OpenCL conformance

Conformance related CMake options

• -DENABLE_CONFORMANCE=ON/OFF Defaults to OFF. This option by itself does not guarantee OpenCL-conformant build; it merely ensures that a build fails if some CMake options which are known to result in non-conformant PoCL build are given. Only applies to CPU driver.

  Changes when ENABLE_CONFORMANCE is ON, the CPU drivers are built with the following changes:

  • read-write images are disabled (some 1D/2D image array tests fail)

  • the list of supported image formats is much smaller

  • SLEEF is always enforced for the builtin library

  • cl_khr_fp16 is disabled

  • cl_khr_subgroup_{ballot,shuffle} are disabled

  • cl_intel_subgroups,cl_intel_required_subgroup_size are disabled

  If ENABLE_CONFORMANCE is OFF, and ENABLE_HOST_CPU_DEVICES is ON, the conformance testsuite is disabled in CMake. This is because some CTS tests will fail on such build.

Supported & Unsupported optional OpenCL 3.0 features

This list is only related to CPU devices (cpu & cpu-minimal drivers). Other drivers (CUDA, TCE etc) only support OpenCL 1.2. Note that 3.0 support on CPU devices requires LLVM 14 or newer.

Supported 3.0 features:

• Shared Virtual Memory

• C11 atomics

• 3D Image Writes

• SPIR-V

• Program Scope Global Variables

• Subgroups

• Generic Address Space

Unsupported 3.0 features:

• Device-side enqueue

• Pipes

• Non-Uniform Work Groups

• Read-Write Images

• Creating 2D Images from Buffers

• sRGB & Depth Images

• Device and Host Timer Synchronization

• Intermediate Language Programs

• Program Initialization and Clean-Up Kernels

• Work Group Collective Functions

How to run the OpenCL 3.0 conformance test suite

You’ll need to build PoCL with enabled ICD, and the ICD must be one that supports OpenCL version 3.0 (for ocl-icd, this is available since version 2.3.0). This is because while the CTS will run with 1.2 devices, it requires 3.0 headers and 3.0 ICD to build. You’ll also need to enable the suite in the pocl’s external test suite set. This is done by adding -DENABLE_TESTSUITES=conformance -DENABLE_CONFORMANCE=ON to the cmake command line. After this make prepare_examples fetches and prepares the conformance suite for testing. After building pocl with make, the CTS can be run with ctest -L <LABEL> where <LABEL> is a CTest label.

There are three different CTest labels for using CTS, one label covers the full set tests in CTS, the other two contain a smaller subset of CTS tests. The fastest is conformance_suite_micro_main label, which takes approx 10-30 minutes on current (desktop) hardware. The medium sized conformance_suite_mini_main can take 1-2 hours on current hardware. The full sized CTS is available with label conformance_suite_full_main. This can take 10-30 hrs on current hardware.

If PoCL is compiled with SPIR-V support, three more labels are available, where _main suffix is replaced by _spirv (e.g. conformance_suite_mini_spirv) These labels will run the same tests as the _main variant, but use offline compilation to produce SPIR-V and use that to create programs, instead of default creating from OpenCL C source.

Note that running ctest -L conformance_suite_micro will run both variants (the online and offline compilation) since the -L option takes a regexp.

Additionally, there is a new cmake label, conformance_30_only to run tests which are only relevant to OpenCL 3.0.

CPU device version 1.2 should also work with CTS 3.0 (tests will be skipped).

Known issues related to CTS

• a few tests from basic/test_basic may fail / segfault because they request a huge amount of memory for buffers.

• a few tests may run much faster if you limit the reported Global memory size with POCL_MEMORY_LIMIT env var. In particular, “kernel_image_methods” test with “max_images” argument.

