vulkan: preprocess mul_mat_id experts and discard workgroups more quickly (#18352)

Run a preprocess to count how many times each expert is used, and use this to
quickly discard workgroups that aren't needed.
This commit is contained in:
Jeff Bolz
2025-12-26 16:12:58 -06:00
committed by GitHub
parent 7ac8902133
commit c9ced4910b
7 changed files with 294 additions and 184 deletions
@@ -92,6 +92,7 @@ layout (binding = 2) writeonly buffer D {D_TYPE data_d[];};
#ifdef MUL_MAT_ID
layout (binding = 3) readonly buffer IDS {int data_ids[];};
layout (binding = 4) readonly buffer Counts {int data_expert_count[];};
shared u16vec4 row_ids[BN];
@@ -134,6 +135,8 @@ void load_row_ids(uint expert_idx, bool nei0_is_pow2, uint ic) {
uint ids[16];
uint iter = 0;
uint expert_count = data_expert_count[expert_idx];
for (uint j = 0; j < num_elements; j += BLOCK_SIZE) {
// prefetch up to 16 elements
if (iter == 0) {
@@ -181,7 +184,7 @@ void load_row_ids(uint expert_idx, bool nei0_is_pow2, uint ic) {
}
_ne1 += total;
iter &= 15;
if (_ne1 >= (ic + 1) * BN) {
if (_ne1 >= (ic + 1) * BN || _ne1 == expert_count) {
break;
}
}
@@ -191,7 +194,13 @@ void load_row_ids(uint expert_idx, bool nei0_is_pow2, uint ic) {
void main() {
const uint tid = gl_LocalInvocationIndex;
const uint ic = gl_WorkGroupID.y;
#ifdef MUL_MAT_ID
const uint expert_idx = gl_GlobalInvocationID.z;
if (ic * BN >= data_expert_count[expert_idx]) {
return;
}
// initialize to row 0 so we don't need to bounds check
if (tid < BN) {
row_ids[tid] = u16vec4(0);
@@ -205,9 +214,7 @@ void main() {
init_iq_shmem(gl_WorkGroupSize);
#endif
#ifdef MUL_MAT_ID
const uint expert_idx = gl_GlobalInvocationID.z;
#else
#ifndef MUL_MAT_ID
const uint batch_idx = gl_GlobalInvocationID.z;
const uint i13 = batch_idx / p.ne12;
@@ -222,7 +229,6 @@ void main() {
const uint blocks_m = (p.M + BM - 1) / BM;
const uint ir = gl_WorkGroupID.x % blocks_m;
const uint ik = gl_WorkGroupID.x / blocks_m;
const uint ic = gl_WorkGroupID.y;
#ifdef MUL_MAT_ID
if (bitCount(p.nei0) == 1) {