Home
last modified time | relevance | path

Searched refs:tiled (Results 1 – 21 of 21) sorted by relevance

/llvm-project-15.0.7/polly/docs/experiments/matmul/
H A Drunall.sh49 -polly-import-jscop-postfix=interchanged+tiled+vector \
57 -polly-import-jscop-postfix=interchanged+tiled -polly-codegen \
59 | opt -O3 -S -o matmul.polly.interchanged+tiled.ll
63 | opt -O3 -S -o matmul.polly.interchanged+tiled+vector.ll
67 | opt -O3 -S -o matmul.polly.interchanged+tiled+vector+openmp.ll
73 llc matmul.polly.interchanged+tiled.ll -o matmul.polly.interchanged+tiled.s -relocation-model=pic
74 gcc matmul.polly.interchanged+tiled.s -o matmul.polly.interchanged+tiled.exe
75 llc matmul.polly.interchanged+tiled+vector.ll -o matmul.polly.interchanged+tiled+vector.s -relocati…
76 gcc matmul.polly.interchanged+tiled+vector.s -o matmul.polly.interchanged+tiled+vector.exe
77 llc matmul.polly.interchanged+tiled+vector+openmp.ll -o matmul.polly.interchanged+tiled+vector+open…
[all …]
/llvm-project-15.0.7/mlir/include/mlir/Interfaces/
H A DTilingInterface.td9 // This file contains an interface to allow operations to generate a tiled
30 tiled implementation is written into. With `tensor`
32 the tiled results are inserted into. With `memref` operands,
33 this will be the operand into which the result of the tiled
75 - `dest` are the Value into which the result of the tiled
86 The method returns the operation that is the tiled
108 used to obtain the tiled implementation of the operation.
131 `getTiledImplementation` which generates the tiled
133 iteration space. This method generates a tiled
137 tiled to generate the result tile. In practical terms this
[all …]
/llvm-project-15.0.7/polly/docs/
H A DHowToManuallyUseTheIndividualPiecesOfPolly.rst400 …-jscop-postfix=interchanged+tiled -polly-codegen -polly-process-unprofitable | opt -S -O3 -o matmu…
410 …interchanged+tiled+vector -polly-codegen -polly-vectorizer=polly -polly-process-unprofitable | opt…
438 …$ llc matmul.polly.interchanged+tiled.ll -o matmul.polly.interchanged+tiled.s -relocation-model=pic
439 $ gcc matmul.polly.interchanged+tiled.s -o matmul.polly.interchanged+tiled.exe
440 …$ llc matmul.polly.interchanged+tiled+vector.ll -o matmul.polly.interchanged+tiled+vector.s -reloc…
441 … $ gcc matmul.polly.interchanged+tiled+vector.s -o matmul.polly.interchanged+tiled+vector.exe
442 …$ llc matmul.polly.interchanged+tiled+vector+openmp.ll -o matmul.polly.interchanged+tiled+vector+o…
443 …$ gcc matmul.polly.interchanged+tiled+vector+openmp.s -lgomp -o matmul.polly.interchanged+tiled+ve…
465 $ time ./matmul.polly.interchanged+tiled.exe
470 $ time ./matmul.polly.interchanged+tiled+vector.exe
[all …]
/llvm-project-15.0.7/mlir/test/Transforms/
H A Dnormalize-memrefs-ops-dynamic.mlir5 // and tiled-layout map.
100 // Same test with maps that are not tiled layout maps in the arguments and the operations in the fu…
101 // This is not normalized since this is not tiled-layout map. No mod and floordiv.
131 // Same test with maps that are not tiled layout maps in the arguments and the operations in the fu…
132 // This is not normalized since this is not tiled-layout map. No floordiv.
162 // Same test with maps that are not tiled layout maps in the arguments and the operations in the fu…
163 // This is not normalized since this is not tiled-layout map. RHS of floordiv is different from tha…
193 // Same test with maps that are not tiled layout maps in the arguments and the operations in the fu…
194 // This is not normalized since this is not tiled-layout map. Multiple mod with the same LHS and RH…
224 // Same test with maps that are not tiled layout maps in the arguments and the operations in the fu…
[all …]
H A Dnormalize-memrefs.mlir70 // A tiled layout.
