On Fri, Sep 15, 2017 at 6:39 PM, Jason Ekstrand <ja...@jlekstrand.net> wrote:
> From: Connor Abbott <cwabbo...@gmail.com> > > v2 (Jason Ekstrand): > - Various whitespace cleanups > - Add helpers for reading/writing objects > - Rework derefs > - [de]serialize nir_shader::num_* > - Fix uses of blob_reserve_bytes > - Use a bitfield struct for packing tex_instr data > --- > src/compiler/Makefile.sources | 2 + > src/compiler/nir/nir_serialize.c | 1195 ++++++++++++++++++++++++++++++ > ++++++++ > src/compiler/nir/nir_serialize.h | 35 ++ > 3 files changed, 1232 insertions(+) > create mode 100644 src/compiler/nir/nir_serialize.c > create mode 100644 src/compiler/nir/nir_serialize.h > I didn't send an R-B with this patch because, to be honest, I didn't actually review it properly. I just fixed it to the point where it's working. The one issue that I think should probably still be considered is error checking. This pass does almost none beyond a couple of asserts. It doesn't even check that you didn't overrun the end of the blob. How robust does it need to be against tampering? For an on-disk shader cache, we're probably safe so long as we have some sort of checksum in the file. For program_binary, do we want to be a bit safer?
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
