Fold constant global variable (#1217)

gWimpyBufferSize is never modified and is actually not used to modify
the number of tests -- gWimpyReductionFactor is used for that purpose by
some tests, but not all.

This patch removes this unnecessary global variable to simplify the
codebase, and reduce differences between tests.

Signed-off-by: Marco Antognini <marco.antognini@arm.com>
This commit is contained in:
Marco Antognini
2021-04-21 10:45:34 +01:00
committed by GitHub
parent 7542ae2443
commit b8d1ea9962
28 changed files with 210 additions and 306 deletions

View File

@@ -300,13 +300,6 @@ int TestFunc_Double_Double_Double(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -291,13 +291,6 @@ int TestFunc_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -302,13 +302,6 @@ int TestFunc_Double_Double_Int(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -294,13 +294,6 @@ int TestFunc_Float_Float_Int(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -298,13 +298,6 @@ int TestFunc_Double_Double_Double_Operator(const Func *f, MTdata d,
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -288,13 +288,6 @@ int TestFunc_Float_Float_Float_Operator(const Func *f, MTdata d,
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -178,8 +178,7 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
int ftz = f->ftz || gForceFTZ;
double maxErrorVal = 0.0f;
double maxErrorVal2 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(double), bufferSize);
uint64_t step = getTestStep(sizeof(double), BUFFER_SIZE);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -204,21 +203,21 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
// Init input array
double *p = (double *)gIn;
double *p2 = (double *)gIn2;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
p[j] = DoubleFromUInt32(genrand_int32(d));
p2[j] = DoubleFromUInt32(genrand_int32(d));
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn2, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2 ***\n", error);
return error;
@@ -228,20 +227,20 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -253,8 +252,8 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_double) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -304,7 +303,7 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
cri.r = (double *)gOut_Ref;
cri.i = (int *)gOut_Ref2;
cri.f_ffpI = f->dfunc.f_ffpI;
cri.lim = bufferSize / sizeof(double);
cri.lim = BUFFER_SIZE / sizeof(double);
cri.count = (cri.lim + threadCount - 1) / threadCount;
ThreadPool_Do(ReferenceD, threadCount, &cri);
}
@@ -312,7 +311,7 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
{
double *r = (double *)gOut_Ref;
int *r2 = (int *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
r[j] = (double)f->dfunc.f_ffpI(s[j], s2[j], r2 + j);
}
@@ -321,14 +320,14 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -340,7 +339,7 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint64_t *t = (uint64_t *)gOut_Ref;
int32_t *t2 = (int32_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -549,7 +548,7 @@ int TestFunc_DoubleI_Double_Double(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -177,8 +177,7 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
int64_t maxError2 = 0;
float maxErrorVal = 0.0f;
float maxErrorVal2 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
cl_uint threadCount = GetThreadCount();
@@ -205,21 +204,21 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
// Init input array
cl_uint *p = (cl_uint *)gIn;
cl_uint *p2 = (cl_uint *)gIn2;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
p[j] = genrand_int32(d);
p2[j] = genrand_int32(d);
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn2, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2 ***\n", error);
return error;
@@ -229,20 +228,20 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -254,8 +253,8 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_float) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -305,7 +304,7 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
cri.r = (float *)gOut_Ref;
cri.i = (int *)gOut_Ref2;
cri.f_ffpI = f->func.f_ffpI;
cri.lim = bufferSize / sizeof(float);
cri.lim = BUFFER_SIZE / sizeof(float);
cri.count = (cri.lim + threadCount - 1) / threadCount;
ThreadPool_Do(ReferenceF, threadCount, &cri);
}
@@ -313,7 +312,7 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
{
float *r = (float *)gOut_Ref;
int *r2 = (int *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = (float)f->func.f_ffpI(s[j], s2[j], r2 + j);
}
@@ -322,14 +321,14 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -341,7 +340,7 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
int32_t *t2 = (int32_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -534,7 +533,7 @@ int TestFunc_FloatI_Float_Float(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -125,9 +125,9 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
cl_program programs[VECTOR_SIZE_COUNT];
cl_kernel kernels[VECTOR_SIZE_COUNT];
int ftz = f->ftz || gForceFTZ;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(cl_double), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(cl_double)) + 1);
uint64_t step = getTestStep(sizeof(cl_double), BUFFER_SIZE);
int scale =
(int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(cl_double)) + 1);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -154,17 +154,17 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
double *p = (double *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j * scale);
}
else
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j);
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -174,10 +174,10 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -188,9 +188,9 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
// Run the kernels
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_double) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t vectorSize = sizeValues[j] * sizeof(cl_double);
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -219,7 +219,7 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
// Calculate the correctly rounded reference result
int *r = (int *)gOut_Ref;
double *s = (double *)gIn;
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
r[j] = f->dfunc.i_f(s[j]);
// Read the data back
@@ -227,7 +227,7 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -238,7 +238,7 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -270,7 +270,7 @@ int TestFunc_Int_Double(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -123,9 +123,8 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
cl_program programs[VECTOR_SIZE_COUNT];
cl_kernel kernels[VECTOR_SIZE_COUNT];
int ftz = f->ftz || gForceFTZ || 0 == (CL_FP_DENORM & gFloatCapabilities);
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(float)) + 1);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
int scale = (int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(float)) + 1);
logFunctionInfo(f->name, sizeof(cl_float), relaxedMode);
@@ -152,17 +151,17 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
cl_uint *p = (cl_uint *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (cl_uint)i + j * scale;
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -172,10 +171,10 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -187,8 +186,8 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -217,7 +216,7 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
// Calculate the correctly rounded reference result
int *r = (int *)gOut_Ref;
float *s = (float *)gIn;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = f->func.i_f(s[j]);
// Read the data back
@@ -225,7 +224,7 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -236,7 +235,7 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -268,7 +267,7 @@ int TestFunc_Int_Float(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -284,13 +284,6 @@ int TestMacro_Int_Double_Double(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -274,13 +274,6 @@ int TestMacro_Int_Float_Float(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -162,13 +162,6 @@ int TestMacro_Int_Double(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -161,13 +161,6 @@ int TestMacro_Int_Float(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -141,8 +141,7 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
double maxErrorVal = 0.0f;
double maxErrorVal2 = 0.0f;
double maxErrorVal3 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(double), bufferSize);
uint64_t step = getTestStep(sizeof(double), BUFFER_SIZE);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -162,7 +161,7 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
double *p = (double *)gIn;
double *p2 = (double *)gIn2;
double *p3 = (double *)gIn3;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
p[j] = DoubleFromUInt32(genrand_int32(d));
p2[j] = DoubleFromUInt32(genrand_int32(d));
@@ -170,21 +169,21 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn3, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer3 ***\n", error);
return error;
@@ -194,10 +193,10 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -209,8 +208,8 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_double) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -253,7 +252,7 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
double *s = (double *)gIn;
double *s2 = (double *)gIn2;
double *s3 = (double *)gIn3;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
r[j] = (double)f->dfunc.f_fff(s[j], s2[j], s3[j]);
// Read the data back
@@ -261,7 +260,7 @@ int TestFunc_mad_Double(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;

