Jason Ekstrand <[email protected]> writes: > On Jul 28, 2015 2:43 AM, "Francisco Jerez" <[email protected]> wrote: >> >> register_coalesce() was considering the exec_size of the MOV >> instruction alone to decide whether the register at offset+1 of the >> source VGRF was being copied to inst->dst.reg_offset+1 of the >> destination VGRF, which is only a valid assumption if the move has a >> 32-bit execution type. Use regs_read() instead to find out the number >> of registers copied by the instruction. >> --- >> src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp > b/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp >> index a253c81..4544122 100644 >> --- a/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp >> +++ b/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp >> @@ -195,7 +195,7 @@ fs_visitor::register_coalesce() >> continue; >> } >> reg_to_offset[offset] = inst->dst.reg_offset; >> - if (inst->exec_size == 16) >> + if (inst->regs_read(0) > 1) >> reg_to_offset[offset + 1] = inst->dst.reg_offset + 1; >> mov[offset] = inst; >> channels_remaining -= inst->regs_written; > > And then we use regs_written directly below... > > Off you want it to be more consistent, we could use regs_written for both. > Either wary, > > Reviewed-by: Jason Ekstrand <[email protected]> > Sure, thanks.
>> -- >> 2.4.6 >>
signature.asc
Description: PGP signature
_______________________________________________ mesa-dev mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/mesa-dev
