SPU improvements [test only ASMJIT] #4504
Conversation
|
Working as expected, then. |
|
got a 8-10% speed increase on unnamed5 on my crapwell processor [not using tsx] nice job catto |
|
GOW 3 |
|
The Orange Box regressed with SPU ASMJIT.
|
Nekotekina
force-pushed
the
master
branch
4 times, most recently
from
0b6d639
to
b23b7ae
Compare
|
Drakengard 3 crash on ASMJIT now for the ingame but boot is good, ni no kuni has huge problem on rendering also. |
| @@ -106,9 +106,6 @@ | |||
| </item> | |||
| <item> | |||
| <widget class="QRadioButton" name="spu_llvm"> | |||
There was a problem hiding this comment.
the checkbox needs an update for the tooltip in the json file
|
Great stuff, but nothing works for now, and ASMJit crashes the emu for almost every game i have |
|
I also verify what @Xcedf said... For every game that I've tested the emu crashes and those who do not crash the emulator give a memory access reading violation |
|
Can also confirm +10-15% improvement in unnamed 5 with my crapwell processor. It's also good to be able to select spu llvm, even if it does nothing for now lol. |
|
Get better test cases as a start, not RDR |
|
Exactly because it doesn't matter is why you should get better test scenarios, not some random difficult to debug game like RDR, otherwise devs will be wasting time by debugging it where they could be debugging something less complex. Removed spam from the PR, avoid off-topic and duplicate posts when the previous one can just be edited. This is going to be a PR with a lot of comments, so avoid useless posts. |
|
I would like to see logs and exact description/screenshots of said ASMJIT crash. |
|
@AniLeo i don't think it's RDR specific @Nekotekina crashes look like this |
|
My Crash looks like this
and the log file is here https://drive.google.com/open?id=1eZL39L9gpAnyImmwosWnXwDMOK1PtwN0 When booting Heavy Rain, the emulator does not crash but gives the following error: F {PPU[0x1000000] Thread (main_thread) [0x0135e384]} MEM: Access violation reading location 0x0 |
|
Hopefully fixed |
|
Confirm, all my cases are fixed, and it seems some games became a little faster than before compare to master, thanks a lot |
|
That was quick, the issue has been resolved and Uncharted works. |
|
Uncharted Trap fixed, Sonic Unleashed regression fixed |
|
Sorry for GoW Ascension false information game is nothing to do with Experimental commit, i just ran it on build full of hacks, to work on normal build it requires WCB+CpuBlit |
Nekotekina
force-pushed
the
master
branch
2 times, most recently
from
2424a78
to
4176b5f
Compare
Remove SPU thread reference from spu_recompiler_base Disable support for far jumps in pathpoints (they were rare and unsafe)
Don't fail if already exists
It was strangely removed...
|
Killzone 3 Demo also ingame, but looks a way better |
Shared between ASMJIT/LLVM recompilers, compiled at startup
Implemented utils::memory_release (not used) Implemented utils::shm class (handler for shared memory) Improved sys_mmapper syscalls Rewritten ppu_patch function Implemented vm::get_super_ptr (ignores memory protection) Minimal allocation alignment increased to 0x10000
|
@Nekotekina thanks this is noticeable speed this fix Walking Dead: michonne sound stutting also do good performance on a lot of games |
|
Borderlands now freezes at startup with this series of commits, throwing the following: F {PPU[0x1000000] Thread (main_thread) [0x01001640]} St13runtime_error thrown: Memory mapping failed - blame Windows (addr=0x25f80000, size=0x20000, flags=0x23) |
|
This got killzone 2's skimirish mode to load in-game for the first time ever, but there's imcompatible formats that are causing it to have graphical bugs, i've put it in the issues. Campaign mode still has same pink text error, and doesn't load in-game. but on the new SPU it has a different pink text error. making the formats compatible could fix the graphical issues (possibly) and probably the same on killzone 1 and 3, who knows. |
No description provided.