Closed GoogleCodeExporter closed 8 years ago
Hmm, lets keep those for now.
They shouldn't hurt performance much (prolly not at all) and having multiple
builds
is a pain :p
Original comment by ramapcsx2
on 4 Nov 2008 at 2:36
If we did do this, if we were #ifdef'ing the compiler features, it wouldn't be
that
difficult to do on the Linux side, just for reference.
In automake, I'd just need to tell it to grep /proc/cpuinfo for each feature,
and add
a define if it's there.
Original comment by arcum42@gmail.com
on 4 Nov 2008 at 2:48
Eh, yeah, having three separate builds is yuck.
There are some potentially better solutions. And yea, this is all code in the
recompiler so it's impact is never as serious as it might first seem. Although
the
VU micro does like to do a lot of recompilation (more than other units) so its
not
exactly something we want to write off entirely.
Assigned low priority for now.
Original comment by Jake.Stine
on 4 Nov 2008 at 3:36
Zeydlitz:
do to the way recompiling works, this check for SSE isn't that big of a deal.
basically it checks for SSE once, but doesn't recompile the SSE checking.
so next time the code is run, it doesn't check for SSE, and just runs the code
it
ran last time without having to check.
as a team, we just decided to get rid of SSE1 and SSE2 checks; since pretty
much all
processors that are going to be able to run PCSX2 at reasonable speed, should
support SSE and SSE2.
Original comment by cottonvibes
on 5 Nov 2008 at 7:07
No real problem.
Original comment by ramapcsx2
on 23 Nov 2008 at 2:02
Original issue reported on code.google.com by
Zeydl...@gmail.com
on 4 Nov 2008 at 2:01