ahmed/OpenCL-CTS
1
0
Fork 0
mirror of https://github.com/KhronosGroup/OpenCL-CTS.git synced 2026-03-22 15:19:02 +00:00
Code Packages Projects Releases Wiki Activity

Reduce difference between files in math_brute_force (#1138)

Browse Source 
* Reduce differences between files

This will help reduce code duplication is future commits.

Some code is moved around, some variables are renamed and some
statements are slightly altered to reduce differences between files in
math_brute_force, yet the semantics remains the same.

The differences were identified using n-way diffs. Many differences
remain however.

Signed-off-by: Marco Antognini <marco.antognini@arm.com>

* Workaround clang-format limitation

Introduces some insignificant spaces to force clang-format to reduce the
indentation and reduce differences between files.

Signed-off-by: Marco Antognini <marco.antognini@arm.com>
This commit is contained in:
Marco Antognini
2021-02-10 10:38:31 +00:00
committed by GitHub
parent f6b501352d
commit 8ad1088af9
13 changed files with 483 additions and 423 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
test_conformance/math_brute_force/unary.cpp
View File

@@ -37,13 +37,14 @@ static int BuildKernel(const char *name, int vectorSize, cl_uint kernel_count,
sizeNames[vectorSize],
"* out, __global float",
sizeNames[vectorSize],
"* in)\n"
"* in )\n"
"{\n"
" int i = get_global_id(0);\n"
" out[i] = ",
name,
"( in[i] );\n"
"}\n" };
const char *c3[] = {
"__kernel void math_kernel",
sizeNames[vectorSize],
@@ -89,7 +90,6 @@ static int BuildKernel(const char *name, int vectorSize, cl_uint kernel_count,
"}\n"
};
const char **kern = c;
size_t kernSize = sizeof(c) / sizeof(c[0]);
@@ -118,7 +118,7 @@ static int BuildKernelDouble(const char *name, int vectorSize,
sizeNames[vectorSize],
"* out, __global double",
sizeNames[vectorSize],
"* in)\n"
"* in )\n"
"{\n"
" int i = get_global_id(0);\n"
" out[i] = ",
@@ -181,7 +181,6 @@ static int BuildKernelDouble(const char *name, int vectorSize,
kernSize = sizeof(c3) / sizeof(c3[0]);
}
char testName[32];
snprintf(testName, sizeof(testName) - 1, "math_kernel%s",
sizeNames[vectorSize]);
@@ -249,7 +248,7 @@ typedef struct TestInfo
int isRangeLimited; // 1 if the function is only to be evaluated over a
// range
float half_sin_cos_tan_limit;
bool relaxedMode; // True if test is to be run in relaxed mode, false
bool relaxedMode; // True if test is running in relaxed mode, false
// otherwise.
} TestInfo;
@@ -269,10 +268,10 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
// Init test_info
memset(&test_info, 0, sizeof(test_info));
test_info.threadCount = GetThreadCount();
test_info.subBufferSize = BUFFER_SIZE
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
@@ -345,8 +344,8 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
&region, &error);
if (error || NULL == test_info.tinfo[i].outBuf[j])
{
vlog_error("Error: Unable to create sub-buffer of gInBuffer "
"for region {%zd, %zd}\n",
vlog_error("Error: Unable to create sub-buffer of "
"gInBuffer for region {%zd, %zd}\n",
region.origin, region.size);
goto exit;
}
@@ -390,6 +389,7 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
goto exit;
}
// Run the kernels
if (!gSkipCorrectnessTesting || skipTestingRelaxed)
{
error = ThreadPool_Do(TestFloat, test_info.jobCount, &test_info);
@@ -443,8 +443,9 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
// Run the kernels
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
size_t vectorSize = sizeof(cl_float) * sizeValues[j];
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(test_info.k[j][0], 0,
sizeof(gOutBuffer[j]), &gOutBuffer[j])))
{
