aboutsummaryrefslogtreecommitdiff
path: root/libpcsxcore
AgeCommit message (Collapse)Author
2011-10-30rcnt: remove gate guess for cnt 0 and 1notaz
at least 1 is wrong according to Ryphecha's info, and breaks at least Wipeout XL.
2011-10-30gte: change gte irq workaroundnotaz
This should be closer to what happens on the real thing. Special thanks to Ryphecha for the idea.
2011-10-30drc: detect dead gte reads toonotaz
2011-10-30drc: add some hack optionsnotaz
intended for caanoo/Wiz, some users want to trade glitches for performance boost.
2011-10-30gte: parametrize remaining opsnotaz
at least the ones from C source, it looks like there could be more.
2011-10-30gte: disable some MAC overflow checks for ARMnotaz
also add some missing ones
2011-10-30gte: eliminate excessive lims and castsnotaz
checking for overflow where it can't occur is a waste of time, get rid of that code.
2011-10-30gte: pass a pointer to registers as argumentnotaz
no more need to load psxRegs in each function and all registers are reachable by ldrh/strh.
2011-10-30drc: try to avoid full constant reloadnotaz
derive from existing ones if possible
2011-10-30drc: fix sub assem_debug printnotaz
2011-10-30drc: avoid invcode checks on close addessesnotaz
yes this makes it possible to miss something, but I hope it's not going to happen and will deal with it if it does.
2011-10-30gte_arm: some asm experimentsnotaz
also removes wrong (?) MAC0 saturation.
2011-10-30pcnt: measure gte toonotaz
2011-10-30drc: do MVMVA decoding; parametrize handlersnotaz
2011-10-30drc: finish GTE reg liveness analysisnotaz
2011-10-30drc: implement literal dupe mergingnotaz
doesn't seem to make much difference, but theoretically should help.
2011-10-30inline/parametrize rootcounter readsnotaz
makes rcnt1 hack impossible though, overclock PSX instead
2011-10-30psxcounters: avoid doing excessive updatesnotaz
this may cause missed interrupts in very rare cases, but performance penalty this makes is surely not worth it.
2011-10-16psxmem: set fixed flag on mappings that really need itnotaz
also fix unmap
2011-10-10psxcounters: change spu update ~2msnotaz
old value ~1.4 was weird anyway
2011-10-10drc: check for interrupts after RFEnotaz
this missing check wasn't causing too many problems because of events on every hsync, which are gone now.
2011-10-10new ability to adjust psx cpu clocknotaz
2011-10-10get rid of old memhandlersnotaz
2011-10-10drc: remove some old inline memhandlersnotaz
generic code handles it better now.
2011-10-10drc: implement memory access speculationnotaz
try to guess RAM mirror we are going to hit
2011-10-08drc: implement shiftimm constant propagationnotaz
it's rare, but it happens.
2011-10-08recompilation-caused frameskip workaroundnotaz
2011-10-08spu/cdda: sync cdda to spu playbacknotaz
this will probably make cdrom/vsync timing desync situation worse (those were never in sync anyway) but fix some cdda related sound stuttering. Those magic values are used trying to preserve other SPU plugin compatibility.
2011-10-08try to support more compilersnotaz
2011-10-08rewrite memhandlers (write)notaz
2011-10-04rewrite memhandlers (read)notaz
2011-10-02yet more random armv5 tweaksnotaz
2011-09-30gte_arm: implement RTPS, RTPTnotaz
2011-09-27psxcounters: avoid update on each hsyncnotaz
no useful work is done on hsync, no reason to interrupt the recompiler that often, set up events as needed.
2011-09-27drc: do gte flag liveness detectionnotaz
just copy-paste existing code and use _nf handlers from previous patch.
2011-09-27gte: split arm code for pollux, generate flagless handlersnotaz
flagless are unused but will be used later, at least I hope so.
2011-09-25gte: switch divider back to tablenotaz
pandora won't be affected because it uses VFP anyway, and pollux prefers the table for some reason. Also send the divider to it's own .c file for future reuse.
2011-09-25drc: fix PCSX HLE hack for armv5notaz
also adds some asserts I did during debugging this
2011-09-23support armv5 buildnotaz
2011-09-23fix build on some ARM toolchainsnotaz
R1 redefined..
2011-09-22drc: do modification check in smaller than page granularitynotaz
There are some games that keep writing in the same 4k page where the code resides (Alien Ressurection is one such example). I've noticed those accesses are usually to a small region, so keep range of addresses that has no code and has been recently accessed and check it when the writes come.
2011-09-20drc: make sure dyna_linker is really called from stubnotaz
in very rare cases add_link() would end up being called with jump to another block and not stub, which would later cause kill_pointer() to crash while cleaning jump_out. I'm guessing wrong pointer was being picked from jump_in by dyna_linker, failing to detect stale cache, and hoping this will fix it.
2011-09-19drc: load ra before SD only when needednotaz
there are far too many assumptions all over the place that it's not loaded early
2011-09-19drc: handle $ra access in JALR DS toonotaz
or at least try to..
2011-09-19drc: fix $ra evicting on JALnotaz
hopefully for good this time
2011-09-19Revert "drc: ujump DS $ra overwrite fix?"notaz
This reverts commit 2adcd6fad4594a18025b4f00c49e43a23f8f8992. That "fix" breaks Delta Force Urban Warfare, supposedly not neded after Ari64's fixes?
2011-08-13add guncon supportnotaz
a bit basic but works
2011-08-01drc: fix page get code in linkage_arm.snotaz
it did not match what was used elsewhere for PCSX, causing rare crashes due to making bad links.. Also comment get_pointer, it's not obvious what it does.
2011-07-26sio: only do IRQ if it's bit has been clearednotaz
tested on the real thing.
2011-07-15cdrom: partially implement attenuationnotaz
based on pcsrx code.