From 8c3a8d17c8a154894895c48a304a04df9ece4328 Mon Sep 17 00:00:00 2001
From: Manupa Karunaratne <manupa.karunaratne@amd.com>
Date: Mon, 15 May 2023 14:41:49 +0000
Subject: [PATCH] [MLIR][ROCDL] add gpu to rocdl erf support

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

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D150355
---
 mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp |  2 ++
 mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir        | 15 +++++++++++++++
 2 files changed, 17 insertions(+)

diff --git a/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp b/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
index 1ac4e8e..bd82013 100644
--- a/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
+++ b/mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
@@ -280,6 +280,8 @@ void mlir::populateGpuToROCDLConversionPatterns(
                                    "__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>>
diff --git a/mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir b/mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir
index 31e7827..aec4403 100644
--- a/mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir
+++ b/mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir
@@ -438,6 +438,21 @@ gpu.module @test_module {
 // -----
 
 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>
-- 
2.7.4