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/lower-alloc-memref-ops.mlir

32 lines
1.0 KiB
MLIR
Raw Normal View History

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
// RUN: npcomp-opt -split-input-file -lower-alloc-memref-ops <%s | FileCheck %s
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
// CHECK-LABEL: func @basic
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
func @basic(%arg0: tensor<?xindex>) -> memref<?xf32> {
// CHECK: %[[I:.*]] = constant 0 : index
// CHECK: %[[E:.*]] = shape.get_extent %arg0, %[[I]]
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
// CHECK: alloc(%[[E]])
%0 = tcp.alloc_memref %arg0 : memref<?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
return %0 : memref<?xf32>
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
}
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
// -----
// CHECK-LABEL: func @all_static
func @all_static(%arg0: tensor<?xindex>) -> memref<3x4x5xf32> {
// CHECK-NOT: shape.get_extent
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
// CHECK: alloc()
%0 = tcp.alloc_memref %arg0 : memref<3x4x5xf32>
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
return %0 : memref<3x4x5xf32>
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
}
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
// -----
// CHECK-LABEL: func @some_static
func @some_static(%arg0: tensor<?xindex>) -> memref<3x?x5x?x7xf32> {
// CHECK-DAG: %[[I1:.*]] = constant 1 : index
// CHECK-DAG: %[[E1:.*]] = shape.get_extent %arg0, %[[I1]]
// CHECK-DAG: %[[I3:.*]] = constant 3 : index
// CHECK-DAG: %[[E3:.*]] = shape.get_extent %arg0, %[[I3]]
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
// CHECK: alloc(%[[E1]], %[[E3]])
%0 = tcp.alloc_memref %arg0 : memref<3x?x5x?x7xf32>
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
return %0 : memref<3x?x5x?x7xf32>
Lower tcp.alloc_memref ops to tcp.get_extent + std.alloc. - tcp.get_extent will be liminated while lowering shapes - std.alloc is supported by the upstream LLVM lowering.
2020-05-19 03:50:16 +08:00
}