View File

@@ -142,8 +142,7 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
float maxErrorVal = 0.0f;
float maxErrorVal2 = 0.0f;
float maxErrorVal3 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
// Init the kernels
{
@@ -161,7 +160,7 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
cl_uint *p = (cl_uint *)gIn;
cl_uint *p2 = (cl_uint *)gIn2;
cl_uint *p3 = (cl_uint *)gIn3;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
p[j] = genrand_int32(d);
p2[j] = genrand_int32(d);
@@ -169,21 +168,21 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn3, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer3 ***\n", error);
return error;
@@ -193,10 +192,10 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -208,8 +207,8 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_float) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -252,7 +251,7 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
float *s = (float *)gIn;
float *s2 = (float *)gIn2;
float *s3 = (float *)gIn3;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = (float)f->func.f_fff(s[j], s2[j], s3[j]);
// Read the data back
@@ -260,7 +259,7 @@ int TestFunc_mad_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;

View File

@@ -100,7 +100,6 @@ cl_mem gOutBuffer2[VECTOR_SIZE_COUNT] = { NULL, NULL, NULL, NULL, NULL, NULL };
static MTdata gMTdata;
cl_device_fp_config gFloatCapabilities = 0;
int gWimpyReductionFactor = 32;
int gWimpyBufferSize = BUFFER_SIZE;
int gVerboseBruteForce = 0;
static int ParseArgs(int argc, const char **argv);

