* Permit half overflow within allowable ULP
Modify the algorithm for calculating half precision ULP error so
that it duplicates the behaviour of the single precision ULP algorithm,
in regards to allowing overflow within the defined ULP error.
In the case where the test value is infinity, but the reference is
finite, pretend the test value is 63336.0 and calculate the ULP error
against that.
Encountered this while testing half precision `hypot()` in PR !529,
for inputs `hypot(-48864.0, 43648.0)` which has reference
`65519.755799`. With RTE rounding this only just rounds to `65504` as half,
and returning INF is currently infinite ULP error. Using the leniency
introduced by this change however the error is `~0.5` within the `2` ULP
bounds defined by the spec.
* Run clang-format over changes
Code now conforms to style guidelines and allows `check-format.sh` to pass.
When running sub-tests explicitly on the command line and that sub-test
fails, the test harness returns `0` from `main()`. This is problematic
in automated testing environments as failed tests can be hidden.
This patch iterates over all the test results in the `resultTestList`
after all sub-tests have completed, if any of the tests failed a
non-zero value is returned from `parseAndCallCommandLineTests()` which
is then propagated up the call stack to `main()`.
- Remove the requirement to detail values for device queries and list
extensions in the submission template. They are printed by computeinfo.
- Require that tests skipped in accordance with the conformance
process be listed in the submission template.
- Remove support for OPENCL_1_0_DEVICE. This variable doesn't control
anything anymore and the updated conformance process no longer refers
to it.
- Require that implementations running via a loader document the loader
that was used.
Signed-off-by: Kevin Petit <kevin.petit@arm.com>
* add test_compiler test for OpenCL C versions
* add checks for required OpenCL C versions
* bugfix: only OpenCL C 2.0 is required for OpenCL 2.1 and 2.2
* pass reference to feature struct vs. pointer
* address review comments regarding C++ identifiers and 3d image writes
* Test feature macro OpenCL 3.0
* Test feature macro OpenCL 3.0 - refactor to template function
* Use test_error and format fix issues
* Fix returned value type for CL_DEVICE_MAX_GLOBAL_VARIABLE_SIZE
* Code format fix
* Feature macro test - refactor code to have one test case
* add consistency checking comparing supported features to CL_DEVICE_OPENCL_C_FEATURES
* fix minor issues
* Feature macro test - fix test version
* Feature macro test - add -cl-std=CL3.0 build option
* Feature macro test - compilation errors
* Feature macro test - compilation errors OSX
* Feature macro test - fix adding build options
* Feature macro test - share function OutputBuildLogs
* Feature macro test - back removed cases
* Feature macro test - review fixes
* Feature macro test - split check image formats
* Feature macro test - fixed version
* Change setup code in `KernelCode()` to use `_explicit` builtin
variants that are common to both OpenCL-2.X and OpenCL-3.0.
* Only test optional supported builtin variants (`_explicit` signature
memory_order/scope) for OpenCL-3.0.
* Disable program scope global variable and generic address space tests
for a OpenCL-3.0 driver which does not optionally support these
features.
* Fix enqueue_flags test to use correct barrier type.
Currently, enqueue_flags test uses CLK_LOCAL_MEM_FENCE.
Use CLK_GLOBAL_MEM_FENCE instead as all threads across work-groups
need to wait here.
* Add check for support for Read-Wrie images
Read-Write images have required OpenCL 2.x.
Read-Write image tests are already being skipped
for 1.x devices.
With OpenCL 3.0, read-write images being optional,
the tests should be run or skipped
depending on the implementation support.
Add a check to decide if Read-Write images are
supported or required to be supported depending
on OpenCL version and decide if the tests should
be run on skipped.
Fixes issue #894
* Fix formatting in case of Read-Write image checks.
Fix formatting in case of Read-write image checks.
Also, combine two ifs into one in case of
kerne_read_write tests
* Fix some more formatting for RW-image checks
Remove unnecessary spaces at various places.
Also, fix lengthy lines.
The test attempts to build a program from source containing invalid code, to
provoke a build failure. In offline compilation mode, this causes the offline
compiler to give an error, failing the test.
Signed-off-by: Einar Hov <einar.hov@arm.com>
* added pipes test
clang-format fixes
* simplify diffs
* added pipes test
clang-format fixes
* simplify diffs
* add Program Scope Global Variables consistency test
* switched other checks to test_assert_error
* add non-uniform work group consistency test
* added read-write images consistency test
* add Creating 2D Images from Buffers consistency test
* add Device and Host Timer Synchronization consistency test
* add Intermediate Language Programs consistency test
* add Subgroups consistency test
* add Program Initialization and Clean-Up Kernels consistency test
* add 3D Image Writes consistency test
* clang-format fixes
* switch the check for 3D image writes extension for clarity
* add Depth Images consistency test
* update test for CL_QUEUE_SIZE, must return CL_INVALID_COMMAND_QUEUE
* formatting fixes
* fix copy-paste typo
* change expected error codes to CL_INVALID_OPERATION
* address review comments
* fix formatting
* address review feedback
Since pipes are cl_mems, We can spec and test that clGetPipeInfo
returns CL_INVALID_OPERATION when pipes are not supported by
passing a non-pipe memory object to the function.
