is phoenix slower on linux than on windows?

agamid · Post by **agamid** » January 23rd, 2003, 1:58 am

I don't know if I'm doing the right things when it comes to optimizations, but it
seems to me that the phoenix that I build on my linux laptop is slower than my
downloaded nightly build of phoenix on a windows 2000 machine at work.

I bring my laptop to work with me and plug it into the net, so they both have the
fast connection, but opening pages on my laptop seems to be more sluggish...
it's pretty zippy on my workstation...

my workstation is a dual pentium 2 400 MHz, while my laptop is a pentium 3 600 MHz.
I have 256 MB RAM on my workstation and 384 MB on my laptop...

anyone running on both platforms noticed this?

ruckerz · Post by **ruckerz** » January 23rd, 2003, 5:13 am

Quantify "Slow" ... some people expect things to reply on a click of a mouse button, which is ok, but how slow is slow? I get maybe a fraction of a second longer delay immediately after parsing html and beginning to display (after the html has been transferred) in linux than in win* .

herman · Post by **herman** » January 23rd, 2003, 6:24 am

agamid wrote:I bring my laptop to work with me and plug it into the net, so they both have the
fast connection, but opening pages on my laptop seems to be more sluggish...
it's pretty zippy on my workstation...

my workstation is a dual pentium 2 400 MHz, while my laptop is a pentium 3 600 MHz.
I have 256 MB RAM on my workstation and 384 MB on my laptop...

You are comparing apples to oranges.
You can´t compare a DUAL CPU workstation to a LAPTOP.
The DUAL CPU has a pretty fast user interface, and beeing a desktop I also think fast memory, i.e. 133 MHz FSB, fast grafics.
Your laptop maybe has slow 66 MHz FSB, has UMA, i.e. the graphic chip competes with the CPU for access to the main memory, because main memory is used for graphics display.

Don´t look at the Megahurts, look at the details of the architecture.

If you really like to compare, install Win2k on your laptop in a dual-boot configuration, then you can compare.

herman

David James · Post by **David James** » January 23rd, 2003, 8:02 am

Yes, Phoenix is noticeably slower in Linux than in Windows. The same goes for Mozilla. This is on a dual boot machine. In fact, if anything, hardware wise for me Phoenix for Linux should be faster as it is running off a faster harddrive. Despite that, and with everything else the same, Phoenix is faster in Windows.

I should qualify this by stating that when I use Phoenix in Linux I usually have a lot of other things running, Mozilla Mail for one. That's not the case in Windows, where it's usually just Phoenix and Windows Explorer and maybe notepad or wordpad. Still, given the speed at which other Linux apps start up and respond to changes, it still seems a big sluggish. Even that behemoth WordPerfect9 for Linux (which uses wine to run) seems relatively fast by comparison.

In truth, if there existed a Linux-native Gecko browser that employed the QT toolkit, I'd probably use that in preference to Phoenix.

Post by **djst** » January 23rd, 2003, 8:12 am

David James wrote:Yes, Phoenix is noticeably slower in Linux than in Windows. The same goes for Mozilla. This is on a dual boot machine. In fact, if anything, hardware wise for me Phoenix for Linux should be faster as it is running off a faster harddrive. Despite that, and with everything else the same, Phoenix is faster in Windows.

I experience the same. I'm running WinXP, WinME, Win2000 and Red Hat 8.0 on this machine, and both Mozilla and Phoenix are slowest on Linux.

AkiAki007 · Post by **AkiAki007** » January 23rd, 2003, 1:13 pm

one reason you might find is slower on linux...

because it is linux, thus no "extra" steps have been taken to optimize your HDD access. Um, trying to remember the name of the program...can't remember, will edit this post tonight if I can, and put the app name here...

anyway, in short, your access speed is probably at like 2.5K/s as opposed to where it "can" be at like 33+K/s, it's all about turning on the right switches for HDD access and stuff, which will make phoenix and ALL other apps significantly faster.

I will post more later. There is also an article on O'Reilly's site about how to do this (which is where I learned about this).

Windows I believe does 1/2 of the optimizations that you want, and you still have the ability do a little bit more using the Xteq Systems X-Setup program.

schapel · Post by **schapel** » January 23rd, 2003, 1:18 pm

Mozilla Linux builds use gcc 2.95 with -O optimization (low optimization), but Windows builds use Visual Studio with a high level of optimization. Developers are working on getting Linux builds to compile using gcc 3.2 with -O2 optimization. This should cause the Linux builds to be about as fast as the Windows builds.

cdn · Post by **cdn** » January 23rd, 2003, 1:38 pm

AkiAki007 wrote:one reason you might find is slower on linux...

because it is linux, thus no "extra" steps have been taken to optimize your HDD access. Um, trying to remember the name of the program...can't remember, will edit this post tonight if I can, and put the app name here...

hdparm ?

Post by **djst** » January 23rd, 2003, 2:02 pm

AkiAki007 wrote:one reason you might find is slower on linux...

because it is linux, thus no "extra" steps have been taken to optimize your HDD access. Um, trying to remember the name of the program...can't remember, will edit this post tonight if I can, and put the app name here...

anyway, in short, your access speed is probably at like 2.5K/s as opposed to where it "can" be at like 33+K/s, it's all about turning on the right switches for HDD access and stuff, which will make phoenix and ALL other apps significantly faster.

