> I haven't been following the 'double' work so dumb questions/comments: > > First, could you document the new opcodes in gallium/docs/source/tgsi.rst?
Yup already done in my tree. > > In the case of micro_u64seq(), etc. why doesn't it do > > dst->u64[0] = src[0].u64[0] == src[1].u64[0] ? ~0UL : 0U; > > Don't we want a 64-bit boolean result? I can imagine wanting to use the > result of a U64SEQ instruction in some subsequent bit-wise masking > instructions with 64-bit operands. A 32-bit result wouldn't work for that. No we've defined the *SEQ family to always return a boolean which is the 32bit integer. You can still use USEQ on these values without 64-bitness as they are stored internally as 2 x 32s to get a mask that would work. > > Finally, it seems this patch and 22/23 could be merged. This patch adds a > bunch of micro_*() functions that the compiler will warn as unused until > patch 22 is applied. Maybe the tgsi_exec.[ch] changes could be one patch > and the other tgsi changes in another. Not a big deal though. Oh I messed up my commits here, the tgsi_exec.c stuff should be in the later softpipe enable patch not here. Dave. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev