[llvm-dev] Loop Unroll

[Florian Hahn via llvm-dev]

> On May 26, 2020, at 18:55, legend xx <legendaryxx7slh at gmail.com> wrote:
> 
> Awesome, thanks! 
> 
> Now I have another question. I have a matrix multiplication code. This is my code:
> 

On a related note, if you are looking to generate high-performance code for matrix operations on matrixes with known dimension you might be interested in Clang’s matrix types extension (https://clang.llvm.org/docs/MatrixTypes.html#matrixtypes <https://clang.llvm.org/docs/MatrixTypes.html#matrixtypes>). This should be less fragile with respect to relying on a number of passes to align for good performance. 

Please note that the implementation in Clang is still incomplete, but the outstanding patches to implement it (almost) completely can be found starting at https://reviews.llvm.org/D76791 <https://reviews.llvm.org/D76791> and the linked patches (‘stack’ in Phabricator terms).

Cheers,
Florian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200527/d6e6fde6/attachment.html>