I read somewhere on HA that some people say that ICL 8 compiled lame is faster than ICL 7.1 compiled one.
I compared my ICL 7.1 (using profile optimization), and ICL 8 (using prof. opt.)
Tested on 5 wavs, and taken average of the maximum encoding speeds.
LAME 3.96 alpha 1 using --preset standard
AMD Athlon XP 2100+
ICL 7.1 - 8.8 play/CPU
ICL 8.0 - 8.1 play/CPU
Am I wrong or ICL 8.0 indeed generates slower code?
Full Version: ICL 8 and LAME
I someone wants to test, I'll post binaries.
My own limited testing of ICL 8.0 gives similar results. I haven't tested it with a LAME compile, but I did test it with oggenc and it was both a little slower and did not yield bit identical results when compared with the 7.1 compile.
I also tested version 8 compared with 7.1 (in linux) and got similar results with oggenc. 7.1 was faster than 8.
john33, how much is the ICL 4.5 compile of LAME faster than the ICL 7.1 compile?
I find ICL 8.0 version of LAME 3.95.1 (own compile) slightly faster than 7.1 on my P4 CPU:
http://www.hydrogenaudio.org/forums/index....ic=17868&st=25&
It's slower on my Athlon machine, though. I haven't tested other versions, but oggenc is also a little faster for me (in both machines I believe). ¿Are sources for 3.90.3 available, btw, so I can test?
http://www.hydrogenaudio.org/forums/index....ic=17868&st=25&
It's slower on my Athlon machine, though. I haven't tested other versions, but oggenc is also a little faster for me (in both machines I believe). ¿Are sources for 3.90.3 available, btw, so I can test?
QUOTE(stipe @ Feb 2 2004, 09:12 PM)
john33, how much is the ICL 4.5 compile of LAME faster than the ICL 7.1 compile?
I'll do some comparative tests and post the results.
QUOTE(Anacondo @ Feb 2 2004, 10:06 PM)
I find ICL 8.0 version of LAME 3.95.1 (own compile) slightly faster than 7.1 on my P4 CPU:
http://www.hydrogenaudio.org/forums/index....ic=17868&st=25&
It's slower on my Athlon machine, though. I haven't tested other versions, but oggenc is also a little faster for me (in both machines I believe).
http://www.hydrogenaudio.org/forums/index....ic=17868&st=25&
It's slower on my Athlon machine, though. I haven't tested other versions, but oggenc is also a little faster for me (in both machines I believe).
Hmmm, maybe there is some benefit for Intel processor users. I am an Athlon user.
QUOTE(Anacondo @ Feb 2 2004, 10:06 PM)
¿Are sources for 3.90.3 available, btw, so I can test?
They're on Rarewares.
I use Athlon too. (athlon xp 2000+).
It seems that ICL 8.0 compiles are faster only for Intel processors.
It seems that ICL 8.0 compiles are faster only for Intel processors.
Results as follows:
AthlonXP 2700+ (clocked at 11 x 200).
QUOTE
D:\testdir>lame45.exe --preset standard 10.wav 10.mp3
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:55/ 0:55| 0:55/ 0:55| 4.6089x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>lame71.exe --preset standard 10.wav 10.mp3
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 1:01/ 1:01| 1:01/ 1:01| 4.1566x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>lame80.exe --preset standard 10.wav 10.mp3
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 1:03/ 1:03| 1:03/ 1:03| 4.0366x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:55/ 0:55| 0:55/ 0:55| 4.6089x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>lame71.exe --preset standard 10.wav 10.mp3
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 1:01/ 1:01| 1:01/ 1:01| 4.1566x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>lame80.exe --preset standard 10.wav 10.mp3
+ Modification for Cuesheet Support Version 0.5.3
+ Modification for Monkey's Audio (APE) Support.
Available at http://www.geocities.com/nyaochi2000/lame/
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 10.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 1:03/ 1:03| 1:03/ 1:03| 4.0366x| 0:00
32 [ 1] *
128 [2429] %%%%%%%%%%%%%%%%%%%%%*****************
160 [4300] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********************
192 [1021] %%%%%%%%%%******
224 [ 494] %%%%%***
256 [ 678] %%%%%%*****
320 [ 861] %%%%%%%%%%****
average: 179.3 kbps LR: 6222 (63.59%) MS: 3562 (36.41%)
Writing LAME Tag...done
D:\testdir>
AthlonXP 2700+ (clocked at 11 x 200).
