From e111237a95969384adffd22d3c8904c4cdd7e8de Mon Sep 17 00:00:00 2001
From: Johannes M Dieterich <Johannes.Dieterich@amd.com>
Date: Mon, 22 Oct 2018 13:01:23 -0500
Subject: [PATCH] Since upstream #9322 by @petrex , hipification resulted in
 illegal code.

The legal function cublasHandle_t cublas_handle() was hipified to the
clearly illegal rocblas_handle rocblas_handle(). It should not work and
correctly fails with gcc as the host compiler as it induces an
ambiguity.

Function now hipifies to rocblas_handle rocblashandle()

Fixes long standing issue we've observed in PyTorch when base compiler is gcc.
---
 caffe2/core/hip/context_hip.h                 |  2 +-
 caffe2/utils/hip/math_hip.cc                  | 28 +++++++++----------
 .../pyHIPIFY/cuda_to_hip_mappings.py          |  2 +-
 3 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/caffe2/core/hip/context_hip.h b/caffe2/core/hip/context_hip.h
index ca2d03b1449d..463303a1f0df 100644
--- a/caffe2/core/hip/context_hip.h
+++ b/caffe2/core/hip/context_hip.h
@@ -175,7 +175,7 @@ class HIPContext final : public BaseContext {
     return hip_objects_.GetStream(gpu_id, stream_id);
   }
 
-  rocblas_handle rocblas_handle() {
+  rocblas_handle rocblashandle() {
     return hip_objects_.GetHandle(gpu_id_, stream_id_);
   }
 
diff --git a/caffe2/utils/hip/math_hip.cc b/caffe2/utils/hip/math_hip.cc
index d499393dbe81..bc6ee92f46e3 100644
--- a/caffe2/utils/hip/math_hip.cc
+++ b/caffe2/utils/hip/math_hip.cc
@@ -759,7 +759,7 @@ void Gemm<float, HIPContext>(
       ? rocblas_operation_none
       : rocblas_operation_transpose;
   ROCBLAS_ENFORCE(rocblas_sgemm(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       cuTransB,
       cuTransA,
       N,
@@ -803,7 +803,7 @@ void Gemm<at::Half, HIPContext>(
       : rocblas_operation_transpose;
   if (math_type == TensorProto_DataType_FLOAT) {
     ROCBLAS_CHECK(rocblas_sgemmEx(
-        context->rocblas_handle(),
+        context->rocblashandle(),
         cuTransB,
         cuTransA,
         N,
@@ -828,7 +828,7 @@ void Gemm<at::Half, HIPContext>(
 
     // call cublasHgemm
     ROCBLAS_CHECK(cublasHgemm(
-        context->rocblas_handle(),
+        context->rocblashandle(),
         cuTransB,
         cuTransA,
         N,
@@ -933,7 +933,7 @@ void GemmStridedBatched<float, HIPContext>(
       ? rocblas_operation_none
       : rocblas_operation_transpose;
   ROCBLAS_ENFORCE(rocblas_sgemm_strided_batched(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       cuTransB,
       cuTransA,
       N,
@@ -1004,7 +1004,7 @@ void GemmStridedBatched<at::Half, HIPContext>(
     __half alpha_fp16 = at::Half(alpha);
     __half beta_fp16 = at::Half(beta);
     ROCBLAS_ENFORCE(cublasHgemmStridedBatched(
-        context->rocblas_handle(),
+        context->rocblashandle(),
         cuTransB,
         cuTransA,
         N,
@@ -1051,7 +1051,7 @@ void GemmEx<float, HIPContext>(
       ? rocblas_operation_none
       : rocblas_operation_transpose;
   ROCBLAS_ENFORCE(rocblas_sgemm(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       cuTransB,
       cuTransA,
       N,
@@ -1083,7 +1083,7 @@ void Gemv<float, HIPContext>(
       ? rocblas_operation_transpose
       : rocblas_operation_none;
   ROCBLAS_ENFORCE(rocblas_sgemv(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       cuTransA,
       N,
       M,
@@ -1170,7 +1170,7 @@ void Gemv<at::Half, HIPContext>(
 
