maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/mlir/test/EDSC/lower-edsc.mlir

45 lines
2.1 KiB
MLIR
Raw Normal View History

Start a testing pass for EDSC lowering. This is mostly plumbing to start allowing testing EDSC lowering. Prototype specifying reference implementation using verbose format without any generation/binding support. Add test pass that dumps the constructed EDSC (of which there can only be one). The idea is to enable iterating from multiple sides, this is wrong on many dimensions at the moment. PiperOrigin-RevId: 229570535
2019-01-17 01:26:16 +08:00
// RUN: mlir-opt -lower-edsc-test %s | FileCheck %s
Port Tablegen'd reference implementation of Add to declarative builders. PiperOrigin-RevId: 238977252
2019-03-18 21:52:23 +08:00
// CHECK-LABEL: func @t1(%arg0: memref<3x4x5x6xvector<4xi8>>, %arg1: memref<3x4x5x6xvector<4xi8>>, %arg2: memref<3x4x5x6xvector<4xi8>>) {
func @t1(%lhs: memref<3x4x5x6xvector<4xi8>>, %rhs: memref<3x4x5x6xvector<4xi8>>, %result: memref<3x4x5x6xvector<4xi8>>) -> () {
// CHECK: for {{.*}} = 0 to 3 {
// CHECK: for {{.*}} = 0 to 4 {
// CHECK: for {{.*}} = 0 to 5 {
// CHECK: for {{.*}}= 0 to 6 {
// CHECK: {{.*}} = load %arg1[{{.*}}] : memref<3x4x5x6xvector<4xi8>>
// CHECK: {{.*}} = load %arg0[{{.*}}] : memref<3x4x5x6xvector<4xi8>>
// CHECK: {{.*}} = addi {{.*}} : vector<4xi8>
// CHECK: store {{.*}}, %arg2[{{.*}}] : memref<3x4x5x6xvector<4xi8>>
return
}
// CHECK-LABEL: func @t2(%arg0: memref<3x4xf32>, %arg1: memref<3x4xf32>, %arg2: memref<3x4xf32>) {
func @t2(%lhs: memref<3x4xf32>, %rhs: memref<3x4xf32>, %result: memref<3x4xf32>) -> () {
// CHECK: for {{.*}} = 0 to 3 {
// CHECK: for {{.*}} = 0 to 4 {
// CHECK: {{.*}} = load %arg1[{{.*}}, {{.*}}] : memref<3x4xf32>
// CHECK: {{.*}} = load %arg0[{{.*}}, {{.*}}] : memref<3x4xf32>
// CHECK: {{.*}} = addf {{.*}}, {{.*}} : f32
// CHECK: store {{.*}}, %arg2[{{.*}}, {{.*}}] : memref<3x4xf32>
return
}
// CHECK-LABEL: func @t3(%arg0: memref<f32>, %arg1: memref<f32>, %arg2: memref<f32>) {
func @t3(%lhs: memref<f32>, %rhs: memref<f32>, %result: memref<f32>) -> () {
// CHECK: {{.*}} = load %arg1[] : memref<f32>
// CHECK: {{.*}} = load %arg0[] : memref<f32>
// CHECK: {{.*}} = addf {{.*}}, {{.*}} : f32
// CHECK: store {{.*}}, %arg2[] : memref<f32>
return
}
Start a testing pass for EDSC lowering. This is mostly plumbing to start allowing testing EDSC lowering. Prototype specifying reference implementation using verbose format without any generation/binding support. Add test pass that dumps the constructed EDSC (of which there can only be one). The idea is to enable iterating from multiple sides, this is wrong on many dimensions at the moment. PiperOrigin-RevId: 229570535
2019-01-17 01:26:16 +08:00
func @fn() {
Cleanup EDSCs and start a functional auto-generated library of custom Ops This CL applies the following simplifications to EDSCs: 1. Rename Block to StmtList because an MLIR Block is a different, not yet supported, notion; 2. Rework Bindable to drop specific storage and just use it as a simple wrapper around Expr. The only value of Bindable is to force a static cast when used by the user to bind into the emitter. For all intended purposes, Bindable is just a lightweight check that an Expr is Unbound. This simplifies usage and reduces the API footprint. After playing with it for some time, it wasn't worth the API cognition overhead; 3. Replace makeExprs and makeBindables by makeNewExprs and copyExprs which is more explicit and less easy to misuse; 4. Add generally useful functionality to MLIREmitter: a. expose zero and one for the ubiquitous common lower bounds and step; b. add support to create already bound Exprs for all function arguments as well as shapes and views for Exprs bound to memrefs. 5. Delete Stmt::operator= and replace by a `Stmt::set` method which is more explicit. 6. Make Stmt::operator Expr() explicit. 7. Indexed.indices assertions are removed to pave the way for expressing slices and views as well as to work with 0-D memrefs. The CL plugs those simplifications with TableGen and allows emitting a full MLIR function for pointwise add. This "x.add" op is both type and rank-agnostic (by allowing ArrayRef of Expr passed to For loops) and opens the door to spinning up a composable library of existing and custom ops that should automate a lot of the tedious work in TF/XLA -> MLIR. Testing needs to be significantly improved but can be done in a separate CL. PiperOrigin-RevId: 231982325
2019-02-02 01:16:31 +08:00
"print"() {op: "x.add", fn: @t1: (memref<3x4x5x6xvector<4xi8>>, memref<3x4x5x6xvector<4xi8>>, memref<3x4x5x6xvector<4xi8>>) -> ()} : () -> ()
Enable specifying the op for which the reference implementation should be printed. Allows emitting reference implementation of multiple ops inside the test lowering pass. PiperOrigin-RevId: 229603494
2019-01-17 04:19:41 +08:00
"print"() {op: "x.add", fn: @t2: (memref<3x4xf32>, memref<3x4xf32>, memref<3x4xf32>) -> ()} : () -> ()
Cleanup EDSCs and start a functional auto-generated library of custom Ops This CL applies the following simplifications to EDSCs: 1. Rename Block to StmtList because an MLIR Block is a different, not yet supported, notion; 2. Rework Bindable to drop specific storage and just use it as a simple wrapper around Expr. The only value of Bindable is to force a static cast when used by the user to bind into the emitter. For all intended purposes, Bindable is just a lightweight check that an Expr is Unbound. This simplifies usage and reduces the API footprint. After playing with it for some time, it wasn't worth the API cognition overhead; 3. Replace makeExprs and makeBindables by makeNewExprs and copyExprs which is more explicit and less easy to misuse; 4. Add generally useful functionality to MLIREmitter: a. expose zero and one for the ubiquitous common lower bounds and step; b. add support to create already bound Exprs for all function arguments as well as shapes and views for Exprs bound to memrefs. 5. Delete Stmt::operator= and replace by a `Stmt::set` method which is more explicit. 6. Make Stmt::operator Expr() explicit. 7. Indexed.indices assertions are removed to pave the way for expressing slices and views as well as to work with 0-D memrefs. The CL plugs those simplifications with TableGen and allows emitting a full MLIR function for pointwise add. This "x.add" op is both type and rank-agnostic (by allowing ArrayRef of Expr passed to For loops) and opens the door to spinning up a composable library of existing and custom ops that should automate a lot of the tedious work in TF/XLA -> MLIR. Testing needs to be significantly improved but can be done in a separate CL. PiperOrigin-RevId: 231982325
2019-02-02 01:16:31 +08:00
"print"() {op: "x.add", fn: @t3: (memref<f32>, memref<f32>, memref<f32>) -> ()} : () -> ()
Start a testing pass for EDSC lowering. This is mostly plumbing to start allowing testing EDSC lowering. Prototype specifying reference implementation using verbose format without any generation/binding support. Add test pass that dumps the constructed EDSC (of which there can only be one). The idea is to enable iterating from multiple sides, this is wrong on many dimensions at the moment. PiperOrigin-RevId: 229570535
2019-01-17 01:26:16 +08:00
return
}
Port Tablegen'd reference implementation of Add to declarative builders. PiperOrigin-RevId: 238977252
2019-03-18 21:52:23 +08:00