mirror of
https://github.com/KhronosGroup/OpenCL-CTS.git
synced 2026-03-19 06:09:01 +00:00
Refactor basic/intmath tests into a single file (#757)
* Refactor basic/intmath tests into a single file These six tests previously duplicated almost all of their sources in separate files. Bring them into a single file with templated routines to perform the tests. This change also makes use of more C++ features and utilities from the harness library to make the code cleaner. * Fix array indices for initialization * Use loop for buffer creation * Refactor to remove duplicate verification code Remove need for hardcoded number of test operations. * Address review comments * Use TEST_SKIPPED_ITSELF
This commit is contained in:
James Price
@@ -3,8 +3,7 @@ set(MODULE_NAME BASIC)
|
||||
set(${MODULE_NAME}_SOURCES
|
||||
main.cpp
|
||||
test_fpmath_float.cpp test_fpmath_float2.cpp test_fpmath_float4.cpp
|
||||
test_intmath_int.cpp test_intmath_int2.cpp test_intmath_int4.cpp
|
||||
test_intmath_long.cpp test_intmath_long2.cpp test_intmath_long4.cpp
|
||||
test_intmath.cpp
|
||||
test_hiloeo.cpp test_local.cpp test_pointercast.cpp
|
||||
test_if.cpp test_loop.cpp
|
||||
test_readimage.cpp test_readimage_int16.cpp test_readimage_fp32.cpp
|
||||
|
||||
240
test_conformance/basic/test_intmath.cpp
Normal file
240
test_conformance/basic/test_intmath.cpp
Normal file
@@ -0,0 +1,240 @@
|
||||
//
|
||||
// Copyright (c) 2020 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
|
||||
#include <functional>
|
||||
#include <string>
|
||||
#include <vector>
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
template <typename T> struct TestDef
|
||||
{
|
||||
const char *name;
|
||||
const char *kernel_code;
|
||||
std::function<T(T, T, T)> ref;
|
||||
};
|
||||
|
||||
template <typename T, unsigned N>
|
||||
int test_intmath(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements, std::string typestr)
|
||||
{
|
||||
TestDef<T> tests[] = {
|
||||
// Test addition
|
||||
{
|
||||
"test_add",
|
||||
R"(
|
||||
__kernel void test_add(__global TYPE *srcA,
|
||||
__global TYPE *srcB,
|
||||
__global TYPE *srcC,
|
||||
__global TYPE *dst)
|
||||
{
|
||||
int tid = get_global_id(0);
|
||||
dst[tid] = srcA[tid] + srcB[tid];
|
||||
};
|
||||
)",
|
||||
[](T a, T b, T c) { return a + b; },
|
||||
},
|
||||
|
||||
// Test subtraction
|
||||
{
|
||||
"test_sub",
|
||||
R"(
|
||||
__kernel void test_sub(__global TYPE *srcA,
|
||||
__global TYPE *srcB,
|
||||
__global TYPE *srcC,
|
||||
__global TYPE *dst)
|
||||
{
|
||||
int tid = get_global_id(0);
|
||||
dst[tid] = srcA[tid] - srcB[tid];
|
||||
};
|
||||
)",
|
||||
[](T a, T b, T c) { return a - b; },
|
||||
},
|
||||
|
||||
// Test multiplication
|
||||
{
|
||||
"test_mul",
|
||||
R"(
|
||||
__kernel void test_mul(__global TYPE *srcA,
|
||||
__global TYPE *srcB,
|
||||
__global TYPE *srcC,
|
||||
__global TYPE *dst)
|
||||
{
|
||||
int tid = get_global_id(0);
|
||||
dst[tid] = srcA[tid] * srcB[tid];
|
||||
};
|
||||
)",
|
||||
[](T a, T b, T c) { return a * b; },
|
||||
},
|
||||
|
||||
// Test multiply-accumulate
|
||||
{
|
||||
"test_mad",
|
||||
R"(
|
||||
__kernel void test_mad(__global TYPE *srcA,
|
||||
__global TYPE *srcB,
|
||||
__global TYPE *srcC,
|
||||
__global TYPE *dst)
|
||||
{
|
||||
int tid = get_global_id(0);
|
||||
dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];
|
||||
};
|
||||
)",
|
||||
[](T a, T b, T c) { return a * b + c; },
|
||||
},
|
||||
};
|
||||
|
||||
clMemWrapper streams[4];
|
||||
cl_int err;
|
||||
|
||||
if (std::is_same<T, cl_ulong>::value && !gHasLong)
|
||||
{
|
||||
log_info("64-bit integers are not supported on this device. Skipping "
|
||||
"test.\n");
|
||||
return TEST_SKIPPED_ITSELF;
|
||||
}
|
||||
|
||||
// Create host buffers and fill with random data.
