC__) || defined(__powerpc__)
// ref: https://github.com/ggml-org/llama.cpp/pull/14146#issuecomment-2972561679
// quants.c
Expand Down Expand Up @@ -138,6 +146,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_0_8x8_q8_0_generic ggml_gemm_q4_0_8x8_q8_0
Expand All @@ -154,6 +164,8 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__loongarch64)
// quants.c
#define quantize_row_q8_K_generic quantize_row_q8_K
Expand Down Expand Up @@ -184,6 +196,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_0_8x8_q8_0_generic ggml_gemm_q4_0_8x8_q8_0
Expand All @@ -200,6 +214,8 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__riscv)
// quants.c
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
Expand All @@ -225,6 +241,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q2_K_8x8_q8_K_generic ggml_gemm_q2_K_8x8_q8_K
Expand All @@ -240,6 +258,8 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__s390x__)
// quants.c
#define quantize_row_q8_K_generic quantize_row_q8_K
Expand Down Expand Up @@ -276,6 +296,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_0_8x8_q8_0_generic ggml_gemm_q4_0_8x8_q8_0
Expand All @@ -292,6 +314,8 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#elif defined(__wasm__)
// quants.c
#define ggml_vec_dot_q4_1_q8_1_generic ggml_vec_dot_q4_1_q8_1
Expand Down Expand Up @@ -330,6 +354,8 @@
#define ggml_gemv_mxfp4_8x8_q8_0_generic ggml_gemv_mxfp4_8x8_q8_0
#define ggml_gemv_q8_0_4x4_q8_0_generic ggml_gemv_q8_0_4x4_q8_0
#define ggml_gemv_q8_0_4x8_q8_0_generic ggml_gemv_q8_0_4x8_q8_0
#define ggml_gemv_q1_0_4x4_q8_0_generic ggml_gemv_q1_0_4x4_q8_0
#define ggml_gemv_q1_0_4x8_q8_0_generic ggml_gemv_q1_0_4x8_q8_0
#define ggml_gemm_q4_0_4x4_q8_0_generic ggml_gemm_q4_0_4x4_q8_0
#define ggml_gemm_q4_0_4x8_q8_0_generic ggml_gemm_q4_0_4x8_q8_0
#define ggml_gemm_q4_0_8x8_q8_0_generic ggml_gemm_q4_0_8x8_q8_0
Expand All @@ -346,4 +372,6 @@
#define ggml_gemm_mxfp4_8x8_q8_0_generic ggml_gemm_mxfp4_8x8_q8_0
#define ggml_gemm_q8_0_4x4_q8_0_generic ggml_gemm_q8_0_4x4_q8_0
#define ggml_gemm_q8_0_4x8_q8_0_generic ggml_gemm_q8_0_4x8_q8_0
#define ggml_gemm_q1_0_4x4_q8_0_generic ggml_gemm_q1_0_4x4_q8_0
Comment thread
pl752 marked this conversation as resolved.
#define ggml_gemm_q1_0_4x8_q8_0_generic ggml_gemm_q1_0_4x8_q8_0
#endif
Loading