* use the CTS framework to choose the OpenCL C version
* address review feedback
In two places, a stray semicolon between an if statement and its body
would cause the body to be executed unconditionally. Fix this.
Signed-off-by: Stuart Brady <stuart.brady@arm.com>
... as required by the OpenCL specification.
Also introduce a utility function to manage polling for changes
via a user-supplied function. Several tests we introduced lately
could have used this.
Signed-off-by: Kévin Petit <kpet@free.fr>
* OpenCL 3.0 test CL_QUEUE_PROPERTIES_ARRAY
* add verification if requested_size <= CL_DEVICE_QUEUE_ON_DEVICE_MAX_SIZE
* remove test_case - set NULL properties, get not empty array with 0 terminator
* add printing test_case description
* change logic of checking if requested properties are supported by device
depending on host/device type queue.
* fix a few bugs, rename test for consistency
* add utility function for comparing properties
Co-authored-by: Grzegorz Wawiorko <grzegorz.wawiorko@intel.com>
* Adds testing for CL_PROGRAM_IL
Added a test to check SpirV as an IL for programs built with IL
which includes a negative test for programs built without IL.
Added a test_fail macro.
Fixes#164
Change-Id: I908241242b369551806e43b90ab414f895d5c8f7
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* COMPUTE-11739 Removing unused platform and device_version
Fixes#164
Change-Id: Ib1168f6396132b69996d07166e1b593fa933d245
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* Dynamically select the `-cl-std` option on the basic `sizeof` and
`wg_barrier` tests to be one of `CL1.X`, `CL2.0` or `CL3.0`.
Use the most recent CL C standard supported on the device.
* Specify GCC flag `-frounding-math` on x86
We have been seeing fails in `test_conversions` on x86_64 when converting to the following integer types from `cl_double` with a non-default rounding mode:
```
char_rtn_double
char_rtp_double
int_rtn_double
int_rtp_double
long_rtn_double
long_rtp_double
long_sat_rtn_double
long_sat_rtp_double
short_rtn_double
short_rtp_double
uchar_rtp_double
uint_rtp_double
ulong_rtp_double
ushort_rtp_double
```
After investigation it was discovered that `rint()` was incorrectly rounding `cl_double` inputs despite the rounding mode being correctly set using `fesetround()` beforehand. E.g For 'char_rtn_double' `-3.5` was getting rounding to `-3.0` rather than `-4.0`.
This is a gcc issue https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92164 when using the builtin `rint()` implementation, rather than `std::rint()`, that only presents itself when compiling a Release build with `-O2` rather than Debug.
Adding the compiler flag `-fno-builtin-rint` to the CMake works around this problem, however based on the discussion in the gcc ticket using `-frounding-math` appears to be a more comprehensive fix. As `-frounding-math` tells gcc that the code will be modifying the rounding mode, removing the assumption that the same rounding mode is in effect everywhere.
* Set FENV_ACCESS ON in rounding_mode.h
Inform compilers which are aware of the `FENV_ACCESS` pragma that TUs
which include `harness/rounding_mode.h` may manipulate the floating
point environment.
* Remove FENV_ACCESS pragma from test_conversions.cpp
This pragma is now set in the included header
`test_common/harness/rounding_mode.h`
Co-authored-by: Kenneth Benzie (Benie) <k.benzie@codeplay.com>
* Use float<->half conversion routines from the OpenCL headers
Fixes#870
Signed-off-by: Kevin Petit <kevin.petit@arm.com>
* Use cl_half_from_double
* Fix windows build errors
* Fix more build errors
* Code formatting
* Remove TEST class
The compiler tries to compile the same source multiple times and more
importantly write out to the same filename. Adding a lock to prevent
simultaneous compilation solves this issue. This is particularly
noticeable in integer_ops testing when using offline compilation,
but just in case, hold the lock for the entire program creation.
Signed-off-by: Mats Petersson <mats.petersson@arm.com>
Signed-off-by: Stuart Brady <stuart.brady@arm.com>
* Fix Tests Assuming Online Compilation in Offline MOde
* Update the list of tests to skip in offline mode. These tests need
to be skipped for offline-binary mode since they make API calls that
rely on a compiler being present in the runtime.
