Hello Martyn! Thanks for the heads up. We have also seen a similar regression on arewefastyet.com for that revision. On x64 on the other hand we have seen an improvement on Sunspider, which was the intended effect. I'll look into this tomorrow and maybe will revert the change for ARM depending on what I'll find out.
Best regards, Michael On Mon, Apr 23, 2012 at 12:37, M Capewell <[email protected]>wrote: > Hi, > This patch is causing a performance regression on the Sunspider 3d-morph > benchmark. On a Cortex-A9 platform: > > r11407 1.39 > r11394 1.39 > r11393 1.40 > r11392 1.39 > r11391 1.39 > r11390 1.00 > r11387 baseline > > Cortex-A8 platforms show even larger regressions. -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