|
||||
std::vector<T> inputA(num_elements * N);
|
||||
std::vector<T> inputB(num_elements * N);
|
||||
std::vector<T> inputC(num_elements * N);
|
||||
std::vector<T> output(num_elements * N);
|
||||
MTdataHolder d(gRandomSeed);
|
||||
for (int i = 0; i < num_elements; i++)
|
||||
{
|
||||
inputA[i] = (T)genrand_int64(d);
|
||||
inputB[i] = (T)genrand_int64(d);
|
||||
inputC[i] = (T)genrand_int64(d);
|
||||
}
|
||||
|
||||
size_t datasize = sizeof(T) * num_elements * N;
|
||||
|
||||
// Create device buffers.
|
||||
for (int i = 0; i < ARRAY_SIZE(streams); i++)
|
||||
{
|
||||
streams[i] =
|
||||
clCreateBuffer(context, CL_MEM_READ_WRITE, datasize, NULL, &err);
|
||||
test_error(err, "clCreateBuffer failed");
|
||||
}
|
||||
|
||||
// Copy input data to device.
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, datasize,
|
||||
inputA.data(), 0, NULL, NULL);
|
||||
test_error(err, "clEnqueueWriteBuffer failed\n");
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, datasize,
|
||||
inputB.data(), 0, NULL, NULL);
|
||||
test_error(err, "clEnqueueWriteBuffer failed\n");
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, datasize,
|
||||
inputC.data(), 0, NULL, NULL);
|
||||
test_error(err, "clEnqueueWriteBuffer failed\n");
|
||||
|
||||
std::string build_options = "-DTYPE=";
|
||||
build_options += typestr;
|
||||
|
||||
// Run test for each operation
|
||||
for (auto test : tests)
|
||||
{
|
||||
log_info("%s... ", test.name);
|
||||
|
||||
// Create kernel and set args
|
||||
clProgramWrapper program;
|
||||
clKernelWrapper kernel;
|
||||
err = create_single_kernel_helper(context, &program, &kernel, 1,
|
||||
&test.kernel_code, test.name,
|
||||
build_options.c_str());
|
||||
test_error(err, "create_single_kernel_helper failed\n");
|
||||
|
||||
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &streams[0]);
|
||||
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &streams[1]);
|
||||
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &streams[2]);
|
||||
err |= clSetKernelArg(kernel, 3, sizeof(cl_mem), &streams[3]);
|
||||
test_error(err, "clSetKernelArgs failed\n");
|
||||
|
||||
// Run kernel
|
||||
size_t threads[1] = { static_cast<size_t>(num_elements) };
|
||||
err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, threads, NULL, 0,
|
||||
NULL, NULL);
|
||||
test_error(err, "clEnqueueNDRangeKernel failed\n");
|
||||
|
||||
// Read results
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, datasize,
|
||||
output.data(), 0, NULL, NULL);
|
||||
test_error(err, "clEnqueueReadBuffer failed\n");
|
||||
|
||||
// Verify results
|
||||
for (int i = 0; i < num_elements * N; i++)
|
||||
{
|
||||
T r = test.ref(inputA[i], inputB[i], inputC[i]);
|
||||
if (r != output[i])
|
||||
{
|
||||
log_error("\n\nverification failed at index %d\n", i);
|
||||
log_error("-> inputs: %llu, %llu, %llu\n",
|
||||
static_cast<cl_uint>(inputA[i]),
|
||||
static_cast<cl_uint>(inputB[i]),
|
||||
static_cast<cl_uint>(inputC[i]));
|
||||
log_error("-> expected %llu, got %llu\n\n",
|
||||
static_cast<cl_uint>(r),
|
||||
static_cast<cl_uint>(output[i]));
|
||||
return TEST_FAIL;
|
||||
}
|
||||
}
|
||||
log_info("passed\n");
|
||||
}
|
||||
|
||||
return TEST_PASS;
|
||||
}
|
||||
|
||||
int test_intmath_int(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_uint, 1>(device, context, queue, num_elements,
|
||||
"uint");
|
||||
}
|
||||
|
||||
int test_intmath_int2(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_uint, 2>(device, context, queue, num_elements,
|
||||
"uint2");
|
||||
}
|
||||
|
||||
int test_intmath_int4(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_uint, 4>(device, context, queue, num_elements,
|
||||
"uint4");
|
||||
}
|
||||
|
||||
int test_intmath_long(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_ulong, 1>(device, context, queue, num_elements,
|
||||
"ulong");
|
||||
}
|
||||
|
||||
int test_intmath_long2(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_ulong, 2>(device, context, queue, num_elements,
|
||||
"ulong2");
|
||||
}
|
||||
|
||||
int test_intmath_long4(cl_device_id device, cl_context context,
|
||||
cl_command_queue queue, int num_elements)
|
||||
{
|
||||
return test_intmath<cl_ulong, 4>(device, context, queue, num_elements,
|
||||
"ulong4");
|
||||
}
|
||||
@@ -1,334 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *int_add_kernel_code =
|
||||
"__kernel void test_int_add(__global int *srcA, __global int *srcB, __global int *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_sub_kernel_code =
|
||||
"__kernel void test_int_sub(__global int *srcA, __global int *srcB, __global int *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mul_kernel_code =
|
||||
"__kernel void test_int_mul(__global int *srcA, __global int *srcB, __global int *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mad_kernel_code =
|
||||
"__kernel void test_int_mad(__global int *srcA, __global int *srcB, __global int *srcC, __global int *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
static const float MAX_ERR = 1e-5f;
|
||||
|
||||
int
|
||||