QUOTE(stipe @ Feb 3 2004, 01:32 AM)
I use Athlon too. (athlon xp 2000+).
It seems that ICL 8.0 compiles are faster only for Intel processors.
It seems that ICL 8.0 compiles are faster only for Intel processors.
Well, the newest Intel compiler is there to promote the newest Intel CPU. They optimize libraries and code generators to let actual Intel CPUs shine. I don't have any Intel P3, but I wouldn't be surprised to see a similar speed drop for the P3 too.
hmm....I'm running a P4 and I found ICL 8 slower, Perhaps, I'm not using all the optimisation switches. I was unimpressed with 8 so I uninstalled it and put 7.1 back on. 
I love 8's new /QxB switch, which greatly speeds things up on a Banias (haven't tried LAME, but it did speed up basically all the other stuff I've tried). Comparing the assembly output of the /QxN switch and that of /QxB, it looks like ICL makes up for the P4's weaknesses by substituting several instructions that are faster on the P4 for a single instructions (e.g. insane uses of the LEA instruction replacing 1 IMUL, iirc). This may speed things up on P4s, but on Banias it tends to slow things down (some things run at 2/3 the speed with /QxN on Banias).
If only they didn't break things on the P4, then they wouldn't have to make a switch to make up for the broken processor (/QxW and /QxN) then release a new version with a new switch (/QxB) to use SSE2 properly on a non-broken processor.
If only they didn't break things on the P4, then they wouldn't have to make a switch to make up for the broken processor (/QxW and /QxN) then release a new version with a new switch (/QxB) to use SSE2 properly on a non-broken processor.
Is anyone using PGO in these tests?
stipe are you the same stipe that makes optimized firebird builds? 
yes, that's me.
QUOTE(kode54 @ Feb 3 2004, 02:28 AM)
Is anyone using PGO in these tests?
i used PGO for both of the compiles, and then compared them.
QUOTE(p0wder @ Feb 2 2004, 08:51 PM)
stipe are you the same stipe that makes optimized firebird builds?
lol, it's a small world.
I thought I recognized that name 
. I'm usually switching between stipe's, scragz', or aebrahim's builds 
. I'm usually switching between stipe's, scragz', or aebrahim's builds
New compiles, new test.
Speed appears to be the only issue here!!
Intel compilers: 45 = 4.5, 71 = 7.1, 80 = 8.0.
CODE
D:\testdir>lame45 --preset standard 10.wav 1045.mp3
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1045.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:44/ 0:44| 0:44/ 0:44| 5.7498x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>lame71 --preset standard 10.wav 1071.mp3
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1071.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:48/ 0:48| 0:48/ 0:48| 5.2613x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>lame80 --preset standard 10.wav 1080.mp3
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1080.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:50/ 0:50| 0:50/ 0:50| 5.0705x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>
All files: 7,026,733 bytes!!
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1045.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:44/ 0:44| 0:44/ 0:44| 5.7498x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>lame71 --preset standard 10.wav 1071.mp3
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1071.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:48/ 0:48| 0:48/ 0:48| 5.2613x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>lame80 --preset standard 10.wav 1080.mp3
LAME version 3.90.3 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), 3DNow!, SIMD
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 10.wav to 1080.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9781/9784 (100%)| 0:50/ 0:50| 0:50/ 0:50| 5.0705x| 0:00
32 [ 1] *
128 [ 56] %
160 [1047] %%%%%%%%%********
192 [4217] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [2108] %%%%%%%%%%%%%%%%%%%%%%%**********
256 [ 876] %%%%%%%%******
320 [1479] %%%%%%%%%%%%%%%%********
average: 220.2 kbps LR: 6386 (65.27%) MS: 3398 (34.73%)
Writing LAME Tag...done
D:\testdir>
All files: 7,026,733 bytes!!
Speed appears to be the only issue here!!
Intel compilers: 45 = 4.5, 71 = 7.1, 80 = 8.0.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.