Merge 'master' into 'amd-master'

Change-Id: I476b12c4e051b47fb2e62dbea709d268d9a2f630

include/hip/hcc_detail/hip_runtime_api.h

@@ -2684,20 +2684,7 @@ public:
         std::tie(*dptr, *bytes) = read_global_description(it0->second.cbegin(),
                 it0->second.cend(), name);
 
-        // HACK for SWDEV-173477
-        //
-        // For code objects with global symbols of length 0, ROCR runtime would
-        // ignore them even though they exist in the symbol table. Therefore the
-        // result from read_agent_globals() can't be trusted entirely.
-        //
-        // As a workaround to tame applications which depend on the existence of
-        // global symbols with length 0, always return hipSuccess here.
-        //
-        // This behavior shall be reverted once ROCR runtime has been fixed to
-        // address SWDEV-173477
-
-        //return *dptr ? hipSuccess : hipErrorNotFound;
-        return hipSuccess;
+        return *dptr ? hipSuccess : hipErrorNotFound;
     }
 
     hipError_t read_agent_global_from_process(hipDeviceptr_t* dptr, size_t* bytes,

src/hip_hcc.cpp

@@ -2469,14 +2469,16 @@ hipError_t hipHccGetAcceleratorView(hipStream_t stream, hc::accelerator_view** a
 namespace hip_impl {
     std::vector<hsa_agent_t> all_hsa_agents() {
         std::vector<hsa_agent_t> r{};
-        for (auto&& acc : hc::accelerator::get_all()) {
+        std::vector<hc::accelerator> visible_accelerators;
+        for (int i=0; i < g_deviceCnt; i++)
+            visible_accelerators.push_back(g_deviceArray[i]->_acc);
+        for (auto&& acc : visible_accelerators) {
             const auto agent = acc.get_hsa_agent();
 
             if (!agent || !acc.is_hsa_accelerator()) continue;
 
             r.emplace_back(*static_cast<hsa_agent_t*>(agent));
         }
-
         return r;
     }
 

