Sunday, May 29, 2011

Tobias gets a beer: Aurora on PowerPC



Tobias gets a beer. This is the G5 build of the Firefox 5 beta, although for some reason it says Aurora. This was built with his 10.4-linked 64-bit ld, and appears to have all of our standard features. And it works!

Unfortunately this is not shippable either, as it is still building that monstrous 750MB+ libxul, so we're going to need to find some way of stripping that library down. Fixed! Good ol' strip -x seems to have done the trick, and I will fix the build system to reenable stripping which was disabled for purposes of linker testing. Doing some more testing before issuing a formal beta. Meanwhile, I am considering options for 4.0.3.

19 comments:

  1. I'm Salivating...!

    Where can I download it, please please please.... ;-)

    ReplyDelete
  2. Well, alas, you can't. The libxul is too big to upload, which is why we can't ship it this way. Mozilla is clearly not issuing binaries anywhere near that size, so we probably need to update the rest of our library toolchain. Tobias, how large did libxul get on your system (ls -l obj-ff-dbg/toolkit/library/XUL) ?

    ReplyDelete
  3. I should note just as proof of functionality that this blog post and my comments are coming from "TenFourFox Aurora," btw.

    ReplyDelete
  4. After a bit of messing around, no additional tools needed: regular old strip -x seems to have done the job nicely. Making sure there are no glitches, but this reduced it down to a G5 optimized XUL of 31MB. Excellent!

    ReplyDelete
  5. Oooo, it's shiny! That's awesome. Great job Tobias! :D

    ReplyDelete
  6. Oh, looks like CHC went and updated his post while I was writing. :P YAY!!! A TFF5 beta is on the way. :D

    Great job Tobias and CHC! It's always inspires me when programmers get together and make cool things for our beloved Macs. :D

    ReplyDelete
  7. TenFourFox is probably the most valuable piece of software made for PowerPC ever. I downloaded it about two weeks ago and i´m following this blog since then. Reading the previous post i begun to lose the hope on a possible TFF 5, but now, barely two days after that... here it is, THE LIGHT! You guys are the best... And now, the question: what about the G3 and G4 versions? Is there any chance? I have a G5, but i want that browser on my PowerBook too! Please don´t give up...

    Cheers from a VERY happy user in Argentina!(please excuse any mistake on my english)

    PPC FOREVER!

    ReplyDelete
  8. So far it looks good, although mostly I've been working on the G5 version. I'll be running off test builds for all four architectures later this week and then we'll do some basic conformance testing on them. If they pass, we'll make it a beta.

    ReplyDelete
  9. Managed to build Aurora 5.0 successfully!

    I built it optimized for G4-7400 using gcc-4.2 (Apple build 5666.3). The resulting binaries are of architecture "ppc" so they should run on any ppc CPU. I noticed that gcc passed "force_cpusubtype_ALL" to the assembler although that option wasn't passed to gcc itself. That's really strange as both "-arch ppc7400" and "-mcpu=7400" were passed to gcc. Had to specify "-faltivec" as well in order to get altivec support.
    I also reenabled "-dead_strip" in the configure script and added "--enable-strip" as well as "--enable-install-strip" to mozconfig. The size of the whole application bundle is 45.9 MB and XUL only is 26.9 MB.
    Also enabled libxul and ipc in mozconfig although I don't know if it's necessary.

    Might we get Aurora instead of firefox because ppc isn't supported any more?

    ReplyDelete
  10. No, it's a branding switch I didn't have set in that bunch of patches. It's corrected now. Thanks for reminding me about the --enable-*strip; we didn't need those in 4.0, but it looks like we'll need them now. I'll also put -faltivec into the mozconfigs.

    So far, most things work. JavaScript is a hair slower, but media decoding and page display is noticeably faster. Plugins do appear to work, amazingly enough, but I'll talk about that later.

    As far as TenFourFox-specific features, I'm having trouble with typed arrays; the code tests okay on the built-in JS conformance tests but makes the Audio API go batty. Loop filtering made little difference to performance unfortunately, so I'm looking at writing YCbCr conversion routines (I've written the piece that does the filtering, and next is the piece to actually do the conversion). However, our custom update system seems to work, and WebM AltiVec acceleration, Pixman VMX acceleration and the nanojit are otherwise operational. More in a few days.

    ReplyDelete
  11. In the Dromae recommended tests the results with TenFourFox 4.0.2 (G4 7450 build) and Aurora 5.0 (G4 7400 build) are practically the same on my PowerBook G4 1.5 GHz running OS X 10.5.8 . I get approx. 29 runs/s. I wonder why that Mac mini G4 1.5 GHz got 50 runs/s ...

    ReplyDelete
  12. That sounds wrong. Even my iBook G4/1.33 manages around 45 (which is roughly proportional to my Sawtooth 450MHz which gets around 10).

    ReplyDelete
  13. I get about 26 runs/sec. with TFF 4.0.2 on my PowerBook G4 1.33 GHz/10.5.8 when I set the processor speed to 'reduced' in the Energy Saver settings.

    ReplyDelete
  14. That's undoubtedly the difference. All these tests were on "highest" (for obvious reasons).

    ReplyDelete
  15. Mine ran on "highest" as well...
    Does enabling tracejit for chrome affect those results?

    ReplyDelete
  16. I reset TenFourFox/Aurora and restarted OS X and now the scores are as follows:

    On a PowerBook G4 12" 1,5 GHz, OS X 10.5.8:
    TenFourFox 4.0.2 (G4 7450): 53 runs/s
    TenFourFox/Aurora 5.0 beta (G4 7400): 55 runs/s

    On a PowerBook Wallstreet/PDQ G4 500 MHz, OS X 10.4.11 (Intel build):
    TenFourFox/Aurora 5.0 beta (G4 7400): 19 runs/s

    So this is now as it should be. That upgraded Wallstreet seems to perform quite well; 38% the score of the G4 1,5 GHz.

    ReplyDelete
  17. Sometimes, when I sleep my PowerBook by closing the lid, after waking up, it gets confused about its processor speed setting. It may say 'reduced', but it's really in 'highest', or vice versa. I can tell by a slightly different buzz it makes when I listen closely. Also the speed of Photoshop filters and the like support my suspicion. I have to switch the speed setting back and forth again to make sure the computer really does what it says.

    ReplyDelete

Due to an increased frequency of spam, comments are now subject to moderation.