@@ -479,9 +480,9 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
}
uint64_t endTime = GetTime();
double current_time = SubtractTime(endTime, startTime);
sum += current_time;
if (current_time < bestTime) bestTime = current_time;
double time = SubtractTime(endTime, startTime);
sum += time;
if (time < bestTime) bestTime = time;
}
if (gReportAverageTimes) bestTime = sum / PERF_LOOP_COUNT;
@@ -497,6 +498,7 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
vlog("\n");
exit:
// Release
for (i = gMinVectorSizeIndex; i < gMaxVectorSizeIndex; i++)
{
clReleaseProgram(test_info.programs[i]);
@@ -553,7 +555,7 @@ static cl_int TestFloat(cl_uint job_id, cl_uint thread_id, void *data)
cl_uint *out[VECTOR_SIZE_COUNT];
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
out[j] = (uint32_t *)clEnqueueMapBuffer(
out[j] = (cl_uint *)clEnqueueMapBuffer(
tinfo->tQueue, tinfo->outBuf[j], CL_FALSE, CL_MAP_WRITE, 0,
buffer_size, 0, NULL, e + j, &error);
if (error || NULL == out[j])
@@ -627,7 +629,7 @@ static cl_int TestFloat(cl_uint job_id, cl_uint thread_id, void *data)
return error;
}
// run the kernel
// Run the kernel
size_t vectorCount =
(buffer_elements + sizeValues[j] - 1) / sizeValues[j];
cl_kernel kernel = job->k[j][thread_id]; // each worker thread has its
@@ -655,7 +657,6 @@ static cl_int TestFloat(cl_uint job_id, cl_uint thread_id, void *data)
}
}
// Get that moving
if ((error = clFlush(tinfo->tQueue))) vlog("clFlush 2 failed\n");
@@ -670,7 +671,7 @@ static cl_int TestFloat(cl_uint job_id, cl_uint thread_id, void *data)
// an in order queue.
for (j = gMinVectorSizeIndex; j + 1 < gMaxVectorSizeIndex; j++)
{
out[j] = (uint32_t *)clEnqueueMapBuffer(
out[j] = (cl_uint *)clEnqueueMapBuffer(
tinfo->tQueue, tinfo->outBuf[j], CL_FALSE, CL_MAP_READ, 0,
buffer_size, 0, NULL, NULL, &error);
if (error || NULL == out[j])
@@ -680,6 +681,7 @@ static cl_int TestFloat(cl_uint job_id, cl_uint thread_id, void *data)
return error;
}
}
// Wait for the last buffer
out[j] = (uint32_t *)clEnqueueMapBuffer(tinfo->tQueue, tinfo->outBuf[j],
CL_TRUE, CL_MAP_READ, 0,
@@ -1246,12 +1248,9 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
/* Qualcomm fix: 9461 read-write flags must be compatible with
* parent buffer */
test_info.tinfo[i].outBuf[j] = clCreateSubBuffer(
gOutBuffer[j], CL_MEM_WRITE_ONLY, CL_BUFFER_CREATE_TYPE_REGION,
&region, &error);
/* Qualcomm fix: end */
if (error || NULL == test_info.tinfo[i].outBuf[j])
{
vlog_error("Error: Unable to create sub-buffer of gInBuffer "
@@ -1281,6 +1280,7 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
goto exit;
}
// Run the kernels
if (!gSkipCorrectnessTesting)
{
error = ThreadPool_Do(TestDouble, test_info.jobCount, &test_info);
@@ -1334,8 +1334,9 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
// Run the kernels
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_double);
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
size_t vectorSize = sizeof(cl_double) * sizeValues[j];
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(test_info.k[j][0], 0,
sizeof(gOutBuffer[j]), &gOutBuffer[j])))
{
@@ -1370,9 +1371,9 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
}
uint64_t endTime = GetTime();
double current_time = SubtractTime(endTime, startTime);
sum += current_time;
if (current_time < bestTime) bestTime = current_time;
double time = SubtractTime(endTime, startTime);
sum += time;
if (time < bestTime) bestTime = time;
}
if (gReportAverageTimes) bestTime = sum / PERF_LOOP_COUNT;
@@ -1393,6 +1394,7 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
vlog("\n");
exit:
// Release
for (i = gMinVectorSizeIndex; i < gMaxVectorSizeIndex; i++)
{
clReleaseProgram(test_info.programs[i]);