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/Dialect/TCP/bufferize.mlir

28 lines
1.4 KiB
MLIR
Raw Normal View History

[RefBackend] Use more idiomatic bufferize pattern for TCP. The time has come for BypassShapes/LowerShapedResultsToMemref to go away :( For the reference backend, being consistent with upstream conventions is the name of the game now. This is a step down in a number of ways, e.g. test clarity and separation of concerns. But it is fewer files and fewer tests, and *does* address the "TODO: This is really fragile". It also eliminates two more ops from the refback dialect (sadly, they are the shaped_results/yield that we were getting kind of fond of, but alas).
2020-10-16 03:26:21 +08:00
// RUN: npcomp-opt -tcp-bufferize <%s | FileCheck %s
// CHECK-LABEL: func @tcp_broadcast_to(
// CHECK-SAME: %[[TENSOR:.*]]: tensor<?xf32>,
// CHECK-SAME: %[[SHAPE:.*]]: tensor<?xindex>) -> tensor<?x?xf32> {
// CHECK: refback.alloc_memref %[[SHAPE]] : memref<?x?xf32>
// Check for two nested loops, but don't look at more detail for now.
// TODO: This pass should not create loops. Instead it should create a
// buffer version of tcp.broadcast_to
// CHECK: scf.for
// CHECK: scf.for
func @tcp_broadcast_to(%arg0: tensor<?xf32>, %arg1: tensor<?xindex>) -> tensor<?x?xf32> {
%0 = tcp.broadcast_to %arg0, %arg1 : (tensor<?xf32>, tensor<?xindex>) -> tensor<?x?xf32>
return %0 : tensor<?x?xf32>
}
[TCP] Replace tcp.matmul with linalg.matmul. This involved adding a `tcp.splatted` op to splat a dynamically sized init tensor. See rationale in TCPOps.td docs. One interesting observation is that when lowering tcf.matmul to linalg.matmul, we need to both 1) create the error checks and 2) calculate a shape transfer function to create the init tensors. Previously, 2) was deferred to bufferizing tcp.matmul later. I'm not sure if this is a conflation of concerns or not. For now, it's not a big burden.
2020-11-10 07:49:22 +08:00
// CHECK-LABEL: func @tcp_splatted(
// CHECK-SAME: %[[SPLAT_VAL:.*]]: f32,
// CHECK-SAME: %[[SHAPE:.*]]: tensor<?xindex>) -> tensor<?x?xf32> {
[RefBackend] Use more idiomatic bufferize pattern for TCP. The time has come for BypassShapes/LowerShapedResultsToMemref to go away :( For the reference backend, being consistent with upstream conventions is the name of the game now. This is a step down in a number of ways, e.g. test clarity and separation of concerns. But it is fewer files and fewer tests, and *does* address the "TODO: This is really fragile". It also eliminates two more ops from the refback dialect (sadly, they are the shaped_results/yield that we were getting kind of fond of, but alas).
2020-10-16 03:26:21 +08:00
// CHECK: %[[RESULT:.*]] = refback.alloc_memref %[[SHAPE]] : memref<?x?xf32>
[TCP] Replace tcp.matmul with linalg.matmul. This involved adding a `tcp.splatted` op to splat a dynamically sized init tensor. See rationale in TCPOps.td docs. One interesting observation is that when lowering tcf.matmul to linalg.matmul, we need to both 1) create the error checks and 2) calculate a shape transfer function to create the init tensors. Previously, 2) was deferred to bufferizing tcp.matmul later. I'm not sure if this is a conflation of concerns or not. For now, it's not a big burden.
2020-11-10 07:49:22 +08:00
// CHECK: linalg.fill(%[[RESULT]], %[[SPLAT_VAL]]) : memref<?x?xf32>, f32
[RefBackend] Use more idiomatic bufferize pattern for TCP. The time has come for BypassShapes/LowerShapedResultsToMemref to go away :( For the reference backend, being consistent with upstream conventions is the name of the game now. This is a step down in a number of ways, e.g. test clarity and separation of concerns. But it is fewer files and fewer tests, and *does* address the "TODO: This is really fragile". It also eliminates two more ops from the refback dialect (sadly, they are the shaped_results/yield that we were getting kind of fond of, but alas).
2020-10-16 03:26:21 +08:00
// CHECK: %[[RESULT_TENSOR:.*]] = tensor_load %[[RESULT]] : memref<?x?xf32>
// CHECK: return %[[RESULT_TENSOR]] : tensor<?x?xf32>
[TCP] Replace tcp.matmul with linalg.matmul. This involved adding a `tcp.splatted` op to splat a dynamically sized init tensor. See rationale in TCPOps.td docs. One interesting observation is that when lowering tcf.matmul to linalg.matmul, we need to both 1) create the error checks and 2) calculate a shape transfer function to create the init tensors. Previously, 2) was deferred to bufferizing tcp.matmul later. I'm not sure if this is a conflation of concerns or not. For now, it's not a big burden.
2020-11-10 07:49:22 +08:00
func @tcp_splatted(%arg0: f32, %arg1: tensor<?xindex>) -> tensor<?x?xf32> {
%0 = tcp.splatted %arg0, %arg1 : (f32, tensor<?xindex>) -> tensor<?x?xf32>
[RefBackend] Use more idiomatic bufferize pattern for TCP. The time has come for BypassShapes/LowerShapedResultsToMemref to go away :( For the reference backend, being consistent with upstream conventions is the name of the game now. This is a step down in a number of ways, e.g. test clarity and separation of concerns. But it is fewer files and fewer tests, and *does* address the "TODO: This is really fragile". It also eliminates two more ops from the refback dialect (sadly, they are the shaped_results/yield that we were getting kind of fond of, but alas).
2020-10-16 03:26:21 +08:00
return %0 : tensor<?x?xf32>
}