tests/hipify-clang/unit_tests/samples/1_hipEvent/hipEvent.cpp

@@ -0,0 +1,181 @@
+// RUN: %run_test hipify "%s" "%t" %hipify_args %clang_args
+/*
+Copyright (c) 2015-2019 Advanced Micro Devices, Inc. All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+*/
+
+#include <iostream>
+
+// CHECK: #include <hip/hip_runtime.h>
+#include <cuda.h>
+
+#define WIDTH 1024
+
+#define NUM (WIDTH * WIDTH)
+
+#define THREADS_PER_BLOCK_X 4
+#define THREADS_PER_BLOCK_Y 4
+#define THREADS_PER_BLOCK_Z 1
+
+// Device (Kernel) function, it must be void
+__global__ void matrixTranspose(float* out, float* in, const int width) {
+    int x = blockDim.x * blockIdx.x + threadIdx.x;
+    int y = blockDim.y * blockIdx.y + threadIdx.y;
+    out[y * width + x] = in[x * width + y];
+}
+
+// CPU implementation of matrix transpose
+void matrixTransposeCPUReference(float* output, float* input, const unsigned int width) {
+    for (unsigned int j = 0; j < width; j++) {
+        for (unsigned int i = 0; i < width; i++) {
+            output[i * width + j] = input[j * width + i];
+        }
+    }
+}
+
+int main() {
+    float* Matrix;
+    float* TransposeMatrix;
+    float* cpuTransposeMatrix;
+
+    float* gpuMatrix;
+    float* gpuTransposeMatrix;
+
+    // CHECK: hipDeviceProp_t devProp;
+    cudaDeviceProp devProp;
+    // CHECK: hipGetDeviceProperties(&devProp, 0);
+    cudaGetDeviceProperties(&devProp, 0);
+
+    std::cout << "Device name " << devProp.name << std::endl;
+
+    // CHECK: hipEvent_t start, stop;
+    cudaEvent_t start, stop;
+    // CHECK: hipEventCreate(&start);
+    cudaEventCreate(&start);
+    // CHECK: hipEventCreate(&stop);
+    cudaEventCreate(&stop);
+    float eventMs = 1.0f;
+
+    int i;
+    int errors;
+
+    Matrix = (float*)malloc(NUM * sizeof(float));
+    TransposeMatrix = (float*)malloc(NUM * sizeof(float));
+    cpuTransposeMatrix = (float*)malloc(NUM * sizeof(float));
+
+    // initialize the input data
+    for (i = 0; i < NUM; i++) {
+        Matrix[i] = (float)i * 10.0f;
+    }
+
+    // allocate the memory on the device side
+    // CHECK: hipMalloc((void**)&gpuMatrix, NUM * sizeof(float));
+    cudaMalloc((void**)&gpuMatrix, NUM * sizeof(float));
+    // CHECK: hipMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(float));
+    cudaMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(float));
+
+    // Record the start event
+    // CHECK: hipEventRecord(start, NULL);
+    cudaEventRecord(start, NULL);
+
+    // Memory transfer from host to device
+    // CHECK: hipMemcpy(gpuMatrix, Matrix, NUM * sizeof(float), hipMemcpyHostToDevice);
+    cudaMemcpy(gpuMatrix, Matrix, NUM * sizeof(float), cudaMemcpyHostToDevice);
+
+    // Record the stop event
+    // CHECK: hipEventRecord(stop, NULL);
+    cudaEventRecord(stop, NULL);
+    // CHECK: hipEventSynchronize(stop);
+    cudaEventSynchronize(stop);
+    // CHECK: hipEventElapsedTime(&eventMs, start, stop);
+    cudaEventElapsedTime(&eventMs, start, stop);
+
+    printf("hipMemcpyHostToDevice time taken  = %6.3fms\n", eventMs);
+
+    // Record the start event
+    // CHECK: hipEventRecord(start, NULL);
+    cudaEventRecord(start, NULL);
+
+    // Lauching kernel from host
+
+    dim3 dimGrid(WIDTH / THREADS_PER_BLOCK_X, WIDTH / THREADS_PER_BLOCK_Y);
+    dim3 dimBlock(THREADS_PER_BLOCK_X, THREADS_PER_BLOCK_Y);
+    // CHECK: hipLaunchKernelGGL(matrixTranspose, dim3(dimGrid), dim3(dimBlock), 0, 0, gpuTransposeMatrix, gpuMatrix, WIDTH);
+    matrixTranspose <<<dimGrid, dimBlock>>>(gpuTransposeMatrix, gpuMatrix, WIDTH);
+
+    // Record the stop event
+    // CHECK: hipEventRecord(stop, NULL);
+    cudaEventRecord(stop, NULL);
+    // CHECK: hipEventSynchronize(stop);
+    cudaEventSynchronize(stop);
+
+    // CHECK: hipEventElapsedTime(&eventMs, start, stop);
+    cudaEventElapsedTime(&eventMs, start, stop);
+
+    printf("kernel Execution time             = %6.3fms\n", eventMs);
+
+    // Record the start event
+    // CHECK: hipEventRecord(start, NULL);
+    cudaEventRecord(start, NULL);
+
+    // Memory transfer from device to host
+    // CHECK: hipMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(float), hipMemcpyDeviceToHost);
+    cudaMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(float), cudaMemcpyDeviceToHost);
+
+    // Record the stop event
+    // CHECK: hipEventRecord(stop, NULL);
+    cudaEventRecord(stop, NULL);
+    // CHECK: hipEventSynchronize(stop);
+    cudaEventSynchronize(stop);
+    // CHECK: hipEventElapsedTime(&eventMs, start, stop);
+    cudaEventElapsedTime(&eventMs, start, stop);
+
+    printf("hipMemcpyDeviceToHost time taken  = %6.3fms\n", eventMs);
+
+    // CPU MatrixTranspose computation
+    matrixTransposeCPUReference(cpuTransposeMatrix, Matrix, WIDTH);
+
+    // verify the results
+    errors = 0;
+    double eps = 1.0E-6;
+    for (i = 0; i < NUM; i++) {
+        if (std::abs(TransposeMatrix[i] - cpuTransposeMatrix[i]) > eps) {
+            errors++;
+        }
+    }
+    if (errors != 0) {
+        printf("FAILED: %d errors\n", errors);
+    } else {
+        printf("PASSED!\n");
+    }
+
+    // free the resources on device side
+    // CHECK: hipFree(gpuMatrix);
+    cudaFree(gpuMatrix);
+    // CHECK: hipFree(gpuTransposeMatrix);
+    cudaFree(gpuTransposeMatrix);
+
+    // free the resources on host side
+    free(Matrix);
+    free(TransposeMatrix);
+    free(cpuTransposeMatrix);
+
+    return errors;
+}

tests/hipify-clang/unit_tests/samples/2_Cookbook/0_MatrixTranspose/MatrixTranspose.cpp

