Thread
Firefox 4 on Tiger PPC
When I try and unpack that .zip file, I get an "operation not permitted" dialog box (header is 'archive utility'). But I did this before, on another Pismo, with no problem. How do I work around that?
Huh, I unzipped it on my Win XP machine and then loaded the .app file just fine. Weird. Something about .zip files that is locked down in OS X?
Shouldn't be. Were you unzipping in a location that was read-only?
Incidentally, for those who want to follow along, I've put up a dev blog: http://tenfourfox.blogspot.com/
I should do the same for Classilla, and will soon.
Incidentally, for those who want to follow along, I've put up a dev blog: http://tenfourfox.blogspot.com/
I should do the same for Classilla, and will soon.
I was just about to post about this on the forum, but glad to see that it's had an early following already. Can't currently try on my iMac G5 because the caps went out on it again. Gonna have to replace those, probably over the Christmas break. Very eager to give it a try, though. My dad has a dual-proc Quicksilver that I'll probably test it on.
Got beta 8 up tonight. Playing with some ideas for better G5 optimization, and there will be a G4/7400 version since the 7450 version is not doing well enough on those machines. Hoping for release in a week or so.
Well, 12 days is close to a week, right? I put up the dev notes on http://tenfourfox.blogspot.com/2010/12/beta-8-now-available.html and you can get it now from http://www.tenfourfox.com/ . Notice new G5 and G4/7400-specific builds (along with G3 and G4/7450). The G5 version has very interesting tuning to maintain stability; the interested can look at about:buildconfig. It had to be done like that because -mcpu=G5 (let alone -fast) makes a very unstable Firefox. I'll have to debug this later, but the tweaks still squeeze out around 15% compared to 7450 on a G5. Lots of good stuff; see the release notes.
For those using this, beta 9 is now available.
I've been using TenFourFox as my sole browser (at home) for a couple weeks now. Very good performance overall, very pleased with it. The only thing I've noticed is that when I use Cmd-+ to scale up the page, it acts multiple times for one press. Sometimes only three increases, sometimes all the way to maximum (eight times). I don't have a sticky key, because this doesn't happen with other browsers.
Is this 10.5? Seems okay on 10.4 G5.
Yes, sorry. Latest 10.5.x, dual 1.8 G5.
Just tried TenFourFox on a G3 iBook after having used it on my G4 Mac Mini and it flies! :O
I can't believe how fast this browser is. I was about to browse sites at about the same speed as my MacBook.
Thank you so much ClassicHasClass
I can't believe how fast this browser is. I was about to browse sites at about the same speed as my MacBook.
Thank you so much ClassicHasClass
TenFourFox beta 9's address bar is acting a little weird on my TiBook. Whenever I hover over a link, it disappears and reappears with the URL of the link, and the same when I take the cursor off the link. Like it's updating slowly. It worked perfectly fine in beta 8.
Just noticed that when I quickly hover over two links, it's fine, it has something to do with transitioning from normal to showing the destination link.
Just noticed that when I quickly hover over two links, it's fine, it has something to do with transitioning from normal to showing the destination link.
There's apparently a performance regression in the version of cairo in Mozilla 2. They're supposed to fix this, but I haven't seen much activity on that front. :-/
agg23, since you've got a G3, how well does the nanojit work for you? (See the release notes on how to enable it.) I don't have a G3 with Tiger, so I'm depending on user reports.
agg23, since you've got a G3, how well does the nanojit work for you? (See the release notes on how to enable it.) I don't have a G3 with Tiger, so I'm depending on user reports.
Well, well. The nanojit may be less effective on G5, but we're getting excellent reports on G4. A user wrote in to report that to my surprise, the G4 really benefits from it -- which makes me wonder where it gets it so wrong on the G5. On my iBook, it cuts SunSpider execution time in half, and matches the G5! So shame on me for assuming the G5 with all its fancy out-of-order hardware would be the last word.
OK. I'll test it...
I just ran sunspider tests and the results are outstanding. Basically, without nanojit it took 15926.9ms and with nanojit it took 8921.4ms...