View File

@@ -218,8 +218,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
double maxErrorVal = 0.0f;
double maxErrorVal2 = 0.0f;
double maxErrorVal3 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(double), bufferSize);
uint64_t step = getTestStep(sizeof(double), BUFFER_SIZE);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -246,7 +245,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
{ // test edge cases
uint32_t x, y, z;
x = y = z = 0;
for (; j < bufferSize / sizeof(double); j++)
for (; j < BUFFER_SIZE / sizeof(double); j++)
{
p[j] = specialValues[x];
p2[j] = specialValues[y];
@@ -261,11 +260,11 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
}
}
}
if (j == bufferSize / sizeof(double))
if (j == BUFFER_SIZE / sizeof(double))
vlog_error("Test Error: not all special cases tested!\n");
}
for (; j < bufferSize / sizeof(double); j++)
for (; j < BUFFER_SIZE / sizeof(double); j++)
{
p[j] = DoubleFromUInt32(genrand_int32(d));
p2[j] = DoubleFromUInt32(genrand_int32(d));
@@ -273,21 +272,21 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn3, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer3 ***\n", error);
return error;
@@ -297,10 +296,10 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -312,8 +311,8 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_double) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -356,7 +355,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
double *s = (double *)gIn;
double *s2 = (double *)gIn2;
double *s3 = (double *)gIn3;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
r[j] = (double)f->dfunc.f_fff(s[j], s2[j], s3[j]);
// Read the data back
@@ -364,7 +363,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -375,7 +374,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
// Verify data
uint64_t *t = (uint64_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -707,7 +706,7 @@ int TestFunc_Double_Double_Double_Double(const Func *f, MTdata d,
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -228,8 +228,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
float maxErrorVal = 0.0f;
float maxErrorVal2 = 0.0f;
float maxErrorVal3 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
cl_uchar overflow[BUFFER_SIZE / sizeof(float)];
@@ -265,7 +264,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
float *fp3 = (float *)gIn3;
uint32_t x, y, z;
x = y = z = 0;
for (; j < bufferSize / sizeof(float); j++)
for (; j < BUFFER_SIZE / sizeof(float); j++)
{
fp[j] = specialValues[x];
fp2[j] = specialValues[y];
@@ -281,11 +280,11 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
}
}
}
if (j == bufferSize / sizeof(float))
if (j == BUFFER_SIZE / sizeof(float))
vlog_error("Test Error: not all special cases tested!\n");
}
for (; j < bufferSize / sizeof(float); j++)
for (; j < BUFFER_SIZE / sizeof(float); j++)
{
p[j] = genrand_int32(d);
p2[j] = genrand_int32(d);
@@ -293,21 +292,21 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, 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)))
BUFFER_SIZE, gIn3, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer3 ***\n", error);
return error;
@@ -317,10 +316,10 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -332,8 +331,8 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeof(cl_float) * sizeValues[j];
size_t localCount = (bufferSize + vectorSize - 1)
/ vectorSize; // bufferSize / vectorSize rounded up
size_t localCount = (BUFFER_SIZE + vectorSize - 1)
/ vectorSize; // BUFFER_SIZE / vectorSize rounded up
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -378,7 +377,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
float *s3 = (float *)gIn3;
if (skipNanInf)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
feclearexcept(FE_OVERFLOW);
r[j] =
@@ -389,7 +388,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] =
(float)f->func.f_fma(s[j], s2[j], s3[j], CORRECTLY_ROUNDED);
}
@@ -399,7 +398,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -410,7 +409,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -842,7 +841,7 @@ int TestFunc_Float_Float_Float_Float(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10u bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -170,12 +170,6 @@ int TestFunc_Double_Double(const Func *f, MTdata d, bool relaxedMode)
test_info.subBufferSize = BUFFER_SIZE
/ (sizeof(cl_double) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_double));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_double)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)

View File

@@ -171,13 +171,6 @@ int TestFunc_Float_Float(const Func *f, MTdata d, bool relaxedMode)
/ (sizeof(cl_float) * RoundUpToNextPowerOfTwo(test_info.threadCount));
test_info.scale = getTestScale(sizeof(cl_float));
if (gWimpyMode)
{
test_info.subBufferSize = gWimpyBufferSize
/ (sizeof(cl_float)
* RoundUpToNextPowerOfTwo(test_info.threadCount));
}
test_info.step = (cl_uint)test_info.subBufferSize * test_info.scale;
if (test_info.step / test_info.subBufferSize != test_info.scale)
{

