[Mesa-dev] [PATCH 1/3] i965/fs: Fix off-by-one in saturate propagation.

[Matt Turner]

On Fri, Apr 4, 2014 at 5:32 PM, Ian Romanick <idr at freedesktop.org> wrote:
> Series is
>
> Reviewed-by: Ian Romanick <ian.d.romanick at intel.com>
>
> Two questions...
>
> Are there piglit tests that exhibit the problems fixed by this series?

Nope. It's the problem of figuring out how to write GLSL to generate
just the right set of instructions, which of course can be affected by
other changes to the compiler. ARB programs might be easier, but they
still have the same problems in general.

I went through about 50 Dota2 shaders that were affected by this
optimization and diff'd before and after this pass made progress to
find the problems fixed by the first and third patches.

Unit testing for the optimization passes is probably the best way to
go, like test_vec4_register_coalesce.cpp.

> Do any of these need to go back to stable? (I'm assuming not...)

Yeah, needs to go back to 10.1. I'll mark them appropriately.