verify_int_add(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_ADD int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_ADD int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_sub(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_SUB int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_SUB int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mul(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MUL int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MUL int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mad(int *inptrA, int *inptrB, int *inptrC, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MAD int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MAD int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_int(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_int *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_int) * num_elements;
|
||||
|
||||
input_ptr[0] = (cl_int*)malloc(length);
|
||||
input_ptr[1] = (cl_int*)malloc(length);
|
||||
input_ptr[2] = (cl_int*)malloc(length);
|
||||
output_ptr = (cl_int*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
|
||||
free_mtdata(d); d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &int_add_kernel_code, "test_int_add");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &int_sub_kernel_code, "test_int_sub");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &int_mul_kernel_code, "test_int_mul");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &int_mad_kernel_code, "test_int_mad");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_int_add(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_int_sub(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_int_mul(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_int_mad(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,333 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *int_add2_kernel_code =
|
||||
"__kernel void test_int_add2(__global int2 *srcA, __global int2 *srcB, __global int2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_sub2_kernel_code =
|
||||
"__kernel void test_int_sub2(__global int2 *srcA, __global int2 *srcB, __global int2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mul2_kernel_code =
|
||||
"__kernel void test_int_mul2(__global int2 *srcA, __global int2 *srcB, __global int2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mad2_kernel_code =
|
||||
"__kernel void test_int_mad2(__global int2 *srcA, __global int2 *srcB, __global int2 *srcC, __global int2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
int
|
||||
verify_int_add2(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_ADD int2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_ADD int2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_sub2(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_SUB int2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_SUB int2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mul2(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MUL int2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MUL int2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mad2(int *inptrA, int *inptrB, int *inptrC, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MAD int2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MAD int2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_int2(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_int *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_int) * 2 * num_elements;
|
||||
|
||||
input_ptr[0] = (cl_int*)malloc(length);
|
||||
input_ptr[1] = (cl_int*)malloc(length);
|
||||
input_ptr[2] = (cl_int*)malloc(length);
|
||||
output_ptr = (cl_int*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements*2; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements*2; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements*2; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
|
||||
free_mtdata( d );
|
||||
d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &int_add2_kernel_code, "test_int_add2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &int_sub2_kernel_code, "test_int_sub2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &int_mul2_kernel_code, "test_int_mul2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &int_mad2_kernel_code, "test_int_mad2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_int_add2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_int_sub2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_int_mul2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_int_mad2(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,332 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *int_add4_kernel_code =
|
||||
"__kernel void test_int_add4(__global int4 *srcA, __global int4 *srcB, __global int4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_sub4_kernel_code =
|
||||
"__kernel void test_int_sub4(__global int4 *srcA, __global int4 *srcB, __global int4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mul4_kernel_code =
|
||||
"__kernel void test_int_mul4(__global int4 *srcA, __global int4 *srcB, __global int4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *int_mad4_kernel_code =
|
||||
"__kernel void test_int_mad4(__global int4 *srcA, __global int4 *srcB, __global int4 *srcC, __global int4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
int
|
||||
