AI/rocm-systems
2
0
Fork 0
Code Issues Pull-Requests Actions Pakete Projekte Releases Wiki Aktivität

Replace hipLaunchKernel -> hipLaunchKernelGGL

Quellcode durchsuchen 
Change-Id: I4d99009e1199811d417becf1e1b934ec4d4e30be
Dieser Commit ist enthalten in:
Maneesh Gupta
2018-10-17 12:01:44 +05:30
Ursprung 3bf8e9ad8e
Commit e1fe095471
78 geänderte Dateien mit 246 neuen und 264 gelöschten Zeilen
Patch-Datei herunterladen Vergleichs-Datei herunterladen Alle Dateien ausklappen Alle Dateien einklappen
hipamd/tests/src/runtimeApi/device/hipDeviceSynchronize.cpp
+2 -2
Datei anzeigen
@@ -33,7 +33,7 @@ THE SOFTWARE.
#define _SIZE sizeof(int) * 1024 * 1024
#define NUM_STREAMS 2
__global__ void Iter(hipLaunchParm lp, int* Ad, int num) {
__global__ void Iter(int* Ad, int num) {
int tx = threadIdx.x + blockIdx.x * blockDim.x;
// Kernel loop designed to execute very slowly... ... ... so we can test timing-related
// behavior below
@@ -58,7 +58,7 @@ int main() {
HIPCHECK(hipMemcpyAsync(Ad[i], A[i], _SIZE, hipMemcpyHostToDevice, stream[i]));
}
for (int i = 0; i < NUM_STREAMS; i++) {
hipLaunchKernel(HIP_KERNEL_NAME(Iter), dim3(1), dim3(1), 0, stream[i], Ad[i], 1 << 30);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Iter), dim3(1), dim3(1), 0, stream[i], Ad[i], 1 << 30);
}
for (int i = 0; i < NUM_STREAMS; i++) {
HIPCHECK(hipMemcpyAsync(A[i], Ad[i], _SIZE, hipMemcpyDeviceToHost, stream[i]));
hipamd/tests/src/runtimeApi/event/hipEventRecord.cpp
+1 -1
Datei anzeigen
@@ -66,7 +66,7 @@ int main(int argc, char* argv[]) {
// Record the start event
HIPCHECK(hipEventRecord(start, NULL));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const float*>(A_d), static_cast<const float*>(B_d), C_d, N);
hipamd/tests/src/runtimeApi/memory/hipArray.cpp
+3 -3
Datei anzeigen
@@ -67,7 +67,7 @@ void memcpy2Dtest(size_t numW, size_t numH, bool usePinnedHost) {
HIPCHECK(hipMemcpy2D(A_d, pitch_A, A_h, width, width, numH, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpy2D(B_d, pitch_B, B_h, width, width, numH, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0, A_d, B_d, C_d,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0, A_d, B_d, C_d,
(pitch_C / sizeof(T)) * numH);
HIPCHECK(hipMemcpy2D(C_h, width, C_d, pitch_C, width, numH, hipMemcpyDeviceToHost));
@@ -117,7 +117,7 @@ void memcpyArraytest(size_t numW, size_t numH, bool usePinnedHost, bool usePitch
HIPCHECK(hipMemcpyToArray(A_d, 0, 0, (void*)A_h, width, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpyToArray(B_d, 0, 0, (void*)B_h, width, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
(T*)A_d->data, (T*)B_d->data, (T*)C_d->data, numW);
HIPCHECK(hipMemcpy(C_h, C_d->data, width, hipMemcpyDeviceToHost));
@@ -156,7 +156,7 @@ void memcpyArraytest(size_t numW, size_t numH, bool usePinnedHost, bool usePitch
hipMemcpyHostToDevice));
}
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
(T*)A_d->data, (T*)B_d->data, (T*)C_d->data, numW * numH);
HIPCHECK(hipMemcpy2D((void*)C_h, width, (void*)C_d->data, width, width, numH,
hipamd/tests/src/runtimeApi/memory/hipHostGetFlags.cpp
+2 -2
Datei anzeigen
@@ -32,7 +32,7 @@ THE SOFTWARE.
