[Mesa-dev] [PATCH v4 0/6] mesa/st: glsl_to_tgsi: improved temp-reg lifetime estimation

[Gert Wollny]

following the comments of Emil and Nicolai I've updated the patch set. 
Because v3 was borked I restarted from v2. 

Changes with respect to the version 2 are: 

- split the changes into more patches 
- correct formatting errors
- rename functions and methods to better clarify what they are used for
- remove unused methods and variables in prog_scope
- eliminate the class tgsi_temp_lifetime
- remove the use  the STL in the core library with one exception though: 
  since in st_glsl_to_tgsi.cpp std::sort is already used and its run-time 
  performance is significantly better than qsort. It is used in the register 
  rename mapping evaluation. It can be disabled by commenting out the define 
  USE_STL_SORT in st_glsl_to_tgsi_temprename.cpp. 
- add more tests and improve the life-time evaluation accordingly
- further reduce memory allocations
- no longer require C++11 for the core library code 
- the tests, however, make use of C++11 and the STL 

The algorithms is the same as described before, with the little exception that 
now initially a dry run over the instructions is used to count the numbers of 
scopes. The run-time overhead of this operation can be neglected.

In order to make it easier to transition to the new code and test it, I tied it 
in parallel to the old code. It can be enabled by setting the environment 
variable MESA_GLSL_TO_TGSI_NEW_MERGE. 

piglit run on the "shader" test set shows no regressions and fixes
spec at glsl-1.50@execution at variable-indexing@gs-input-array-vec2-index-rd

The performance numbers estimated by running the shader-db are given in the 
commit message of the last patch, the trend is the same like reported before,
i.e. the all-over run-times are a bit lower, mostly because the new evaluation 
for the mapping uses a binary search. However, because of the stocastical 
sampling measuring these numbers with perf borders at statistical 
insignificance, the influence on the all-over run-time is just too low. 

I've also run a few programs (GPUtest benchmarks, Unigine-valley, Unigine-heaven) 
and couldn't seen any visually indications that registers would be megrged 
wrongly.

Many thanks for any commenst, 
Gert 
** BLURB HERE ***

Gert Wollny (6):
  mesa/st: glsl_to_tgsi move some helper classes to  extra  files
  mesa/st: glsl_to_tgsi: implement new temporary  register lifetime
    tracker
  mesa/st: glsl_to_tgsi: add tests for the new temporary lifetime
    tracker
  mesa/st: glsl_to_tgsi: add register renamame mapping  evaluator
  mesa/st: glsl_to_tgsi: Add test set for evaluation of  rename mapping
  mesa/st: glsl_to_tgsi: tie in new temporary register  merge approach

 configure.ac                                       |    1 +
 src/mesa/Makefile.am                               |    2 +-
 src/mesa/Makefile.sources                          |    4 +
 src/mesa/state_tracker/st_glsl_to_tgsi.cpp         |  315 +-----
 src/mesa/state_tracker/st_glsl_to_tgsi_private.cpp |  207 ++++
 src/mesa/state_tracker/st_glsl_to_tgsi_private.h   |  165 +++
 .../state_tracker/st_glsl_to_tgsi_temprename.cpp   |  761 ++++++++++++++
 .../state_tracker/st_glsl_to_tgsi_temprename.h     |   36 +
 src/mesa/state_tracker/tests/Makefile.am           |   40 +
 .../tests/test_glsl_to_tgsi_lifetime.cpp           | 1070 ++++++++++++++++++++
 10 files changed, 2313 insertions(+), 288 deletions(-)
 create mode 100644 src/mesa/state_tracker/st_glsl_to_tgsi_private.cpp
 create mode 100644 src/mesa/state_tracker/st_glsl_to_tgsi_private.h
 create mode 100644 src/mesa/state_tracker/st_glsl_to_tgsi_temprename.cpp
 create mode 100644 src/mesa/state_tracker/st_glsl_to_tgsi_temprename.h
 create mode 100644 src/mesa/state_tracker/tests/Makefile.am
 create mode 100644 src/mesa/state_tracker/tests/test_glsl_to_tgsi_lifetime.cpp

-- 
2.13.0