Merge branch 'master' into amd-master

Change-Id: Ieac16a172dacdcd6d57e253a82bb13e6632f5cab

Author: mangupta

.gitignore

@@ -1,101 +1,3 @@
 *.o
 
-rodinia_3.0/hip/b+tree/b+tree
-rodinia_3.0/hip/b+tree/output.txt
-rodinia_3.0/hip/b+tree/run0.diff
-rodinia_3.0/hip/b+tree/run0.out
-rodinia_3.0/hip/backprop/backprop
-rodinia_3.0/hip/backprop/out_hip.txt
-rodinia_3.0/hip/backprop/run0.diff
-rodinia_3.0/hip/backprop/run0.out
-rodinia_3.0/hip/bfs/bfs
-rodinia_3.0/hip/bfs/run0.diff
-rodinia_3.0/hip/bfs/run0.out
-rodinia_3.0/hip/bfs/run1.diff
-rodinia_3.0/hip/bfs/run1.out
-rodinia_3.0/hip/cfd/density
-rodinia_3.0/hip/cfd/density_energy
-rodinia_3.0/hip/cfd/euler3d
-rodinia_3.0/hip/cfd/momentum
-rodinia_3.0/hip/cfd/run0.diff
-rodinia_3.0/hip/cfd/run0.out
-rodinia_3.0/hip/cfd/run1.diff
-rodinia_3.0/hip/cfd/run1.out
-rodinia_3.0/hip/dwt2d/192.bmp.dwt.b
-rodinia_3.0/hip/dwt2d/192.bmp.dwt.g
-rodinia_3.0/hip/dwt2d/192.bmp.dwt.r
-rodinia_3.0/hip/dwt2d/dwt2d
-rodinia_3.0/hip/dwt2d/rgb.bmp.dwt.b
-rodinia_3.0/hip/dwt2d/rgb.bmp.dwt.g
-rodinia_3.0/hip/dwt2d/rgb.bmp.dwt.r
-rodinia_3.0/hip/dwt2d/run0.diff
-rodinia_3.0/hip/dwt2d/run0.out
-rodinia_3.0/hip/dwt2d/run1.diff
-rodinia_3.0/hip/dwt2d/run1.out
-rodinia_3.0/hip/gaussian/gaussian
-rodinia_3.0/hip/gaussian/run0.diff
-rodinia_3.0/hip/gaussian/run0.out
-rodinia_3.0/hip/gaussian/run1.diff
-rodinia_3.0/hip/gaussian/run1.out
-rodinia_3.0/hip/gaussian/run2.diff
-rodinia_3.0/hip/gaussian/run2.out
-rodinia_3.0/hip/gaussian/run3.diff
-rodinia_3.0/hip/gaussian/run3.out
-rodinia_3.0/hip/gaussian/run4.diff
-rodinia_3.0/hip/gaussian/run4.out
-rodinia_3.0/hip/heartwall/heartwall
-rodinia_3.0/hip/heartwall/run0.diff
-rodinia_3.0/hip/heartwall/run0.out
-rodinia_3.0/hip/hotspot/hotspot
-rodinia_3.0/hip/hotspot/run0.diff
-rodinia_3.0/hip/hotspot/run0.out
-rodinia_3.0/hip/hybridsort/hybridsort
-rodinia_3.0/hip/hybridsort/run0.diff
-rodinia_3.0/hip/hybridsort/run0.out
-rodinia_3.0/hip/kmeans/kmeans
-rodinia_3.0/hip/kmeans/run0.diff
-rodinia_3.0/hip/kmeans/run0.out
-rodinia_3.0/hip/kmeans/run1.diff
-rodinia_3.0/hip/kmeans/run1.out
-rodinia_3.0/hip/kmeans/run2.diff
-rodinia_3.0/hip/kmeans/run2.out
-rodinia_3.0/hip/kmeans/run3.diff
-rodinia_3.0/hip/kmeans/run3.out
-rodinia_3.0/hip/lavaMD/lavaMD
-rodinia_3.0/hip/lavaMD/run0.diff
-rodinia_3.0/hip/lavaMD/run0.out
-rodinia_3.0/hip/lavaMD/run1.diff
-rodinia_3.0/hip/lavaMD/run1.out
-rodinia_3.0/hip/lavaMD/run2.diff
-rodinia_3.0/hip/lavaMD/run2.out
-rodinia_3.0/hip/lavaMD/run3.diff
-rodinia_3.0/hip/lavaMD/run3.out
-rodinia_3.0/hip/lavaMD/run4.diff
-rodinia_3.0/hip/lavaMD/run4.out
-rodinia_3.0/hip/lud/lud_cuda
-rodinia_3.0/hip/lud/run0.diff
-rodinia_3.0/hip/lud/run0.out
-rodinia_3.0/hip/myocyte/myocyte
-rodinia_3.0/hip/myocyte/output.txt
-rodinia_3.0/hip/myocyte/run0.diff
-rodinia_3.0/hip/myocyte/run0.out
-rodinia_3.0/hip/nn/nn
-rodinia_3.0/hip/nn/run0.diff
-rodinia_3.0/hip/nn/run0.out
-rodinia_3.0/hip/nw/nw
-rodinia_3.0/hip/nw/run0.diff
-rodinia_3.0/hip/nw/run0.out
-rodinia_3.0/hip/pathfinder/pathfinder
-rodinia_3.0/hip/pathfinder/run0.diff
-rodinia_3.0/hip/pathfinder/run0.out
-rodinia_3.0/hip/srad/image_out.pgm
-rodinia_3.0/hip/srad/run0.diff
-rodinia_3.0/hip/srad/run0.out
-rodinia_3.0/hip/srad/srad_v1/srad
-rodinia_3.0/hip/srad/srad_v2/sradCPU
-rodinia_3.0/hip/srad/srad_v2/sradGPU
-rodinia_3.0/hip/streamcluster/output.txt
-rodinia_3.0/hip/streamcluster/run0.diff
-rodinia_3.0/hip/streamcluster/run0.out
-rodinia_3.0/hip/streamcluster/streamcluster
 vectorAdd/vectoradd_hip.exe

