[Mesa-users] osmesa state tracker
Burlen Loring
burlen.loring at gmail.com
Wed Oct 14 13:35:13 PDT 2015
Hi, A while back (9.2.2) I tested the osmesa gallium llvmpipe state
tracker and saw a big speed up compared to the older mesa. I went to
repeat the test on a shader based algorithm with osmesa in 11.0.0 and
found there was no speed up compared to osmesa in mesa 7.10. I was
expecting a big speed up, it takes 40 sec in mesa 7.10 and only 5 sec on
my gpu with nvidia driver.
Was there any change in how one should build or run the llvmpipe state
tracker?
Thanks
Burlen
here is runtime info:
GLVendor = VMware, Inc.
GLVersion = 3.0 Mesa 11.0.0
GLRenderer = Gallium 0.4 on llvmpipe (LLVM 3.4, 256 bits)
here is how I configured:
../mesa-11.0.0/configure CXXFLAGS="-DNDEBUG -O3 -mavx -ffast-math -Wall
-Wextra -Wno-unknown-pragmas -Wno-unused-local-typedefs
-fopt-info-optimized -fPIC -DDEFAULT_SOFTWARE_DEPTH_BITS=31"
CFLAGS="-DNDEBUG -O3 -mavx -ffast-math -Wall -Wextra
-Wno-unknown-pragmas -Wno-unused-local-typedefs -fopt-info-optimized
-fPIC -DDEFAULT_SOFTWARE_DEPTH_BITS=31" --enable-static=no
--enable-shared=yes --disable-xvmc --disable-glx --disable-dri
--with-dri-drivers= --with-gallium-drivers=swrast --enable-texture-float
--disable-shared-glapi --disable-egl --disable-gles1 --disable-gles2
--with-egl-platforms= --enable-gallium-osmesa --enable-gallium-llvm=yes
--prefix=/work/apps//mesa/11.0.0
and the table produced:
prefix: /work/apps//mesa/11.0.0
exec_prefix: ${prefix}
libdir: ${exec_prefix}/lib
includedir: ${prefix}/include
OpenGL: yes (ES1: no ES2: no)
OSMesa: libOSMesa (Gallium)
GLX: no
EGL: no
llvm: yes
llvm-config: /bin/llvm-config
llvm-version: 3.4.2
Gallium: yes
Shared libs: yes
Static libs: no
Shared-glapi: no
CFLAGS: -DNDEBUG -O3 -mavx -ffast-math -Wall -Wextra
-Wno-unknown-pragmas -Wno-unused-local-typedefs -fopt-info-optimized
-fPIC -DDEFAULT_SOFTWARE_DEPTH_BITS=31 -Wall -std=c99
-Werror=implicit-function-declaration -Werror=missing-prototypes
-fno-strict-aliasing -fno-math-errno -fno-trapping-math -fno-builtin-memcmp
CXXFLAGS: -DNDEBUG -O3 -mavx -ffast-math -Wall -Wextra
-Wno-unknown-pragmas -Wno-unused-local-typedefs -fopt-info-optimized
-fPIC -DDEFAULT_SOFTWARE_DEPTH_BITS=31 -Wall -fno-strict-aliasing
-fno-builtin-memcmp
Macros: -D__STDC_LIMIT_MACROS -D_GNU_SOURCE
-DUSE_SSE41 -DNDEBUG -DTEXTURE_FLOAT_ENABLED -DUSE_X86_64_ASM
-DHAVE_XLOCALE_H -DHAVE_SYS_SYSCTL_H -DHAVE_STRTOF -DHAVE_MKOSTEMP
-DHAVE_DLOPEN -DHAVE_POSIX_MEMALIGN -DMESA_EGL_NO_X11_HEADERS
-DHAVE_LLVM=0x0304 -DMESA_LLVM_VERSION_PATCH=2
LLVM_CFLAGS: -I/usr/include -D__STDC_CONSTANT_MACROS
-D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS
LLVM_CXXFLAGS: -I/usr/include -D__STDC_CONSTANT_MACROS
-D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -pipe
-Wp,-D_FORTIFY_SOURCE=2 -fexceptions --param=ssp-buffer-size=4
-grecord-gcc-switches -m64 -mtune=generic -std=c++11
LLVM_CPPFLAGS: -I/usr/include -D__STDC_CONSTANT_MACROS
-D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS
LLVM_LDFLAGS: -L/usr/lib64/llvm -lz -lpthread -lffi -ltinfo
-ldl -lm
PYTHON2: python2
Run 'make' to build Mesa
More information about the mesa-users
mailing list