- Jul 31, 2025
-
-
Mark Horvath authored
-
- Jul 30, 2025
-
-
Mark Horvath authored
-
Mark Horvath authored
As the library avoids branching as much as possible there is no reason to share functions in the binary between the SVE2 and SME/SME2 paths. That kind of sharing is made possible by the __arm_streaming_compatible attribute. In addition, functions using SME2 instructions must be attributed with the __arm_streaming (or __arm_locally_streaming) keyword.
-
Mark Horvath authored
-
- Jul 28, 2025
-
-
Mark Horvath authored
Specify CMake's 'Release' build type on CI for the coverage runs to not to cause coverage loss for asserts, as in most of the cases those assert paths cannot be driven through the API tests. Update the list of long vector tests for the local coverage script.
-
- Jul 25, 2025
-
-
Mark Horvath authored
* Make it work with SME2 sources. * Print output to disasm.txt as well. * Demangle C++ symbols.
-
- Jul 23, 2025
-
-
Denes Tarjan authored
-
Daniil Kulikov authored
-
- Jul 11, 2025
-
-
Noureldin Abdelfattah authored
-
- Jul 10, 2025
-
-
Mark Horvath authored
-
Mark Horvath authored
So far all SME2 implementations only used the SME feature set, so it makes sense to rename it to just SME and compile it for the SME extension. That way future implementations using SME2 features can be distinguished. Changes related to filenames are part of a separate commit.
-
Noureldin Abdelfattah authored
-
- Jul 08, 2025
-
-
Mark Horvath authored
Add an example for extracting one operation from KleidiCV as well.
-
Mark Horvath authored
As it is known at compilation time whether the streaming or non-streaming variants should be used. In some environments it causes issues if the non-streaming routines are called from streaming-compatible functions. (The correct behaviour would be that the compiler automatically links to the streaming variants in such a case.)
-
- Jul 04, 2025
-
-
Denes Tarjan authored
-
- Jul 03, 2025
-
-
- Jul 01, 2025
-
-
Mark Horvath authored
-
Mark Horvath authored
-
Mark Horvath authored
-
- Jun 30, 2025
-
-
Noureldin Abdelfattah authored
-
- Jun 26, 2025
-
-
Denes Tarjan authored
-
Denes Tarjan authored
Gaussian Blur 15x15 kernels always use the custom sigma variant, with much better performance and near the same accuracy. Custom Sigma kernels are uniformized and simplified using std::reference_wrapper in the SVE variant.
-
- Jun 23, 2025
-
-
Noureldin Abdelfattah authored
-
- Jun 10, 2025
-
-
Mark Horvath authored
-
Mark Horvath authored
-
- Jun 04, 2025
-
-
Noureldin Abdelfattah authored
-
Mark Horvath authored
-
- Jun 03, 2025
-
-
Mark Horvath authored
-
- Jun 02, 2025
-
-
Noureldin Abdelfattah authored
-
Denes Tarjan authored
-
Noureldin Abdelfattah authored
-
Mark Horvath authored
-
Mark Horvath authored
Update documentation with oldest supported GCC version as well.
-
- May 30, 2025
-
-
Mark Horvath authored
-
Mark Horvath authored
-
Noureldin Abdelfattah authored
-
Mark Horvath authored
Benchmarks added for Gaussian blur 21x21 and Medan blur benchmarks are extended.
-
Mark Horvath authored
-