vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

CUDA: refactor and optimize IQ MMVQ (#8215)

Browse source 
* CUDA: refactor and optimize IQ MMVQ

* uint -> uint32_t

* __dp4a -> ggml_cuda_dp4a

* remove MIN_CC_DP4A checks

* change default

* try CI fix
This commit is contained in:
Johannes Gäßler 2024-07-01 20:39:06 +02:00 • committed by GitHub
parent dae57a1ebc
commit cb5fad4c6c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 406 additions and 487 deletions
Download patch file Download diff file Expand all files Collapse all files

21
ggml/src/ggml-sycl/vecdotq.hpp
View file

@ -820,7 +820,6 @@ vec_dot_iq2_xxs_q8_1(const void *__restrict__ vbq,
#if QK_K == 256
const block_iq2_xxs * bq2 = (const block_iq2_xxs *) vbq;
#if QR2_XXS == 8
const int ib32 = iqs;
const uint16_t * q2 = bq2->qs + 4*ib32;
const uint8_t * aux8 = (const uint8_t *)q2;
@ -838,26 +837,6 @@ vec_dot_iq2_xxs_q8_1(const void *__restrict__ vbq,
}
const float d = (float)bq2->d * (0.5f + aux32) * bq8_1[ib32].ds[0] * 0.25f;
return d * sumi;
#else
// iqs is 0...15
const int ib32 = iqs/2;
const int il = iqs%2;
const uint16_t * q2 = bq2->qs + 4*ib32;
const uint8_t * aux8 = (const uint8_t *)q2;
const uint8_t * grid1 = (const uint8_t *)(iq2xxs_grid + aux8[2*il+0]);
const uint8_t * grid2 = (const uint8_t *)(iq2xxs_grid + aux8[2*il+1]);
const uint32_t aux32 = q2[2] | (q2[3] << 16);
const float d = (float)bq2->d * (0.5f + (aux32 >> 28)) * bq8_1[ib32].ds[0] * 0.25f;
const uint8_t signs1 = ksigns_iq2xs[(aux32 >> 14*il) & 127];
const uint8_t signs2 = ksigns_iq2xs[(aux32 >> (14*il + 7)) & 127];
const int8_t * q8 = bq8_1[ib32].qs + 16*il;
int sumi1 = 0, sumi2 = 0;
for (int j = 0; j < 8; ++j) {
sumi1 += q8[j+0] * grid1[j] * (signs1 & kmask_iq2xs[j] ? -1 : 1);
sumi2 += q8[j+8] * grid2[j] * (signs2 & kmask_iq2xs[j] ? -1 : 1);
}
return d * (sumi1 + sumi2);
#endif
#else
assert(false);
return 0.f;