[Mesa-dev] MR: NIR: fsign optimizations and a couple 1-bit Boolean optimizations too.

[Ian Romanick]

https://gitlab.freedesktop.org/mesa/mesa/merge_requests/27

This series mostly improves code generation for cases of x*fsign(y).  I
had some difficulty getting everything working after rebasing on the
1-bit Boolean changes.  In that process, I noticed a couple things that
led to the last two patches.

shader-db results across the whole series for Skylake:

total instructions in shared programs: 14956605 -> 14941001 (-0.10%)
instructions in affected programs: 991239 -> 975635 (-1.57%)
helped: 4710
HURT: 0
helped stats (abs) min: 1 max: 221 x̄: 3.31 x̃: 2
helped stats (rel) min: 0.07% max: 10.00% x̄: 1.85% x̃: 1.38%
95% mean confidence interval for instructions value: -3.49 -3.14
95% mean confidence interval for instructions %-change: -1.89% -1.82%
Instructions are helped.

total cycles in shared programs: 359972474 -> 359757117 (-0.06%)
cycles in affected programs: 16235692 -> 16020335 (-1.33%)
helped: 4723
HURT: 473
helped stats (abs) min: 1 max: 2086 x̄: 50.91 x̃: 26
helped stats (rel) min: 0.01% max: 29.14% x̄: 2.81% x̃: 2.34%
HURT stats (abs)   min: 1 max: 1836 x̄: 53.03 x̃: 10
HURT stats (rel)   min: <.01% max: 34.36% x̄: 2.54% x̃: 0.57%
95% mean confidence interval for cycles value: -44.91 -37.98
95% mean confidence interval for cycles %-change: -2.40% -2.24%
Cycles are helped.

total spills in shared programs: 10056 -> 10055 (<.01%)
spills in affected programs: 54 -> 53 (-1.85%)
helped: 1
HURT: 0

total fills in shared programs: 21788 -> 21787 (<.01%)
fills in affected programs: 151 -> 150 (-0.66%)
helped: 1
HURT: 0