[PATCH] D36130: [SLP] Vectorize jumbled memory loads.

[Evgeny Astigeevich via llvm-commits]

+llvm-commits

On 15/12/2017, 17:25, "Evgeny Astigeevich" <Evgeny.Astigeevich at arm.com> wrote:

    Hi Mohammad,
    
    I created a problem report: https://bugs.llvm.org/show_bug.cgi?id=35673
    
    I think the patch has the following issues:
    
    1. No assessment of performance impact on other targets. 
    2. No tests for other targets. I see it's tested only for x86.
    3. In the patch I don't see any checks if the optimization is profitable.
    
    What if we revert the patch to fix these issues?
    
    Thanks,
    Evgeny Astigeevich
    
    On 14/12/2017, 13:32, "Evgeny Astigeevich via Phabricator" <reviews at reviews.llvm.org> wrote:
    
        eastig added a comment.
        
        Hi Shahid,
        
        These changes caused 27.7% and 30.2% regressions on an AArch64 Juno board (http://lnt.llvm.org/db_default/v4/nts/83681):
        
        MultiSource/Benchmarks/mediabench/gsm/toast/toast: 30.20%
        MultiSource/Benchmarks/MiBench/telecomm-gsm/telecomm-gsm: 27.73%
        
        We have the same benchmarks regressed on our AArch64 boards (Cortex-A53, Cortex-A57).
        
        -Evgeny Astigeevich
        The ARM Compiler Optimisation team
        
        
        https://reviews.llvm.org/D36130