   if (math_type == TensorProto_DataType_FLOAT) {
     ROCBLAS_CHECK(cublasSgemmEx(
-        context->rocblas_handle(),
+        context->rocblashandle(),
         cuTransA,
         rocblas_operation_none,
         m,
@@ -1192,7 +1192,7 @@ void Gemv<at::Half, HIPContext>(
     __half beta_fp16 = at::Half(beta);
 
     ROCBLAS_CHECK(cublasHgemm(
-        context->rocblas_handle(),
+        context->rocblashandle(),
         cuTransA,
         rocblas_operation_none,
         m,
@@ -1390,7 +1390,7 @@ void Dot<float, HIPContext>(
     HIPContext* context) {
   float result;
   ROCBLAS_ENFORCE(
-      rocblas_sdot(context->rocblas_handle(), n, a, 1, b, 1, &result));
+      rocblas_sdot(context->rocblashandle(), n, a, 1, b, 1, &result));
   context->CopyFromCPU<float>(1, &result, y);
 }
 
@@ -1406,7 +1406,7 @@ void Dot<at::Half, HIPContext>(
   at::Half result;
   // execute with 32-bit math
   ROCBLAS_CHECK(cublasDotEx(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       n,
       a,
       CUDA_R_16F,
@@ -1879,7 +1879,7 @@ void Axpy<float, HIPContext>(
     float* Y,
     HIPContext* context) {
   ROCBLAS_ENFORCE(
-      rocblas_saxpy(context->rocblas_handle(), N, &alpha, X, 1, Y, 1));
+      rocblas_saxpy(context->rocblashandle(), N, &alpha, X, 1, Y, 1));
 }
 
 template <>
@@ -1891,7 +1891,7 @@ void Axpy<double, HIPContext>(
     HIPContext* context) {
   double alpha_d{alpha};
   ROCBLAS_ENFORCE(
-      rocblas_daxpy(context->rocblas_handle(), N, &alpha_d, X, 1, Y, 1));
+      rocblas_daxpy(context->rocblashandle(), N, &alpha_d, X, 1, Y, 1));
 }
 
 template <>
@@ -1904,7 +1904,7 @@ void Axpy<at::Half, HIPContext>(
   CAFFE_THROW("Unsupported math type");
 #if ROCBLAS_FP16
   ROCBLAS_CHECK(cublasAxpyEx(
-      context->rocblas_handle(),
+      context->rocblashandle(),
       N,
       &alpha,
       CUDA_R_16F,
diff --git a/tools/amd_build/pyHIPIFY/cuda_to_hip_mappings.py b/tools/amd_build/pyHIPIFY/cuda_to_hip_mappings.py
index 2d1f722c8db2..4a8aef36eaae 100644
--- a/tools/amd_build/pyHIPIFY/cuda_to_hip_mappings.py
+++ b/tools/amd_build/pyHIPIFY/cuda_to_hip_mappings.py
@@ -2230,7 +2230,7 @@
     ("HasCudaGPU" , ("HasHipGPU", API_CAFFE2)),
     ("__expf" , ("expf", API_CAFFE2)),
     ("CUBLAS_ENFORCE" , ("ROCBLAS_ENFORCE", API_CAFFE2)),
-    ("cublas_handle" , ("rocblas_handle", API_CAFFE2)),
+    ("cublas_handle" , ("rocblashandle", API_CAFFE2)),
     ("CURAND_ENFORCE" ,("HIPRAND_ENFORCE", API_CAFFE2)),
     ("curandGenerateUniform" , ("hiprandGenerateUniform", API_CAFFE2)),
     ("curand_generator" , ("hiprand_generator", API_CAFFE2)),