[mlir][GPUToROCDL] Lower arith.remf to GPU intrinsic.

Differential Revision: https://reviews.llvm.org/D159423

Author: akuegel

mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp

@@ -317,8 +317,9 @@ void mlir::configureGpuToROCDLConversionLegality(ConversionTarget &target) {
   target.addLegalDialect<ROCDL::ROCDLDialect>();
   target.addIllegalDialect<gpu::GPUDialect>();
   target.addIllegalOp<LLVM::CosOp, LLVM::ExpOp, LLVM::Exp2Op, LLVM::FAbsOp,
-                      LLVM::FCeilOp, LLVM::FFloorOp, LLVM::LogOp, LLVM::Log10Op,
-                      LLVM::Log2Op, LLVM::PowOp, LLVM::SinOp, LLVM::SqrtOp>();
+                      LLVM::FCeilOp, LLVM::FFloorOp, LLVM::FRemOp, LLVM::LogOp,
+                      LLVM::Log10Op, LLVM::Log2Op, LLVM::PowOp, LLVM::SinOp,
+                      LLVM::SqrtOp>();
 
   // TODO: Remove once we support replacing non-root ops.
   target.addLegalOp<gpu::YieldOp, gpu::GPUModuleOp, gpu::ModuleEndOp>();
@@ -386,6 +387,8 @@ void mlir::populateGpuToROCDLConversionPatterns(
                                     "__ocml_expm1_f64");
   populateOpPatterns<math::FloorOp>(converter, patterns, "__ocml_floor_f32",
                                     "__ocml_floor_f64");
+  populateOpPatterns<arith::RemFOp>(converter, patterns, "__ocml_fmod_f32",
+                                    "__ocml_fmod_f64");
   populateOpPatterns<math::LogOp>(converter, patterns, "__ocml_log_f32",
                                   "__ocml_log_f64");
   populateOpPatterns<math::Log10Op>(converter, patterns, "__ocml_log10_f32",

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

@@ -574,6 +574,21 @@ gpu.module @test_module {
 
 // -----
 
+gpu.module @test_module {
+  // CHECK: llvm.func @__ocml_fmod_f32(f32, f32) -> f32
+  // CHECK: llvm.func @__ocml_fmod_f64(f64, f64) -> f64
+  // CHECK-LABEL: func @gpu_fmod
+  func.func @gpu_fmod(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = arith.remf %arg_f32, %arg_f32 : f32
+    // CHECK: llvm.call @__ocml_fmod_f32(%{{.*}}, %{{.*}}) : (f32, f32) -> f32
+    %result64 = arith.remf %arg_f64, %arg_f64 : f64
+    // CHECK: llvm.call @__ocml_fmod_f64(%{{.*}}, %{{.*}}) : (f64, f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
 gpu.module @test_module {
   // CHECK-LABEL: func @gpu_shuffle()
   func.func @gpu_shuffle() -> (f32, f32) {
@@ -612,4 +627,4 @@ gpu.module @test_module {
     %shfli, %predi = gpu.shuffle idx %arg0, %arg1, %arg2 : f32
     func.return %shfl, %shfli : f32, f32
   }
-}
+}