@@ -0,0 +1,130 @@
+// RUN: %run_test hipify "%s" "%t" %hipify_args %clang_args
+/*
+Copyright (c) 2015-2019 Advanced Micro Devices, Inc. All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+*/
+
+#include <iostream>
+
+// CHECK: #include <hip/hip_runtime.h>
+#include <cuda.h>
+
+#define WIDTH 1024
+
+#define NUM (WIDTH * WIDTH)
+
+#define THREADS_PER_BLOCK_X 4
+#define THREADS_PER_BLOCK_Y 4
+#define THREADS_PER_BLOCK_Z 1
+
+// Device (Kernel) function, it must be void
+__global__ void matrixTranspose(float* out, float* in, const int width) {
+    int x = blockDim.x * blockIdx.x + threadIdx.x;
+    int y = blockDim.y * blockIdx.y + threadIdx.y;
+    out[y * width + x] = in[x * width + y];
+}
+
+// CPU implementation of matrix transpose
+void matrixTransposeCPUReference(float* output, float* input, const unsigned int width) {
+    for (unsigned int j = 0; j < width; j++) {
+        for (unsigned int i = 0; i < width; i++) {
+            output[i * width + j] = input[j * width + i];
+        }
+    }
+}
+
+int main() {
+    float* Matrix;
+    float* TransposeMatrix;
+    float* cpuTransposeMatrix;
+
+    float* gpuMatrix;
+    float* gpuTransposeMatrix;
+
+    // CHECK: hipDeviceProp_t devProp;
+    cudaDeviceProp devProp;
+    // CHECK: hipGetDeviceProperties(&devProp, 0);
+    cudaGetDeviceProperties(&devProp, 0);
+
+    std::cout << "Device name " << devProp.name << std::endl;
+
+    int i;
+    int errors;
+
+    Matrix = (float*)malloc(NUM * sizeof(float));
+    TransposeMatrix = (float*)malloc(NUM * sizeof(float));
+    cpuTransposeMatrix = (float*)malloc(NUM * sizeof(float));
+
+    // initialize the input data
+    for (i = 0; i < NUM; i++) {
+        Matrix[i] = (float)i * 10.0f;
+    }
+
+    // allocate the memory on the device side
+    // CHECK: hipMalloc((void**)&gpuMatrix, NUM * sizeof(float));
+    cudaMalloc((void**)&gpuMatrix, NUM * sizeof(float));
+    // CHECK: hipMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(float));
+    cudaMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(float));
+
+    // Memory transfer from host to device
+    // CHECK: hipMemcpy(gpuMatrix, Matrix, NUM * sizeof(float), hipMemcpyHostToDevice);
+    cudaMemcpy(gpuMatrix, Matrix, NUM * sizeof(float), cudaMemcpyHostToDevice);
+
+    // Lauching kernel from host
+
+    dim3 dimGrid(WIDTH / THREADS_PER_BLOCK_X, WIDTH / THREADS_PER_BLOCK_Y);
+    dim3 dimBlock(THREADS_PER_BLOCK_X, THREADS_PER_BLOCK_Y);
+    // CHECK: hipLaunchKernelGGL(matrixTranspose, dim3(dimGrid), dim3(dimBlock), 0, 0, gpuTransposeMatrix, gpuMatrix, WIDTH);
+    matrixTranspose <<<dimGrid, dimBlock>>>(gpuTransposeMatrix, gpuMatrix, WIDTH);
+
+    // Memory transfer from device to host
+    // CHECK: hipMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(float), hipMemcpyDeviceToHost);
+    cudaMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(float), cudaMemcpyDeviceToHost);
+
+    // CPU MatrixTranspose computation
+    matrixTransposeCPUReference(cpuTransposeMatrix, Matrix, WIDTH);
+
+    // verify the results
+    errors = 0;
+    double eps = 1.0E-6;
+    for (i = 0; i < NUM; i++) {
+        if (std::abs(TransposeMatrix[i] - cpuTransposeMatrix[i]) > eps) {
+            errors++;
+        }
+    }
+    if (errors != 0) {
+        printf("FAILED: %d errors\n", errors);
+    } else {
+        printf("PASSED!\n");
+    }
+
+    // free the resources on device side
+    // CHECK: hipFree(gpuMatrix);
+    cudaFree(gpuMatrix);
+    // CHECK: hipFree(gpuTransposeMatrix);
+    cudaFree(gpuTransposeMatrix);
+
+    // free the resources on host side
+    free(Matrix);
+    free(TransposeMatrix);
+    free(cpuTransposeMatrix);
+
+    return errors;
+}