View File

@@ -136,9 +136,9 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
int ftz = f->ftz || gForceFTZ;
double maxErrorVal0 = 0.0f;
double maxErrorVal1 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(cl_double), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(cl_double)) + 1);
uint64_t step = getTestStep(sizeof(cl_double), BUFFER_SIZE);
int scale =
(int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(cl_double)) + 1);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -160,16 +160,16 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
double *p = (double *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j * scale);
}
else
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j);
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -179,20 +179,20 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -204,7 +204,7 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_double);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -240,7 +240,7 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
double *r = (double *)gOut_Ref;
double *r2 = (double *)gOut_Ref2;
double *s = (double *)gIn;
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
{
long double dd;
r[j] = (double)f->dfunc.f_fpf(s[j], &dd);
@@ -252,14 +252,14 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -271,7 +271,7 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint64_t *t = (uint64_t *)gOut_Ref;
uint64_t *t2 = (uint64_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -413,7 +413,7 @@ int TestFunc_Double2_Double(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -136,9 +136,8 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
int ftz = f->ftz || gForceFTZ || 0 == (CL_FP_DENORM & gFloatCapabilities);
float maxErrorVal0 = 0.0f;
float maxErrorVal1 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(float)) + 1);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
int scale = (int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(float)) + 1);
cl_uchar overflow[BUFFER_SIZE / sizeof(float)];
int isFract = 0 == strcmp("fract", f->nameInCode);
int skipNanInf = isFract && !gInfNanSupport;
@@ -162,7 +161,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
uint32_t *p = (uint32_t *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
p[j] = (uint32_t)i + j * scale;
if (relaxedMode && strcmp(f->name, "sincos") == 0)
@@ -174,7 +173,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
p[j] = (uint32_t)i + j;
if (relaxedMode && strcmp(f->name, "sincos") == 0)
@@ -186,7 +185,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -196,20 +195,20 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -221,7 +220,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -273,7 +272,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
if (skipNanInf)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
double dd;
feclearexcept(FE_OVERFLOW);
@@ -290,7 +289,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
double dd;
if (relaxedMode)
@@ -309,14 +308,14 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -332,7 +331,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
uint32_t *t2 = (uint32_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -548,7 +547,7 @@ int TestFunc_Float2_Float(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -144,9 +144,9 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
double maxErrorVal = 0.0f;
double maxErrorVal2 = 0.0f;
cl_ulong maxiError = f->double_ulps == INFINITY ? CL_ULONG_MAX : 0;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(cl_double), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(cl_double)) + 1);
uint64_t step = getTestStep(sizeof(cl_double), BUFFER_SIZE);
int scale =
(int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(cl_double)) + 1);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -168,16 +168,16 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
double *p = (double *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j * scale);
}
else
{
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
p[j] = DoubleFromUInt32((uint32_t)i + j);
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -187,20 +187,20 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -212,7 +212,7 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_double);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -248,7 +248,7 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
double *r = (double *)gOut_Ref;
int *r2 = (int *)gOut_Ref2;
double *s = (double *)gIn;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
r[j] = (double)f->dfunc.f_fpI(s[j], r2 + j);
// Read the data back
@@ -256,14 +256,14 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -275,7 +275,7 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint64_t *t = (uint64_t *)gOut_Ref;
int32_t *t2 = (int32_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -385,7 +385,7 @@ int TestFunc_DoubleI_Double(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -141,9 +141,8 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
int ftz = f->ftz || gForceFTZ || 0 == (CL_FP_DENORM & gFloatCapabilities);
float maxErrorVal = 0.0f;
float maxErrorVal2 = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(float)) + 1);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
int scale = (int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(float)) + 1);
cl_ulong maxiError;
logFunctionInfo(f->name, sizeof(cl_float), relaxedMode);
@@ -172,16 +171,16 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
uint32_t *p = (uint32_t *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j * scale;
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -191,20 +190,20 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
goto exit;
}
memset_pattern4(gOut2[j], &pattern, bufferSize);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
