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

ggml : use WARP_SIZE/2 for argmax reduction offset

This commit is contained in:
Aadeshveer Singh 2025-12-16 17:00:18 +05:30
parent 7b1db3d3b7
commit c2f3f7a20e
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/argmax.cu
View File

@ -21,7 +21,7 @@ static __global__ void argmax_f32(const float * __restrict__ x, int32_t * __rest
}
#pragma unroll
for (int offset = 16; offset > 0; offset >>= 1) {
for (int offset = WARP_SIZE/2; offset > 0; offset >>= 1) {
const float val = __shfl_xor_sync(0xFFFFFFFF, maxval, offset, WARP_SIZE);
const int col = __shfl_xor_sync(0xFFFFFFFF, argmax, offset, WARP_SIZE);
if (val > maxval) {
@ -50,7 +50,7 @@ static __global__ void argmax_f32(const float * __restrict__ x, int32_t * __rest
argmax = shared_argmax[lane_id];
}
#pragma unroll
for (int offset = 16; offset > 0; offset >>= 1) {
for (int offset = WARP_SIZE/2; offset > 0; offset >>= 1) {
const float val = __shfl_xor_sync(0xFFFFFFFF, maxval, offset, WARP_SIZE);
const int col = __shfl_xor_sync(0xFFFFFFFF, argmax, offset, WARP_SIZE);
if (val > maxval) {