[MLIR][ROCDL] add gpu to rocdl erf support

[manupak]

Summary

This commit adds lowering of lib func
call to support erf in rocdl.

[manupak]

• clang-format

[manupak]

Thanks for the review @ThomasRaoux ! Can someone help me land this patch ?
also cc : @krzysz00

mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp

void mlir::populateGpuToROCDLConversionPatterns(
   populateOpPatterns<math::SinOp>(converter, patterns, "__ocml_sin_f32",
                                   "__ocml_sin_f64");
   populateOpPatterns<math::SqrtOp>(converter, patterns, "__ocml_sqrt_f32",
                                    "__ocml_sqrt_f64");
   populateOpPatterns<math::TanhOp>(converter, patterns, "__ocml_tanh_f32",
                                    "__ocml_tanh_f64");
   populateOpPatterns<math::TanOp>(converter, patterns, "__ocml_tan_f32",
                                   "__ocml_tan_f64");
+  populateOpPatterns<math::ErfOp>(converter, patterns, "__ocml_erf_f32",
+                                  "__ocml_erf_f64");
 }
 
 std::unique_ptr<OperationPass<gpu::GPUModuleOp>>
 mlir::createLowerGpuOpsToROCDLOpsPass(const std::string &chipset,
                                       unsigned indexBitwidth,
                                       bool useBarePtrCallConv,
                                       gpu::amd::Runtime runtime) {
   return std::make_unique<LowerGpuOpsToROCDLOpsPass>(
       chipset, indexBitwidth, useBarePtrCallConv, runtime);
 }

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

func.func @gpu_pow(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
     // CHECK: llvm.call @__ocml_pow_f64(%{{.*}}, %{{.*}}) : (f64, f64) -> f64
     func.return %result32, %result64 : f32, f64
   }
 }
 
 // -----
 
 gpu.module @test_module {
+  // CHECK: llvm.func @__ocml_erf_f32(f32) -> f32
+  // CHECK: llvm.func @__ocml_erf_f64(f64) -> f64
+  // CHECK-LABEL: func @gpu_erf
+  func.func @gpu_erf(%arg_f32 : f32, %arg_f64 : f64) -> (f32, f64) {
+    %result32 = math.erf %arg_f32 : f32
+    // CHECK: llvm.call @__ocml_erf_f32(%{{.*}}) : (f32) -> f32
+    %result64 = math.erf %arg_f64 : f64
+    // CHECK: llvm.call @__ocml_erf_f64(%{{.*}}) : (f64) -> f64
+    func.return %result32, %result64 : f32, f64
+  }
+}
+
+// -----
+
+gpu.module @test_module {
   // CHECK-LABEL: func @gpu_unroll
   func.func @gpu_unroll(%arg0 : vector<4xf32>) -> vector<4xf32> {
     %result = math.exp %arg0 : vector<4xf32>
     // CHECK: llvm.call @__ocml_exp_f32(%{{.*}}) : (f32) -> f32
     // CHECK: llvm.call @__ocml_exp_f32(%{{.*}}) : (f32) -> f32
     // CHECK: llvm.call @__ocml_exp_f32(%{{.*}}) : (f32) -> f32
     // CHECK: llvm.call @__ocml_exp_f32(%{{.*}}) : (f32) -> f32
     func.return %result : vector<4xf32>