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

small improvements

This commit is contained in:
Ruben Ortlam 2026-03-19 15:09:51 +01:00
parent 2df9eca0cb
commit cea4c85a97
1 changed files with 10 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp
View File

@ -67,7 +67,7 @@ shared FLOAT_TYPEV4 kvsh[SHMEM_STAGING != 0 ? Bc * kvsh_stride : 1];
#ifdef MMQ
shared block_a_cache kblocksh[Bc * qf_stride];
shared block_a_cache kblocksh[SHMEM_STAGING != 0 ? Bc * qf_stride : 1];
#endif
shared vec4 occupancy_limiter[LIMIT_OCCUPANCY_SHMEM > 0 ? LIMIT_OCCUPANCY_SHMEM : 1];
@ -147,12 +147,19 @@ void main() {
Qf[buf_ib].qs[buf_iqs] = pack32(i8vec4(vals));
#ifdef DATA_A_Q8_0
if (buf_iqs == 0) {
// sum is only needed for q4_0
Qf[buf_ib].ds = FLOAT_TYPEV2(qd, 0.0);
}
#else // DATA_A_Q4_0
const FLOAT_TYPE thread_sum = vals.x + vals.y + vals.z + vals.w;
const FLOAT_TYPE sum = subgroupClusteredAdd(thread_sum, 8);
if (buf_iqs == 0) {
Qf[buf_ib].ds = FLOAT_TYPEV2(qd, sum * qd);
}
#endif
#endif
}
barrier();
@ -431,8 +438,8 @@ void main() {
}
} else {
[[unroll]] for (uint32_t d = 0; d < 4; d++) {
uint vui = (uint(k_packed.k_data_packed16[k_offset + ib].qs[iqs / 2 + d * 2 + 1]) << 16) |
uint(k_packed.k_data_packed16[k_offset + ib].qs[iqs / 2 + d * 2 + 0]);
uint vui = pack32(u16vec2(k_packed.k_data_packed16[k_offset + ib].qs[iqs / 2 + d * 2 + 0],
k_packed.k_data_packed16[k_offset + ib].qs[iqs / 2 + d * 2 + 1]));
k_quants[d ] = int32_t( vui & 0x0F0F0F0F);
k_quants[d + 4] = int32_t((vui >> 4) & 0x0F0F0F0F);