hexu
1 Followers · 0 Following
  • Joined on 2025-12-08
Repositories
3
 Projects Packages Public Activity Starred Repositories
hexu created branch feature/scalar-tmem-softmax-vector in wu-arch/kernels 2026-07-10 16:29:30 +08:00
hexu pushed to wu-blackwell-fp8-e4m3 at wu-arch/radiance 2026-07-10 16:29:07 +08:00
ff825123de chore: update vortex scalar tmem softmax
hexu pushed to wu-blackwell-fp8-e4m3 at wu-arch/vortex 2026-07-10 16:28:57 +08:00
dff1107bf5 feat: add scalar tmem softmax pipeline
9251ba0a24 feat: add scalar fexp support
97a1eff701 Add scalar TMEM load-store path
Compare 3 commits »
hexu created branch wu-blackwell-fp8-e4m3 in wu-arch/vortex 2026-07-10 16:28:56 +08:00
hexu pushed to wu-tmem-bank-scalar-softmax at wu-arch/chipyard 2026-07-03 16:54:05 +08:00
hexu created branch wu-tmem-bank-scalar-softmax in wu-arch/chipyard 2026-07-03 16:54:05 +08:00
hexu pushed to wu-blackwell-4lane-tc at wu-arch/radiance 2026-07-03 16:53:12 +08:00
f88085331e Save pre-TMEM-bank Radiance changes
hexu pushed to wu-tmem-bank-scalar-softmax at wu-arch/radiance 2026-07-03 16:52:15 +08:00
hexu created branch wu-tmem-bank-scalar-softmax in wu-arch/radiance 2026-07-03 16:52:15 +08:00
hexu pushed to wu-blackwell-fp8-e4m3 at wu-arch/chipyard 2026-07-03 16:48:41 +08:00
5c88e9b67f chore: update radiance fp8 blackwell pointer
55b7fa2007 chore: update radiance fexp support
bdca34b2cc Update Radiance for scalar TMEM support
b0a79c26ee Update radiance pre-TMEM-bank baseline
Compare 4 commits »
hexu created branch wu-blackwell-fp8-e4m3 in wu-arch/chipyard 2026-07-03 16:48:41 +08:00
hexu pushed to wu-blackwell-fp8-e4m3 at wu-arch/kernels 2026-07-03 16:48:18 +08:00
8a15e5251e fix: match blackwell fp8 fragment width
3f7ce1f1c9 feat: add blackwell fp8 e4m3 kernel
f1aa1303d2 feat: add flash pipeline kernel cases
d6fbd447c3 Add Wu TMEM FlashAttention validation cases
Compare 4 commits »
hexu created branch wu-blackwell-fp8-e4m3 in wu-arch/kernels 2026-07-03 16:48:18 +08:00
hexu pushed to wu-blackwell-fp8-e4m3 at wu-arch/radiance 2026-07-03 16:47:45 +08:00
ca4c48251d feat: use fp8 multiply for blackwell bwgmma
12f5c6d92d feat: switch blackwell bwgmma inputs to fp8
68a7f66046 feat: add fp8 tensor dot product path
2afb96bb14 feat: add fp8 e4m3 decode support
007350fd5a feat: include vortex fexp RTL
Compare 7 commits »
hexu created branch wu-blackwell-fp8-e4m3 in wu-arch/radiance 2026-07-03 16:47:45 +08:00
hexu pushed to main at hexu/tatu 2026-06-29 15:05:17 +08:00
b6afa61e66 feat: implement privileged mode support
hexu pushed to main at hexu/tatu 2026-06-27 15:18:32 +08:00
a32db39c80 fix: pass remaining riscv isa tests
a2e0126199 fix: resolve OoO simulation timeout
502803c37f Initial Chisel core implementation
hexu created branch main in hexu/tatu 2026-06-27 15:18:22 +08:00
hexu created repository hexu/tatu 2026-06-27 15:14:59 +08:00
hexu pushed to wu-blackwell-4lane-tc at wu-arch/chipyard 2026-05-27 16:47:53 +08:00
d764de5115 Update radiance Blackwell SMEM alignment fix