Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
cee7971
feat(gdn): add native SM100 CP MN precompute kernel
guangyunh-nv Aug 3, 2026
c002253
feat(gdn): add native SM100 CP prefill kernel
guangyunh-nv Aug 3, 2026
8fd34e0
feat(gdn): add SM100 CP launch pipeline
guangyunh-nv Aug 3, 2026
e5ef7aa
feat(gdn): enable SM100 context-parallel prefill
guangyunh-nv Aug 3, 2026
8d98849
test(gdn): cover SM100 context-parallel prefill
guangyunh-nv Aug 3, 2026
74e65b0
perf(gdn): simplify SM100 CP dispatch heuristic
guangyunh-nv Aug 3, 2026
d118fbe
feat(gdn): support typed SM100 CP states
guangyunh-nv Aug 3, 2026
02648d5
perf(gdn): account for batched CP work
guangyunh-nv Aug 3, 2026
cfe91cb
perf(gdn): add UTCMMA CP state fixup kernels
guangyunh-nv Aug 3, 2026
a2605a9
refactor(gdn): rename CP state index variables
guangyunh-nv Aug 3, 2026
58af7b7
feat(gdn): complete SM100 CP state fixup dispatch
guangyunh-nv Aug 3, 2026
57f79d7
fix(gdn): restore SM100 CP kernel-level contracts
guangyunh-nv Aug 3, 2026
c3ed114
perf(gdn): prefetch SM100 UTC64 fixup transfers
guangyunh-nv Aug 3, 2026
29f6e06
fix(gdn): only forward indexed state when present
guangyunh-nv Aug 4, 2026
f4d1a87
fix(gdn): make SM120 CP final state optional
guangyunh-nv Aug 4, 2026
bb694ec
fix(gdn): make SM90 CP final state optional
guangyunh-nv Aug 4, 2026
66c515d
test(gdn): skip SM100 CP kernels before CUDA 13
guangyunh-nv Aug 4, 2026
b147c75
chore(gdn): satisfy pre-commit checks
guangyunh-nv Aug 4, 2026
11616db
fix(gdn): address SM100 CP dispatch review
guangyunh-nv Aug 4, 2026
6cb2e70
test(gdn): cover SM100 CP state indices
guangyunh-nv Aug 4, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion flashinfer/gdn_kernels/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -71,9 +71,10 @@
get_mtp_config = None # type: ignore

try:
from .blackwell import chunk_gated_delta_rule_sm100
from .blackwell import chunk_gated_delta_rule_sm100, cp_delta_rule_dsl_sm100
except (ImportError, RuntimeError):
chunk_gated_delta_rule_sm100 = None # type: ignore
cp_delta_rule_dsl_sm100 = None # type: ignore

try:
from .delta_rule_dsl import (
Expand Down Expand Up @@ -104,5 +105,6 @@
"chunk_gated_delta_rule_sm100",
"chunk_gated_delta_rule_sm120",
"cp_delta_rule_dsl_sm90",
"cp_delta_rule_dsl_sm100",
"cp_delta_rule_dsl_sm120",
]
6 changes: 6 additions & 0 deletions flashinfer/gdn_kernels/blackwell/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,12 @@
except (ImportError, RuntimeError):
chunk_gated_delta_rule_sm100 = None # type: ignore

try:
from .gdn_cp_prefill import cp_delta_rule_dsl_sm100
except (ImportError, RuntimeError):
cp_delta_rule_dsl_sm100 = None # type: ignore

__all__ = [
"chunk_gated_delta_rule_sm100",
"cp_delta_rule_dsl_sm100",
]
Loading
Loading