iq2_kl: Metal GEMV - slightly better (46.5 t/s -> 47.2 t/s)

This commit is contained in:
Iwan Kawrakow
2025-07-12 07:46:09 +02:00
parent 1693cc6e60
commit e27708b341

View File

@@ -7265,7 +7265,7 @@ void kernel_mul_mv_iq2_kl_f32_impl(
device const char * cx0 = (device const char *) src0 + (first_row + offset0)*row_size;
device const float * y = (device const float *) src1 + r1*ne10 + im*ne00*ne1;
float yl[32];
float2 yl[16];
float sumf[N_DST]={0.f};
float drow[N_DST];
@@ -7298,9 +7298,10 @@ void kernel_mul_mv_iq2_kl_f32_impl(
for (int ib = ix; ib < nb; ib += 4) {
for (int i = 0; i < 16; ++i) {
yl[i+ 0] = y4[i+ 0];
yl[i+16] = y4[i+32];
device const float2 * y2 = (device const float2 *)y4;
for (int i = 0; i < 8; ++i) {
yl[i+0] = y2[i+ 0];
yl[i+8] = y2[i+16];
}
device const char * cx = cx0;
@@ -7316,7 +7317,7 @@ void kernel_mul_mv_iq2_kl_f32_impl(
device const uint16_t * ql = (device const uint16_t *)x->qs + 8*iq + 4*ir;
device const uint16_t * qh = (device const uint16_t *)x->qh + 4*ir;
float2 acc = {0.f};
float2 acc[2] = {0.f};
for (int l = 0; l < 4; ++l) {
uint16_t h = qh[l] >> 2*iq;
aux16[0] = ((ql[l] >> 0) & 0x0f0f) | ((h & 0x0101) << 4);
@@ -7324,12 +7325,12 @@ void kernel_mul_mv_iq2_kl_f32_impl(
for (int j = 0; j < 2; ++j) {
threadgroup const float2 & val1 = all_values[aux8[j+0]];
threadgroup const float2 & val2 = all_values[aux8[j+2]];
acc[0] += yl[4*l+2*j+ 0] * val1[0] + yl[4*l+2*j+ 1] * val1[1];
acc[1] += yl[4*l+2*j+16] * val2[0] + yl[4*l+2*j+17] * val2[1];
acc[0] += yl[2*l+j+0] * val1;
acc[1] += yl[2*l+j+8] * val2;
}
}
sumf[row] += drow[row] * (acc[0] * ls1 + acc[1] * ls2);
sumf[row] += drow[row] * ((acc[0][0] + acc[0][1]) * ls1 + (acc[1][0] + acc[1][1]) * ls2);
cx += row_size;