memset_pattern4(gOut2[j], &pattern, BUFFER_SIZE);
if ((error = clEnqueueWriteBuffer(gQueue, gOutBuffer2[j], CL_FALSE,
0, BUFFER_SIZE, gOut2[j], 0, NULL,
NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2b(%d) ***\n",
error, j);
@@ -216,7 +215,7 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -252,7 +251,7 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
float *r = (float *)gOut_Ref;
int *r2 = (int *)gOut_Ref2;
float *s = (float *)gIn;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = (float)f->func.f_fpI(s[j], r2 + j);
// Read the data back
@@ -260,14 +259,14 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
}
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer2[j], CL_TRUE, 0,
bufferSize, gOut2[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut2[j], 0, NULL, NULL)))
{
vlog_error("ReadArray2 failed %d\n", error);
goto exit;
@@ -279,7 +278,7 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
int32_t *t2 = (int32_t *)gOut_Ref2;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -384,7 +383,7 @@ int TestFunc_FloatI_Float(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -134,8 +134,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
float maxError = 0.0f;
int ftz = f->ftz || gForceFTZ;
double maxErrorVal = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(cl_double), bufferSize);
uint64_t step = getTestStep(sizeof(cl_double), BUFFER_SIZE);
logFunctionInfo(f->name, sizeof(cl_double), relaxedMode);
@@ -155,10 +154,10 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
{
// Init input array
cl_ulong *p = (cl_ulong *)gIn;
for (j = 0; j < bufferSize / sizeof(cl_ulong); j++) p[j] = random64(d);
for (j = 0; j < BUFFER_SIZE / sizeof(cl_ulong); j++) p[j] = random64(d);
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -168,10 +167,10 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -183,7 +182,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_double);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -212,7 +211,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
// Calculate the correctly rounded reference result
double *r = (double *)gOut_Ref;
cl_ulong *s = (cl_ulong *)gIn;
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
r[j] = (double)f->dfunc.f_u(s[j]);
// Read the data back
@@ -220,7 +219,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -231,7 +230,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint64_t *t = (uint64_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(cl_double); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(cl_double); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -283,7 +282,7 @@ int TestFunc_Double_ULong(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -126,9 +126,8 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
float maxError = 0.0f;
int ftz = f->ftz || gForceFTZ || 0 == (CL_FP_DENORM & gFloatCapabilities);
float maxErrorVal = 0.0f;
size_t bufferSize = (gWimpyMode) ? gWimpyBufferSize : BUFFER_SIZE;
uint64_t step = getTestStep(sizeof(float), bufferSize);
int scale = (int)((1ULL << 32) / (16 * bufferSize / sizeof(double)) + 1);
uint64_t step = getTestStep(sizeof(float), BUFFER_SIZE);
int scale = (int)((1ULL << 32) / (16 * BUFFER_SIZE / sizeof(double)) + 1);
int isRangeLimited = 0;
float half_sin_cos_tan_limit = 0;
@@ -172,16 +171,16 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
uint32_t *p = (uint32_t *)gIn;
if (gWimpyMode)
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j * scale;
}
else
{
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
p[j] = (uint32_t)i + j;
}
if ((error = clEnqueueWriteBuffer(gQueue, gInBuffer, CL_FALSE, 0,
bufferSize, gIn, 0, NULL, NULL)))
BUFFER_SIZE, gIn, 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer ***\n", error);
return error;
@@ -191,10 +190,10 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
uint32_t pattern = 0xffffdead;
memset_pattern4(gOut[j], &pattern, bufferSize);
memset_pattern4(gOut[j], &pattern, BUFFER_SIZE);
if ((error =
clEnqueueWriteBuffer(gQueue, gOutBuffer[j], CL_FALSE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("\n*** Error %d in clEnqueueWriteBuffer2(%d) ***\n",
error, j);
@@ -206,7 +205,7 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
for (j = gMinVectorSizeIndex; j < gMaxVectorSizeIndex; j++)
{
size_t vectorSize = sizeValues[j] * sizeof(cl_float);
size_t localCount = (bufferSize + vectorSize - 1) / vectorSize;
size_t localCount = (BUFFER_SIZE + vectorSize - 1) / vectorSize;
if ((error = clSetKernelArg(kernels[j], 0, sizeof(gOutBuffer[j]),
&gOutBuffer[j])))
{
@@ -235,7 +234,7 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
// Calculate the correctly rounded reference result
float *r = (float *)gOut_Ref;
cl_uint *s = (cl_uint *)gIn;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
r[j] = (float)f->func.f_u(s[j]);
// Read the data back
@@ -243,7 +242,7 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
{
if ((error =
clEnqueueReadBuffer(gQueue, gOutBuffer[j], CL_TRUE, 0,
bufferSize, gOut[j], 0, NULL, NULL)))
BUFFER_SIZE, gOut[j], 0, NULL, NULL)))
{
vlog_error("ReadArray failed %d\n", error);
goto exit;
@@ -255,7 +254,7 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
// Verify data
uint32_t *t = (uint32_t *)gOut_Ref;
for (j = 0; j < bufferSize / sizeof(float); j++)
for (j = 0; j < BUFFER_SIZE / sizeof(float); j++)
{
for (k = gMinVectorSizeIndex; k < gMaxVectorSizeIndex; k++)
{
@@ -316,7 +315,7 @@ int TestFunc_Float_UInt(const Func *f, MTdata d, bool relaxedMode)
if (gVerboseBruteForce)
{
vlog("base:%14u step:%10zu bufferSize:%10zd \n", i, step,
bufferSize);
BUFFER_SIZE);
}
else
{

View File

@@ -34,7 +34,6 @@
struct Func;
extern int gWimpyBufferSize;
extern int gWimpyReductionFactor;
#define VECTOR_SIZE_COUNT 6