56w ago - Shortly following the
Alpha 2 Module release, today
KDSBest has updated their PS3 / PC SPU emulation project which allows your x86 machine to emulate the PlayStation 3 SPU to Beta status.
Download:
PS3 SPU Emulation Beta Module
To quote: Let's call it beta (SPU emulation). Tested with my own elf a lot commands if they work properly. Disabled buttons depending on state.
And last but not least, if you press run a new thread is spawned. Now you are able to press pause and it stops the run ofc. The performance get a super boost with the threading.
Project Update: alpha2 - pre alpha so it is not in pre alpha stage anymore. It detects code modifications on the fly too. The new feature list is not complete that's my fault I guess.
Thanks for the news update. I need some infos from some reversers to create the next stunning feature:
Crypto dumping... dumps all crypto calls... Let's call it function parameter dumping cause it will work for everything and will implement a callback on known functions which will make dumping at certain calls possible.
I try to implement this and GET/PUT DMA commands this weekend. I think I can't finish it this weekend, but this will be the beta stage. V0.9 should then just implement the rest of the commands that anergistic implements.
After that it is unknown what I do and so on. Since then my reversing starts to get more focus.
Update: From
mr_wicked: I was trying to run an SPU module in anergistic with no success. In IBM systemsim (in combination with libspe2) my code ran just fine..
Finding a broken instruction in anergistic is like finding a needle in a haystack. So I ended up throwing IBM systemsim in IDA and exported all SPU instructions to .asm. After writing a small emulator around these instructions I was able to find two nasty errors. One in "andhi" and one in the "cgx" instruction. Here are the corrected Anergistic instructions:
00010101,ri10,andhi,half,signed
{
int i;
for (i = 0; i < 8; ++i)
rth[i] = rah[i] & i10;
}
01101000010,rr,cgx
{
int i;
for (i = 0; i < 4; ++i)
{
u64 r = (u64)(rtw[i] & 1) + (u64) raw[i] + (u64) rbw[i];
rtw[i] = (r >> 32) & 1;
}
} My application also needed rotqbybi and rotqbi so I implemented those as well:
00111001100,rr,rotqbybi
{
u32 shift_count = (rbw[0] & 0x7f) >> 3;
rtw[0] = raw[0];
rtw[1] = raw[1];
rtw[2] = raw[2];
rtw[3] = raw[3];
while (shift_count--)
{
rtw[0] = (rtw[0] << 8) | (rtw[1] >> 24);
rtw[1] = (rtw[1] << 8) | (rtw[2] >> 24);
rtw[2] = (rtw[2] << 8) | (rtw[3] >> 24);
rtw[3] = (rtw[3] << 8) | (rtw[0] >> 24);
}
}
00111011000,rr,rotqbi
{
u32 shift_count = (rbw[0] & 0x7f) & 7;
rtw[0] = raw[0];
rtw[1] = raw[1];
rtw[2] = raw[2];
rtw[3] = raw[3];
while (shift_count--)
{
u32 t = (rtw[1] >> 31) | 2 * rtw[0];
rtw[1] = (rtw[2] >> 31) | 2 * rtw[1];
rtw[2] = (rtw[3] >> 31) | 2 * rtw[2];
rtw[3] = (rtw[0] >> 31) | 2 * rtw[3];
rtw[0] = t;
}
} Although trivial to fix it was really nasty to find these. I couldn't find a repository with active development, so I'm posting it here.
22 Comments - Go to Forum Thread »
, great development !
Thank you, rodq. I'll try the update and see what happens.
Also, for everyone else, it seems this is only happening on 500GB+ drives, or maybe larger...
