On Thursday, June 7, 2018 4:46:24 PM PDT Jason Ekstrand wrote: > On Thu, Jun 7, 2018 at 4:02 PM, Kenneth Graunke <kenn...@whitecape.org> > wrote: > > > We were enabling undefined memory checking for genxml values based on > > Valgrind being installed at build time, even for release builds. This > > generates piles and piles of assembly whenever you touch genxml. > > > > With gcc 7.3.1 and -O3 and -march=native on a Kabylake with Valgrind > > installed at build time: > > > > text data bss dec hex filename > > 5978385 262884 13488 6254757 5f70a5 libvulkan_intel.so > > 3799377 262884 13488 4075749 3e30e5 libvulkan_intel.so > > > > That's a 36% reduction in text size. > > --- > > src/intel/vulkan/anv_private.h | 2 ++ > > 1 file changed, 2 insertions(+) > > > > Eric, > > > > You probably will want to make the equivalent change in > > src/broadcom/cle/v3d_packet_helpers.h > > > > --Ken > > > > diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_ > > private.h > > index b47420f89ec..b035bf1c943 100644 > > --- a/src/intel/vulkan/anv_private.h > > +++ b/src/intel/vulkan/anv_private.h > > @@ -36,7 +36,9 @@ > > #include <valgrind.h> > > #include <memcheck.h> > > #define VG(x) x > > +#ifndef NDEBUG > > #define __gen_validate_value(x) VALGRIND_CHECK_MEM_IS_DEFINED(&(x), > > sizeof(x)) > > +#endif > > > > We could probably be fancier if we wanted to and disable all valgrind > checks in release builds but this should take care of the hot-paths and > leave us with at least something in release builds. > > Reviewed-by: Jason Ekstrand <ja...@jlekstrand.net>
Yeah, we could...I think we probably still want to record mappings so that valgrind works on people's applications (without reporting a million errors from the driver). But we don't need to do our own internal validation... --Ken
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev