forked from buddy-compiler/buddy-mlir
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[examples] Add transpose vectorization example.
- Loading branch information
Showing
4 changed files
with
207 additions
and
3 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,3 +1 @@ | ||
| log.mlir | ||
| log.ll | ||
| log.s | ||
| log.* |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,68 @@ | ||
| // RUN: buddy-opt %s \ | ||
| // RUN: -convert-linalg-to-affine-loops \ | ||
| // RUN: -affine-loop-fusion \ | ||
| // RUN: -lower-affine \ | ||
| // RUN: -func-bufferize \ | ||
| // RUN: -arith-bufferize \ | ||
| // RUN: -tensor-bufferize \ | ||
| // RUN: -buffer-deallocation \ | ||
| // RUN: -finalizing-bufferize \ | ||
| // RUN: -convert-vector-to-scf \ | ||
| // RUN: -expand-strided-metadata \ | ||
| // RUN: -convert-vector-to-llvm \ | ||
| // RUN: -memref-expand \ | ||
| // RUN: -arith-expand \ | ||
| // RUN: -convert-arith-to-llvm \ | ||
| // RUN: -finalize-memref-to-llvm \ | ||
| // RUN: -convert-scf-to-cf \ | ||
| // RUN: -convert-openmp-to-llvm \ | ||
| // RUN: -convert-arith-to-llvm \ | ||
| // RUN: -convert-math-to-llvm \ | ||
| // RUN: -convert-math-to-libm \ | ||
| // RUN: -convert-func-to-llvm \ | ||
| // RUN: -reconcile-unrealized-casts \ | ||
| // RUN: | mlir-cpu-runner -e main -entry-point-result=void \ | ||
| // RUN: -shared-libs=%mlir_runner_utils_dir/libmlir_runner_utils%shlibext \ | ||
| // RUN: -shared-libs=%mlir_runner_utils_dir/libmlir_c_runner_utils%shlibext \ | ||
| // RUN: | FileCheck %s | ||
|
|
||
| module { | ||
| memref.global "private" constant @__constant_1x32x40x128xf32 : memref<1x32x40x128xf32> = dense<3.000000e+00> {alignment = 64 : i64} | ||
| func.func private @rtclock() -> f64 | ||
| func.func private @printMemrefF32(memref<*xf32>) | ||
| func.func @kernel(%arg0: memref<1x32x40x128xf32>) { | ||
| %0 = call @rtclock() : () -> f64 | ||
| %alloc = memref.alloc() {alignment = 64 : i64} : memref<1x40x32x128xf32> | ||
| affine.for %arg1 = 0 to 1 { | ||
| affine.for %arg2 = 0 to 40 { | ||
| affine.for %arg3 = 0 to 32 { | ||
| affine.for %arg4 = 0 to 128 step 64 { | ||
| %3 = vector.load %arg0[%arg1, %arg3, %arg2, %arg4] : memref<1x32x40x128xf32>, vector<64xf32> | ||
| vector.store %3, %alloc[%arg1, %arg2, %arg3, %arg4] : memref<1x40x32x128xf32>, vector<64xf32> | ||
| } | ||
| } | ||
| } | ||
| } | ||
| %1 = call @rtclock() : () -> f64 | ||
| %2 = arith.subf %1, %0 : f64 | ||
| %cast = memref.cast %alloc : memref<1x40x32x128xf32> to memref<*xf32> | ||
|
|
||
| // All the elements of the MemRef are the same, | ||
| // only check the first line to verify the correctness. | ||
| // CHECK: Unranked Memref base@ = {{.*}} rank = 4 offset = 0 sizes = [1, 40, 32, 128] strides = [163840, 4096, 128, 1] data = | ||
| // CHECK-NEXT: [ | ||
| // CHECK-SAME: [ | ||
| // CHECK-SAME: [ | ||
| // CHECK-SAME: [3{{(, 3)*}}], | ||
|
|
||
| call @printMemrefF32(%cast) : (memref<*xf32>) -> () | ||
| vector.print %2 : f64 | ||
| return | ||
| } | ||
| func.func @main() { | ||
