Did you see what happened today on the CVS ? I've received this from PearPC dev mail-list :
from Jens von der Heydt <heydt@pietzpluswild.de>
to Pearpc-users@lists.sourceforge.net,
date Mon, Mar 31, 2008 at 1:38 PM
subject [Pearpc-users] I created an opcode-tester for pearpc
mailing list <pearpc-users.lists.sourceforge.net> Filter messages from this mailing list
Hi there,
ok, the subject isn't actually true since I used Gwenole Beauchesne's code for Kheperix and integraded it into the latest cvs (x86_64 only) to test PearPC's quality and stability against a large number of test results gathered on a G4 cpu.
It's only working for the jitc right now because that was what actually intrested me to do but it could easily be applied to the generic branch as well.
I basically installed a new mutex / semaphore to the jitc so that I could run the tests. The original code was more advanced and compatible to qemu and some other emulators but wasn't really usefull to implement
it into pearpc.
I installed a new configure - switch (--enable-opcodemode=yes) and brutaly #if-ed some init - stuff so that pearpc wouldn't do more than was needed to get the opcode tester running. It's in a quite useable state right now and maybe somebody can use it to optimize the ways pearpc handles G4 - code (some few missing opcode implementations here and there) and altivec missing in latest branch.
The Testfile is quite large so I uploaded the whole thing to : http://www.vdh-webservice.de/ppc_opcodetest.zip
Just configure and build with --enable-opcodemode=yes . When you run it, please copy the testresults file from /TESTFILE into the DIR where you run pearpc form. Currently this is strictly linux -only.
Please test and reply,
Jens
Cool stuff, not ?
Thanks to all contributors !
