mirror of
https://github.com/KhronosGroup/OpenCL-CTS.git
synced 2026-03-19 06:09:01 +00:00
5b7af4b36d
The code base uses a mix of 'device' and 'deviceID'. I suggest we standardise on 'device' which is shorter and slightly more prevalent. Contributes to #2181 Signed-off-by: Kevin Petit <kevin.petit@arm.com>
198 lines
8.5 KiB
C++
198 lines
8.5 KiB
C++
//
|
|
// Copyright (c) 2016-2023 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 "testBase.h"
|
|
#include "types.hpp"
|
|
|
|
#include <sstream>
|
|
#include <string>
|
|
|
|
template<typename T>
|
|
int test_fmath(cl_device_id deviceID,
|
|
cl_context context,
|
|
cl_command_queue queue,
|
|
const char *spvName,
|
|
const char *funcName,
|
|
const char *Tname,
|
|
bool fast_math,
|
|
std::vector<T> &h_lhs,
|
|
std::vector<T> &h_rhs)
|
|
{
|
|
|
|
if(std::string(Tname).find("double") != std::string::npos) {
|
|
if(!is_extension_available(deviceID, "cl_khr_fp64")) {
|
|
log_info("Extension cl_khr_fp64 not supported; skipping double tests.\n");
|
|
return 0;
|
|
}
|
|
}
|
|
cl_int err = CL_SUCCESS;
|
|
int num = (int)h_lhs.size();
|
|
size_t bytes = num * sizeof(T);
|
|
|
|
clMemWrapper lhs = clCreateBuffer(context, CL_MEM_READ_ONLY, bytes, NULL, &err);
|
|
SPIRV_CHECK_ERROR(err, "Failed to create lhs buffer");
|
|
|
|
err = clEnqueueWriteBuffer(queue, lhs, CL_TRUE, 0, bytes, &h_lhs[0], 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to copy to lhs buffer");
|
|
|
|
clMemWrapper rhs = clCreateBuffer(context, CL_MEM_READ_ONLY, bytes, NULL, &err);
|
|
SPIRV_CHECK_ERROR(err, "Failed to create rhs buffer");
|
|
|
|
err = clEnqueueWriteBuffer(queue, rhs, CL_TRUE, 0, bytes, &h_rhs[0], 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to copy to rhs buffer");
|
|
|
|
std::string kernelStr;
|
|
|
|
{
|
|
std::stringstream kernelStream;
|
|
|
|
if (is_double<T>::value) {
|
|
kernelStream << "#pragma OPENCL EXTENSION cl_khr_fp64 : enable\n";
|
|
} else if (sizeof(T) == sizeof(cl_half)) {
|
|
kernelStream << "#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n";
|
|
}
|
|
|
|
kernelStream << "#define spirv_fadd(a, b) (a) + (b) \n";
|
|
kernelStream << "#define spirv_fsub(a, b) (a) - (b) \n";
|
|
kernelStream << "#define spirv_fmul(a, b) (a) * (b) \n";
|
|
kernelStream << "#define spirv_fdiv(a, b) (a) / (b) \n";
|
|
kernelStream << "#define spirv_frem(a, b) fmod(a, b) \n";
|
|
kernelStream << "#define spirv_fmod(a, b) copysign(fmod(a,b),b) \n";
|
|
kernelStream << "#define T " << Tname << "\n";
|
|
kernelStream << "#define FUNC spirv_" << funcName << "\n";
|
|
kernelStream << "__kernel void fmath_cl(__global T *out, \n";
|
|
kernelStream << "const __global T *lhs, const __global T *rhs) \n";
|
|
kernelStream << "{ \n";
|
|
kernelStream << " int id = get_global_id(0); \n";
|
|
kernelStream << " out[id] = FUNC(lhs[id], rhs[id]); \n";
|
|
kernelStream << "} \n";
|
|
kernelStr = kernelStream.str();
|
|
}
|
|
|
|
const char *kernelBuf = kernelStr.c_str();
|
|
|
|
std::vector<T> h_ref(num);
|
|
|
|
{
|
|
// Run the cl kernel for reference results
|
|
clProgramWrapper prog;
|
|
clKernelWrapper kernel;
|
|
err = create_single_kernel_helper(context, &prog, &kernel, 1,
|
|
&kernelBuf, "fmath_cl");
|
|
SPIRV_CHECK_ERROR(err, "Failed to create cl kernel");
|
|
|
|
clMemWrapper ref = clCreateBuffer(context, CL_MEM_READ_WRITE, bytes, NULL, &err);
|
|
SPIRV_CHECK_ERROR(err, "Failed to create ref buffer");
