Benchmarking The New Radeon R300 Register Allocator

[phoronix]

Phoronix: Benchmarking The New Radeon R300 Register Allocator

This morning I mentioned that work from last year's Google Summer of Code project on improving the ATI Radeon R300 compiler for the Mesa Gallium3D driver had never stopped. Tom Stellard's latest R300 GLSL compiler work has been focused on improving the register allocator. Tom's initial figures showed roughly a 10% boost for software with intensive OpenGL shaders, but here is what my test results show for this yet-to-be-merged code.

http://www.phoronix.com/vr.php?view=15852

[»John«]

Considering John Bridgman's claim that most driver optimizations that yield a real performance boost are supposed to be rather low-level, this is actually pretty cool! Good job, Tom and thanks a lot

[DanL]

All improvements are welcome, even if they're snagged by other bottlenecks on some cards. There's not a whole lot of data/benchmarking in this article, but it clearly shows a step in the right direction. Thank you, Tom.

[bug77]

Even for the same performance, less register allocations, less power used. So that's still a gain.

[V!NCENT]

These optimisations make me happy ^_^

Thanks!

[cb88]

meh... doesn't seem to be a great improvemet but every little bit helps