[PATCH] [AArch64] Refines the Cortex-A57 Machine Model

[Andrew Trick]

> On Sep 26, 2014, at 11:00 AM, Dave Estes <cestes at codeaurora.org> wrote:
> 
> Excellent. Thanks, Jiangning. Your new numbers show a regressions in gap and equake. I'll try to get an equake number, but I do know that we're seeing ~1% gain. Interestingly enough one of the regressions that we're seeing is on twolf, but your device shows a gain. :) Despite the differences, I too think this latest patch looks like a good foundation for future work.
> 
> If I can get a fresh LGTM, I'll get it committed.

Some of the SPEC2k floating-point benchmarks are extremely sensitive to seemingly innocuous scheduler changes. equake and twolf in particular. Without low-level hardware profiling it could be difficult to know what causes a regression within the 10% range and determine whether it’s something the scheduler could account for. Often the issue is not something the scheduler can account for--code ends up landing at a different instruction address. Tweaking scheduler heuristics to compensate only sets yourself up for future regressions.

-Andy