verify_int_add4(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_ADD int4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_ADD int4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_sub4(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_SUB int4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_SUB int4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mul4(int *inptrA, int *inptrB, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MUL int4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MUL int4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_int_mad4(int *inptrA, int *inptrB, int *inptrC, int *outptr, int n)
|
||||
{
|
||||
int r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("INT_MAD int4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("INT_MAD int4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_int4(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_int *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_int) * 4 * num_elements;
|
||||
|
||||
input_ptr[0] = (cl_int*)malloc(length);
|
||||
input_ptr[1] = (cl_int*)malloc(length);
|
||||
input_ptr[2] = (cl_int*)malloc(length);
|
||||
output_ptr = (cl_int*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements*4; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements*4; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements*4; i++)
|
||||
p[i] = (int)genrand_int32(d);
|
||||
|
||||
free_mtdata(d); d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &int_add4_kernel_code, "test_int_add4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &int_sub4_kernel_code, "test_int_sub4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &int_mul4_kernel_code, "test_int_mul4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &int_mad4_kernel_code, "test_int_mad4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_int_add4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_int_sub4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_int_mul4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_int_mad4(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,342 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *long_add_kernel_code =
|
||||
"__kernel void test_long_add(__global long *srcA, __global long *srcB, __global long *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_sub_kernel_code =
|
||||
"__kernel void test_long_sub(__global long *srcA, __global long *srcB, __global long *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mul_kernel_code =
|
||||
"__kernel void test_long_mul(__global long *srcA, __global long *srcB, __global long *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mad_kernel_code =
|
||||
"__kernel void test_long_mad(__global long *srcA, __global long *srcB, __global long *srcC, __global long *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
static const float MAX_ERR = 1e-5f;
|
||||
|
||||
int
|
||||
verify_long_add(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_ADD int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_ADD int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_sub(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_SUB int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_SUB int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mul(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MUL int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MUL int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mad(cl_long *inptrA, cl_long *inptrB, cl_long *inptrC, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MAD int test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MAD int test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_long(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_long *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
|
||||
if(! gHasLong )
|
||||
{
|
||||
log_info("64-bit integers are not supported by this device. Skipping test.\n");
|
||||
return CL_SUCCESS;
|
||||
}
|
||||
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_long) * num_elements;
|
||||
|
||||
input_ptr[0] = (cl_long*)malloc(length);
|
||||
input_ptr[1] = (cl_long*)malloc(length);
|
||||
input_ptr[2] = (cl_long*)malloc(length);
|
||||
output_ptr = (cl_long*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
|
||||
free_mtdata(d);
|
||||
d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &long_add_kernel_code, "test_long_add");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &long_sub_kernel_code, "test_long_sub");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &long_mul_kernel_code, "test_long_mul");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &long_mad_kernel_code, "test_long_mad");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_long_add(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_long_sub(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_long_mul(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_long_mad(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,340 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *long_add2_kernel_code =
|
||||
"__kernel void test_long_add2(__global long2 *srcA, __global long2 *srcB, __global long2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_sub2_kernel_code =
|
||||
"__kernel void test_long_sub2(__global long2 *srcA, __global long2 *srcB, __global long2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mul2_kernel_code =
