I tested this with uploading 1024x1024 656 textures in a loop for 10 seconds. With glTexImage2D on SNB I get 17% better performance, mobile IVB (interestingly only) 0..1% better performance and BDW 3% better performance. For all these tests Mesa was compiled with -O2 -march=native and no Piglit regressions.
/Juha-Pekka Juha-Pekka Heikkila (3): configure.ac: Add detection for sse2 compilation support mesa/main: Add sse2 streaming clamping mesa/main: Clamp rgba with streamed sse configure.ac | 7 +++ src/mesa/Makefile.am | 8 +++ src/mesa/main/pixeltransfer.c | 62 +++++++++++++------ src/mesa/main/sse2_clamping.c | 138 ++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/sse2_clamping.h | 49 +++++++++++++++ 5 files changed, 245 insertions(+), 19 deletions(-) create mode 100644 src/mesa/main/sse2_clamping.c create mode 100644 src/mesa/main/sse2_clamping.h -- 1.8.5.1 _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
