happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Added if condition to support only vector length 256.

This commit is contained in:
Vithule, Prashant 2026-01-21 09:28:15 +00:00 committed by Vithulep
parent 57b218d447
commit 6363036f2b
1 changed files with 298 additions and 296 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cpu/arch/arm/repack.cpp
View File

@ -3227,6 +3227,7 @@ void ggml_gemm_q4_K_8x8_q8_K(int n,
UNUSED(blocklen); UNUSED(blocklen);
#if defined(__aarch64__) && defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8) #if defined(__aarch64__) && defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8)
if (svcntb()*8 == 256) {
constexpr int q8_k_blocklen = 4; constexpr int q8_k_blocklen = 4;
const svuint8_t m4b_1 = svdup_n_u8(0x0f); const svuint8_t m4b_1 = svdup_n_u8(0x0f);
// 8 accumulators: 2 row pairs × 4 col pairs // 8 accumulators: 2 row pairs × 4 col pairs
@ -3551,6 +3552,7 @@ void ggml_gemm_q4_K_8x8_q8_K(int n,
} // for x } // for x
} // for y } // for y
return; return;
}
#elif defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8) #elif defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
constexpr int q8_k_blocklen = 4; constexpr int q8_k_blocklen = 4;