You can see more detailed results below:
Without nanojit
With nanojit
You can see more detailed results below:
Without nanojit
With nanojit
excuse my ignorance but what is nanojit?
agg23, this is a G3, right? If so, that's fabulous. This means the only CPU that the nanojit doesn't broadly benefit is the G5.
highlandcattle, the nanojit is a component of Mozilla's JavaScript engine that translates "traces" of JavaScript execution into native assembly code. Previously, the only working nanojits were for ARM, x86, x86_64 and SPARC, and the SPARC one was "unofficial." PPC and MIPS had code present, but it was incomplete for Firefox. So, I finished it and now we have native code compilation for JavaScript for PowerPC for the first time. I submitted it back to Mozilla to get it into the source tree (bug 624164), but I put it in TenFourFox so people can use and benefit from it now. There is at least one bug I know of already, but it's so far been remarkably functional.
highlandcattle, the nanojit is a component of Mozilla's JavaScript engine that translates "traces" of JavaScript execution into native assembly code. Previously, the only working nanojits were for ARM, x86, x86_64 and SPARC, and the SPARC one was "unofficial." PPC and MIPS had code present, but it was incomplete for Firefox. So, I finished it and now we have native code compilation for JavaScript for PowerPC for the first time. I submitted it back to Mozilla to get it into the source tree (bug 624164), but I put it in TenFourFox so people can use and benefit from it now. There is at least one bug I know of already, but it's so far been remarkably functional.
Yes, it is a G3.
Excellent.
b9 is definitely a step backward from b8 on my G5. Tab bar is screwy, and I've had two system panics since I've been using it. (That may be completely coincidental, but ...) So for now I've reinstalled b8.
The tab bar was a Mozilla bug, allegedly fixed in b10. However, I'm skipping b10 since b11 is supposed to be out next week. Not sure what to say about the crashes, although I did test drive it on my dad's 10.5.8 G4 mini and it seemed to be okay.
In better news, I think I cracked the performance issues with the nanojit on G5. I'm running off a test build right now, but I think I've got it benching faster than stock interpreter for the first time. The performance improvement isn't as dramatic as the G3 or G4, but it's still an improvement.
In better news, I think I cracked the performance issues with the nanojit on G5. I'm running off a test build right now, but I think I've got it benching faster than stock interpreter for the first time. The performance improvement isn't as dramatic as the G3 or G4, but it's still an improvement.
This works great on my 'new' iBook G3 600Mhz! No problems with it.
Sites I tested so far are Youtube, MSN.com, Hotmail, 68kmla offcourse and Macintosh Garden. They all seem to work flawlessly, Flash does also work.
TFF itself does also look beautiful. Even HTML5 works, but the performance is not really good to be honest (but still works better than I thought), I used http://9elements.com/io/projects/html5/canvas/ for the HTML5 test.
But I hope that you will keep developing this wonderful application!
Sites I tested so far are Youtube, MSN.com, Hotmail, 68kmla offcourse and Macintosh Garden. They all seem to work flawlessly, Flash does also work.
TFF itself does also look beautiful. Even HTML5 works, but the performance is not really good to be honest (but still works better than I thought), I used http://9elements.com/io/projects/html5/canvas/ for the HTML5 test.
But I hope that you will keep developing this wonderful application!
I suspect the big problem with that test is that TenFourFox doesn't have hardware acceleration, and unless Mozilla gets rid of the NPOT texture size requirement, it's not going to. 
That said, beta 11 will have the JavaScript accelerator enabled by default for all systems, including G5s (although it has the most dramatic improvement on G3 and G4). This will be a big difference if you haven't already turned it on.
That said, beta 11 will have the JavaScript accelerator enabled by default for all systems, including G5s (although it has the most dramatic improvement on G3 and G4). This will be a big difference if you haven't already turned it on.
What kind of JS acceleration are you seeing, overall, on G3/G4/G5? Which operations are seeing the biggest boost?
agg23 posted his G3 numbers, but by and large for G3s and G4s it's about 2x speed. My father's 1.5GHz G4 mini, for example, benches at 6185ms without it but 2811ms with it in SunSpider.
For G5s, however, it seems to hit a wall with the extra hardware. SunSpider actually takes a hit, but with tuning to blacklist expensive JSOPs, Dromaeo and V8 both get wins. It's not as dramatic, however, and a very fast G4 will actually bench faster. I'm trying to figure out why; it seems that the G5's longer pipeline takes a bigger penalty for stalls. That said, the base interpreter without acceleration on my quad is around 3300ms. The accelerator "accelerates" that to 3700ms, but Dromaeo gains several runs/sec and the V8 score jumps, so I'm calling it an overall win.
