Linearer Modus

11.09.2011 21:31

pi_css5 auf einer Sun Ultra5 mit 270MHz Ultrasparc IIi

Standard-CFLAGS: 272,23 Sekunden

Mit

Zitat:CFLAGS = -O4 -mcpu=ultrasparc -pipe -fomit-frame-pointer -ffast-math -finline-functions -fweb -frename-registers -mapp-regs -mvis -Wa,-Av8plusa

sind es dann 94,94 Sekunden.

Die Ultra5 schneidet somit deutlich schlechter ab als die Ultra2.

***Alpha*** · 11.09.2011 21:33

Seit wann gibts O-Flags bis -O4? dachte die gehen nur bis -O3?

11.09.2011 21:42

Kannst bis O5 oder so gehen auf Sparc.

**oreissig** · 21.10.2011 21:58

so, bevor meine RS/6000 44P-170 an TNC geht, arbeite ich nochmal fix mein programm damit durch

AIX 4.3.3

Code:

-bash-4.1$ /usr/java131/bin/java pi_fftcs 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.006286438480884726 calculating 1048576 digits of PI... AGM iteration precision= 48: 3.20 sec precision= 80: 1.88 sec precision= 176: 1.79 sec precision= 352: 1.80 sec precision= 688: 1.79 sec precision= 1392: 1.79 sec precision= 2784: 1.79 sec precision= 5584: 1.78 sec precision= 11168: 1.78 sec precision= 22336: 1.78 sec precision= 44688: 1.78 sec precision= 89408: 1.77 sec precision= 178816: 1.78 sec precision= 357648: 1.78 sec precision= 715312: 1.78 sec precision= 1430640: 1.87 sec writing pi1048576.txt 42.03 sec. (real time) -bash-4.1$ /usr/java131/bin/java -version java version "1.3.1" Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.1) Classic VM (build 1.3.1, J2RE 1.3.1 IBM AIX build ca131-20070913 (SR11) (JIT enabled: jitc))

Code:

-bash-4.1$ ./pi_css5 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.00587898 calculating 1048576 digits of PI... AGM iteration precision= 48: 1.42 sec precision= 80: 1.42 sec precision= 176: 1.43 sec precision= 352: 1.42 sec precision= 688: 1.42 sec precision= 1392: 1.42 sec precision= 2784: 1.42 sec precision= 5584: 1.41 sec precision= 11168: 1.41 sec precision= 22336: 1.41 sec precision= 44688: 1.41 sec precision= 89408: 1.41 sec precision= 178816: 1.40 sec precision= 357648: 1.41 sec precision= 715312: 1.41 sec precision= 1430640: 1.41 sec writing pi1048576.txt... 26.61 sec. (real time)

Code:

BYTEmark* Native Mode Benchmark ver. 2 (10/95) Index-split by Andrew D. Balsa (11/97) Linux/Unix* port by Uwe F. Mayer (12/96,11/97) TEST : Iterations/sec. : Old Index : New Index : : Pentium 90* : AMD K6/233* --------------------:------------------:-------------:------------ NUMERIC SORT : 351.84 : 9.02 : 2.96 STRING SORT : 28.816 : 12.88 : 1.99 BITFIELD : 5.105e+07 : 8.76 : 1.83 FP EMULATION : 17.708 : 8.50 : 1.96 FOURIER : 7683.2 : 8.74 : 4.91 ASSIGNMENT : 5.3466 : 20.34 : 5.28 IDEA : 928.12 : 14.20 : 4.21 HUFFMAN : 408.4 : 11.32 : 3.62 NEURAL NET : 8.3532 : 13.42 : 5.64 LU DECOMPOSITION : 267.04 : 13.83 : 9.99 ==========================ORIGINAL BYTEMARK RESULTS========================== INTEGER INDEX : 11.601 FLOATING-POINT INDEX: 11.749 Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0 ==============================LINUX DATA BELOW=============================== CPU : POWER3-II 450MHz L2 Cache : 8MB OS : AIX 4.3.3 C compiler : gcc version 3.4.3 libc : MEMORY INDEX : 2.679 INTEGER INDEX : 3.068 FLOATING-POINT INDEX: 6.517 Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38 * Trademarks are property of their respective holder.

