[Mesa-dev] [PATCH 00/12] RFC: Combine shader input/output enums.

Paul Berry stereotype441 at gmail.com
Mon Mar 11 15:51:04 PDT 2013


This patch series combines the enums gl_vert_result, gl_geom_attrib,
gl_geom_result, and gl_frag_attrib into a single enum.

These four enums serve very similar purposes: they all identify data
that flows through the pipeline from vertex shader to fragment shader,
but their definitions don't match.  For example, the enum value
representing CLIP_DIST0 is:

- 17 in gl_vert_result,
- 9 in gl_geom_attrib,
- 16 in gl_geom_result, and
- 14 in gl_frag_attrib.

This isn't so much of a problem for Gallium drivers, since
src/mesa/state_tracker/st_program.c translates these enums into
TGSI_SEMANTIC_* values and indices, which are common to all pipeline
stages.  But it would be a major obstacle to implementing geometry
shaders in the i965 driver.  And IMHO it's a source of unnecessary
confusion in Mesa core which is only going to get worse as we add more
pipeline stages.

This patch series replaces all four enums with a single enum called
gl_varying_slot, and updates all the code that refers to those enums.
The new enum is slightly larger than any of the enums that it
replaces, since some of the varying slots aren't used by all pipeline
stages.

Patch 01 changes some of the bitfields that are used to refer to
gl_frag_attrib from 32-bit to 64-bit, so that they won't run out of
bits once the change is made.

Patch 02 adds the new enum (and bitfield macros that refer to it).

Patches 03-12 replace the old enums with the new enum.  To make code
review easier, and to ease future bisections, I've tried to break this
down into (a) patches that renumber the old enums to match the new
enum and (b) search-and-replace patches that replace uses of the old
enum with the new enum.  The former should be easy to code review
(since they just affect the definition of the enums), and the latter
should be less likely to introduce regressions (since they are simple
search-and-replace patches).

This series has been tested on i965 Gen5-7, swrast, and Gallium
softpipe.

Note: patches 04 and 12 are very large, so for purposes of mailing
list discussion I'm going to trim them down to representative example
hunks.  The complete series can be found in branch
"combine_varying_slot_enums" of
git://github.com/stereotype441/mesa.git.


[PATCH 01/12] i965: Change fragment input related bitfields to 64-bit.
[PATCH 02/12] mtypes.h: Add new gl_varying_slot enum, and bitfield defines.
[PATCH 03/12] mtypes.h: Modify gl_vert_result to refer to new gl_varying_slot enum.
[PATCH 04/12] Replace gl_vert_result enum with gl_varying_slot.
[PATCH 05/12] mtypes.h: Modify gl_geom_attrib to refer to new gl_varying_slot enum.
[PATCH 06/12] Replace gl_geom_attrib enum with gl_varying_slot.
[PATCH 07/12] mtypes.h: Modify gl_geom_result to refer to new gl_varying_slot enum.
[PATCH 08/12] Replace gl_geom_result enum with gl_varying_slot.
[PATCH 09/12] mtypes.h: Modify gl_frag_attrib to refer to new gl_varying_slot enum.
[PATCH 10/12] Get rid of _mesa_vert_result_to_frag_attrib().
[PATCH 11/12] Get rid of _mesa_frag_attrib_to_vert_result().
[PATCH 12/12] Replace gl_frag_attrib enum with gl_varying_slot.


More information about the mesa-dev mailing list