• With LLVM 15 and 16, when running CTS with the offline compilation mode (= via SPIR-V), Clang + SPIR-V translator produce invalid SPIR-V for several tests. PoCL bugreport: https://github.com/pocl/pocl/issues/1232 Related Khronos issues: https://github.com/KhronosGroup/SPIRV-LLVM-Translator/issues/2008 https://github.com/KhronosGroup/SPIRV-LLVM-Translator/issues/2024 https://github.com/KhronosGroup/SPIRV-LLVM-Translator/issues/2025

• Integer division by zero. OpenCL 1.2 specification requires that division by zero on integers results in undefined values, instead of raising exceptions. This requires pocl to install a handler of SIGFPE. The handler is per-process, but it checks the thread ID, so that it only ignores the error for the CPU driver threads, not the user program’s threads. This might not work on every system. The handler can be disabled completely by setting the env variable POCL_SIGFPE_HANDLER to 0. Note that this is currently only relevant for x86(-64) + Linux, on all other systems this issue is not handled in any way (thus Pocl is likely non-conformant there).

• Many of native_ and half_ variants of kernel library functions are mapped to the “full” variants.

• clSetUserEventStatus() called with negative status. The Spec leaves the behaviour in this case as “implementation defined”, and this part of pocl is only very lightly tested by the conformance tests. clSetUserEventStatus() called with CL_COMPLETE works as expected, and is heavily used by the conversions conformance test.

Conformance tests results (precision of builtin math library functions)

Note that it’s impossible to test double precision on the entire range, therefore the results may vary.