This does sound interesting. So why doesn't such a modern distribution as Red Hat 8.0 optimize the hard drive access automatically? I mean why would anyone want to use a slower speed that possible?

agamid · Post by **agamid** » January 23rd, 2003, 3:56 pm

herman wrote:You are comparing apples to oranges.

I know it's not the same machine. I'm not that dumb. I'm just wondering if anyone had the
same feeling, and it seems like they do. first of all, my workstation is about 5 years old. my laptop is now getting to be 3. the graphics card in my laptop probably sucks, whereas the one on my desktop is probably much better for 2D, but the difference I experienced seemed to be drastic.

most other apps I open on my laptop are really fast. even opera on my laptop seems very fast.
by slow, it's not just a fraction of a second... it's seconds. I have tuned my hdparm to some degree, turned on UDMA and stuff. but it's still not even close to what opera was on my laptop. I don't expect it to beat or be the same, I'm just wondering why it's so much slower.

is it really just due to hard drive speeds?
according to hdparm, my buffered disk reads are 14.51 MB/s
UDMA is on and IO-support is 16-bit (should this be 32?)

I compiled phoenix myself, so I'm just wondering if some of the libraries I have are known to be slow and that I should change them... (orbit, libidl)... are there better flags I should compile with?

I can't run the nightlies. I get errors about missing symbols. I know I can probably fix those by some links or hacks, but I'd rather run my own compiled ones.

AkiAki007 · Post by **AkiAki007** » January 23rd, 2003, 9:00 pm

cdn wrote:
AkiAki007 wrote:one reason you might find is slower on linux...

because it is linux, thus no "extra" steps have been taken to optimize your HDD access. Um, trying to remember the name of the program...can't remember, will edit this post tonight if I can, and put the app name here...

hdparm ?

yup
hdparm
here is what i do in my rc file

hdparm -X66 -d1 -u1 -m16 -c3 /dev/hda

speeds me up nicely

daihard · Post by **daihard** » January 23rd, 2003, 9:18 pm

AkiAki007 wrote:yup
hdparm
here is what i do in my rc file

hdparm -X66 -d1 -u1 -m16 -c3 /dev/hda

speeds me up nicely

What are your values from "hdparm -t -T /dev/hda"? My RH 7.3 kernel enables DMA on my hard disk by default (-d1), and uses 16 for the sector count (-m16). Here's my results:

Code: Select all

Timing buffer-cache reads:   128 MB in  0.34 seconds =378.85 MB/sec
Timing buffered disk reads:  64 MB in  1.45 seconds = 44.22 MB/sec

I changed other values to match yours, but it didn't seem to help much.

daihard · Post by **daihard** » January 23rd, 2003, 9:23 pm

djst wrote:This does sound interesting. So why doesn't such a modern distribution as Red Hat 8.0 optimize the hard drive access automatically? I mean why would anyone want to use a slower speed that possible?

AFAIK, the official explanation is that not all the hard disks can be optimised like that safely. For instance, turning on 32-bit I/O access will definitely speed up the disk I/O but may increase the risk of data corruption.

That said, RedHat 7.3 seems to know how to optimise my disk pretty well.

As I posted above, the performance on my hard disk seems rather satisfactory. Another thing you might want to keep in mind is that RH 8 seems slower than 7.3. I used to have 7.3 on a P3-500MHz system, on which I now have 8.0. Almost always, 8.0 is much slower than 7.3.

daihard · Post by **daihard** » January 23rd, 2003, 9:25 pm

schapel wrote:Mozilla Linux builds use gcc 2.95 with -O optimization (low optimization), but Windows builds use Visual Studio with a high level of optimization. Developers are working on getting Linux builds to compile using gcc 3.2 with -O2 optimization. This should cause the Linux builds to be about as fast as the Windows builds.

I sure hope they will still build Phoenix using GCC 2.9x. Otherwise, I will not be able to run it on my RH 7.3 system... will I? On a side note, Intel's new compiler seems to do much better in optimisation than GCC (I read a comparison article a few months ago).

AkiAki007 · Post by **AkiAki007** » January 23rd, 2003, 10:10 pm

daihard wrote:
AkiAki007 wrote:yup
hdparm
here is what i do in my rc file

hdparm -X66 -d1 -u1 -m16 -c3 /dev/hda

speeds me up nicely

What are your values from "hdparm -t -T /dev/hda"? My RH 7.3 kernel enables DMA on my hard disk by default (-d1), and uses 16 for the sector count (-m16). Here's my results:
Code: Select all
Timing buffer-cache reads:   128 MB in  0.34 seconds =378.85 MB/sec
Timing buffered disk reads:  64 MB in  1.45 seconds = 44.22 MB/sec
I changed other values to match yours, but it didn't seem to help much.

Mine actually came out significantly slower than yours. at about 1/2 the speed, but that's probably my hdd compared to yours :-P

Code: Select all

/dev/hda:
 Timing buffer-cache reads:   128 MB in  0.86 seconds =148.84 MB/sec
 Timing buffered disk reads:  64 MB in  2.60 seconds = 24.62 MB/sec

is phoenix slower on linux than on windows?

is phoenix slower on linux than on windows?

Re: is phoenix slower on linux than on windows?

Build Configuration

Re: Build Configuration