#define LEN 1024 * 1024
#define SIZE LEN * sizeof(float)
__global__ void Add(hipLaunchParm lp, float* Ad, float* Bd, float* Cd) {
__global__ void Add(float* Ad, float* Bd, float* Cd) {
int tx = threadIdx.x + blockIdx.x * blockDim.x;
Cd[tx] = Ad[tx] + Bd[tx];
}
@@ -74,7 +74,7 @@ int main() {
dim3 dimGrid(LEN / 512, 1, 1);
dim3 dimBlock(512, 1, 1);
hipLaunchKernel(HIP_KERNEL_NAME(Add), dimGrid, dimBlock, 0, 0, Ad, Bd, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Add), dimGrid, dimBlock, 0, 0, Ad, Bd, Cd);
HIPCHECK(
hipMemcpy(C, Cd, SIZE, hipMemcpyDeviceToHost)); // Note this really HostToHost not
hipamd/tests/src/runtimeApi/memory/hipHostRegister.cpp
+2 -2
Datei anzeigen
@@ -28,7 +28,7 @@ THE SOFTWARE.
#include "test_common.h"
#include <malloc.h>
__global__ void Inc(hipLaunchParm lp, float* Ad) {
__global__ void Inc(float* Ad) {
int tx = threadIdx.x + blockIdx.x * blockDim.x;
Ad[tx] = Ad[tx] + float(1);
}
@@ -99,7 +99,7 @@ int main(int argc, char* argv[]) {
// Reference the registered device pointer Ad from inside the kernel:
for (int i = 0; i < num_devices; i++) {
HIPCHECK(hipSetDevice(i));
hipLaunchKernel(Inc, dim3(N / 512), dim3(512), 0, 0, Ad[i]);
hipLaunchKernelGGL(Inc, dim3(N / 512), dim3(512), 0, 0, Ad[i]);
HIPCHECK(hipDeviceSynchronize());
}
hipamd/tests/src/runtimeApi/memory/hipMemcpy.cpp
+1 -1
Datei anzeigen
@@ -230,7 +230,7 @@ void memcpytest2(DeviceMemory<T>* dmem, HostMemory<T>* hmem, size_t numElements,
useMemkindDefault ? hipMemcpyDefault : hipMemcpyHostToDevice));
}
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const T*>(dmem->A_d()), static_cast<const T*>(dmem->B_d()),
dmem->C_d(), numElements);
hipamd/tests/src/runtimeApi/memory/hipMemcpyDtoD.cpp
+2 -2
Datei anzeigen
@@ -51,7 +51,7 @@ int main() {
HIPCHECK(hipSetDevice(0));
HIPCHECK(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpy(B_d, B_h, Nbytes, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(A_d), static_cast<const int*>(B_d), C_d, N);
HIPCHECK(hipMemcpy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
@@ -62,7 +62,7 @@ int main() {
HIPCHECK(hipMemcpyDtoD((hipDeviceptr_t)X_d, (hipDeviceptr_t)A_d, Nbytes));
HIPCHECK(hipMemcpyDtoD((hipDeviceptr_t)Y_d, (hipDeviceptr_t)B_d, Nbytes));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(X_d), static_cast<const int*>(Y_d), Z_d, N);
HIPCHECK(hipMemcpyDtoH(C_h, (hipDeviceptr_t)Z_d, Nbytes));
HIPCHECK(hipDeviceSynchronize());
hipamd/tests/src/runtimeApi/memory/hipMemcpyDtoDAsync.cpp
+2 -2
Datei anzeigen
@@ -52,7 +52,7 @@ int main() {
HIPCHECK(hipSetDevice(0));
HIPCHECK(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpy(B_d, B_h, Nbytes, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(A_d), static_cast<const int*>(B_d), C_d, N);
HIPCHECK(hipMemcpy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
@@ -63,7 +63,7 @@ int main() {
HIPCHECK(hipMemcpyDtoDAsync((hipDeviceptr_t)X_d, (hipDeviceptr_t)A_d, Nbytes, s));
HIPCHECK(hipMemcpyDtoDAsync((hipDeviceptr_t)Y_d, (hipDeviceptr_t)B_d, Nbytes, s));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(X_d), static_cast<const int*>(Y_d), Z_d, N);
HIPCHECK(hipMemcpyDtoHAsync(C_h, (hipDeviceptr_t)Z_d, Nbytes, s));
HIPCHECK(hipStreamSynchronize(s));
hipamd/tests/src/runtimeApi/memory/hipMemcpyPeer.cpp
+2 -2
Datei anzeigen
@@ -50,7 +50,7 @@ int main() {
HIPCHECK(hipSetDevice(0));
HIPCHECK(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpy(B_d, B_h, Nbytes, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(A_d), static_cast<const int*>(B_d), C_d, N);
HIPCHECK(hipMemcpy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
@@ -62,7 +62,7 @@ int main() {
Nbytes); // this call is eqv to hipMemcpy(hipMemcpyD2D) which goes via stg bufs.