| %0 = memref.get_global @__constant_1x32x40x128xf32 : memref<1x32x40x128xf32> | ||
| call @kernel(%0) : (memref<1x32x40x128xf32>) -> () | ||
| return | ||
| } | ||
| } | ||
|
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,70 @@ | ||
| // RUN: buddy-opt %s \ | ||
| // RUN: -pass-pipeline "builtin.module(func.func(tosa-to-linalg-named),func.func(tosa-to-linalg),func.func(tosa-to-tensor),func.func(tosa-to-arith))" \ | ||
| // RUN: | buddy-opt \ | ||
| // RUN: -arith-expand \ | ||
| // RUN: -eliminate-empty-tensors \ | ||
| // RUN: -empty-tensor-to-alloc-tensor \ | ||
| // RUN: -one-shot-bufferize \ | ||
| // RUN: -convert-linalg-to-affine-loops \ | ||
| // RUN: -affine-loop-fusion \ | ||
| // RUN: -lower-affine \ | ||
| // RUN: -func-bufferize \ | ||
| // RUN: -arith-bufferize \ | ||
| // RUN: -tensor-bufferize \ | ||
| // RUN: -buffer-deallocation \ | ||
| // RUN: -finalizing-bufferize \ | ||
| // RUN: -convert-vector-to-scf \ | ||
| // RUN: -expand-strided-metadata \ | ||
| // RUN: -convert-vector-to-llvm \ | ||
| // RUN: -memref-expand \ | ||
| // RUN: -arith-expand \ | ||
| // RUN: -convert-arith-to-llvm \ | ||
| // RUN: -finalize-memref-to-llvm \ | ||
| // RUN: -convert-scf-to-cf \ | ||
| // RUN: -convert-openmp-to-llvm \ | ||
| // RUN: -convert-arith-to-llvm \ | ||
| // RUN: -convert-math-to-llvm \ | ||
| // RUN: -convert-math-to-libm \ | ||
| // RUN: -convert-func-to-llvm \ | ||
| // RUN: -reconcile-unrealized-casts \ | ||
| // RUN: | mlir-cpu-runner -e main -entry-point-result=void \ | ||
| // RUN: -shared-libs=%mlir_runner_utils_dir/libmlir_runner_utils%shlibext \ | ||
| // RUN: -shared-libs=%mlir_runner_utils_dir/libmlir_c_runner_utils%shlibext \ | ||
| // RUN: | FileCheck %s | ||
|
|
||
| func.func private @rtclock() -> f64 | ||
| func.func private @printMemrefF32(%ptr : tensor<*xf32>) | ||
|
|
||
| func.func @kernel(%t0 : tensor<1x32x40x128xf32>) { | ||
| %t_start = call @rtclock() : () -> f64 | ||
|
|
||
| %idx = "tosa.const"() <{value = dense<[0, 2, 1, 3]> : tensor<4xi32>}> : () -> tensor<4xi32> | ||
| %t1 = tosa.transpose %t0, %idx : (tensor<1x32x40x128xf32>, tensor<4xi32>) -> tensor<1x40x32x128xf32> | ||
|
|
||
| %t_end = call @rtclock() : () -> f64 | ||
| %time = arith.subf %t_end, %t_start : f64 | ||
|
|
||
| %tensor_unranked = tensor.cast %t1 : tensor<1x40x32x128xf32> to tensor<*xf32> | ||
|
|
||
| // All the elements of the MemRef are the same, | ||
| // only check the first line to verify the correctness. | ||
| // CHECK: Unranked Memref base@ = {{.*}} rank = 4 offset = 0 sizes = [1, 40, 32, 128] strides = [163840, 4096, 128, 1] data = | ||
| // CHECK-NEXT: [ | ||
| // CHECK-SAME: [ | ||
| // CHECK-SAME: [ | ||
| // CHECK-SAME: [3{{(, 3)*}}], | ||
|
|
||
| // Print results. | ||
| call @printMemrefF32(%tensor_unranked) : (tensor<*xf32>) -> () | ||
| // Print timings. | ||
| vector.print %time : f64 | ||
|
|
||
| return | ||
| } | ||
|
|
||
| func.func @main() { | ||
| %c0 = arith.constant dense<3.0> : tensor<1x32x40x128xf32> | ||
| call @kernel(%c0) : (tensor<1x32x40x128xf32>) -> () | ||
|
|
||
| return | ||
| } |