x86-64 CPU with AVX2+FMA, LLVM 4.0, tested on Nov 1, 2017

NAME	Worst ULP	WHERE
add	0.00	{0x0p+0, 0x0p+0}
addD	0.00	{0x0p+0, 0x0p+0}
assignment	0.00	0x0p+0
assignmentD	0.00	0x0p+0
cbrt	0.50	-0x1.5629d2p+116
cbrtD	0.59	0x1.0000000000136p+1022
ceil	0.00	0x0p+0
ceilD	0.00	0x0p+0
copysign	0.00	{0x0p+0, 0x0p+0}
copysignD	0.00	{0x0p+0, 0x0p+0}
cos	2.37	0x1.1338ccp+20
cosD	2.27	-0x1.d10000000074p+380
cosh	2.41	-0x1.602166p+2
coshD	1.43	-0x1.98000000003efp+5
cospi	1.94	0x1.d73b56p-2
cospiD	2.46	-0x1.adffffffffa91p-2
divide	0.00	{0x0p+0, 0x0p+0}
divideD	0.00	{0x0p+0, 0x0p+0}
exp	0.95	-0x1.762532p+2
expD	0.94	0x1.2f0000000023dp+7
exp10	0.79	-0x1.309022p+5
exp10D	0.64	-0x1.34ffffffffcc9p+8
exp2	0.79	-0x1.fa3d0ep+6
exp2D	0.75	-0x1.ff00000000417p+9
expm1	1.00	-0x1.7a0002p-25
expm1D	0.99	-0x1.26p+5
fabs	0.00	0x0p+0
fabsD	0.00	0x0p+0
fdim	0.00	{0x0p+0, 0x0p+0}
fdimD	0.00	{0x0p+0, 0x0p+0}
floor	0.00	0x0p+0
floorD	0.00	0x0p+0
fma	0.00	{0x0p+0, 0x0p+0, 0x0p+0}
fmaD	0.00	{0x0p+0, 0x0p+0, 0x0p+0}
fmax	0.00	{0x0p+0, 0x0p+0}
fmaxD	0.00	{0x0p+0, 0x0p+0}
fmin	0.00	{0x0p+0, 0x0p+0}
fminD	0.00	{0x0p+0, 0x0p+0}
fmod	0.00	{0x0p+0, 0x0p+0}
fmodD	0.00	{0x0p+0, 0x0p+0}
fract	{ 0.00, 0.00}	{0x0p+0, 0x0p+0}
fractD	{ 0.00, 0.00}	{0x0p+0, 0x0p+0}
frexp	{ 0.00, 0}	0x0p+0
frexpD	{ 0.00, 0}	0x0p+0
hypot	1.93	{0x1.17c998p-127, -0x1.5fedb8p-127}
hypotD	1.73	{0x1.5d2ebeed7663cp-1022, 0x1.67457048a2318p-1022}
ldexp	0.00	{0x0p+0, 0}
ldexpD	0.00	{0x0p+0, 0}
log10	0.50	0x1.7fee2ep-1
log10D	0.50	0x1.9100000000639p+1022
log	0.63	0x1.7fcb3ep-1
logD	0.75	0x1.7d00000000381p+0
log1p	1.00	-0x1.fa0002p-126
log1pD	1.00	-0x1.e000000000001p-1022
log2	0.59	0x1.1107a2p+0
log2D	0.72	0x1.120000000063dp+0
logb	0.00	0x0p+0
logbD	0.00	0x0p+0
mad	0.00	{0x0p+0, 0x0p+0, 0x0p+0} no ULP check
madD	0.00	{0x0p+0, 0x0p+0, 0x0p+0} no ULP check
maxmag	0.00	{0x0p+0, 0x0p+0}
maxmagD	0.00	{0x0p+0, 0x0p+0}
minmag	0.00	{0x0p+0, 0x0p+0}
minmagD	0.00	{0x0p+0, 0x0p+0}
modf	{ 0.00, 0.00}	{0x0p+0, 0x0p+0}
modfD	{ 0.00, 0.00}	{0x0p+0, 0x0p+0}
multiply	0.00	{0x0p+0, 0x0p+0}
multiplyD	0.00	{0x0p+0, 0x0p+0}
nan	0.00	0x0p+0
nanD	0.00	0x0p+0
nextafter	0.00	{0x0p+0, 0x0p+0}
nextafterD	0.00	{0x0p+0, 0x0p+0}
pow	0.82	{0x1.91237cp-1, 0x1.4da146p+8}
powD	0.80	{0x1.2bfb4b18164c9p+65, -0x1.b78438ae9c3bdp-8}
pown	0.65	{-0x1.9p+6, -2}
pownD	0.62	{-0x1.7ffffffffffffp+1, 3}
powr	0.82	{0x1.91237cp-1, 0x1.4da146p+8}
powrD	0.80	{0x1.2bfb4b18164c9p+65, -0x1.b78438ae9c3bdp-8}
remainder	0.00	{0x0p+0, 0x0p+0}
remainderD	0.00	{0x0p+0, 0x0p+0}
remquo	{ 0.00, 0}	0x0p+0
remquoD	{ 0.00, 0}	0x0p+0
rint	0.00	0x0p+0
rintD	0.00	0x0p+0
rootn	0.69	{-0x1.e2fe6ep-74, -141}
rootnD	0.68	{-0x1.8000000000001p+1, 3}
round	0.00	0x0p+0
roundD	0.00	0x0p+0
rsqrt	1.49	0x1.019566p+124
rsqrtD	1.49	0x1.01ffffffffa39p+1016
sin	2.48	-0x1.09f07ap+21
sinD	1.87	-0x1.f2fffffffffbap+32
sincos	{ 2.48, 2.37}	{0x1.09f07ap+21, 0x1.1338ccp+20}
sincosD	{ 1.87, 2.27}	{0x1.f2fffffffffbap+32, 0x1.d10000000074p+380}
sinh	2.32	0x1.e76078p+2
sinhD	1.53	-0x1.3100000000278p+4
sinpi	2.13	-0x1.45f3ep-9
sinpiD	2.50	-0x1.46000000000dap-7
sqrt	0.00	0x0p+0
sqrtD	0.00	0x0p+0
subtract	0.00	{0x0p+0, 0x0p+0}
subtractD	0.00	{0x0p+0, 0x0p+0}
tan	4.35	-0x1.b4eba2p+22
tanD	4.00	-0x1.2f000000003edp+333
tanh	1.18	-0x1.ca742ap-1
tanhD	1.19	0x1.f400000000395p-1
tanpi	4.21	-0x1.f99d16p-3
tanpiD	4.09	0x1.f6000000001d3p-3
trunc	0.00	0x0p+0
truncD	0.00	0x0p+0