Merge branch 'develop_v2.3' into 'main'

[HIPIFY] Add rules for hipify

See merge request dcutoolkit/deeplearing/TransformerEngine!6

hipify_custom_map.json

 {
     "custom_map" : {
+           "common/util/cuda_runtime.h": "common/util/hip_runtime.h",
+           "../util/cuda_runtime.h": "../util/hip_runtime.h",
+           "common/util/cuda_driver.h": "common/util/hip_driver.h",
+           "../util/cuda_driver.h": "../util/hip_driver.h",
+           "common/util/cuda_nvml.h": "common/util/hip_nvml.h",
            "common/utils.cuh" : "common/utils_hip.cuh",
            "common/transpose/cast_transpose.h" : "common/transpose/cast_transpose_hip.h",
            "common/recipe/recipe_common.cuh" : "common/recipe/recipe_common_hip.cuh",

transformer_engine/common/comm_gemm_overlap/comm_gemm_overlap.cpp

@@ -12,13 +12,8 @@
 #include <numeric>
 
 #include "common/common.h"
-#ifdef USE_ROCM
-#include "common/util/hip_driver.h"
-#include "common/util/hip_runtime.h"
-#else
 #include "common/util/cuda_driver.h"
 #include "common/util/cuda_runtime.h"
-#endif
 #include "common/util/logging.h"
 #include "common/util/system.h"
 #include "userbuffers/userbuffers.h"

transformer_engine/common/comm_gemm_overlap/userbuffers/userbuffers-host.cpp

@@ -19,15 +19,9 @@
 #include <map>
 #include <utility>
 
-#ifdef USE_ROCM
-#include "common/util/hip_driver.h"
-#include "common/util/hip_nvml.h"
-#include "common/util/hip_runtime.h"
-#else
 #include "common/util/cuda_driver.h"
 #include "common/util/cuda_nvml.h"
 #include "common/util/cuda_runtime.h"
-#endif
 #include "common/util/logging.h"
 #include "common/util/system.h"
 #include "ipcsocket.h"

transformer_engine/common/common.cu

@@ -10,11 +10,7 @@
 
 #include "./common.h"
 #include "./utils.cuh"
-#ifdef __HIP_PLATFORM_AMD__
-#include "common/util/hip_runtime.h"
-#else
 #include "common/util/cuda_runtime.h"
-#endif
 #include "common/util/logging.h"
 
 namespace transformer_engine {

transformer_engine/common/gemm/cublaslt_gemm.cu

@@ -22,11 +22,7 @@
 #include "../common.h"
 #include "../util/handle_manager.h"
 #include "../util/logging.h"
-#ifdef __HIP_PLATFORM_AMD__
-#include "common/util/hip_runtime.h"
-#else
 #include "common/util/cuda_runtime.h"
-#endif
 
 #ifndef __HIP_PLATFORM_AMD__
 namespace {

transformer_engine/common/recipe/delayed_scaling.cu

@@ -11,11 +11,7 @@
 #include <string>
 
 #include "../common.h"
-#ifdef __HIP_PLATFORM_AMD__
-#include "../util/hip_runtime.h"
-#else
 #include "../util/cuda_runtime.h"
-#endif
 #include "../util/logging.h"
 
 namespace transformer_engine {

transformer_engine/common/util/cuda_driver.cpp

@@ -7,11 +7,7 @@
 #include <filesystem>
 
 #include "../common.h"
-#ifdef USE_ROCM
-#include "../util/hip_runtime.h"
-#else
 #include "../util/cuda_runtime.h"
-#endif
 
 namespace transformer_engine {
 

transformer_engine/common/util/cuda_runtime.cpp

@@ -10,15 +10,9 @@
 #include <mutex>
 
 #include "../common.h"
-#ifdef USE_ROCM
-#include "../util/hip_driver.h"
-#include "../util/system.h"
-#include "common/util/hip_runtime.h"
-#else
 #include "../util/cuda_driver.h"
 #include "../util/system.h"
 #include "common/util/cuda_runtime.h"
-#endif
 
 namespace transformer_engine {
 

transformer_engine/common/util/rtc.h

@@ -19,13 +19,8 @@
 #include <vector>
 
 #include "../common.h"
-#ifdef __HIP_PLATFORM_AMD__
-#include "../util/hip_driver.h"
-#include "../util/hip_runtime.h"
-#else
 #include "../util/cuda_driver.h"
 #include "../util/cuda_runtime.h"
-#endif
 
 namespace transformer_engine {
 

transformer_engine/pytorch/csrc/extensions/gemm.cpp

@@ -12,11 +12,7 @@
 
 #include "../common.h"
 #include "common.h"
-#ifdef USE_ROCM
-#include "common/util/hip_runtime.h"
-#else
 #include "common/util/cuda_runtime.h"
-#endif
 #include "common/util/system.h"
 #include "extensions.h"
 #include "pybind.h"