All existing tests in `test_image_streams`, that are capable of testing
image formats using the `CL_UNORM_INT_2_101010_EXT` data type, do so.
Signed-off-by: Gorazd Sumkovski <gorazd.sumkovski@arm.com>
The kernel_read_write tests have a lot of duplicate code. These are the
next steps to reducing the duplication, by using the functions in
test_common.cpp as common for 1D, 1D array and 2D array.
---------
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
Co-authored-by: Ahmed Hesham <117350656+ahesham-arm@users.noreply.github.com>
* Don't recalculate image parameters repeatedly in `test_read_image()`
We've already done this in the loop. There's no need to recalculate
those parameters over and over again in `sample_image_pixel*()` and
`read_image_pixel*()`. This should save some work during the image
streams test.
This only affects the 3D tests for now, but my time profiles indicate
this is where we spend the most time anyway.
* Vectorize read_image_pixel_float() and sample_image_pixel_float() for SSE/AVX
This shortens the image streams test time from 45 minutes without it to
37 minutes. Unfortunately, most of the time is now spent waiting for
memory, particularly in the 3D tests, because the 3D image doesn't
neatly fit in the cache, especially in the linear sampling case, where
pixels from two 2D slices must be sampled. Software prefetching won't
help; it only helps when execution time is dominated by operations, but
this is dominated by memory access. Randomized offsets are likely a
factor, because they throw off the hardware prefetcher.
One possible further optimization is, in the linear sampling case, to
load two sampled pixels at once. This is easy to do using AVX, which
extends SSE with 256-bit vectors.
Obviously, this only applies to x86 CPUs with SSE2. The greatest
performance gains, however, are seen with SSE4.1. Most modern x86 CPus
have SSE4. Work is needed to support other CPUs' vector units--ARM
Advanced SIMD/NEON is probably the most important one. Another
possibility is arranging the code so that the compiler's
autovectorization will kick in and do what I did here manually.
* Make InitFloatCoords suitable for all image types
Contributes #616
* Create common functions neutral for image types
Remove 3D specific code from common test_read_image so using
it for other image types is simpler in following patches
Contributes #616
* Removing unused code
Tidying commented out or unnecessary code
Contributes #616
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* Restoring 'lod' variable name
Contributes #616
* Default cases to handle unsupported image types
Contributes #616
* Resolving build issues
Contributes #616
* Fix formatting
Contributes #616
* Using TEST_FAIL as an error code.
Contributes #616
* Add static keyword, improve error handling
Contributes #616
* Fix build errors with least disruption
Contributes #616
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* images: Stop checking gDeviceType != CL_DEVICE_TYPE_GPU
If the device type also advertises CL_DEVICE_TYPE_DEFAULT (which should
be valid), this causes it to be considered a CPU device and the tests
enforce different precision and rounding expectations.
* Fix clang-format
* Drop redundant NORM_OFFSET checks
* Minor fixes for CL_UNORM_SHORT_565, CL_UNORM_SHORT_555
* Fix verification for undefined bit
* Relax current infinitely precision requirement for these formats
and move check in common function.
* Add proper debug output.
Signed-off-by: John Kesapides <john.kesapides@arm.com>
* Minor Formating fix.
Signed-off-by: John Kesapides <john.kesapides@arm.com>
* Move common global variable and functions to header
InitFloatCoords for 3D read images has also been renamed
so it can later be used as a common function
Contributes #616
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* Set-up for using 3D functions as a base
test_read_image_3D had been moved to common.cpp (and renamed
test_read_image) along with corresponding
determine_validation_error_offset and InitFloatCoords.
Only function names and the formatting have been changed.
Contributes #616
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
