main

Ccandyhongfix: sync with upstream commit 523a1b2

03c21c85创建于 1月29日历史提交

文件	最后提交记录	最后更新时间
Pipeliner	[Backend] Bump to llvm/llvm-project@bc773632355b (#7881)	9 个月前
WarpSpecialization	[WS] reorder partition-loops and lower-aref (#7927)	9 个月前
AccelerateMatmul.cpp	[BACKEND] Don't use mmav5 with num warps < 4 (#7928)	9 个月前
CMakeLists.txt	[WS] Reimplement PartitionLoops pass supporting general control flow (#7415)	10 个月前
Coalesce.cpp	[NFC] Use RankedTensorType's clone and cloneWithEncoding member functions (#7464)	10 个月前
CoalesceAsyncCopy.cpp	[AMD] NFC: simplify pass/pattern constructor declaration (#7665)	10 个月前
CombineTensorSelectAndIf.cpp	Add support for Nvidia Blackwell GPUs (#5724)	1 年前
DecomposeScaledBlocked.cpp	[AMD] Support scaled dot for gfx12 (#7644)	10 个月前
F32DotTC.cpp	[BACKEND] bump llvm to 2f7ade4b (#5599)	1 年前
FuseNestedLoops.cpp	[RELAND][Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (#6514)	1 年前
HoistTMEMAlloc.cpp	[BACKEND] Fix wrong check in tmem_alloc canonicalization pattern (#7719)	10 个月前
OptimizeAccumulatorInit.cpp	[Utility] fix pass-by-reference addIterArgsToLoop API (#7029)	1 年前
OptimizeDotOperands.cpp	[LAYOUTS] Fix memdesc_subviews when we don't slice along the swizzling pattern (#7480)	10 个月前
OptimizeThreadLocality.cpp	[NFC] Use RankedTensorType's clone and cloneWithEncoding member functions (#7464)	10 个月前
Prefetch.cpp	[TritonGPU] Split MemDescSubview into MemDescIndex and MemDescSubslice (#7622)	10 个月前
ReduceDataDuplication.cpp	[LAYOUTS] Move order to LinearEncoding implementation (#6243)	1 年前
RemoveLayoutConversions.cpp	[NFC] Use RankedTensorType's clone and cloneWithEncoding member functions (#7464)	10 个月前
ReorderInstructions.cpp	[BACKEND] Fix dereference nullptr (#5944)	1 年前
Utility.cpp	[NVWS] Add pass to insert aref for TMA load (#7581)	10 个月前