AIX 5.1

Code:

-bash-3.00$ /usr/java14/jre/bin/java pi_fftcs 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.006286438480884726 calculating 1048576 digits of PI... AGM iteration precision= 48: 1.84 sec precision= 80: 1.65 sec precision= 176: 1.66 sec precision= 352: 1.65 sec precision= 688: 1.65 sec precision= 1392: 1.65 sec precision= 2784: 1.65 sec precision= 5584: 1.65 sec precision= 11168: 1.64 sec precision= 22336: 1.64 sec precision= 44688: 1.68 sec precision= 89408: 1.64 sec precision= 178816: 1.64 sec precision= 357648: 1.64 sec precision= 715312: 1.64 sec precision= 1430640: 1.64 sec writing pi1048576.txt 32.80 sec. (real time) -bash-3.00$ /usr/java14/jre/bin/java -version java version "1.4.2" Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2) Classic VM (build 1.4.2, J2RE 1.4.2 IBM AIX build ca142ifx-20110707 (SR13 FP10) (JIT enabled: jitc))

Code:

-bash-3.00$ ./pi_css5 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.00587898 calculating 1048576 digits of PI... AGM iteration precision= 48: 1.42 sec precision= 80: 1.43 sec precision= 176: 1.43 sec precision= 352: 1.43 sec precision= 688: 1.43 sec precision= 1392: 1.42 sec precision= 2784: 1.37 sec precision= 5584: 1.43 sec precision= 11168: 1.42 sec precision= 22336: 1.41 sec precision= 44688: 1.41 sec precision= 89408: 1.41 sec precision= 178816: 1.41 sec precision= 357648: 1.41 sec precision= 715312: 1.41 sec precision= 1430640: 1.41 sec writing pi1048576.txt... 26.62 sec. (real time)

Code:

BYTEmark* Native Mode Benchmark ver. 2 (10/95) Index-split by Andrew D. Balsa (11/97) Linux/Unix* port by Uwe F. Mayer (12/96,11/97) TEST : Iterations/sec. : Old Index : New Index : : Pentium 90* : AMD K6/233* --------------------:------------------:-------------:------------ NUMERIC SORT : 356 : 9.13 : 3.00 STRING SORT : 28.532 : 12.75 : 1.97 BITFIELD : 5.0986e+07 : 8.75 : 1.83 FP EMULATION : 25.728 : 12.35 : 2.85 FOURIER : 8204.9 : 9.33 : 5.24 ASSIGNMENT : 9.0516 : 34.44 : 8.93 IDEA : 1013.5 : 15.50 : 4.60 HUFFMAN : 530.24 : 14.70 : 4.70 NEURAL NET : 8.3866 : 13.47 : 5.67 LU DECOMPOSITION : 281.36 : 14.58 : 10.53 ==========================ORIGINAL BYTEMARK RESULTS========================== INTEGER INDEX : 13.868 FLOATING-POINT INDEX: 12.237 Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0 ==============================LINUX DATA BELOW=============================== CPU : POWER3-II 450MHz L2 Cache : 8MB OS : AIX 5.1 C compiler : gcc version 4.0.0 libc : MEMORY INDEX : 3.182 INTEGER INDEX : 3.686 FLOATING-POINT INDEX: 6.787 Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38 * Trademarks are property of their respective holder.

AIX 5.2

Code:

BYTEmark* Native Mode Benchmark ver. 2 (10/95) Index-split by Andrew D. Balsa (11/97) Linux/Unix* port by Uwe F. Mayer (12/96,11/97) TEST : Iterations/sec. : Old Index : New Index : : Pentium 90* : AMD K6/233* --------------------:------------------:-------------:------------ NUMERIC SORT : 345.04 : 8.85 : 2.91 STRING SORT : 34.198 : 15.28 : 2.37 BITFIELD : 5.1137e+07 : 8.77 : 1.83 FP EMULATION : 59.6 : 28.60 : 6.60 FOURIER : 7473.4 : 8.50 : 4.77 ASSIGNMENT : 8.4318 : 32.08 : 8.32 IDEA : 1044.1 : 15.97 : 4.74 HUFFMAN : 526.73 : 14.61 : 4.66 NEURAL NET : 9.37 : 15.05 : 6.33 LU DECOMPOSITION : 276.96 : 14.35 : 10.36 ==========================ORIGINAL BYTEMARK RESULTS========================== INTEGER INDEX : 15.873 FLOATING-POINT INDEX: 12.244 Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0 ==============================LINUX DATA BELOW=============================== CPU : POWER3-II 450MHz L2 Cache : 8MB OS : AIX 5.2 C compiler : gcc version 4.1.1 libc : MEMORY INDEX : 3.304 INTEGER INDEX : 4.538 FLOATING-POINT INDEX: 6.791 Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38 * Trademarks are property of their respective holder.

