biondizzle/nvfp4-megamoe-kernel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

test: add 1 tmem_load back

Browse Source
This commit is contained in:
biondizzle
2026-05-28 09:51:21 +00:00
parent 61f19ce891
commit 4b129c146e
1 changed files with 8 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
tests/unit/test_tmem_cols.cu
View File

@@ -49,16 +49,14 @@ __global__ void test_tmem_loop(float* out) {
asm volatile("tcgen05.wait::st.sync.aligned;" ::: "memory");
__syncthreads();
// Read back — skip for now
// if (threadIdx.x < 32) {
// for (int c = 0; c < 4; c++) {
// uint32_t u0, u1, u2, u3;
// tmem_load(tb + c, u0, u1, u2, u3);
// float v0; memcpy(&v0, &u0, 4);
// if (lane == 0) out[c] = v0;
// }
// }
// __syncthreads();
// Read back — 1 column only
if (threadIdx.x < 32) {
uint32_t u0, u1, u2, u3;
tmem_load(tb + 0, u0, u1, u2, u3);
float v0; memcpy(&v0, &u0, 4);
if (lane == 0) out[0] = v0;
}
__syncthreads();
if (threadIdx.x < 32) tmem_dealloc(tb, 32);
}