Thanks to visit codestin.com
Credit goes to github.com

Skip to content

core(OpenCL): optimize convertTo() with CV_16F (convertFp16() replacement) #24918

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Jan 26, 2024
+353 −89
Merged

core(OpenCL): optimize convertTo() with CV_16F (convertFp16() replacement) #24918

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
37743ef
core: convert FP16 perf tests
alalek Jan 23, 2024
3346d10
core(ocl): fix convertTo() perfomance
alalek Jan 23, 2024
05b5684
core(ocl): add Device::hasFP64() / hasFP16()
alalek Jan 24, 2024
fc552ea
core(ocl): fix FP16/FP64 checks in convertTo()
alalek Jan 24, 2024
b9b3860
core: deprecate convertFp16()
alalek Jan 24, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions modules/core/include/opencv2/core.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -556,6 +556,8 @@ The format of half precision floating point is defined in IEEE 754-2008.

@param src input array.
@param dst output array.

@deprecated Use Mat::convertTo with CV_16F instead.
*/
CV_EXPORTS_W void convertFp16(InputArray src, OutputArray dst);

Expand Down
5 changes: 5 additions & 0 deletions modules/core/include/opencv2/core/ocl.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -127,6 +127,11 @@ class CV_EXPORTS_W_SIMPLE Device
CV_WRAP int singleFPConfig() const;
CV_WRAP int halfFPConfig() const;

/// true if 'cl_khr_fp64' extension is available
CV_WRAP bool hasFP64() const;
/// true if 'cl_khr_fp16' extension is available
CV_WRAP bool hasFP16() const;

CV_WRAP bool endianLittle() const;
CV_WRAP bool errorCorrectionSupport() const;

Expand Down
8 changes: 4 additions & 4 deletions modules/core/include/opencv2/core/opencl/opencl_info.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -141,13 +141,13 @@ static void dumpOpenCLInformation()
DUMP_MESSAGE_STDOUT(" Max memory allocation size = " << maxMemAllocSizeStr);
DUMP_CONFIG_PROPERTY("cv_ocl_current_maxMemAllocSize", device.maxMemAllocSize());

const char* doubleSupportStr = device.doubleFPConfig() > 0 ? "Yes" : "No";
const char* doubleSupportStr = device.hasFP64() ? "Yes" : "No";
Copy link
Member

@fengyuentau fengyuentau Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So what is going to happen with doubleFPConfig and halfFPConfig. Are they deprecating as well?

Copy link
Contributor Author

@opencv-alalek opencv-alalek Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No, they are still needed if we want to compute with proper inf/nans support.

Copy link
Member

@fengyuentau fengyuentau Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I checked doubleFPConfig in the whole opencv project and it is basically used like this,

bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0

they are still needed if we want to compute with proper inf/nans support

Did I miss anything here? Or it is in the user code instead?

Copy link
Contributor Author

@opencv-alalek opencv-alalek Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Usage is not correct.

Copy link
Member

@fengyuentau fengyuentau Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What is the correct way? All these code is wrong?

Copy link
Contributor Author

@opencv-alalek opencv-alalek Jan 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All are subject for revising.

Copy link
Member

@fengyuentau fengyuentau Jan 26, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Okay.

DUMP_MESSAGE_STDOUT(" Double support = " << doubleSupportStr);
DUMP_CONFIG_PROPERTY("cv_ocl_current_haveDoubleSupport", device.doubleFPConfig() > 0);
DUMP_CONFIG_PROPERTY("cv_ocl_current_haveDoubleSupport", device.hasFP64());

const char* halfSupportStr = device.halfFPConfig() > 0 ? "Yes" : "No";
const char* halfSupportStr = device.hasFP16() ? "Yes" : "No";
DUMP_MESSAGE_STDOUT(" Half support = " << halfSupportStr);
DUMP_CONFIG_PROPERTY("cv_ocl_current_haveHalfSupport", device.halfFPConfig() > 0);
DUMP_CONFIG_PROPERTY("cv_ocl_current_haveHalfSupport", device.hasFP16());

const char* isUnifiedMemoryStr = device.hostUnifiedMemory() ? "Yes" : "No";
DUMP_MESSAGE_STDOUT(" Host unified memory = " << isUnifiedMemoryStr);
Expand Down
181 changes: 181 additions & 0 deletions modules/core/perf/opencl/perf_matop.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,187 @@ OCL_PERF_TEST_P(ConvertToFixture, ConvertTo,
SANITY_CHECK(dst);
}


//#define RUN_CONVERTFP16
static Size convertFP16_srcSize(4000, 4000);

OCL_PERF_TEST(Core, ConvertFP32FP16MatMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_32F;
const int dtype = CV_16F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

Mat src(srcSize, type);
Mat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP32FP16MatUMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_32F;
const int dtype = CV_16F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

Mat src(srcSize, type);
UMat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP32FP16UMatMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_32F;
const int dtype = CV_16F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

UMat src(srcSize, type);
Mat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP32FP16UMatUMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_32F;
const int dtype = CV_16F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

UMat src(srcSize, type);
UMat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP16FP32MatMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_16F;
const int dtype = CV_32F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

Mat src(srcSize, type);
Mat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP16FP32MatUMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_16F;
const int dtype = CV_32F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

Mat src(srcSize, type);
UMat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP16FP32UMatMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_16F;
const int dtype = CV_32F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

UMat src(srcSize, type);
Mat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}

OCL_PERF_TEST(Core, ConvertFP16FP32UMatUMat)
{
const Size srcSize = convertFP16_srcSize;
const int type = CV_16F;
const int dtype = CV_32F;

checkDeviceMaxMemoryAllocSize(srcSize, type);
checkDeviceMaxMemoryAllocSize(srcSize, dtype);

UMat src(srcSize, type);
UMat dst(srcSize, dtype);
declare.in(src, WARMUP_RNG).out(dst);

#ifdef RUN_CONVERTFP16
OCL_TEST_CYCLE() convertFp16(src, dst);
#else
OCL_TEST_CYCLE() src.convertTo(dst, dtype);
#endif

SANITY_CHECK_NOTHING();
}


///////////// CopyTo ////////////////////////

typedef Size_MatType CopyToFixture;
Expand Down
Loading