hipMemcpyPeer(Y_d, 1, B_d, 0, Nbytes);
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(X_d), static_cast<const int*>(Y_d), Z_d, N);
HIPCHECK(hipMemcpy(C_h, Z_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
hipamd/tests/src/runtimeApi/memory/hipMemcpyPeerAsync.cpp
+2 -2
Datei anzeigen
@@ -54,7 +54,7 @@ int main() {
HIPCHECK(hipSetDevice(0));
HIPCHECK(hipMemcpy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIPCHECK(hipMemcpy(B_d, B_h, Nbytes, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(A_d), static_cast<const int*>(B_d), C_d, N);
HIPCHECK(hipMemcpy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
@@ -65,7 +65,7 @@ int main() {
HIPCHECK(hipMemcpyPeerAsync(X_d, 1, A_d, 0, Nbytes, s));
HIPCHECK(hipMemcpyPeerAsync(Y_d, 1, B_d, 0, Nbytes, s));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(X_d), static_cast<const int*>(Y_d), Z_d, N);
HIPCHECK(hipMemcpy(C_h, Z_d, Nbytes, hipMemcpyDeviceToHost));
HIPCHECK(hipDeviceSynchronize());
hipamd/tests/src/runtimeApi/memory/hipMemcpy_simple.cpp
+1 -1
Datei anzeigen
@@ -61,7 +61,7 @@ void simpleTest1() {
HIPCHECK(memcopy(A_d, A_h, Nbytes, hipMemcpyHostToDevice));
HIPCHECK(memcopy(B_d, B_h, Nbytes, hipMemcpyHostToDevice));
hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const int*>(A_d), static_cast<const int*>(B_d), C_d, N);
HIPCHECK(memcopy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));
hipamd/tests/src/runtimeApi/memory/hipMemoryAllocateCoherent.cpp
+2 -2
Datei anzeigen
@@ -29,7 +29,7 @@ THE SOFTWARE.
#include <cstdio>
#include "hip/hip_runtime.h"
__global__ void Kernel(hipLaunchParm lp, volatile float* hostRes) {
__global__ void Kernel(volatile float* hostRes) {
int tid = threadIdx.x + blockIdx.x * blockDim.x;
hostRes[tid] = tid + 1;
__threadfence_system();
@@ -45,7 +45,7 @@ int main() {
hipHostMalloc((void**)&hostRes, blocks * sizeof(float), hipHostMallocMapped);
hostRes[0] = 0;
hostRes[1] = 0;
hipLaunchKernel(HIP_KERNEL_NAME(Kernel), dim3(1), dim3(blocks), 0, 0, hostRes);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Kernel), dim3(1), dim3(blocks), 0, 0, hostRes);
int eleCounter = 0;
while (eleCounter < blocks) {
// blocks until the value changes
hipamd/tests/src/runtimeApi/multiThread/hipMultiThreadStreams1.cpp
+2 -2
Datei anzeigen
@@ -82,9 +82,9 @@ void simpleVectorAdd(size_t numElements, int iters, hipStream_t stream) {
// HIPCHECK(hipStreamSynchronize(stream));
// This is the null stream?
