Address reviwer comments

Author: nirvedhmeshram

mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp

@@ -26,6 +26,7 @@
 #include "mlir/Conversion/LLVMCommon/LoweringOptions.h"
 #include "mlir/Conversion/LLVMCommon/Pattern.h"
 #include "mlir/Conversion/LLVMCommon/TypeConverter.h"
+#include "mlir/Conversion/MathToLLVM/MathToLLVM.h"
 #include "mlir/Conversion/MathToROCDL/MathToROCDL.h"
 #include "mlir/Conversion/MemRefToLLVM/MemRefToLLVM.h"
 #include "mlir/Conversion/VectorToLLVM/ConvertVectorToLLVM.h"
@@ -290,6 +291,7 @@ struct LowerGpuOpsToROCDLOpsPass
     populateAMDGPUToROCDLConversionPatterns(converter, llvmPatterns,
                                             *maybeChipset);
     populateVectorToLLVMConversionPatterns(converter, llvmPatterns);
+    populateMathToLLVMConversionPatterns(converter, llvmPatterns);
     cf::populateControlFlowToLLVMConversionPatterns(converter, llvmPatterns);
     populateFuncToLLVMConversionPatterns(converter, llvmPatterns);
     populateFinalizeMemRefToLLVMConversionPatterns(converter, llvmPatterns);
@@ -333,13 +335,11 @@ void mlir::configureGpuToROCDLConversionLegality(ConversionTarget &target) {
                       LLVM::FFloorOp, LLVM::FRemOp, LLVM::LogOp, LLVM::Log10Op,
                       LLVM::Log2Op, LLVM::PowOp, LLVM::SinOp>();
   // These ops are legal for f32 type.
-  target.addDynamicallyLegalOp<LLVM::ExpOp, LLVM::LogOp>(
-      [](mlir::Operation *op) {
-        return llvm::any_of(op->getOperandTypes(), [](Type type) {
-          return llvm::isa<FloatType>(type) &&
-                 type.getIntOrFloatBitWidth() == 32;
-        });
-      });
+  target.addDynamicallyLegalOp<LLVM::ExpOp, LLVM::LogOp>([](Operation *op) {
+    return any_of(op->getOperandTypes(), [](Type type) {
+      return isa<FloatType>(type) && type.getIntOrFloatBitWidth() == 32;
+    });
+  });
   // TODO: Remove once we support replacing non-root ops.
   target.addLegalOp<gpu::YieldOp, gpu::GPUModuleOp>();
 }

mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir

@@ -131,6 +131,60 @@ gpu.module @test_module {
 
 // -----
 
+gpu.module @test_module {
+  // CHECK-LABEL: func @gpu_sqrt
+  func.func @gpu_sqrt(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = math.sqrt %arg_f32 : f32
+    // CHECK: llvm.intr.sqrt(%{{.*}})  : (f32) -> f32
+    %result64 = math.sqrt %arg_f64 : f64
+    // CHECK: llvm.intr.sqrt(%{{.*}})  : (f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
+gpu.module @test_module {
+  // CHECK-LABEL: func @gpu_fabs
+  func.func @gpu_fabs(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = math.absf %arg_f32 : f32
+    // CHECK: llvm.intr.fabs(%{{.*}})  : (f32) -> f32
+    %result64 = math.absf %arg_f64 : f64
+    // CHECK: llvm.intr.fabs(%{{.*}})  : (f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
+gpu.module @test_module {
+  // CHECK: llvm.func @__ocml_exp_f64(f64) -> f64
+  // CHECK-LABEL: func @gpu_exp
+  func.func @gpu_exp(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = math.exp %arg_f32 : f32
+    // CHECK: llvm.intr.exp(%{{.*}})  : (f32) -> f32
+    %result64 = math.exp %arg_f64 : f64
+    // CHECK: llvm.call @__ocml_exp_f64(%{{.*}}) : (f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
+gpu.module @test_module {
+  // CHECK: llvm.func @__ocml_log_f64(f64) -> f64
+  // CHECK-LABEL: func @gpu_log
+  func.func @gpu_log(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = math.log %arg_f32 : f32
+    // CHECK: llvm.intr.log(%{{.*}})  : (f32) -> f32
+    %result64 = math.log %arg_f64 : f64
+    // CHECK: llvm.call @__ocml_log_f64(%{{.*}}) : (f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
 gpu.module @test_module {
   // CHECK: llvm.func @__ocml_cbrt_f32(f32) -> f32
   // CHECK: llvm.func @__ocml_cbrt_f64(f64) -> f64