add4/Makefile

@@ -6,7 +6,7 @@ common.o: common.cpp common.h Makefile
 
 HIP_PATH?= $(wildcard /opt/rocm/hip)
 ifeq (,$(HIP_PATH))
-	HIP_PATH=../../..
+	HIP_PATH=/opt/rocm
 endif
 HIPCC=$(HIP_PATH)/bin/hipcc
 

cuda-stream/Makefile

@@ -1,6 +1,6 @@
 HIP_PATH?= $(wildcard /opt/rocm/hip)
 ifeq (,$(HIP_PATH))
-	HIP_PATH=../../..
+	HIP_PATH=/opt/rocm
 endif
 HIPCC=$(HIP_PATH)/bin/hipcc
 

gpu-burn/Makefile

@@ -1,4 +1,9 @@
-HIP_PATH ?= /opt/rocm/hip
+HIP_PATH?= $(wildcard /opt/rocm/hip)
+ifeq (,$(HIP_PATH))
+        HIP_PATH=/opt/rocm
+endif
+
+
 HCC_PATH ?= /opt/rocm/hcc
 HIP_PLATFORM = $(shell $(HIP_PATH)/bin/hipconfig --platform)
 HIP_INCLUDE = -I${HIP_PATH}/include -I${HCC_PATH}/include

mini-nbody/hip/HIP-nbody-block.sh

@@ -17,7 +17,7 @@ if [ -d /opt/rocm/hip ]
 then
     HIP_PATH=/opt/rocm/hip
 else
-    echo "Please install rocm package"
+    HIP_PATH=/opt/rocm
 fi
 
 fi

mini-nbody/hip/HIP-nbody-orig.sh

@@ -14,7 +14,7 @@ if [ -d /opt/rocm/hip ]
 then
     HIP_PATH=/opt/rocm/hip
 else
-    echo "Please install rocm package"
+    HIP_PATH=/opt/rocm
 fi
 
 fi

mini-nbody/hip/HIP-nbody-soa.sh

@@ -17,7 +17,7 @@ if [ -d /opt/rocm/hip ]
 then
     HIP_PATH=/opt/rocm/hip
 else
-    echo "Please install rocm package"
+    HIP_PATH=/opt/rocm
 fi
 
 fi

openmp-helloworld/CMakeLists.txt

@@ -0,0 +1,50 @@
+project(openmp_helloworld)
+
+cmake_minimum_required(VERSION 3.16)
+
+# Search for rocm in common locations
+if(WIN32)
+  list(APPEND CMAKE_PREFIX_PATH "C:/hip")
+  list(APPEND CMAKE_PREFIX_PATH "C:/Program Files/AMD HIP SDK/hip")
+else()
+  list(APPEND CMAKE_PREFIX_PATH /opt/rocm/hip /opt/rocm)
+endif()
+
+# Find HIP.
+# The user may override AMDGPU_TARGETS defined in the HIP config file
+# to select the AMDGPU archs to compile for.
+# ex. set(AMDGPU_TARGETS "gfx803;gfx900;gfx906")
+find_package(hip REQUIRED)
+
+# Find OpenMP.
+find_package(OpenMP REQUIRED)
+
+# Set compiler and linker.
+if(NOT WIN32)
+  set(CMAKE_CXX_COMPILER ${HIP_HIPCC_EXECUTABLE})
+  set(CMAKE_CXX_LINKER   ${HIP_HIPCC_EXECUTABLE})
+endif()
+
+set(CMAKE_BUILD_TYPE Release)
+
+if(WIN32)
+  # Compile for OpenMP code (Windows requires this).
+  set(OpenMP_CXX_FLAGS "-Xclang -fopenmp")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
+  # Tell CMake where to find the OpenMP libraries (libomp.lib).
+  link_directories("C:\\Program Files (x86)\\Microsoft Visual Studio\\2019\\Community\\VC\\Tools\\Llvm\\x64\\lib")
+endif()
+
+# Source files.
+set(CPP_SOURCES ${CMAKE_SOURCE_DIR}/openmp_helloworld.cpp)
+
+# Preparing the executable.
+add_executable(test_openmp_helloworld ${CPP_SOURCES})
+
+# Link Libraries - HIP Device and OpenMP.
+target_compile_options(test_openmp_helloworld PRIVATE ${OpenMP_CXX_FLAGS})
+target_link_libraries(test_openmp_helloworld PRIVATE hip::device ${OpenMP_CXX_FLAGS})
+
+if(WIN32)
+  target_link_libraries(test_openmp_helloworld PRIVATE OpenMP::OpenMP_CXX)
+endif()