- [x] Skip `get_kernel_arg_info_compatibility` since it makes calls to
`clBuildProgram`. (these tests cannot be run in offline mode since:
*Kernel argument information is only available if the program object
associated with kernel is created with `clCreateProgramWithSource` and
the program executable was built with the `-cl-kernel-arg-info` option
specified in options argument to `clBuildProgram` or `clCompileProgram`
.*)
- [x] Skip `compiler` tests that make calls to `clCompileProgram`. These
tests could still be run in offline spirv mode if there is a compiler in
the driver.
- [x] Use offline compilation path in `contractions` in the case that
CTS is run in offline.
* Avoid shadowing `error` variable
Co-authored-by: Ewan Crawford <ewan@codeplay.com>
- [x] Add `CL_DEVICE_NAME` to list of device properties printed into the
clDeviceInfo file for offline compilation testing.
- [x] Add `get_device_name` helper function.
- [x] Update offline compiler interface explanation file with
`CL_DEVICE_NAME` and `CL_DEVICE_IMAGE_SUPPORT` which was missed from
this file when added.
- [x] Add `CL_DEVICE_IMAGE_SUPPORT` to the clDeviceInfo file
which contains
device specific information for the purposes of offline compilation. For
devices with images support `CL_DEVICE_IMAGE_SUPPORT=1` is printed,
otherwise `CL_DEVICE_IMAGE_SUPPORT=0`.
* Remove duplicate IsXSubnormal functions
Is{Double,FLoat,Half}Subnormal function duplicates are currently in the codebase.
Fixes#511
Change-Id: Ibe97f20a6e38db6aed00c9ba397cfa72036bd1c5
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* Update conversions.h
* Enable -Werror for GCC/Clang builds
Fixes many of the errors this produces, and disables a handful that
didn't have solutions that were obvious (to me).
* Check for `-W*` flags empirically
* Remove cl_APPLE_fp64_basic_ops support
* Undo NAN conversion fix
* Add comments to warning override flags
* Remove unneeded STRINGIFY definition
* Fix tautological compare issue in basic
* Use ABS_ERROR macro in image tests
* Use fabs for ABS_ERROR macro
* Move ABS_ERROR definition to common header
* Accept OpenCL 3.0 in version parsing code and use where appropriate
There were a number of tests against 2.2 that are clearer against 3.0.
Fixes#751
Signed-off-by: Kévin Petit <kpet@free.fr>
* Remove CL_EXPERIMENTAL guards
Signed-off-by: Kévin Petit <kpet@free.fr>
* formatting
* Configure the headers for OpenCL 3.0
* more format fixes
* Use std::string for get_device_*_string helpers
Removes need to use BufferOwningPtr to avoid memory leaks.
* Rename get_device_info to get_device_info_string
Makes it more obvious that it should only be called for device queries
that return strings.
* Tokenize extensions in is_extension_available
Avoids the potential issue where one extension name is a prefix of
another.
* Throw exception when device info size is 0
* Allow CL_HALF_FLOAT denorm flushing for write tests (#452)
* On mismatch, add relaxation when denormal half result is expected
* Refactor to use common validation function
* Clean up some diagnostics
* Fix review comments
- use cl_half
- remove extraneous casts
- replace literals with sizeof()
* Document rollover trick for IsHalfSubnormal
Removing all references to check_opencl_version as similar
get_device_cl_version() can be used instead.
Fixes#527
Change-Id: I474b6f536033707e1beb9b5b39410de24672c040
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
They are duplicate functions
Fixes#709
Change-Id: I8f7d6b8254047adb9e09ede4951dedc3ec5c1099
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
is_extension_available is modified to handle errors by throwing
exceptions.
Contributes to #627
Change-Id: Ie9423df588892a0f8effa03d1cb48bf12400b983
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
- Introduce MTdataHolder
- Use BufferOwningPtr to manage allocations (fixes a few leaks in error cases)
- Introduce variable for common expressions
- Remove image format support check as the format is required by OpenCL 1.0
Contributes to #700
Signed-off-by: Kévin Petit <kpet@free.fr>
(Patch1)
A number of tests have got their own code for checking the presence of
extensions. This change replaces that code with is_extension_available
function.
Contributes to #627
Change-Id: I8dd2233719aa8c84841ac61776437d7f6e3fafe6
Signed-off-by: Ellen Norris-Thompson <ellen.norris-thompson@arm.com>
* Harness: Added an option for a test to skip itself.
New condtion in callSingleTestFunction. This allows a test to return TEST_SKIP if the device does not support the requested test.
* Split generic_ptr_to_host_mem into two tests.
The old generic_ptr_to_host_mem test tests two different device capabilities (SVM and not SVM). This is a prerequisite for the following commit.
* Generic Address Space: Skip tests utlilising SVM on devices that do not support SVM.
Where a device does not support SVM, do not run the generic address space test(s) that rely on SVM.
