AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-397576 Adding protection for HIP calls to understand MI failures

Browse Source 
Change-Id: Icd6a2cb3be5520ee57fdd76bfafa06801b7061c0
This commit is contained in:
gobhardw
2023-05-03 19:36:05 +05:30
committed by Gopesh Bhardwaj
parent 55c6a2d4ad
commit 0b401d71c6
5 changed files with 116 additions and 116 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/featuretests/profiler/apps/hello_world_hip.cpp
+15 -19
View File
@@ -25,6 +25,7 @@ THE SOFTWARE.
#include <string.h>
#include <hip/hip_runtime.h>
#include "utils/test_helper.h"
#include <fstream>
#include <iostream>
@@ -33,44 +34,39 @@ THE SOFTWARE.
#define SUCCESS 0
#define FAILURE 1
__global__ void helloworld(char *in, char *out) {
__global__ void helloworld(char* in, char* out) {
int num = hipThreadIdx_x + hipBlockDim_x * hipBlockIdx_x;
out[num] = in[num] + 1;
}
int main(int argc, char *argv[]) {
int main(int argc, char* argv[]) {
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
HIP_RC(hipGetDeviceProperties(&devProp, 0));
std::cout << " System minor " << devProp.minor << std::endl;
std::cout << " System major " << devProp.major << std::endl;
std::cout << " agent prop name " << devProp.name << std::endl;
/* Initial input,output for the host and create memory objects for the
* kernel*/
const char *input = "GdkknVnqkc";
const char* input = "GdkknVnqkc";
size_t strlength = strlen(input);
std::cout << "input string:" << std::endl;
std::cout << input << std::endl;
char *output = reinterpret_cast<char *>(malloc(strlength + 1));
char* output = reinterpret_cast<char*>(malloc(strlength + 1));
char *inputBuffer;
char *outputBuffer;
hipMalloc(reinterpret_cast<void **>(&inputBuffer),
(strlength + 1) * sizeof(char));
hipMalloc(reinterpret_cast<void **>(&outputBuffer),
(strlength + 1) * sizeof(char));
char* inputBuffer;
char* outputBuffer;
HIP_RC(hipMalloc(reinterpret_cast<void**>(&inputBuffer), (strlength + 1) * sizeof(char)));
HIP_RC(hipMalloc(reinterpret_cast<void**>(&outputBuffer), (strlength + 1) * sizeof(char)));
hipMemcpy(inputBuffer, input, (strlength + 1) * sizeof(char),
hipMemcpyHostToDevice);
HIP_RC(hipMemcpy(inputBuffer, input, (strlength + 1) * sizeof(char), hipMemcpyHostToDevice));
hipLaunchKernelGGL(helloworld, dim3(1), dim3(strlength), 0, 0, inputBuffer,
outputBuffer);
HIP_KL(hipLaunchKernelGGL(helloworld, dim3(1), dim3(strlength), 0, 0, inputBuffer, outputBuffer));
hipMemcpy(output, outputBuffer, (strlength + 1) * sizeof(char),
hipMemcpyDeviceToHost);
HIP_RC(hipMemcpy(output, outputBuffer, (strlength + 1) * sizeof(char), hipMemcpyDeviceToHost));
hipFree(inputBuffer);
hipFree(outputBuffer);
HIP_RC(hipFree(inputBuffer));
HIP_RC(hipFree(outputBuffer));
output[strlength] = '\0'; // Add the terminal character to the end of output.
std::cout << "\noutput string:" << std::endl;
tests/featuretests/profiler/apps/hello_world_omp.cpp
+11 -20
View File
@@ -31,39 +31,31 @@ THE SOFTWARE.
// HIP header
#include <hip/hip_runtime.h>
#include "utils/test_helper.h"
#define NUM_THREADS 16
#define CHECK(cmd) \
{ \
hipError_t error = cmd; \
if (error != hipSuccess) { \
fprintf(stderr, "error: '%s'(%d) at %s:%d\n", hipGetErrorString(error), \
error, __FILE__, __LINE__); \
exit(EXIT_FAILURE); \
} \
}
__global__ void hip_helloworld(unsigned omp_id, int *A_d) {
__global__ void hip_helloworld(unsigned omp_id, int* A_d) {
// Note: the printf command will only work if printf is enabled in your build.
