biondizzle

0 Followers · 0 Following

• Joined on 2025-12-10

Repositories

25

Projects Packages Public Activity Starred Repositories

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 18:26:17 +00:00

593584fc8d 🎉 Mark D1.3 as SOLVED! SMEM-P rank mismatch fixed, enables hd>64 support

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:36:23 +00:00

0bee39d2d5 Fix rP scope issue: use rP_bf16.iterator instead of rP.iterator

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:35:59 +00:00

0b09e7e4a2 Fix duplicate else: line in SMEM-P block

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:35:27 +00:00

018a961c01 SMEM-P: Use QK C-fragment layout instead of TMEM layout to fix rank mismatch

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:34:52 +00:00

7a74fac11f Fix sP_2d definition for tSMEM_CPYsP

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:34:11 +00:00

ffafd47d07 Remove debug print lines referencing deleted sP_2d

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:33:41 +00:00

6a078b88d9 Remove duplicate sP_2d line causing indentation error

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:33:26 +00:00

1fd3670ca4 SMEM-P: Implement rank mismatch fix by reshaping source tensor

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:32:41 +00:00

1ede7a4c1f Fix indentation and variable reference in SMEM-P debug code

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:32:02 +00:00

ff004a9c55 Fix duplicate sP_2d definition causing indentation error

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:31:26 +00:00

303c2c5275 Clean debug SMEM-P path to understand rank mismatch

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:30:49 +00:00

ef0ac2c187 Fix syntax error in debug code

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:30:32 +00:00

e400c935a9 Add comprehensive debugging to SMEM-P path to diagnose rank mismatch

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:27:49 +00:00

fe3b1abf22 Update STAGE_D.md checklist with current progress and lessons learned

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:26:32 +00:00

77b0f5824b Add more debug prints for sP shapes

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:25:51 +00:00

303df9b8c4 Add debug prints to SMEM-P path to understand rank mismatch

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:21:15 +00:00

1f4fe3e404 Fix SMEM-P copy rank mismatch (use rP_bf16 directly instead of group_modes)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:20:38 +00:00

162bf51d64 D1.3: Implement SMEM-P path (write P to SMEM via tiled_smem_copy instead of zeroing sP)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:19:37 +00:00

9346a81d43 D1.3: Implement SMEM-P path (write P to SMEM via tiled_smem_copy instead of zeroing sP)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 09:04:10 +00:00

1d1de22775 Stage D1: Multi-PV-tile support for hd>256 (tcgen05 MMA max N=256)