|
||||
"__kernel void test_long_mul2(__global long2 *srcA, __global long2 *srcB, __global long2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mad2_kernel_code =
|
||||
"__kernel void test_long_mad2(__global long2 *srcA, __global long2 *srcB, __global long2 *srcC, __global long2 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
int
|
||||
verify_long_add2(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_ADD long2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_ADD long2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_sub2(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_SUB long2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_SUB long2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mul2(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MUL long2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MUL long2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mad2(cl_long *inptrA, cl_long *inptrB, cl_long *inptrC, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MAD long2 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MAD long2 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_long2(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_long *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
|
||||
if(! gHasLong)
|
||||
{
|
||||
log_info("64-bit integers are not supported in this device. Skipping test.\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_long) * 2* num_elements;
|
||||
|
||||
input_ptr[0] = (cl_long*)malloc(length);
|
||||
input_ptr[1] = (cl_long*)malloc(length);
|
||||
input_ptr[2] = (cl_long*)malloc(length);
|
||||
output_ptr = (cl_long*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements * 2; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements * 2; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements * 2; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
|
||||
free_mtdata(d);
|
||||
d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &long_add2_kernel_code, "test_long_add2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &long_sub2_kernel_code, "test_long_sub2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &long_mul2_kernel_code, "test_long_mul2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &long_mad2_kernel_code, "test_long_mad2");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_long_add2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_long_sub2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_long_mul2(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_long_mad2(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
@@ -1,340 +0,0 @@
|
||||
//
|
||||
// Copyright (c) 2017 The Khronos Group Inc.
|
||||
//
|
||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||
// you may not use this file except in compliance with the License.
|
||||
// You may obtain a copy of the License at
|
||||
//
|
||||
// http://www.apache.org/licenses/LICENSE-2.0
|
||||
//
|
||||
// Unless required by applicable law or agreed to in writing, software
|
||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
// See the License for the specific language governing permissions and
|
||||
// limitations under the License.
|
||||
//
|
||||
#include "harness/compat.h"
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/stat.h>
|
||||
|
||||
|
||||
#include "procs.h"
|
||||
|
||||
const char *long_add4_kernel_code =
|
||||
"__kernel void test_long_add4(__global long4 *srcA, __global long4 *srcB, __global long4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] + srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_sub4_kernel_code =
|
||||
"__kernel void test_long_sub4(__global long4 *srcA, __global long4 *srcB, __global long4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] - srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mul4_kernel_code =
|
||||
"__kernel void test_long_mul4(__global long4 *srcA, __global long4 *srcB, __global long4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid];\n"
|
||||
"}\n";
|
||||
|
||||
const char *long_mad4_kernel_code =
|
||||
"__kernel void test_long_mad4(__global long4 *srcA, __global long4 *srcB, __global long4 *srcC, __global long4 *dst)\n"
|
||||
"{\n"
|
||||
" int tid = get_global_id(0);\n"
|
||||
"\n"
|
||||
" dst[tid] = srcA[tid] * srcB[tid] + srcC[tid];\n"
|
||||
"}\n";
|
||||
|
||||
int
|
||||
verify_long_add4(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] + inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_ADD long4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_ADD long4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_sub4(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] - inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_SUB long4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_SUB long4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mul4(cl_long *inptrA, cl_long *inptrB, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MUL long4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MUL long4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
verify_long_mad4(cl_long *inptrA, cl_long *inptrB, cl_long *inptrC, cl_long *outptr, int n)
|
||||
{
|
||||
cl_long r;
|
||||
int i;
|
||||
|
||||
for (i=0; i<n; i++)
|
||||
{
|
||||
r = inptrA[i] * inptrB[i] + inptrC[i];