// printf("Hello World... from HIP thread = %u\n", omp_id);
A_d[omp_id] = omp_id;
}
int main(int argc, char *argv[]) {
int main(int argc, char* argv[]) {
int *A_h, *A_d;
size_t Nbytes = NUM_THREADS * sizeof(int);
hipDeviceProp_t props;
CHECK(hipGetDeviceProperties(&props, 0 /*deviceID*/));
HIP_RC(hipGetDeviceProperties(&props, 0 /*deviceID*/));
// printf("info: running on device %s\n", props.name);
A_h = reinterpret_cast<int *>(malloc(Nbytes));
CHECK(hipMalloc(&A_d, Nbytes));
A_h = reinterpret_cast<int*>(malloc(Nbytes));
HIP_RC(hipMalloc(&A_d, Nbytes));
for (int i = 0; i < NUM_THREADS; i++) {
A_h[i] = 0;
}
CHECK(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIP_RC(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
// Beginning of parallel region
#pragma omp parallel num_threads(NUM_THREADS)
@@ -71,13 +63,12 @@ int main(int argc, char *argv[]) {
// fprintf(stderr, "Hello World... from OMP thread = %d\n",
// omp_get_thread_num());
hipLaunchKernelGGL(hip_helloworld, dim3(1), dim3(1), 0, 0,
omp_get_thread_num(), A_d);
HIP_KL(hipLaunchKernelGGL(hip_helloworld, dim3(1), dim3(1), 0, 0, omp_get_thread_num(), A_d));
}
// Ending of parallel region
hipStreamSynchronize(0);
CHECK(hipMemcpy(A_h, A_d, Nbytes, hipMemcpyDeviceToHost));
HIP_RC(hipStreamSynchronize(0));
HIP_RC(hipMemcpy(A_h, A_d, Nbytes, hipMemcpyDeviceToHost));
// printf("Device Results:\n");
for (int i = 0; i < NUM_THREADS; i++) {
// printf(" A_d[%d] = %d\n", i, A_h[i]);
@@ -86,6 +77,6 @@ int main(int argc, char *argv[]) {
printf("PASSED!\n");
free(A_h);
CHECK(hipFree(A_d));
HIP_RC(hipFree(A_d));
return 0;
}
tests/featuretests/profiler/apps/vector_add_hip.cpp
+26 -36
View File
@@ -27,8 +27,7 @@ THE SOFTWARE.
#include <iostream>
#include "hip/hip_runtime.h"
#define HIP_ASSERT(x) (assert((x) == hipSuccess))
#include "utils/test_helper.h"
#define WIDTH 1024
#define HEIGHT 1024
@@ -39,10 +38,8 @@ THE SOFTWARE.