|
|
|
|
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &ref);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 0");
|
|
|
|
err = clSetKernelArg(kernel, 1, sizeof(cl_mem), &lhs);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 1");
|
|
|
|
err = clSetKernelArg(kernel, 2, sizeof(cl_mem), &rhs);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 2");
|
|
|
|
size_t global = num;
|
|
err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global, NULL, 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to enqueue cl kernel");
|
|
|
|
err = clEnqueueReadBuffer(queue, ref, CL_TRUE, 0, bytes, &h_ref[0], 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to read from ref");
|
|
}
|
|
|
|
clProgramWrapper prog;
|
|
err = get_program_with_il(prog, deviceID, context, spvName);
|
|
SPIRV_CHECK_ERROR(err, "Failed to build program");
|
|
|
|
clKernelWrapper kernel = clCreateKernel(prog, "fmath_spv", &err);
|
|
SPIRV_CHECK_ERROR(err, "Failed to create spv kernel");
|
|
|
|
clMemWrapper res = clCreateBuffer(context, CL_MEM_READ_WRITE, bytes, NULL, &err);
|
|
SPIRV_CHECK_ERROR(err, "Failed to create res buffer");
|
|
|
|
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &res);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 0");
|
|
|
|
err = clSetKernelArg(kernel, 1, sizeof(cl_mem), &lhs);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 1");
|
|
|
|
err = clSetKernelArg(kernel, 2, sizeof(cl_mem), &rhs);
|
|
SPIRV_CHECK_ERROR(err, "Failed to set arg 2");
|
|
|
|
size_t global = num;
|
|
err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global, NULL, 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to enqueue cl kernel");
|
|
|
|
std::vector<T> h_res(num);
|
|
err = clEnqueueReadBuffer(queue, res, CL_TRUE, 0, bytes, &h_res[0], 0, NULL, NULL);
|
|
SPIRV_CHECK_ERROR(err, "Failed to read from ref");
|
|
|
|
for (int i = 0; i < num; i++) {
|
|
if (h_res[i] != h_ref[i]) {
|
|
log_error("Values do not match at location %d\n", i);
|
|
return -1;
|
|
}
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
#define TEST_FMATH_FUNC(TYPE, FUNC, MODE) \
|
|
REGISTER_TEST(op_##FUNC##_##TYPE##_##MODE) \
|
|
{ \
|
|
if (sizeof(cl_##TYPE) == 2) \
|
|
{ \
|
|
PASSIVE_REQUIRE_FP16_SUPPORT(device); \
|
|
} \
|
|
const int num = 1 << 20; \
|
|
std::vector<cl_##TYPE> lhs(num); \
|
|
std::vector<cl_##TYPE> rhs(num); \
|
|
\
|
|
RandomSeed seed(gRandomSeed); \
|
|
\
|
|
for (int i = 0; i < num; i++) \
|
|
{ \
|
|
lhs[i] = genrandReal<cl_##TYPE>(seed); \
|
|
rhs[i] = genrandReal<cl_##TYPE>(seed); \
|
|
} \
|
|
\
|
|
const char *mode = #MODE; \
|
|
return test_fmath(device, context, queue, #FUNC "_" #TYPE, #FUNC, \
|
|
#TYPE, mode[0] == 'f', lhs, rhs); \
|
|
}
|
|
|
|
#define TEST_FMATH_MODE(TYPE, MODE) \
|
|
TEST_FMATH_FUNC(TYPE, fadd, MODE) \
|
|
TEST_FMATH_FUNC(TYPE, fsub, MODE) \
|
|
TEST_FMATH_FUNC(TYPE, fmul, MODE) \
|
|
TEST_FMATH_FUNC(TYPE, fdiv, MODE) \
|
|
// disable those tests until we figure out what the precision requirements
|
|
// are
|
|
// TEST_FMATH_FUNC(TYPE, frem, MODE)
|
|
// TEST_FMATH_FUNC(TYPE, fmod, MODE)
|
|
|
|
#define TEST_FMATH_TYPE(TYPE) \
|
|
TEST_FMATH_MODE(TYPE, regular) \
|
|
TEST_FMATH_MODE(TYPE, fast) \
|
|
|
|
TEST_FMATH_TYPE(float)
|
|
TEST_FMATH_TYPE(double)
|
|
|
|
TEST_FMATH_TYPE(float4)
|
|
TEST_FMATH_TYPE(double2)
|
|
|
|
TEST_FMATH_TYPE(half)
|