For G5s, however, it seems to hit a wall with the extra hardware. SunSpider actually takes a hit, but with tuning to blacklist expensive JSOPs, Dromaeo and V8 both get wins. It's not as dramatic, however, and a very fast G4 will actually bench faster. I'm trying to figure out why; it seems that the G5's longer pipeline takes a bigger penalty for stalls. That said, the base interpreter without acceleration on my quad is around 3300ms. The accelerator "accelerates" that to 3700ms, but Dromaeo gains several runs/sec and the V8 score jumps, so I'm calling it an overall win.
Hello,
I'm using TenFourFox on my G3 Power Mac (B&W) and it works great, It's a lot faster than firefox 3.6 and more responsive.
Thanks Cameron :beige:
I'm using TenFourFox on my G3 Power Mac (B&W) and it works great, It's a lot faster than firefox 3.6 and more responsive.
Thanks Cameron :beige:
You're welcome, but now you have to upgrade it, because beta 11 is out This version turns on JavaScript acceleration for everybody, even G5 users. G5 users will see less improvement, but will still see some. G4 and G3 owners will see the most improvement. There are also some bug fixes and the browser is overall quicker thanks to rendering improvements.
Just for yuks, I put a chart on the main page of a browser JavaScript speed comparison I did when I was at my folks' last (see above) since Dad runs 10.5.8 on his G4 mini so that he can use iTunes 10 (I'm the Nexus One heathen, my folks use their iPhone 3GSes still). This gave me a chance to test Safari 5. No points for guessing that TenFourFox won by a large margin, but I think people will be very surprised by which browser came in 2nd. I didn't expect this either.
This version turns on JavaScript acceleration for everybody, even G5 users. G5 users will see less improvement, but will still see some. G4 and G3 owners will see the most improvement. There are also some bug fixes and the browser is overall quicker thanks to rendering improvements.Just for yuks, I put a chart on the main page of a browser JavaScript speed comparison I did when I was at my folks' last (see above) since Dad runs 10.5.8 on his G4 mini so that he can use iTunes 10 (I'm the Nexus One heathen, my folks use their iPhone 3GSes still). This gave me a chance to test Safari 5. No points for guessing that TenFourFox won by a large margin, but I think people will be very surprised by which browser came in 2nd. I didn't expect this either.
I've encountered a glitch on the G5. Using beta 11, sometimes when I attempt to login to this forum with the correct user name/password the browser claims "You have exceeded the maximum number of attempts blah, blah". This happens on the first login attempt. As I recall this did not happen with previous builds, but my memory may be faulty.
I've gotten that randomly with other browsers, so I don't think it's a 10.4Fx issue specifically. I've seen it on my office PC with Fx 3.6, for example.
But since we're bumping this thread, watch for beta 12 to come down the wire (labeled b13pre due to an issue with checkout I didn't notice until the build was done, minor issue). This fixes most of the rest of the 10.4-specific bugs, and also is a lot more aggressive about memory consumption. However, it also really wrecked the plugin code and I hope I got all the gaps. QuickDraw-based plugins are dying, and it's going to be a cold hard CoreAnimation world, so it will be harder and harder to support PPC plugins that are no longer maintained (which is why I am deprecating them, and disabling them in the future).
Anyway, I've been using it on my G5 and iBook and looks good, so push out to beta channel tonight when I get back from work.
But since we're bumping this thread, watch for beta 12 to come down the wire (labeled b13pre due to an issue with checkout I didn't notice until the build was done, minor issue). This fixes most of the rest of the 10.4-specific bugs, and also is a lot more aggressive about memory consumption. However, it also really wrecked the plugin code and I hope I got all the gaps. QuickDraw-based plugins are dying, and it's going to be a cold hard CoreAnimation world, so it will be harder and harder to support PPC plugins that are no longer maintained (which is why I am deprecating them, and disabling them in the future).
Anyway, I've been using it on my G5 and iBook and looks good, so push out to beta channel tonight when I get back from work.