#define THREADS_PER_BLOCK_Y 16
#define THREADS_PER_BLOCK_Z 1
__global__ void vectoradd_float(float *__restrict__ a,
const float *__restrict__ b,
const float *__restrict__ c, int width,
int height) {
__global__ void vectoradd_float(float* __restrict__ a, const float* __restrict__ b,
const float* __restrict__ c, int width, int height) {
int x = hipBlockDim_x * hipBlockIdx_x + hipThreadIdx_x;
int y = hipBlockDim_y * hipBlockIdx_y + hipThreadIdx_y;
@@ -53,16 +50,16 @@ __global__ void vectoradd_float(float *__restrict__ a,
}
int main() {
float *hostA;
float *hostB;
float *hostC;
float* hostA;
float* hostB;
float* hostC;
float *deviceA;
float *deviceB;
float *deviceC;
float* deviceA;
float* deviceB;
float* deviceC;
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
HIP_RC(hipGetDeviceProperties(&devProp, 0));
std::cout << " System minor " << devProp.minor << std::endl;
std::cout << " System major " << devProp.major << std::endl;
std::cout << " agent prop name " << devProp.name << std::endl;
@@ -72,9 +69,9 @@ int main() {
int i;
int errors;
hostA = reinterpret_cast<float *>(malloc(NUM * sizeof(float)));
hostB = reinterpret_cast<float *>(malloc(NUM * sizeof(float)));
hostC = reinterpret_cast<float *>(malloc(NUM * sizeof(float)));
hostA = reinterpret_cast<float*>(malloc(NUM * sizeof(float)));
hostB = reinterpret_cast<float*>(malloc(NUM * sizeof(float)));
hostC = reinterpret_cast<float*>(malloc(NUM * sizeof(float)));
// initialize the input data
for (i = 0; i < NUM; i++) {
@@ -82,26 +79,19 @@ int main() {
hostC[i] = static_cast<float>(i) * 100.0f;
}
HIP_ASSERT(
hipMalloc(reinterpret_cast<void **>(&deviceA), NUM * sizeof(float)));
HIP_ASSERT(
hipMalloc(reinterpret_cast<void **>(&deviceB), NUM * sizeof(float)));
HIP_ASSERT(
hipMalloc(reinterpret_cast<void **>(&deviceC), NUM * sizeof(float)));
HIP_RC(hipMalloc(reinterpret_cast<void**>(&deviceA), NUM * sizeof(float)));
HIP_RC(hipMalloc(reinterpret_cast<void**>(&deviceB), NUM * sizeof(float)));
HIP_RC(hipMalloc(reinterpret_cast<void**>(&deviceC), NUM * sizeof(float)));
HIP_ASSERT(
hipMemcpy(deviceB, hostB, NUM * sizeof(float), hipMemcpyHostToDevice));
HIP_ASSERT(
hipMemcpy(deviceC, hostC, NUM * sizeof(float), hipMemcpyHostToDevice));
HIP_RC(hipMemcpy(deviceB, hostB, NUM * sizeof(float), hipMemcpyHostToDevice));
HIP_RC(hipMemcpy(deviceC, hostC, NUM * sizeof(float), hipMemcpyHostToDevice));
hipLaunchKernelGGL(
vectoradd_float,
dim3(WIDTH / THREADS_PER_BLOCK_X, HEIGHT / THREADS_PER_BLOCK_Y),
dim3(THREADS_PER_BLOCK_X, THREADS_PER_BLOCK_Y), 0, 0, deviceA, deviceB,
deviceC, WIDTH, HEIGHT);
HIP_KL(hipLaunchKernelGGL(vectoradd_float,
dim3(WIDTH / THREADS_PER_BLOCK_X, HEIGHT / THREADS_PER_BLOCK_Y),
dim3(THREADS_PER_BLOCK_X, THREADS_PER_BLOCK_Y), 0, 0, deviceA, deviceB,
deviceC, WIDTH, HEIGHT));
HIP_ASSERT(
hipMemcpy(hostA, deviceA, NUM * sizeof(float), hipMemcpyDeviceToHost));
HIP_RC(hipMemcpy(hostA, deviceA, NUM * sizeof(float), hipMemcpyDeviceToHost));
// verify the results
errors = 0;
@@ -116,9 +106,9 @@ int main() {
printf("PASSED!\n");
}
HIP_ASSERT(hipFree(deviceA));
HIP_ASSERT(hipFree(deviceB));
HIP_ASSERT(hipFree(deviceC));
HIP_RC(hipFree(deviceA));
HIP_RC(hipFree(deviceB));
HIP_RC(hipFree(deviceC));
free(hostA);
free(hostB);
tests/featuretests/profiler/apps/vector_add_mpi.cpp
+15 -41
View File
@@ -27,44 +27,21 @@ THE SOFTWARE.
