I'm guessing that the pipeline supports late-forwarding of the accumulate
operand, meaning that the instruction issues when the multiply operands are
available, regardless of the availability of the accumulate operands. The
accumulate operands can come in late, which means that there is more
flexibility in scheduling.
On Wed, Dec 5, 2012 at 10:23 AM, Sebastien DELDON-GNB <
sebastien.deldon at st.com> wrote:
> Hi all,****
>
> ** **
>
> Can someone explain me why is vmlx forwarding option enabled for
> cortex-a9 ? and what the purpose of it ?****
>
> ** **
>
> Best Regards****
>
> Seb****
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20121205/4a87421e/attachment.html>