// hipLaunchKernel(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0, A_d, B_d,
// hipLaunchKernelGGL(HipTest::vectorADD, dim3(blocks), dim3(threadsPerBlock), 0, 0, A_d, B_d,
// C_d, numElements);
hipLaunchKernel(HipTest::vectorADDReverse, dim3(blocks), dim3(threadsPerBlock), 0, 0,
hipLaunchKernelGGL(HipTest::vectorADDReverse, dim3(blocks), dim3(threadsPerBlock), 0, 0,
static_cast<const T*>(A_d), static_cast<const T*>(B_d), C_d, numElements);
MemTraits<C>::Copy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost, stream);
hipamd/tests/src/runtimeApi/multiThread/hipMultiThreadStreams2.cpp
+4 -4
Datei anzeigen
@@ -33,7 +33,7 @@ THE SOFTWARE.
template <typename T>
__global__ void Inc(hipLaunchParm lp, T* Array) {
__global__ void Inc(T* Array) {
int tx = threadIdx.x + blockIdx.x * blockDim.x;
Array[tx] = Array[tx] + T(1);
}
@@ -53,7 +53,7 @@ void run1(size_t size, hipStream_t stream) {
HIPCHECK(hipMemcpyAsync(Bh, Ah, size, hipMemcpyHostToHost, stream));
HIPCHECK(hipMemcpyAsync(Cd, Bh, size, hipMemcpyHostToDevice, stream));
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream, Cd);
HIPCHECK(hipMemcpyAsync(Dd, Cd, size, hipMemcpyDeviceToDevice, stream));
HIPCHECK(hipMemcpyAsync(Eh, Dd, size, hipMemcpyDeviceToHost, stream));
HIPCHECK(hipDeviceSynchronize());
@@ -80,8 +80,8 @@ void run(size_t size, hipStream_t stream1, hipStream_t stream2) {
HIPCHECK(hipMemcpyAsync(Bhh, Ahh, size, hipMemcpyHostToHost, stream2));
HIPCHECK(hipMemcpyAsync(Cd, Bh, size, hipMemcpyHostToDevice, stream1));
HIPCHECK(hipMemcpyAsync(Cdd, Bhh, size, hipMemcpyHostToDevice, stream2));
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream1, Cd);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream2, Cdd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream1, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 500), dim3(500), 0, stream2, Cdd);
HIPCHECK(hipMemcpyAsync(Dd, Cd, size, hipMemcpyDeviceToDevice, stream1));
HIPCHECK(hipMemcpyAsync(Ddd, Cdd, size, hipMemcpyDeviceToDevice, stream2));
HIPCHECK(hipMemcpyAsync(Eh, Dd, size, hipMemcpyDeviceToHost, stream1));
hipamd/tests/src/runtimeApi/stream/hipAPIStreamDisable.cpp
+4 -4
Datei anzeigen
@@ -28,7 +28,7 @@ THE SOFTWARE.
const int NN = 1 << 21;
__global__ void kernel(hipLaunchParm lp, float* x, float* y, int n) {
__global__ void kernel(float* x, float* y, int n) {
int tid = threadIdx.x;
if (tid < 1) {
for (int i = 0; i < n; i++) {
@@ -38,7 +38,7 @@ __global__ void kernel(hipLaunchParm lp, float* x, float* y, int n) {
}
}
__global__ void nKernel(hipLaunchParm lp, float* y) {
__global__ void nKernel(float* y) {
int tid = threadIdx.x;
y[tid] = y[tid] + 1.0f;
}
@@ -55,8 +55,8 @@ int main() {
for (int i = 0; i < num_streams; i++) {
HIPCHECK(hipStreamCreate(&streams[i]));
HIPCHECK(hipMalloc(&data[i], NN * sizeof(float)));
hipLaunchKernel(HIP_KERNEL_NAME(kernel), dim3(1), dim3(1), 0, streams[i], data[i], xd, N);
hipLaunchKernel(HIP_KERNEL_NAME(nKernel), dim3(1), dim3(1), 0, 0, yd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(kernel), dim3(1), dim3(1), 0, streams[i], data[i], xd, N);
hipLaunchKernelGGL(HIP_KERNEL_NAME(nKernel), dim3(1), dim3(1), 0, 0, yd);
}
HIPCHECK(hipMemcpy(&x, xd, sizeof(float), hipMemcpyDeviceToHost));
hipamd/tests/src/runtimeApi/stream/hipAPIStreamEnable.cpp
+4 -4
Datei anzeigen
@@ -30,7 +30,7 @@ THE SOFTWARE.