openmp-helloworld/Makefile

@@ -0,0 +1,30 @@
+HIP_PATH?= $(wildcard /opt/rocm/hip)
+ifeq (,$(HIP_PATH))
+	HIP_PATH=/opt/rocm
+endif
+
+HIPCC=$(HIP_PATH)/bin/hipcc
+CXX=$(HIPCC)
+CXXFLAGS =-fopenmp
+
+SOURCES = openmp_helloworld.cpp
+
+EXECUTABLE=./openmp_helloworld.exe
+
+.PHONY: test
+
+
+all: $(EXECUTABLE) test
+
+
+$(EXECUTABLE):
+	$(CXX) $(CXXFLAGS) $(SOURCES) -o $@
+
+
+test: $(EXECUTABLE)
+	$(EXECUTABLE)
+
+
+clean:
+	rm -f $(EXECUTABLE) *.o
+

openmp-helloworld/README.md

@@ -0,0 +1,87 @@
+# Simple OpenMP hello world example written directly to the HIP interface.
+
+## Requirements
+* Installed ROCm 3.9 or newer. See  [ROCm Installation Guide](https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation-Guide.html).
+
+## Windows Requirements
+* Set HIP_DIR to the HIP installation location.
+* libamdhip64.dll and amd_comgr.dll must be in PATH or in System32.
+* Install MS Visual Studio 2019 for C++ development with Optional C++ Clang tools for Windows.
+* Ensure libomp.dll from MSVC C++ Clang tools is in PATH (by default, location is C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\Llvm\x64\bin).
+* Modify the CMakeLists.txt of this project to the corresponding libomp.lib location.
+
+## How to run this code:
+
+### Using Make on Linux:
+* To build and run: `make`.
+* To clean the environment: `make clean`.
+
+
+### Using CMake on Linux:
+* To build: `mkdir -p build; cd build; cmake ..; make`
+* To run the test: `./test_openmp_helloworld`
+* To clean the build environment: `make clean`
+
+### Using CMake on Windows:
+* CMake Command: `cmake -G Ninja -DCMAKE_C_COMPILER=<HIP_DIR>/bin/clang.exe -DCMAKE_CXX_COMPILER=<HIP_DIR>/bin/clang++.exe`
+* To build: `ninja`
+* To run the test: `./test_openmp_helloworld`
+
+**Note:** You may override `AMDGPU_TARGETS` in the HIP config file by modifying the CMakeLists.txt.
+
+## Expected Results:
+```
+info: running on device Device 66a3
+Hello World... from OMP thread = 0
+Hello World... from OMP thread = 15
+Hello World... from OMP thread = 3
+Hello World... from OMP thread = 13
+Hello World... from OMP thread = 11
+Hello World... from OMP thread = 8
+Hello World... from OMP thread = 4
+Hello World... from OMP thread = 1
+Hello World... from OMP thread = 10
+Hello World... from OMP thread = 9
+Hello World... from OMP thread = 7
+Hello World... from OMP thread = 12
+Hello World... from OMP thread = 6
+Hello World... from OMP thread = 14
+Hello World... from OMP thread = 5
+Hello World... from OMP thread = 2
+Hello World... from HIP thread = 0
+Hello World... from HIP thread = 2
+Hello World... from HIP thread = 5
+Hello World... from HIP thread = 14
+Hello World... from HIP thread = 6
+Hello World... from HIP thread = 12
+Hello World... from HIP thread = 7
+Hello World... from HIP thread = 9
+Hello World... from HIP thread = 1
+Hello World... from HIP thread = 11
+Hello World... from HIP thread = 10
+Hello World... from HIP thread = 4
+Hello World... from HIP thread = 8
+Hello World... from HIP thread = 13
+Hello World... from HIP thread = 15
+Hello World... from HIP thread = 3
+Device Results:
+  A_d[0] = 0
+  A_d[1] = 1
+  A_d[2] = 2
+  A_d[3] = 3
+  A_d[4] = 4
+  A_d[5] = 5
+  A_d[6] = 6
+  A_d[7] = 7
+  A_d[8] = 8
+  A_d[9] = 9
+  A_d[10] = 10
+  A_d[11] = 11
+  A_d[12] = 12
+  A_d[13] = 13
+  A_d[14] = 14
+  A_d[15] = 15
+PASSED!
+```
+
+**Note:** HIP thread's printf may not display on builds with printf support disabled.