2466 replies to this topic

[DaMadOne]

Sinetek, don't be sad! we have 3 versions of OSX running on 64bit AMD largely thanks to you! 64bit OSX on AMD seemed like a pipe dream not long ago. I know it seems like you/we have hit a brick wall with the nVidia stuff but as the old saying goes.. "Rome wasn't built in a day" but you sure did get A LOT done in a very short time. Look at it this way, as more and more people come to ML on AMD because of you, its only a matter of time before either you or someone else figures it out. It may just be that negative depressive thinking keeping you from figuring it out! So cheer up! You have plenty of people here ready and willing to help out in any way we can.

[Andy Vandijck]

Sinetek, on 10 January 2013 - 11:12 PM, said:

Ähm. So I haven't posted in a while. The nVidia problem has had me depressed because I took it WAY personal; still don't understand what's going on with it.
So thanks for understanding, I in the first place am mad at that problem. Or maybe it's teaching me humility.
over, out, sinetek

Don't take it personal.
We all have been looking into it and none of us have found a solution but I do have an hypothesis for the ATI and NVidia issues.
Someone reported that it worked fine on Bulldozer/Piledriver cpus.
I guess the delay is too long causing command buffer corruption.
We'll get this tackled, don't worry

[SS01]

Anyone tested trinity?

[Deltac0]

SS01, on 11 January 2013 - 12:08 PM, said:

Anyone tested trinity?

Shouldn't it work pretty much same like Piledriver? I think Trinity is based on Piledriver, but no idea about how the graphics would work...

[Mad Marcsen]

Did you mean http://www.insanelym...d/#entry1879250 and http://www.insanelym...s/#entry1879441 post from bitcore? He's running ML on AMD FX with Radeon and reported it as working.

So, if i understand this right, you mean that the ssse emluator introduces a delay which might causes the troubles?

[Andy Vandijck]

nanjuah, on 11 January 2013 - 01:36 PM, said:

Did you mean http://www.insanelym...d/#entry1879250 and http://www.insanelym...s/#entry1879441 post from bitcore? He's running ML on AMD FX with Radeon and reported it as working.

So, if i understand this right, you mean that the ssse emluator introduces a delay which might causes the troubles?

Yes I think that is the reason
The emulator itself works nicely...

[Mad Marcsen]

But how could one get around this? There is always a litte delay when he needs to calculate, or do you think that, if you decrease the delay time, it might work?

Or is it possible to find the ssse calls in the kext's and (don't know if this is possible, i'm ot a developer ) reroute them to sse?

[Andy Vandijck]

In Tiger times somebody once did something similar with the dvd player driver.
He patched the SSE3 functions to SSE2 ones to make the video playback go smooth so it is possible but not easy

[Mad Marcsen]

so, the kernel needs ssse3 too boot up the base system, right?

but what if someone made a "container", that catches all ssse3 calls and redirect them to sse3? is this possible?

is there no "easy" solution?

i have absolutely no clue about c. all i can do is a little bash scripting, so i'm not really helpfull here. but if i can do something, please let me know.

thanks.

[Andy Vandijck]

nanjuah, on 11 January 2013 - 04:28 PM, said:

so, the kernel needs ssse3 too boot up the base system, right?

but what if someone made a "container", that catches all ssse3 calls and redirect them to sse3? is this possible?

is there no "easy" solution?

i have absolutely no clue about c. all i can do is a little bash scripting, so i'm not really helpfull here. but if i can do something, please let me know.

thanks.

SSSE3 is nothing like SSE3
The only option I see is making software that replaces the functions, expands the __text page size and modifies instruction pointers to pull this off...

[IiTz ShAnE]

Andy Vandijck, on 11 January 2013 - 05:11 PM, said:

SSSE3 is nothing like SSE3
The only option I see is making software that replaces the functions, expands the __text page size and modifies instruction pointers to pull this off...

Is it not possible to modify the files needed? If a person gets ML installed but experiences graphics problems, they could install a package for the files.

[Andy Vandijck]

IiTz ShAnE, on 11 January 2013 - 06:52 PM, said:

Is it not possible to modify the files needed? If a person gets ML installed but experiences graphics problems, they could install a package for the files.

It is possible ofcourse but I'll have to look into it when I have a lot of time

[IiTz ShAnE]

Andy Vandijck, on 11 January 2013 - 07:13 PM, said:

It is possible ofcourse but I'll have to look into it when I have a lot of time

Of course. By all means take all the time you need. If I could learn all the code easily enough I would be helping lol

[Sinetek]

Mad Marcsen, on 11 January 2013 - 02:07 PM, said:

But how could one get around this? There is always a litte delay when he needs to calculate, or do you think that, if you decrease the delay time, it might work?

Or is it possible to find the ssse calls in the kext's and (don't know if this is possible, i'm ot a developer ) reroute them to sse?

Oh, I tried that already. I patched all the critical system binaries by hand to reroute the SSSE3 and removed the opemu. The system crashes in precisely the same manner.

Well, what I'm left with is the idea to try Mountain Lion DP1 and it's 32-bit kexts. I was going to try it, but the image I downloaded is corrupted :/

[Mad Marcsen]

ok,

tried to search the web for ati kext's from dp1 and found nothing. but i try to get hands on dp1

edit: maybe someone in the forums has got a dp1 and could upload a few kexts? would go faster

[IiTz ShAnE]

Sinetek, on 11 January 2013 - 08:04 PM, said:

Oh, I tried that already. I patched all the critical system binaries by hand to reroute the SSSE3 and removed the opemu. The system crashes in precisely the same manner.

Well, what I'm left with is the idea to try Mountain Lion DP1 and it's 32-bit kexts. I was going to try it, but the image I downloaded is corrupted :/

I've got a working DP1 dmg. I can upload kexts if needed.

[Mad Marcsen]

excellent

so what do we need? i don't know. could you make zip files? one with all the ati stuff, one with all the nvidia stuff. and what else do we need..... sinetek?!

edit: but can one run those kexts if they are 32bit or are they 32/64?

[instant idiot]

IiTz ShAnE, on 11 January 2013 - 08:39 PM, said:

I've got a working DP1 dmg. I can upload kexts if needed.

So do I, if it helps.

[IiTz ShAnE]

Here you go. Kexts are x86_64 and i386

http://www.mediafire.com/?p691gftr61pb2df

[Mad Marcsen]

thank you very much Iitz ShAnE

Downloading now. Let's see what happens