maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/mlir/test/Dialect/Linalg
History
Nicolas Vasilache 54fafd17a7 [mlir][Linalg] Introduce canonicalization to remove dead LinalgOps 
When any of the memrefs in a structured linalg op has a zero dimension, it becomes dead.
This is consistent with the fact that linalg ops deduce their loop bounds from their operands.

Note however that this is not the case for the `tensor<0xelt_type>` which is a special convention
that must be lowered away into either `memref<elt_type>` or just `elt_type` before this
canonicalization can kick in.

Differential Revision: https://reviews.llvm.org/D85413
 		2020-08-06 06:08:46 -04:00
..
affine.mlir [MLIR][Linalg] Generate the right type of load/store when lowering max/min pooling ops 2020-07-04 14:55:02 +05:30
canonicalize.mlir [mlir][Linalg] Introduce canonicalization to remove dead LinalgOps 2020-08-06 06:08:46 -04:00
drop-unit-extent-dims.mlir Enable FileCheck -enable-var-scope by default in MLIR test 2020-06-12 00:43:09 +00:00
fold-unit-trip-loops.mlir Enable FileCheck -enable-var-scope by default in MLIR test 2020-06-12 00:43:09 +00:00
forward-vector-transfers.mlir [mlir][Vector] Degrade masking information when forwarding linalg.copy to vector.transfer 2020-07-15 02:32:45 -04:00
fusion-2-level.mlir [mlir][Linalg] Retire C++ MatmulOp in favor of a linalg-ods-gen'd op. 2020-06-16 10:46:35 -04:00
fusion-tensor.mlir [mlir][Linalg] Enable fusion of std.constant (producer) with 2020-07-27 09:51:20 -07:00
fusion.mlir [mlir][Linalg] Retire C++ MatmulOp in favor of a linalg-ods-gen'd op. 2020-06-16 10:46:35 -04:00
fusion_indexed_generic.mlir [MLIR][Standard] Make the `dim` operation index an operand. 2020-06-10 13:54:47 +00:00
hoisting.mlir [mlir][linalg] Improve aliasing approximation for hoisting transfer read/write 2020-07-10 14:55:04 -07:00
invalid.mlir [mlir][Linalg] Conv {1,2,3}D ops defined with TC syntax 2020-07-31 13:20:17 +02:00
llvm.mlir [mlir] switch the modeling of LLVM types to use the new mechanism 2020-08-04 14:29:25 +02:00
loops.mlir [mlir][Linalg] Conv {1,2,3}D ops defined with TC syntax 2020-07-31 13:20:17 +02:00
parallel_loops.mlir [MLIR][Standard] Make the `dim` operation index an operand. 2020-06-10 13:54:47 +00:00
promote.mlir [mlir][Linalg] Add an option to use Alloca instead of malloc/free pairs. 2020-07-01 09:44:01 -04:00
promotion_options.mlir [mlir][Linalg] Retire C++ MatmulOp in favor of a linalg-ods-gen'd op. 2020-06-16 10:46:35 -04:00
roundtrip.mlir [MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op 2020-07-28 12:34:19 +02:00
standard.mlir [MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op 2020-07-28 12:34:19 +02:00
tensors-to-buffers.mlir Change filecheck default to dump input on failure 2020-06-09 18:57:46 +00:00
tile.mlir [MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op 2020-07-28 12:34:19 +02:00
tile_conv.mlir [mlir][Linalg] Fixing bug in subview size computation in Linalg tiling. 2020-07-23 11:09:55 -07:00
tile_conv_padding.mlir Enable FileCheck -enable-var-scope by default in MLIR test 2020-06-12 00:43:09 +00:00
tile_indexed_generic.mlir [mlir][Linalg] Refactor linalg tiling 2020-05-20 09:39:56 -04:00
tile_parallel.mlir [MLIR][Standard] Make the `dim` operation index an operand. 2020-06-10 13:54:47 +00:00
tile_parallel_reduce.mlir [mlir][Linalg] Retire C++ MatmulOp in favor of a linalg-ods-gen'd op. 2020-06-16 10:46:35 -04:00
tile_simple_conv.mlir [mlir][Linalg] Fixing bug in subview size computation in Linalg tiling. 2020-07-23 11:09:55 -07:00
transform-patterns-matmul-to-vector.mlir [MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op 2020-07-28 12:34:19 +02:00
transform-patterns.mlir [MLIR][Linalg] Retire C++ DotOp in favor of a linalg-ods-gen'd op 2020-07-28 12:34:19 +02:00