const int NN = 1 << 21;
__global__ void kernel(hipLaunchParm lp, float* x, float* y, int n) {
__global__ void kernel(float* x, float* y, int n) {
int tid = threadIdx.x;
if (tid < 1) {
for (int i = 0; i < n; i++) {
@@ -40,7 +40,7 @@ __global__ void kernel(hipLaunchParm lp, float* x, float* y, int n) {
}
}
__global__ void nKernel(hipLaunchParm lp, float* y) {
__global__ void nKernel(float* y) {
int tid = threadIdx.x;
y[tid] = y[tid] + 1.0f;
}
@@ -57,8 +57,8 @@ int main() {
for (int i = 0; i < num_streams; i++) {
HIPCHECK(hipStreamCreate(&streams[i]));
HIPCHECK(hipMalloc(&data[i], NN * sizeof(float)));
hipLaunchKernel(HIP_KERNEL_NAME(kernel), dim3(1), dim3(1), 0, streams[i], data[i], xd, N);
hipLaunchKernel(HIP_KERNEL_NAME(nKernel), dim3(1), dim3(1), 0, 0, yd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(kernel), dim3(1), dim3(1), 0, streams[i], data[i], xd, N);
hipLaunchKernelGGL(HIP_KERNEL_NAME(nKernel), dim3(1), dim3(1), 0, 0, yd);
}
HIPCHECK(hipMemcpy(&x, xd, sizeof(float), hipMemcpyDeviceToHost));
hipamd/tests/src/runtimeApi/stream/hipNullStream.cpp
+4 -4
Datei anzeigen
@@ -33,7 +33,7 @@ int p_db = 0;
template <typename T>
__global__ void vectorADDRepeat(hipLaunchParm lp, const T* A_d, const T* B_d, T* C_d, size_t NELEM,
__global__ void vectorADDRepeat(const T* A_d, const T* B_d, T* C_d, size_t NELEM,
int repeat) {
size_t offset = (blockIdx.x * blockDim.x + threadIdx.x);
size_t stride = blockDim.x * gridDim.x;
@@ -113,7 +113,7 @@ void Streamer<T>::enqueAsync() {
printf("testing: %s numElements=%zu size=%6.2fMB\n", __func__, _numElements,
_numElements * sizeof(T) / 1024.0 / 1024.0);
unsigned blocks = HipTest::setNumBlocks(blocksPerCU, threadsPerBlock, _numElements);
hipLaunchKernel(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0, _stream,
hipLaunchKernelGGL(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0, _stream,
static_cast<const T*>(_A_d), static_cast<const T*>(_B_d), _C_d, _numElements,
p_repeat);
}
@@ -206,7 +206,7 @@ int main(int argc, char* argv[]) {
// Dispatch to NULL stream, should wait for prior async activity to complete before
// beginning:
hipLaunchKernel(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0,
hipLaunchKernelGGL(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0,
0 /*nullstream*/, static_cast<const int*>(lastStreamer->_C_d),
static_cast<const int*>(lastStreamer->_C_d), nullStreamer->_C_d,
numElements, 1 /*repeat*/);
@@ -242,7 +242,7 @@ int main(int argc, char* argv[]) {
// Dispatch to NULL stream, should wait for prior async activity to complete before
// beginning:
hipLaunchKernel(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0,
hipLaunchKernelGGL(vectorADDRepeat, dim3(blocks), dim3(threadsPerBlock), 0,
0 /*nullstream*/, static_cast<const int*>(lastStreamer->_C_d),
static_cast<const int*>(lastStreamer->_C_d), nullStreamer->_C_d,
numElements, 1 /*repeat*/);
hipamd/tests/src/runtimeApi/stream/hipStream.h
+1 -1
Datei anzeigen
@@ -72,7 +72,7 @@ void D2H(T* Dst, T* Src, size_t size) {
}
template <typename T>
__global__ void Inc(hipLaunchParm lp, T* In) {
__global__ void Inc(T* In) {
int tx = threadIdx.x + blockIdx.x * blockDim.x;
In[tx] = In[tx] + 1;
}
hipamd/tests/src/runtimeApi/stream/hipStreamL5.cpp
+20 -20
Datei anzeigen
@@ -76,7 +76,7 @@ void test12345() {
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Ad, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Ch, Bd, size, stream);
HIPCHECK(hipDeviceSynchronize());
@@ -111,7 +111,7 @@ void test13452() {
H2D(Ad, Dh, size);
H2HAsync(Bh, Ah, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Ch, Bd, size, stream);
H2DAsync(Cd, Ch, size, stream);
@@ -152,7 +152,7 @@ void test14523() {
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Ch, Bd, size, stream);
