Hongtao Yu 315528f349 Use full-vectorized load instructions for load vectorization (#445) ...

* Stablize load vectorization

* fix test failures

* Shared one mask check when decomposing a load

* Revert "fix test failures"

This reverts commit 75a461ae3ea4fdd5105dc73675582368eda80bc6.

* Emit vectorized loads

* Fix test failures due to using vectorized load

2024-01-18 13:34:05 -06:00

..

ConvertLayoutOpToLLVM

[Backend] Refactor mfma selection (#441)

2024-01-16 21:05:35 -06:00

DotOpToLLVM

[Backend] Refactor mfma selection (#441)

2024-01-16 21:05:35 -06:00

BarrierOpToLLVM.cpp

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

BarrierOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

ClusterOpsToLLVM.cpp

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

ClusterOpsToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

CMakeLists.txt

Merge commit '721897fcc4f942aa97d2e9ba3787a5e213758177' into ifu-231108

2023-11-08 18:51:23 +00:00

ConvertLayoutOpToLLVM.cpp

Merge pull request #410 from ROCmSoftwarePlatform/ifu-231117

2023-12-15 09:09:40 -06:00

ConvertLayoutOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

DotOpToLLVM.cpp

Merge commit 'cb3d79a185e40c9d8a579bea07747a8a8d157d52' into ifu-231117

2023-11-17 20:42:12 +00:00

DotOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

ElementwiseOpToLLVM.cpp

support type conversion between fp8 formats and bf16/fp32 with HW instructions on MI300 (#414)

2024-01-15 17:14:49 -06:00

ElementwiseOpToLLVM.h

[BACKEND] Remove dead code (#2263)

2023-09-07 20:40:48 -07:00

GCNAsmFormat.cpp

Add licenses to AMD related files (#351)

2023-10-16 15:18:01 -05:00

LoadStoreOpToLLVM.cpp

Use full-vectorized load instructions for load vectorization (#445)

2024-01-18 13:34:05 -06:00

LoadStoreOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

PTXAsmFormat.cpp

Merge commit '36fc54b6f28168d3644808bfe299f1ba06a36272' into ifu230908-2

2023-10-02 18:01:04 +00:00

ReduceOpToLLVM.cpp

fixed warp size in lowering reduce op (#471)

2024-01-18 09:38:41 -06:00

ReduceOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

RegReallocOpToLLVM.cpp

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

RegReallocOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

ScanOpToLLVM.cpp

resolve some merge conflicts

2023-11-06 23:13:10 +00:00

ScanOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

TensorPtrOpsToLLVM.cpp

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

TensorPtrOpsToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00

TritonGPUToLLVM.cpp

Add support for MFMA layout to view_slice op (#442)

2024-01-03 12:13:36 -06:00

TritonGPUToLLVM.h

[BACKEND] Improve printf. (#2532)

2023-10-25 08:47:55 +00:00

TritonGPUToLLVMBase.h

[Backend] Refactor mfma selection (#441)

2024-01-16 21:05:35 -06:00

TritonGPUToLLVMPass.cpp

[Backend] Refactor mfma selection (#441)

2024-01-16 21:05:35 -06:00

TypeConverter.cpp

Merge pull request #410 from ROCmSoftwarePlatform/ifu-231117

2023-12-15 09:09:40 -06:00

TypeConverter.h

support gemm fp8/fp16 mixed input (#333)

2023-09-27 08:00:31 -05:00

Utility.cpp

Add shortcut for creation fp16/bfp16 (#468)

2024-01-17 11:29:19 -06:00

Utility.h

Add shortcut for creation fp16/bfp16 (#468)

2024-01-17 11:29:19 -06:00

ViewOpToLLVM.cpp

[BACKEND] Fix unsupported view op created during optimizations (#2510)

2023-10-18 16:37:13 +01:00

ViewOpToLLVM.h

Initial code merge of Hopper support (#2036)

2023-08-07 09:53:04 +08:00