[Piglit] [PATCH] OpenCL: Add a custom test case for build-in function get_work_dim.

Aaron Watry awatry at gmail.com
Tue Jun 4 06:17:39 PDT 2013


Sounds good.  I've pushed the simplified test case.

--Aaron

On Sun, Jun 2, 2013 at 8:40 PM, Sun, Yi <yi.sun at intel.com> wrote:
> Hi Aaron,
>
> Thanks you for your review and comments.
> Actually, I like the test cases you provided, and I prefer it, a pretty
> good case. :)
>
> Thanks
>   --Sun, Yi
>
> On Thu, 2013-05-30 at 11:48 -0500, Aaron Watry wrote:
>> I'm not really sure if I like this approach.  Couldn't this be tested
>> with the attached test case instead?  Note that I can't actually
>> execute that test case currently, since I'm getting an error from the
>> R600 back-end on the machines that I have available to me, and my
>> Nvidia machine is not with me at the moment.
>>
>> If we do have to go with this approach, I have the following comments:
>>
>> On Thu, May 30, 2013 at 3:06 AM, Yi Sun <yi.sun at intel.com> wrote:
>> > Invoked function get_work_dim in OpenCL, and read back the return value,
>> > then validate it with the value which passed to the function clEnqueueNDRangeKernel.
>> >
>> > Signed-off-by: Yi Sun <yi.sun at intel.com>
>> > ---
>> >  tests/all_cl.tests                |    3 +-
>> >  tests/cl/custom/CMakeLists.cl.txt |    1 +
>> >  tests/cl/custom/get_work_dim.c    |   88 +++++++++++++++++++++++++++++++++++++
>> >  3 files changed, 91 insertions(+), 1 deletions(-)
>> >  create mode 100644 tests/cl/custom/get_work_dim.c
>> >
>> > diff --git a/tests/all_cl.tests b/tests/all_cl.tests
>> > index 8cefa43..b0b950f 100644
>> > --- a/tests/all_cl.tests
>> > +++ b/tests/all_cl.tests
>> > @@ -38,6 +38,7 @@ profile.tests['Program'] = program
>> >
>> >  # Custom
>> >  add_plain_test(custom, 'Run simple kernel', ['cl-custom-run-simple-kernel'])
>> > +add_plain_test(custom, 'Buildin function get_work_dim', ['cl-custom-get_work_dim'])
>>
>> Buildin should be changed to Builtin (here and everywhere else that
>> uses this spelling).
>>
>> >
>> >  # API
>> >  #  Platform
>> > diff --git a/tests/cl/custom/CMakeLists.cl.txt b/tests/cl/custom/CMakeLists.cl.txt
>> > index 4f93e2a..9f8d5cb 100644
>> > --- a/tests/cl/custom/CMakeLists.cl.txt
>> > +++ b/tests/cl/custom/CMakeLists.cl.txt
>> > @@ -1 +1,2 @@
>> >  piglit_cl_add_custom_test (run-simple-kernel run-simple-kernel.c)
>> > +piglit_cl_add_custom_test (get_work_dim get_work_dim.c)
>> > diff --git a/tests/cl/custom/get_work_dim.c b/tests/cl/custom/get_work_dim.c
>> > new file mode 100644
>> > index 0000000..583b5a4
>> > --- /dev/null
>> > +++ b/tests/cl/custom/get_work_dim.c
>> > @@ -0,0 +1,88 @@
>> > +/*
>> > + * Copyright © 2013 Yi Sun <yi.sun at intel.com>
>> > + *
>> > + * Permission is hereby granted, free of charge, to any person obtaining a
>> > + * copy of this software and associated documentation files (the "Software"),
>> > + * to deal in the Software without restriction, including without limitation
>> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
>> > + * and/or sell copies of the Software, and to permit persons to whom the
>> > + * Software is furnished to do so, subject to the following conditions:
>> > + *
>> > + * The above copyright notice and this permission notice (including the next
>> > + * paragraph) shall be included in all copies or substantial portions of the
>> > + * Software.
>> > + *
>> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
>> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
>> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
>> > + * DEALINGS IN THE SOFTWARE.
