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

More int mult, less float mult, worse performance

Browse source
This commit is contained in:
JohannesGaessler 2023-05-12 09:11:47 +02:00
parent d882d1c2fe
commit e7b9d97bae
1 changed files with 12 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

22
ggml-cuda.cu
View file

@ -235,8 +235,8 @@ template <int block_size> static __global__ void dequantize_mul_mat_q4_0(const v
__shared__ float tmp[block_size]; // separate sum for each thread __shared__ float tmp[block_size]; // separate sum for each thread
tmp[tid] = 0; tmp[tid] = 0;
for (int i = 0; i < ncols/block_size; i += 2) { for (int i = 0; i < ncols/block_size; i += 4) {
const int col = i*block_size + 2*tid; const int col = i*block_size + 4*tid;
// dequantize // dequantize
const float d0 = x[(row*ncols + col)/QK4_0].d; const float d0 = x[(row*ncols + col)/QK4_0].d;
@ -245,19 +245,21 @@ template <int block_size> static __global__ void dequantize_mul_mat_q4_0(const v
const uint8_t * p0 = x[(row*ncols + col)/QK4_0].qs; const uint8_t * p0 = x[(row*ncols + col)/QK4_0].qs;
const int8_t * p1 = y[col/QK8_0].qs; const int8_t * p1 = y[col/QK8_0].qs;
const uint8_t vui0 = p0[((row*ncols + col)%QK4_0)/2]; const uint8_t vui00 = p0[((row*ncols + col)%QK4_0)/2];
const uint8_t vui01 = p0[((row*ncols + col + 2)%QK4_0)/2];
const int vi10 = p1[(col + 0)%QK8_0]; const int vi10 = p1[(col + 0)%QK8_0];
const int vi11 = p1[(col + 1)%QK8_0]; const int vi11 = p1[(col + 1)%QK8_0];
const int vi12 = p1[(col + 2)%QK8_0];
const int vi13 = p1[(col + 3)%QK8_0];
const int vi00 = vui0 & 0xF; const int vi00 = vui00 & 0xF;
const int vi01 = vui0 >> 4; const int vi01 = vui00 >> 4;
const int vi02 = vui01 & 0xF;
const float v0 = (vi00 - 8)*vi10*d0*d1; const int vi03 = vui01 >> 4;
const float v1 = (vi01 - 8)*vi11*d0*d1;
// matrix multiplication // matrix multiplication
tmp[tid] += v0; const int sumi = (vi00 - 8)*vi10 + (vi01 - 8)*vi11 + (vi02 - 8)*vi12 + (vi03 - 8)*vi13;
tmp[tid] += v1; tmp[tid] += sumi*d0*d1;
} }
// sum up partial sums and write back result // sum up partial sums and write back result