|
||||
if (r != outptr[i])
|
||||
{
|
||||
log_error("LONG_MAD long4 test failed\n");
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
log_info("LONG_MAD long4 test passed\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
int
|
||||
test_intmath_long4(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
|
||||
{
|
||||
cl_mem streams[4];
|
||||
cl_program program[4];
|
||||
cl_kernel kernel[4];
|
||||
|
||||
cl_long *input_ptr[3], *output_ptr, *p;
|
||||
size_t threads[1];
|
||||
int err, i;
|
||||
|
||||
if(! gHasLong )
|
||||
{
|
||||
log_info("64-bit integers are not supported by this device. Skipping test.\n");
|
||||
return 0;
|
||||
}
|
||||
|
||||
MTdata d = init_genrand( gRandomSeed );
|
||||
|
||||
size_t length = sizeof(cl_long) * 4 * num_elements;
|
||||
|
||||
input_ptr[0] = (cl_long*)malloc(length);
|
||||
input_ptr[1] = (cl_long*)malloc(length);
|
||||
input_ptr[2] = (cl_long*)malloc(length);
|
||||
output_ptr = (cl_long*)malloc(length);
|
||||
|
||||
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[0])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[1])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[2] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[2])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
streams[3] = clCreateBuffer(context, CL_MEM_READ_WRITE, length, NULL, NULL);
|
||||
if (!streams[3])
|
||||
{
|
||||
log_error("clCreateBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
p = input_ptr[0];
|
||||
for (i=0; i<num_elements * 4; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[1];
|
||||
for (i=0; i<num_elements * 4; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
p = input_ptr[2];
|
||||
for (i=0; i<num_elements * 4; i++)
|
||||
p[i] = (cl_long)genrand_int32(d) | ((cl_long) genrand_int32(d) << 32);
|
||||
|
||||
free_mtdata(d);
|
||||
d = NULL;
|
||||
|
||||
err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, 0, length, input_ptr[0], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[1], CL_TRUE, 0, length, input_ptr[1], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
err = clEnqueueWriteBuffer(queue, streams[2], CL_TRUE, 0, length, input_ptr[2], 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueWriteBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[0], &kernel[0], 1, &long_add4_kernel_code, "test_long_add4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[1], &kernel[1], 1, &long_sub4_kernel_code, "test_long_sub4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[2], &kernel[2], 1, &long_mul4_kernel_code, "test_long_mul4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = create_single_kernel_helper(context, &program[3], &kernel[3], 1, &long_mad4_kernel_code, "test_long_mad4");
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("create_single_kernel_helper failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[0], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[0], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[0], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[1], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[1], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[1], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[2], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[2], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[2], 2, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clSetKernelArg(kernel[3], 0, sizeof streams[0], &streams[0]);
|
||||
err |= clSetKernelArg(kernel[3], 1, sizeof streams[1], &streams[1]);
|
||||
err |= clSetKernelArg(kernel[3], 2, sizeof streams[2], &streams[2]);
|
||||
err |= clSetKernelArg(kernel[3], 3, sizeof streams[3], &streams[3]);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clSetKernelArgs failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
threads[0] = (unsigned int)num_elements;
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
err = clEnqueueNDRangeKernel(queue, kernel[i], 1, NULL, threads, NULL, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueNDRangeKernel failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
err = clEnqueueReadBuffer(queue, streams[3], CL_TRUE, 0, length, output_ptr, 0, NULL, NULL);
|
||||
if (err != CL_SUCCESS)
|
||||
{
|
||||
log_error("clEnqueueReadBuffer failed\n");
|
||||
return -1;
|
||||
}
|
||||
|
||||
switch (i)
|
||||
{
|
||||
case 0:
|
||||
err = verify_long_add4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 1:
|
||||
err = verify_long_sub4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 2:
|
||||
err = verify_long_mul4(input_ptr[0], input_ptr[1], output_ptr, num_elements);
|
||||
break;
|
||||
case 3:
|
||||
err = verify_long_mad4(input_ptr[0], input_ptr[1], input_ptr[2], output_ptr, num_elements);
|
||||
break;
|
||||
}
|
||||
if (err)
|
||||
break;
|
||||
}
|
||||
|
||||
// cleanup
|
||||
clReleaseMemObject(streams[0]);
|
||||
clReleaseMemObject(streams[1]);
|
||||
clReleaseMemObject(streams[2]);
|
||||
clReleaseMemObject(streams[3]);
|
||||
for (i=0; i<4; i++)
|
||||
{
|
||||
clReleaseKernel(kernel[i]);
|
||||
clReleaseProgram(program[i]);
|
||||
}
|
||||
free(input_ptr[0]);
|
||||
free(input_ptr[1]);
|
||||
free(input_ptr[2]);
|
||||
free(output_ptr);
|
||||
return err;
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user