ahmed/OpenCL-CTS
1
0
Fork 0
mirror of https://github.com/KhronosGroup/OpenCL-CTS.git synced 2026-03-19 06:09:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add fp16 testing to conversions and bruteforce (#1975)

Browse Source 
Merge the `fp16-staging` branch into `main`, adding fp16 (`half`)
testing to the conversions and math bruteforce tests.

---------

Signed-off-by: Ahmed Hesham <ahmed.hesham@arm.com>
Signed-off-by: Sven van Haastregt <sven.vanhaastregt@arm.com>
Signed-off-by: Guo, Yilong <yilong.guo@intel.com>
Signed-off-by: John Kesapides <john.kesapides@arm.com>
Co-authored-by: Marcin Hajder <marcin.hajder@gmail.com>
Co-authored-by: Ewan Crawford <ewan@codeplay.com>
Co-authored-by: Wawiorko, Grzegorz <grzegorz.wawiorko@intel.com>
Co-authored-by: Sreelakshmi Haridas Maruthur <sharidas@quicinc.com>
Co-authored-by: Harald van Dijk <harald@gigawatt.nl>
Co-authored-by: Ben Ashbaugh <ben.ashbaugh@intel.com>
Co-authored-by: Haonan Yang <haonan.yang@intel.com>
Co-authored-by: Ahmed Hesham <117350656+ahesham-arm@users.noreply.github.com>
Co-authored-by: niranjanjoshi121 <43807392+niranjanjoshi121@users.noreply.github.com>
Co-authored-by: Wenwan Xing <wenwan.xing@intel.com>
Co-authored-by: Yilong Guo <yilong.guo@intel.com>
Co-authored-by: Romaric Jodin <89833130+rjodinchr@users.noreply.github.com>
Co-authored-by: joshqti <127994991+joshqti@users.noreply.github.com>
Co-authored-by: Pekka Jääskeläinen <pekka.jaaskelainen@tuni.fi>
Co-authored-by: imilenkovic00 <155085410+imilenkovic00@users.noreply.github.com>
Co-authored-by: John Kesapides <46718829+JohnKesapidesARM@users.noreply.github.com>
Co-authored-by: Aharon Abramson <aharon.abramson@mobileye.com>
This commit is contained in:
Sven van Haastregt
2024-06-18 18:43:11 +02:00
committed by GitHub
parent b3c89ebde0
commit b6941b6c61
30 changed files with 7149 additions and 350 deletions
Download Patch File Download Diff File Expand all files Collapse all files

201
test_conformance/math_brute_force/mad_half.cpp Normal file
View File

@@ -0,0 +1,201 @@
//
// Copyright (c) 2017-2024 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 "common.h"
#include "function_list.h"
#include "test_functions.h"
#include "utility.h"
#include <cstring>
namespace {
cl_int BuildKernel_HalfFn(cl_uint job_id, cl_uint thread_id UNUSED, void *p)
{
BuildKernelInfo &info = *(BuildKernelInfo *)p;
auto generator = [](const std::string &kernel_name, const char *builtin,
cl_uint vector_size_index) {
return GetTernaryKernel(kernel_name, builtin, ParameterType::Half,
ParameterType::Half, ParameterType::Half,
ParameterType::Half, vector_size_index);
};
return BuildKernels(info, job_id, generator);
}
} // anonymous namespace
int TestFunc_mad_Half(const Func *f, MTdata d, bool relaxedMode)
{
int error;
Programs programs;
KernelMatrix kernels;
const unsigned thread_id = 0; // Test is currently not multithreaded.
float maxError = 0.0f;
float maxErrorVal = 0.0f;
float maxErrorVal2 = 0.0f;
float maxErrorVal3 = 0.0f;
size_t bufferSize = BUFFER_SIZE;
logFunctionInfo(f->name, sizeof(cl_half), relaxedMode);
uint64_t step = getTestStep(sizeof(cl_half), bufferSize);
// Init the kernels
{
BuildKernelInfo build_info = { 1, kernels, programs, f->nameInCode };
if ((error = ThreadPool_Do(BuildKernel_HalfFn,
gMaxVectorSizeIndex - gMinVectorSizeIndex,
&build_info)))
return error;
}
for (uint64_t i = 0; i < (1ULL << 32); i += step)
{
// Init input array
cl_ushort *p = (cl_ushort *)gIn;
cl_ushort *p2 = (cl_ushort *)gIn2;
cl_ushort *p3 = (cl_ushort *)gIn3;
for (size_t j = 0; j < bufferSize / sizeof(cl_ushort); j++)
{
p[j] = (cl_ushort)genrand_int32(d);
p2[j] = (cl_ushort)genrand_int32(d);
p3[j] = (cl_ushort)genrand_int32(d);
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer2, CL_FALSE, 0,
bufferSize, gIn2, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2 ***\n", error);
return error;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer3, CL_FALSE, 0,
bufferSize, gIn3, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer3 ***\n", error);
return error;
}
// write garbage into output arrays
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xacdcacdc;
if (gHostFill)
{
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer[j],
CL_FALSE, 0, BUFFER_SIZE,
gOut[j], 0, NULL, NULL)))
{
vlog_error(
"\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
return error;
}
}
else
{
error = clEnqueueFillBuffer(gQueue, gOutBuffer[j], &pattern,
sizeof(pattern), 0, BUFFER_SIZE, 0,
NULL, NULL);
test_error(error, "clEnqueueFillBuffer failed!\n");
}
}
// Run the kernels
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_half) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j][thread_id], 0,
sizeof(gOutBuffer[j]), &gOutBuffer[j])))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clSetKernelArg(kernels[j][thread_id], 1,
sizeof(gInBuffer), &gInBuffer)))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clSetKernelArg(kernels[j][thread_id], 2,
sizeof(gInBuffer2), &gInBuffer2)))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clSetKernelArg(kernels[j][thread_id], 3,
sizeof(gInBuffer3), &gInBuffer3)))
{
LogBuildError(programs[j]);
return error;
}
if ((error = clEnqueueNDRangeKernel(gQueue, kernels[j][thread_id],
1, NULL, &localCount, NULL, 0,
NULL, NULL)))
{
vlog_error("FAILED -- could not execute kernel\n");
return error;
}
}
// Get that moving
if ((error = clFlush(gQueue))) vlog("clFlush failed\n");
// Read the data back
for (auto j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
return error;
}
}
if (gSkipCorrectnessTesting) break;
// Verify data - no verification possible. MAD is a random number
// generator.
if (0 == (i & 0x0fffffff))
{
vlog(".");
fflush(stdout);
}
}
if (!gSkipCorrectnessTesting)
{
if (gWimpyMode)
vlog("Wimp pass");
else
vlog("pass");
vlog("\t%8.2f @ {%a, %a, %a}", maxError, maxErrorVal, maxErrorVal2,
maxErrorVal3);
}
vlog("\n");
return error;
}