gcc 4?
- Neil Stevens
- Posts: 79
- Joined: Thu Jan 27, 2005 2:22 pm
- Location: California
- Contact:
You can find the info about gcc 4 here:
http://gcc.gnu.org/gcc-4.0/changes.html
http://gcc.gnu.org/gcc-4.0/changes.html
I guess gcc 3.2.3 or 3.3 or whatever the number was has a new mips core that should give a lot better results, however it also means our patches would no longer work. In other words, it'd be a lot of work to get it working.
As far as I'm concerned, gcc 4 is still too much of an unknown to worry about it for now. We spend too much time on tools and not enough on better projects anyways ;)
As far as I'm concerned, gcc 4 is still too much of an unknown to worry about it for now. We spend too much time on tools and not enough on better projects anyways ;)
Shoot Pixels Not People!
Makeshift Development
Makeshift Development
- Neil Stevens
- Posts: 79
- Joined: Thu Jan 27, 2005 2:22 pm
- Location: California
- Contact:
During development, yeah, it's horrible. Slowdowns for optimized release code are fine, but not for debug versions.ooPo wrote:Slower compiles for faster code? Yeah, that sucks.
Edit: Looks like it's actually faster, though, fortunately:
It's about time.When compiling without optimizations (-O0), the C++ frontend is much faster than in any previous versions of GCC. Independent testers have measured speed-ups up to 25% in real-world production code, compared to the 3.4 family (which was already the fastest version to date). Upgrading from older versions might show even bigger improvements.
-
- Posts: 564
- Joined: Sat Jan 17, 2004 10:22 am
- Location: Sweden
- Contact:
I came quite far, new datatypes, mmi and cop2 was okish, but and this is a reason for seppuku, cause any coder knows the value of backing up his/hers work, yet in a crititcal moment I failed to comply to the tradition set before us by the great masters of this universe, and all the stances and techniques are now lost.
Atm I have only downloaded the sources, but I intend to start again. doing it over cvs would be a good start I reckon.
Atm I have only downloaded the sources, but I intend to start again. doing it over cvs would be a good start I reckon.
Kung VU
-
- Posts: 564
- Joined: Sat Jan 17, 2004 10:22 am
- Location: Sweden
- Contact:
3.4 generated alot better code than 3.2.2 and I saw some major possibilities with the DFA engine, 4.x has taken it all to a new level, so yes alot better code is indeed possible, and for sure reason enough to bother about 4.x, it will take some time but its interesting work :)Neil Stevens wrote:Looks like it's actually faster, though, fortunately:
When compiling without optimizations (-O0), the C++ frontend is much faster than in any previous versions of GCC. Independent testers have measured speed-ups up to 25% in real-world production code, compared to the 3.4 family (which was already the fastest version to date). Upgrading from older versions might show even bigger improvements.
Kung VU
I've done that before. Did that three times on one project. The fourth time through, it goes much faster and you're a lot more careful about saving for some weird reason. ;) :-Dblackdroid wrote:I came quite far, new datatypes, mmi and cop2 was okish, but and this is a reason for seppuku, cause any coder knows the value of backing up his/hers work, yet in a crititcal moment I failed to comply to the tradition set before us by the great masters of this universe, and all the stances and techniques are now lost.
Atm I have only downloaded the sources, but I intend to start again. doing it over cvs would be a good start I reckon.
Hello !
I am new over there, and I may help retrofitting gcc 3.2 stuff into the new 4.0 version. If you have already done some work, I can try to complete it as far as I can.
Regards,
Ixar
PS: I have already read the gcc 3.2 patches. And a lot of messages on this forum. If you have any pointer I should refer to, don't hesitate.
I am new over there, and I may help retrofitting gcc 3.2 stuff into the new 4.0 version. If you have already done some work, I can try to complete it as far as I can.
Regards,
Ixar
PS: I have already read the gcc 3.2 patches. And a lot of messages on this forum. If you have any pointer I should refer to, don't hesitate.
SVN (dunno if you are aware of it), for the last patches release :
http://svn.ps2dev.org/listing.php?repna ... rev=0&sc=0
then Oopo, the toolchain master !
It would be really great if you could update the toolchain to GCC 4.0 !
http://svn.ps2dev.org/listing.php?repna ... rev=0&sc=0
then Oopo, the toolchain master !
It would be really great if you could update the toolchain to GCC 4.0 !
Yes I already had a look at the SVN files. I have read most of the patches to understand the goal of the patches.
However I need to read more documentation (I am still reading) to understand how to implement properly the code into gcc4.
I will do a summary of the different things that needs to be patched, and were I will do the patch to make sure I don't forget anything, and that I have clearly understood.
However I need to read more documentation (I am still reading) to understand how to implement properly the code into gcc4.
I will do a summary of the different things that needs to be patched, and were I will do the patch to make sure I don't forget anything, and that I have clearly understood.
According to the patch comments, it seems that their is some differences between PS2 Linux kit toolchain and the homebrew ps2dev toolchain.
I need more information on that. Is it related to the RTE ?
For my own test, I have ps2 linux kit dvd. (Just bought yesterday).
If something can be done to manage differences between RTE/non-RTE compilation, I may do it properly if I have the information.
I need more information on that. Is it related to the RTE ?
For my own test, I have ps2 linux kit dvd. (Just bought yesterday).
If something can be done to manage differences between RTE/non-RTE compilation, I may do it properly if I have the information.
I have studied the binutils part of the toolchain.
I will implement the r5900 starting from mips r1000 as it's the only way to do it properly even if it's more similar to the r3000. Since we can't disable instruction properly from the r3000... I am relying on binutils 1.14 patches for that part of the work. It should be pretty fast since it's almost good in the patch.
gcc stuff seems more tricky. But I will see that when binutils is ready.
I will implement the r5900 starting from mips r1000 as it's the only way to do it properly even if it's more similar to the r3000. Since we can't disable instruction properly from the r3000... I am relying on binutils 1.14 patches for that part of the work. It should be pretty fast since it's almost good in the patch.
gcc stuff seems more tricky. But I will see that when binutils is ready.