/llvm-project-15.0.7/mlir/lib/Dialect/Linalg/TransformOps/
H A DLinalgTransformOps.cpp116 FailureOr<TiledLinalgOp> tiled = applyFn(linalgOp); in applyTilingToAll() local
117 if (failed(tiled)) in applyTilingToAll()
120 tiledLinalgOps.push_back(tiled->op); in applyTilingToAll()
121 if (tiled->loops.size() != numLoops) in applyTilingToAll()
127 loopOps[i].push_back(tiled->loops[i]); in applyTilingToAll()
352 auto tiled = tileAndFuse(producerOp, containingOp, rewriter); in apply() local
353 if (succeeded(tiled)) in apply()
354 fusedOps.append(*tiled); in apply()
360 if (failed(tiled) && failed(cloned)) { in apply()
891 SmallVector<Operation *> tiled; in apply() local
[all …]
/llvm-project-15.0.7/mlir/include/mlir/Dialect/Linalg/TransformOps/
H A DLinalgTransformOps.td90 currently tiled and/or cloned multiple times into the containing op.
382 scalarized (i.e. their dynamic dimensions tiled by 1).
393 tiled-by-1 operations, which can be empty.
395 This operation does not return handles to the tiled loop.
609 This transform generates a loop nest with a smaller ("tiled") target
631 tiled operations and loops around them.
635 associated with the `target` handle will have been tiled.
668 Tile a TilingInterface op to a tiled `scf.foreach_thread`. Tiling is
674 Zero tile sizes indicate that the dimension is not tiled and can be
690 tiled operations, which can all be empty.
[all …]
/llvm-project-15.0.7/mlir/test/Dialect/Linalg/
H A Dtransform-op-scalarize.mlir6 // The op is first tiled by 10 in the first dimension, which creates a
H A Dtransform-op-fuse-into-containing.mlir46 // linalg.fill is tileable. The op is tiled and fused.
H A Dtile-scalarize-dynamic-dims.mlir31 // The input IR of this test case is a tiled and peeled linalg.matmul op.
/llvm-project-15.0.7/mlir/test/Integration/Dialect/Vector/GPU/CUDA/
H A Dtest-reduction-distribute.mlir13 // Run a tiled reduction fused with an elementwise op.
/llvm-project-15.0.7/polly/test/ScheduleOptimizer/
H A Dtile_after_fusion.ll17 ; checks whether they are tiled after being fused when polly-opt-fusion equals
/llvm-project-15.0.7/mlir/include/mlir/Dialect/MemRef/Transforms/
H A DPasses.td50 can express tiled or linearized access to multi-dimensional data
/llvm-project-15.0.7/mlir/include/mlir/IR/
H A DBuiltinTypes.td380 // A 2-d tiled layout with tiles of size 128 x 256.
383 // A tiled data layout with non-constant tile sizes.
407 // A 2-d dynamic shaped memref that also has a dynamically sized tiled
489 // MxN matrix stored in a 2-d blocked/tiled layout with 64x64 tiles.
/llvm-project-15.0.7/mlir/test/Dialect/Affine/
H A Ddma-generate.mlir178 // DMA on tiled loop nest. This also tests the case where the bounds are
276 // If this loop nest isn't tiled, the access requires a non-constant DMA
577 // This a 3-d loop nest tiled by 4 x 4 x 4. Under %i, %j, %k, the size of a
/llvm-project-15.0.7/mlir/test/Conversion/SCFToGPU/
H A Dparallel_loop.mlir41 // tiled 2-d parallel loop mapped to block.y and block.x and thread.y and thread.x.
141 // tiled 2-d parallel loop mapped to block.y and seq. and thread.y and seq.
/llvm-project-15.0.7/polly/lib/External/ppcg/
H A DREADME103 specifies that in kernel 0, two loops should be tiled with a tile
/llvm-project-15.0.7/mlir/include/mlir/Dialect/Transform/IR/
H A DTransformDialect.td242 (e.g., the first loop is tiled and the second unrolled) with the necessary
/llvm-project-15.0.7/mlir/docs/Dialects/Linalg/
H A D_index.md235 implement transformations such as tiling, tiled producer-consumer fusion, and
/llvm-project-15.0.7/mlir/docs/Rationale/
H A DRationale.md753 // A tiled matmul code (128x128x128) represented in schedule tree form
/llvm-project-15.0.7/polly/lib/External/isl/doc/
H A Duser.pod10590 A band node can be tiled using the following function.