Hi guys,
This doesn't build on sparc64.
Also, this section in snappy-internal.h may need some review:
// Potentially unaligned loads and stores.
#if defined(__i386__) || defined(__x86_64__) || defined(__powerpc__)
Miod tells me we don't do unaligned accesses on powerpc.
Updated version attached fixes build on BE machines, the tests do
successfully complete on macppc even with unaligned loads and stores.
So I'm tempted to leave them enabled as doing the memcpy() to avoid
unaligned access has a 25-30% speed penalty on my Mac mini (but have
added a comment to the patch to make it easier to find if people
notice problems later).
Any progress in getting this in?
Best regards,
Piotr Sikora < piotr.sik...@frickle.com >
