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

metal : fix hardcoded constants in mul_vec_q_n_f32

Browse source
This commit is contained in:
Cebtenzzre 2023-09-29 18:20:19 -04:00
parent 40e07a60f9
commit f12863ed38
1 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

14
ggml-metal.metal
View file

@ -435,31 +435,31 @@ void mul_vec_q_n_f32(device const void * src0, device const float * src1, device
const uint offset0 = first_row * nb + im/gqa*(nb*ne0);
device const block_q_type * x = (device const block_q_type *) src0 + offset0;
device const float * y = (device const float *) src1 + r1*ne10 + im*ne00*ne1;
float yl[16]; // src1 vector cache
float yl[QK4_0/2]; // src1 vector cache
float sumf[nr]={0.f};
const int ix = tiisg/2;
const int il = 8*(tiisg%2);
const int il = QK4_0/4*(tiisg%2);
device const float * yb = y + ix * QK4_0 + il;
// each thread in a SIMD group deals with half a block.
for (int ib = ix; ib < nb; ib += nw/2) {
float sumy = 0;
for (int i = 0; i < 8; i += 2) {
for (int i = 0; i < QK4_0/4; i += 2) {
sumy += yb[i] + yb[i+1];
yl[i+0] = yb[i+ 0];
yl[i+1] = yb[i+ 1]/256.f;
sumy += yb[i+16] + yb[i+17];
yl[i+8] = yb[i+16]/16.f;
yl[i+9] = yb[i+17]/4096.f;
sumy += yb[i+QK4_0/2] + yb[i+QK4_0/2+1];
yl[i+QK4_0/4] = yb[i+QK4_0/2] /16.f;
yl[i+QK4_0/4+1] = yb[i+QK4_0/2+1]/4096.f;
}
for (int row = 0; row < nr; row++) {
sumf[row] += block_q_n_dot_y(x+ib+row*nb, sumy, yl, il);
}
yb += QK4_0 * 16;
yb += QK4_0 * nw/2;
}
for (int row = 0; row < nr; ++row) {