block_q8_K Struct Reference

#include <ckernel_quant.h>

Data Fields
int16_t	bsums [256/16]
float	d
int8_t	qs [256]

Detailed Description

Definition at line 207 of file ckernel_quant.h.

Field Documentation

bsums

int16_t block_q8_K::bsums[256/16]

Definition at line 210 of file ckernel_quant.h.

Referenced by accum_q4_k_packed_meta_x16_q8_k_block(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse_chunk4(), accum_q4_k_packed_meta_x8_q8_k_block(), accum_q4_k_packed_meta_x8_q8_k_block_mreuse(), accum_q4_k_packed_meta_x8_q8_k_gemv_block(), accum_q4_k_packed_meta_x8_q8_k_superblock(), accum_q4_k_packed_meta_x8_q8_k_superblock_rows(), accum_q4_k_packed_u8_x16_q8_k_block(), accum_q4_k_packed_vnni_x16_q8_k_16m_superblock(), accum_q4_k_packed_vnni_x16_q8_k_gemv_block(), accum_q4_k_packed_vnni_x8_q8_k_4m_superblock(), dot_q4_k_packed_meta_q8_k_block(), dot_q4_k_packed_u8_q8_k_block(), dot_q4_k_packed_vnni_x8_q8_k_compact_order(), dot_q5_0_q8_k_32_sse(), dot_q5_k_q8_k_row(), fused_rmsnorm_linear_q4k(), gemm_nt_q4_k_packed_meta_x16_q8_k_llama_order(), gemv_q4_k_q8_k_sse(), quantize_row_q8_k_ref(), quantize_row_q8_k_sse(), and rmsnorm_q8_k_fused().

d

float block_q8_K::d

Definition at line 208 of file ckernel_quant.h.

Referenced by accum_q4_k_packed_meta_x16_q8_k_block(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse_chunk4(), accum_q4_k_packed_meta_x8_q8_k_block(), accum_q4_k_packed_meta_x8_q8_k_block_mreuse(), accum_q4_k_packed_meta_x8_q8_k_gemv_block(), accum_q4_k_packed_meta_x8_q8_k_superblock(), accum_q4_k_packed_u8_x16_q8_k_block(), accum_q4_k_packed_vnni_x16_q8_k_16m_superblock(), accum_q4_k_packed_vnni_x16_q8_k_gemv_block(), ck_debug_check_q8k(), dot_q4_k_packed_meta_q8_k_block(), dot_q4_k_packed_u8_q8_k_block(), dot_q4_k_q8_k_ref(), dot_q5_0_q8_k_32_sse(), dot_q5_k_q8_k_row(), dot_q6_k_q8_k_256_sse(), dot_q6_k_q8_k_ref(), fused_rmsnorm_linear_q4k(), gemm_nt_q4_k_packed_meta_x16_q8_k_llama_order(), gemv_q4_k_q8_k_sse(), quantize_row_q8_k_ref(), quantize_row_q8_k_sse(), and rmsnorm_q8_k_fused().

qs

int8_t block_q8_K::qs[256]

Definition at line 209 of file ckernel_quant.h.

Referenced by accum_q4_k_packed_meta_x16_q8_k_block(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse(), accum_q4_k_packed_meta_x16_q8_k_block_mreuse_chunk4(), accum_q4_k_packed_meta_x8_q8_k_block(), accum_q4_k_packed_meta_x8_q8_k_block_mreuse(), accum_q4_k_packed_meta_x8_q8_k_gemv_block(), accum_q4_k_packed_meta_x8_q8_k_superblock(), accum_q4_k_packed_u8_x16_q8_k_block(), accum_q4_k_packed_vnni_x16_q8_k_16m_superblock(), accum_q4_k_packed_vnni_x16_q8_k_gemv_block(), dot_q4_k_packed_meta_q8_k_block(), dot_q4_k_packed_u8_q8_k_block(), dot_q4_k_q8_k_ref(), dot_q5_0_q8_k_32_sse(), dot_q5_k_q8_k_row(), dot_q6_k_q8_k_256_sse(), dot_q6_k_q8_k_ref(), fused_rmsnorm_linear_q4k(), gemm_nt_q4_k_packed_meta_x16_q8_k_llama_order(), gemv_q4_k_q8_k_sse(), quantize_row_q8_k_ref(), quantize_row_q8_k_sse(), and rmsnorm_q8_k_fused().

---

The documentation for this struct was generated from the following file:

• ckernel_quant.h