Ben Ashbaugh

icon indicating an email [email protected]

icon location I work on GPUs for Intel, but the projects on my Github are my own. Interested in parallel and heterogeneous compute.

Results 189 issues of Ben Ashbaugh

Semaphore plus Barrier plus Out-of-Order Queue Test

See related OpenCL spec issue discussion: https://github.com/KhronosGroup/OpenCL-Docs/issues/1178#issuecomment-2148210553 We should have a targeted semaphore test that mixes semaphores, barriers, and out-of-order queues, since this is likely to be a common pattern...

mobica-backlog

Missing tests for work-item functions with out-of-range arguments

The OpenCL C [work-item functions](https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_C.html#work-item-functions) have defined results for all values of their _dimidx_ argument, even those that are out-of-range. For example, for **get_global_size**: > Valid values of dimindx are...

mobica-backlog

Thread dimensions optimization

2 comment

Implements the optimization described by #1397. This doesn't affect most of our devices with typical maximum work-group sizes, but for our FPGA emulation device with a very large maximum work-group...

Address review comments from Add negative test for test_svm size check

See comments: https://github.com/KhronosGroup/OpenCL-CTS/pull/1802#pullrequestreview-1917402959

Disable gcc-11 on macos

There is a compiler issue with __has_cpp_attribute that is causing these CI configs to fail, so disable these configs (for now, at least).

add queries for cl_intel_required_subgroup_size

add device queries for cl_intel_unified_shared_memory

add queries for cl_intel_command_queue_families

add queries for cl_intel_device_attribute_query

minimum values for CL_DEVICE_HALF_FP_CONFIG should not be tied to an OpenCL version

The current minimum value for [CL_DEVICE_HALF_FP_CONFIG](https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_API.html#CL_DEVICE_HALF_FP_CONFIG) has a dependency on OpenCL 2.0 or newer: > If half-precision is supported by the device, then the minimum half-precision floating-point capability for OpenCL...