llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
torch-mlir/test/E2E/resolve-tensor-load-store-o...

28 lines
1.0 KiB
MLIR
Raw Normal View History

"Finish" tensor -> memref conversion. There's a lot of details to flesh out here, but the basic approach seems promising (see comments in createE2ELoweringPipeline). This approach will be put to the test when we try to do our first fusions since that tickles some of the nasty phase ordering issues involved here. But we're not there yet.
2020-05-12 05:24:57 +08:00
// RUN: npcomp-opt -resolve-tensor-load-store-ops <%s | FileCheck %s --dump-input=fail
// CHECK-LABEL: func @basic
func @basic(%arg0: tensor<?xf32>) -> tensor<?xf32> {
Rework reference shape lowering based on upstream shape dialect changes. * Primarily, the upstream shape dialect now uses tensor<?xindex> for non-erroring, immediate shape calculations (and will return this for shape_of of a tensor or memref). * In addition, upstream passes do not yet exist for fully lowering to standard ops, so the passes here need to be extended to handle this new convention. * This should be seen as an intermediate state, necessary to integrate a new LLVM version and needs more work and cleanup for generality. * There is a good deal of awkwardness in these conversions. The hope is that additional upstream work will yield better defined conversion paths once out of this intermediate state.
2020-08-03 13:06:12 +08:00
%shape = "shape.shape_of"(%arg0) : (tensor<?xf32>) -> tensor<?xindex>
"Finish" tensor -> memref conversion. There's a lot of details to flesh out here, but the basic approach seems promising (see comments in createE2ELoweringPipeline). This approach will be put to the test when we try to do our first fusions since that tickles some of the nasty phase ordering issues involved here. But we're not there yet.
2020-05-12 05:24:57 +08:00
Add custom assembly format for tcp.alloc_memref/tcp.get_extent This makes the IR a bit easier to scan.
2020-05-12 06:20:49 +08:00
// CHECK: %[[SRCMEMREF:.+]] = tcp.alloc_memref
%src_memref = tcp.alloc_memref %shape : memref<?xf32>
"Finish" tensor -> memref conversion. There's a lot of details to flesh out here, but the basic approach seems promising (see comments in createE2ELoweringPipeline). This approach will be put to the test when we try to do our first fusions since that tickles some of the nasty phase ordering issues involved here. But we're not there yet.
2020-05-12 05:24:57 +08:00
// tensor_store of argument remains.
// CHECK: tensor_store %arg0, %[[SRCMEMREF]]
tensor_store %arg0, %src_memref : memref<?xf32>
%src = tensor_load %src_memref : memref<?xf32>
Add custom assembly format for tcp.alloc_memref/tcp.get_extent This makes the IR a bit easier to scan.
2020-05-12 06:20:49 +08:00
// CHECK: %[[DSTMEMREF:.+]] = tcp.alloc_memref
%dst_memref = tcp.alloc_memref %shape : memref<?xf32>
"Finish" tensor -> memref conversion. There's a lot of details to flesh out here, but the basic approach seems promising (see comments in createE2ELoweringPipeline). This approach will be put to the test when we try to do our first fusions since that tickles some of the nasty phase ordering issues involved here. But we're not there yet.
2020-05-12 05:24:57 +08:00
// tensor_store of internally created tensor is eliminated.
// CHECK-NOT: tensor_store
// CHECK: linalg.copy(%[[SRCMEMREF]], %[[DSTMEMREF]])
tensor_store %src, %dst_memref : memref<?xf32>
%ret = tensor_load %dst_memref : memref<?xf32>
// The tensor_load feeding into the return remains.
// %[[RET:.+]] = tensor_load %[[DSTMEMREF]]
// return %[[RET]]
return %ret : tensor<?xf32>
}