[Mesa-dev] [PATCH 00/28] update swr rasterizer

[Emil Velikov]

Hi Tim,

On 16 March 2017 at 00:12, Tim Rowley <timothy.o.rowley at intel.com> wrote:
> Hightlights include: lots of simd16 work, assert rework, and autogen
> changes (scripts centralized, one file added, two removed).
>
> Tim Rowley (28):
>   swr: [rasterizer core] Support sparse numa id values on all OSes
>   swr: [rasterizer core] Finish SIMD16 PA OPT except tesselation
>   swr: [rasterizer core] Finish SIMD16 PA OPT including tesselation
>   swr: [rasterizer core/scripts] Autogen backend initialization
>     function(s)
>   swr: [rasterizer archrast] Add additional API events
>   swr: [rasterizer core] Implement SIMD16 GS and STREAMOUT
>   swr: [rasterizer archrast] Fix performance issue with archrast stats
>   swr: [rasterizer common] Add InterpolateComponentFlat utility
>   swr: [rasterizer core] Fix RECT_LIST primitive assembly
>   swr: [rasterizer archrast/scripts] Further archrast cleanups
>   swr: [rasterizer archrast] Remove redundant data from archrast files
>   swr: [rasterizer archrast/core/scripts] Fix archrast multithreading
>     issue
>   swr: [rasterizer core] Implement double pumped SIMD16 TESS
>   swr: [rasterizer archrast] Fix the early and late depthstencil events
>   swr: [rasterizer] Backend code adjustments
>   swr: [rasterizer] Slight assert refactoring
>   swr: [rasterizer core] Allow no arguments to SWR_INVALID macro
>   swr: [rasterizer core/common] Fix the native AVX512 build under ICC
>   swr: [rasterizer core] Fix typo in SIMD16 code path
>   swr: [rasterizer] Convert more SWR_ASSERT(false, ...) to
>     SWR_INVALID(...)
>   swr: [rasterizer jitter] Fix LogicOp blend jit after assert changes
>   swr: [rasterizer core] SIMD16 Frontend WIP - fix tesselation crashes
>   swr: [rasterizer core] fix trifan regression from 52f9f54dce
>   swr: [rasterizer scripts] Put codegen scripts into a separate
>     directory
>   swr: [rasterizer codegen] Quiet gen_backends.py execution
>   swr: [rasterizer codegen] Rewrite gen_llvm_ir_macros.py to use mako
>   swr: [rasterizer codegen] Fix generation of knobs
>   swr: [rasterizer codegen] Rewrite gen_llvm_types.py to use mako
>
Above all - thanks for omitting the execute bit/shebang for the new
python script !

A few high-level comments/suggestions.

 - seems like this series will break the scons build
You really do _not_ want to do that, please ensure that things build
[always] with both scons and autoconf.
 - there's at least one patch that mixes how licence text is commented
alongside other changes
Please keep that to a separate patch ?
 - some variables (gBackendPixelRateTable comes to mind) are declared
in various places, and during the series not all instances are updated
at once
Just move them to a header, in a prep patch ?
 - renaming actual mako templates to .cpp/.hpp is confusing/misleading
Worth leaving as-is and/or checking how other generators in-tree name
such files ?

I would strongly urge you to address the first issue, but at the end
of the day it's wishful thinking.

Thanks
Emil