maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[mlir] Add std.powf to ROCDL lowering. 

Differential Revision: https://reviews.llvm.org/D93313
This commit is contained in:
Tres Popp 2020-12-15 18:26:11 +01:00
parent 32e7a9cab6
commit 9adc64539f
2 changed files with 18 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
View File

@ -84,7 +84,7 @@ void mlir::configureGpuToROCDLConversionLegality(ConversionTarget &target) {
target.addIllegalDialect<gpu::GPUDialect>();
target.addIllegalOp<LLVM::CosOp, LLVM::ExpOp, LLVM::FAbsOp, LLVM::FCeilOp,
LLVM::FFloorOp, LLVM::LogOp, LLVM::Log10Op, LLVM::Log2Op,
LLVM::SinOp, LLVM::SqrtOp>();
LLVM::PowOp, LLVM::SinOp, LLVM::SqrtOp>();
// TODO: Remove once we support replacing non-root ops.
target.addLegalOp<gpu::YieldOp, gpu::GPUModuleOp, gpu::ModuleEndOp>();
@ -123,6 +123,8 @@ void mlir::populateGpuToROCDLConversionPatterns(
"__ocml_log10_f64");
patterns.insert<OpToFuncCallLowering<Log2Op>>(converter, "__ocml_log2_f32",
"__ocml_log2_f64");
patterns.insert<OpToFuncCallLowering<PowFOp>>(converter, "__ocml_pow_f32",
"__ocml_pow_f64");
patterns.insert<OpToFuncCallLowering<RsqrtOp>>(converter, "__ocml_rsqrt_f32",
"__ocml_rsqrt_f64");
patterns.insert<OpToFuncCallLowering<SinOp>>(converter, "__ocml_sin_f32",

15
mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir
View File

@ -310,3 +310,18 @@ gpu.module @test_module {
std.return %result32, %result64 : f32, f64
}
}
// -----
gpu.module @test_module {
// CHECK: llvm.func @__ocml_pow_f32(!llvm.float, !llvm.float) -> !llvm.float
// CHECK: llvm.func @__ocml_pow_f64(!llvm.double, !llvm.double) -> !llvm.double
// CHECK-LABEL: func @gpu_pow
func @gpu_pow(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
%result32 = std.powf %arg_f32, %arg_f32 : f32
// CHECK: llvm.call @__ocml_pow_f32(%{{.*}}, %{{.*}}) : (!llvm.float, !llvm.float) -> !llvm.float
%result64 = std.powf %arg_f64, %arg_f64 : f64
// CHECK: llvm.call @__ocml_pow_f64(%{{.*}}, %{{.*}}) : (!llvm.double, !llvm.double) -> !llvm.double
std.return %result32, %result64 : f32, f64
}
}