H2DAsync(Cd, Ch, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
HIPCHECK(hipDeviceSynchronize());
@@ -190,7 +190,7 @@ void test15234() {
H2HAsync(Bh, Ah, size, stream);
D2HAsync(Ch, Ad, size, stream);
H2DAsync(Bd, Ch, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2DAsync(Cd, Bd, size, stream);
D2H(Eh, Cd, size);
@@ -217,7 +217,7 @@ void test23451() {
setArray(Ah, N, T(1));
H2DAsync(Ad, Ah, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Bh, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
@@ -254,7 +254,7 @@ void test24513() {
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Bh, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
HIPCHECK(hipDeviceSynchronize());
D2H(Eh, Cd, size);
@@ -291,7 +291,7 @@ void test25134() {
H2DAsync(Ad, Ah, size, stream);
D2HAsync(Bh, Ad, size, stream);
H2HAsync(Ch, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2DAsync(Cd, Bd, size, stream);
D2H(Eh, Cd, size);
@@ -324,7 +324,7 @@ void test21345() {
H2DAsync(Ad, Ah, size, stream);
H2HAsync(Ch, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Dh, Bd, size, stream);
@@ -358,7 +358,7 @@ void test34512() {
H2D(Ad, Ah, size);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Bh, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
@@ -393,7 +393,7 @@ void test35124() {
H2D(Ad, Dh, size);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
D2HAsync(Ah, Ad, size, stream);
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Bd, Bh, size, stream);
@@ -430,7 +430,7 @@ void test31245() {
H2D(Ad, Dh, size);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Bd, Bh, size, stream);
D2DAsync(Cd, Bd, size, stream);
@@ -469,7 +469,7 @@ void test32451() {
setArray(Eh, N, T(2));
H2D(Ad, Eh, size);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Ad);
H2DAsync(Bd, Ah, size, stream);
D2DAsync(Cd, Bd, size, stream);
D2HAsync(Bh, Cd, size, stream);
@@ -507,7 +507,7 @@ void test45123() {
D2HAsync(Ah, Bd, size, stream);
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Cd, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
D2H(Ch, Cd, size);
HIPCHECK(hipDeviceSynchronize());
@@ -539,7 +539,7 @@ void test41235() {
D2DAsync(Bd, Ad, size, stream);
D2HAsync(Ah, Bd, size, stream);
H2DAsync(Cd, Ah, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
D2HAsync(Bh, Cd, size, stream);
HIPCHECK(hipDeviceSynchronize());
@@ -574,7 +574,7 @@ void test42351() {
D2DAsync(Bd, Ad, size, stream);
H2DAsync(Cd, Ah, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Cd);
D2HAsync(Bh, Cd, size, stream);
H2HAsync(Ch, Bh, size, stream);
@@ -609,7 +609,7 @@ void test43512() {
H2D(Ad, Dh, size);
D2DAsync(Bd, Ad, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2HAsync(Ah, Bd, size, stream);
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Cd, Bh, size, stream);
@@ -645,7 +645,7 @@ void test51234() {
D2HAsync(Ah, Ad, size, stream);
H2HAsync(Bh, Ah, size, stream);
H2DAsync(Bd, Bh, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2DAsync(Cd, Bd, size, stream);
D2H(Ch, Cd, size);
@@ -681,7 +681,7 @@ void test52341() {
D2HAsync(Ah, Ad, size, stream);
H2DAsync(Bd, Ah, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2DAsync(Cd, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
@@ -723,7 +723,7 @@ void test53412() {
H2D(Bd, Eh, size);
D2HAsync(Ah, Ad, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Bd);
D2DAsync(Cd, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
H2DAsync(Dd, Ch, size, stream);
@@ -770,7 +770,7 @@ void test54123() {
D2DAsync(Cd, Bd, size, stream);
H2HAsync(Ch, Bh, size, stream);
H2DAsync(Dd, Ch, size, stream);
hipLaunchKernel(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Dd);
hipLaunchKernelGGL(HIP_KERNEL_NAME(Inc), dim3(N / 512), dim3(512), 0, stream, Dd);
D2H(Fh, Cd, size);
D2H(Gh, Dd, size);