>> > + */
>> > +
>> > +#include "piglit-framework-cl-custom.h"
>> > +
>> > +PIGLIT_CL_CUSTOM_TEST_CONFIG_BEGIN
>> > +
>> > +       config.name = "Buildin Function get_work_dim";
>> > +       config.run_per_device = true;
>> > +
>> > +PIGLIT_CL_CUSTOM_TEST_CONFIG_END
>> > +
>> > +//char* source = "kernel void test(global int* out){ *out = -1; }";
>>
>> Remove this commented out source line
>>
>> > +char* source ="                                          \n"\
>> > +"kernel void buildin_work_dim( __global int *ret ) {     \n"\
>> > +"  *ret = get_work_dim();                                \n"\
>> > +"}                                                       \n"\
>> > +"\n";
>> > +
>> > +enum piglit_result
>> > +piglit_cl_test(const int argc,
>> > +               const char** argv,
>> > +               const struct piglit_cl_custom_test_config* config,
>> > +               const struct piglit_cl_custom_test_env* env)
>> > +{
>> > +       enum piglit_result result = PIGLIT_PASS;
>> > +
>> > +       size_t global_size[] = {256, 256};
>> > +       size_t local_size[] = {16,16};
>>
>> This won't work on all valid OpenCL implementations.  local_size
>> should be changed to {1,1}, or you should use the
>> clGetKernelWorkGroupInfo API function to get the maximum local size
>> and then make sure that 16x16 fits within those bounds.
>>
>> I think that one example where this won't work would be the Apple
>> MacOS 10.6/10.7 CPU-based CL implementation...  I think that this
>> implementation defined the max local size to be either 1 or the number
>> of CPU cores available.
>>
>> > +       int i;
>> > +       cl_int data = 0;
>> > +
>> > +       piglit_cl_context context = NULL;
>> > +       cl_mem buffer = NULL;
>> > +       cl_program program = NULL;
>> > +       cl_kernel kernel = NULL;
>> > +
>> > +       /* Create objects up to the kernel */
>> > +       context = piglit_cl_create_context(env->platform_id, &env->device_id, 1);
>> > +       buffer = piglit_cl_create_buffer(context, CL_MEM_READ_WRITE, sizeof(cl_int));
>> > +       program = piglit_cl_build_program_with_source(context, 1, &source, "");
>> > +       kernel = piglit_cl_create_kernel(program, "buildin_work_dim");
>> > +
>> > +       /* Set kernel arguments and run the kernel */
>> > +       piglit_cl_set_kernel_buffer_arg(kernel, 0, &buffer);
>> > +
>> > +       for( i=1; i <= 2; i++ )
>> > +       {
>>
>> Why are we only testing up to two dimensions?  Why not test all 3?
>>
>> --Aaron
>>
>> > +               // Run the kernel
>> > +               piglit_cl_execute_ND_range_kernel(context->command_queues[0], kernel, i, global_size, local_size);
>> > +
>> > +               /* Read the buffer and check the result */
>> > +               piglit_cl_read_buffer(context->command_queues[0], buffer, 0, sizeof(cl_int), &data);
>> > +
>> > +               if(data != i) {
>> > +                       fprintf(stderr, "get_work_dim return wrong value(%d), should be %d!.\n", data, i);
>> > +                       result = PIGLIT_FAIL;
>> > +               }
>> > +       }
>> > +
>> > +       /* Release resources */
>> > +       clReleaseKernel(kernel);
>> > +       clReleaseProgram(program);
>> > +       clReleaseMemObject(buffer);
>> > +       piglit_cl_release_context(context);
>> > +
>> > +       return result;
>> > +}
>> > --
>> > 1.7.6.4
>> >
>> > _______________________________________________
>> > Piglit mailing list
>> > Piglit at lists.freedesktop.org
>> > http://lists.freedesktop.org/mailman/listinfo/piglit
>


More information about the Piglit mailing list