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

debug: more indices for SF layout dump

Browse Source
This commit is contained in:
biondizzle
2026-05-14 16:26:15 +00:00
parent c4b5b52a33
commit 1ef2fbc2fd
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/nvfp4_megamoe_kernel/cutlass_nvfp4_gemm/cutlass_nvfp4_gemm.cu
View File

@@ -136,7 +136,7 @@ __global__ void remap_sf_to_cutlass_kernel(
// This avoids garbled output from concurrent threads
if (threadIdx.x == 0 && blockIdx.x == 0) {
// Print a series of indices to understand the decomposition
int debug_indices[] = {0, 1, 2, 3, 4, 15, 16, 31, 32, 63, 64, 127, 128, 255, 256, 511, 512, 512*448-1};
int debug_indices[] = {0, 1, 4, 16, 64, 128, 256, 511, 512, 513, 516, 1024, 2048, 4096, 8192, 65536, 131072};
int n_debug = sizeof(debug_indices) / sizeof(debug_indices[0]);
for (int di = 0; di < n_debug; di++) {
int didx = debug_indices[di];