Bug#701561: Please test ioquake3 on PowerPC, particularly without Altivec

[Simon McVittie]

On 04/06/13 13:58, thinkbrown at gmail.com wrote:
> One thing I'd suggest is that
> altivec is enabled based on the CPU itself, and not just PPC vs PPC64.

Sorry, I'm not going to do significant development on this myself, but
if you or another powerpc porter want to put a patchset together,
upstream would probably take it.  There seems to be some support for
runtime detection already, but AIUI, compiling the whole thing with
-maltivec means that's useless, because gcc is within its rights to emit
Altivec instructions anyway?

Another possibility would be to have a separate "lowest common
denominator" Q3 binary for PowerPC-without-Altivec, and have it
transparently exec the Altivec binary (with the same command line) if it
determines that we do actually have Altivec. It would probably need to
behave like its own mini-architecture within the Quake 3 engine, with a
separate suffix or directory for loadable modules (perhaps *ppc.so for
Altivec and *oldppc.so for non-Altivec?)

What I'm more concerned about is whether the PowerPC JIT for the Quake 3
bytecode language (code/qcommon/vm_powerpc*) emits Altivec opcodes - I
have no idea how I'd determine that. That's why I'm specifically asking
for testing with the bytecode game logic.

> Is Rage128 still a supported 3D accelerator? I know it'll be slow as
> heck, but it's the only card I have for my PowerMac G3s.

Your guess is as good as mine :-)

If all G3s are likely to be this low-spec, perhaps ioquake3/powerpc
should just continue to require Altivec, and put a note in xmessage(1)
on older machines...

    S