#include <iostream>
#include "hip/hip_runtime.h"
#define HIP_RC(call) \
do { \
hipError_t err = call; \
if (hipSuccess != err) { \
printf("HIP ERROR (code = %d, %s) at %s:%d\n", err, \
hipGetErrorString(err), __FILE__, __LINE__); \
assert(0); \
exit(1); \
} \
} while (0)
#define HIP_KL(call) \
do { \
call; \
hipError_t err = hipGetLastError(); \
if (hipSuccess != err) { \
printf("HIP ERROR (code = %d, %s) at %s:%d\n", err, \
hipGetErrorString(err), __FILE__, __LINE__); \
assert(0); \
exit(1); \
} \
} while (0)
#include "utils/test_helper.h"
// CUDA kernel to add elements of two arrays
__global__ void add(int n, float *x, float *y) {
__global__ void add(int n, float* x, float* y) {
int index = blockIdx.x * blockDim.x + threadIdx.x;
int stride = blockDim.x * gridDim.x;
for (int i = index; i < n; i += stride)
y[i] = x[i] + y[i];
for (int i = index; i < n; i += stride) y[i] = x[i] + y[i];
}
int main(int argc, char *argv[]) {
int main(int argc, char* argv[]) {
int N = 1 << 20;
float *x = new float[N];
float *y = new float[N];
float *d_x;
float *d_y;
float* x = new float[N];
float* y = new float[N];
float* d_x;
float* d_y;
int myId;
int devId;
@@ -75,24 +52,22 @@ int main(int argc, char *argv[]) {
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &myId);
MPI_Comm_size(MPI_COMM_WORLD, &numRank);
hipGetDeviceCount(&deviceCount);
HIP_RC(hipGetDeviceCount(&deviceCount));
std::cout << "device count and rank is" << deviceCount << ": " << numRank
<< std::endl;
std::cout << "device count and rank is" << deviceCount << ": " << numRank << std::endl;
// set the device ID to the rank ID mod deviceCount (in this case 4 since
// there are 4 devices on a node)
devId = myId % deviceCount;
// set the device ID
hipSetDevice(devId);
HIP_RC(hipSetDevice(devId));
printf("Rank Id: %d | Device Id : %d | Num Devices: %d\n", myId, devId,
deviceCount);
printf("Rank Id: %d | Device Id : %d | Num Devices: %d\n", myId, devId, deviceCount);
fflush(stdout);
// Allocate Unified Memory -- accessible from CPU or GPU
hipMallocManaged(&d_x, N * sizeof(float));
hipMallocManaged(&d_y, N * sizeof(float));
HIP_RC(hipMallocManaged(&d_x, N * sizeof(float)));
HIP_RC(hipMallocManaged(&d_y, N * sizeof(float)));
// initialize x and y arrays on the host
for (int i = 0; i < N; i++) {
@@ -116,8 +91,7 @@ int main(int argc, char *argv[]) {
// Check for errors (all values should be 3.0f)
float maxError = 0.0f;
for (int i = 0; i < N; i++)
maxError = fmax(maxError, fabs(y[i] - 3.0f));
for (int i = 0; i < N; i++) maxError = fmax(maxError, fabs(y[i] - 3.0f));
printf("Max error: %f\n", maxError);
// Free memory
tests/featuretests/utils/test_helper.h
+49
View File
@@ -0,0 +1,49 @@
/*
Copyright (c) 2015-2016 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.
*/
#ifndef TESTS_FEATURETESTS_PROFILER_UTILS_TEST_HELPER_H_
#define TESTS_FEATURETESTS_PROFILER_UTILS_TEST_HELPER_H_
#include <cstdio>
#include <unistd.h>
#define HIP_RC(call) \
do { \
hipError_t err = call; \
if (hipSuccess != err) { \
fprintf(stderr, "HIP ERROR (%s) at %s:%d\n", hipGetErrorString(err), __FILE__, __LINE__); \
assert(0); \
exit(1); \
} \
} while (0)
#define HIP_KL(call) \
do { \
call; \
hipError_t err = hipGetLastError(); \
if (hipSuccess != err) { \
fprintf(stderr, "HIP ERROR (%s) at %s:%d\n", hipGetErrorString(err), __FILE__, __LINE__); \
assert(0); \
exit(1); \
} \
} while (0)
#endif // TESTS_FEATURETESTS_PROFILER_UTILS_TEST_HELPER_H_