Code:

$ make gcc -Wall -O3 -mcpu=power3 -c pi_fftcs.c -o pi_fftcs.o gcc -Wall -O3 -mcpu=power3 -c fftsg_h.c -o fftsg_h.o gcc -Wall -O3 -mcpu=power3 pi_fftcs.o fftsg_h.o -lm -static -o pi_css5 Target "all" is up to date. $ ./pi_css5 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.00587898 calculating 1048576 digits of PI... AGM iteration precision= 48: 1.37 sec precision= 80: 1.37 sec precision= 176: 1.36 sec precision= 352: 1.37 sec precision= 688: 1.37 sec precision= 1392: 1.37 sec precision= 2784: 1.36 sec precision= 5584: 1.36 sec precision= 11168: 1.37 sec precision= 22336: 1.35 sec precision= 44688: 1.34 sec precision= 89408: 1.31 sec precision= 178816: 1.34 sec precision= 357648: 1.35 sec precision= 715312: 1.36 sec precision= 1430640: 1.36 sec writing pi1048576.txt... 25.57 sec. (real time)

Code:

$ java pi_fftcs 1000000 Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave initializing... nfft= 262144 radix= 10000 error_margin= 0.006286438480884726 calculating 1048576 digits of PI... AGM iteration precision= 48: 6.08 sec precision= 80: 3.56 sec precision= 176: 2.48 sec precision= 352: 1.89 sec precision= 688: 2.21 sec precision= 1392: 1.75 sec precision= 2784: 1.73 sec precision= 5584: 1.58 sec precision= 11168: 1.70 sec precision= 22336: 1.70 sec precision= 44688: 1.58 sec precision= 89408: 1.57 sec precision= 178816: 1.76 sec precision= 357648: 1.54 sec precision= 715312: 1.54 sec precision= 1430640: 1.55 sec writing pi1048576.txt 47.25 sec. (real time)

AIX 5.3

**oreissig** · 06.11.2011 02:47

heute: Sun SPARCclassic unter Solaris 2.5.1

Code:

BYTEmark* Native Mode Benchmark ver. 2 (10/95)

Index-split by Andrew D. Balsa (11/97)

Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index

                    :                  : Pentium 90* : AMD K6/233*

--------------------:------------------:-------------:------------

NUMERIC SORT        :          12.988  :       0.33  :       0.11

STRING SORT         :          0.3383  :       0.15  :       0.02

BITFIELD            :      4.0981e+06  :       0.70  :       0.15

FP EMULATION        :          1.1411  :       0.55  :       0.13

FOURIER             :            99.7  :       0.11  :       0.06

ASSIGNMENT          :          0.1261  :       0.48  :       0.12

IDEA                :          26.645  :       0.41  :       0.12

HUFFMAN             :          22.883  :       0.63  :       0.20

NEURAL NET          :         0.10571  :       0.17  :       0.07

LU DECOMPOSITION    :          4.4169  :       0.23  :       0.17

==========================ORIGINAL BYTEMARK RESULTS==========================

INTEGER INDEX       : 0.423

FLOATING-POINT INDEX: 0.164

Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0

==============================LINUX DATA BELOW===============================

CPU                 : 50MHz microSPARC

L2 Cache            :

OS                  : SunOS 5.5.1

C compiler          : gcc 3.4.2

libc                : /usr/lib/libdl.so.1

MEMORY INDEX        : 0.075

INTEGER INDEX       : 0.136

FLOATING-POINT INDEX: 0.091

Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38

* Trademarks are property of their respective holder.

pi_css5 hatte iteration times um die 65.30sec, darauf wollt ich jetz netmehr warten
morgen gehts weiter

EDIT: so

Code:

$ ./pi_css5 1000000

Calculation of PI using FFT and AGM, ver. LG1.1.2-MP1.5.2a.memsave

initializing...

nfft= 262144

radix= 10000

error_margin= 0.00628644

calculating 1048576 digits of PI...

AGM iteration

precision= 48: 65.46 sec

precision= 80: 65.16 sec

precision= 176: 65.25 sec

precision= 352: 65.28 sec

precision= 688: 65.12 sec

precision= 1392: 65.12 sec

precision= 2784: 65.17 sec

precision= 5584: 65.17 sec

precision= 11168: 65.18 sec

precision= 22336: 65.17 sec

precision= 44688: 65.17 sec

precision= 89408: 65.18 sec

precision= 178816: 65.13 sec

precision= 357648: 65.13 sec

precision= 715312: 65.14 sec

precision= 1430640: 65.16 sec

writing pi1048576.txt...

1219.19 sec. (real time)

**oreissig** · 06.11.2011 13:55

so, nochmal SPARCclassic unter Solaris 2.5.1, diesmal mit dem letzten 2er gcc

Code:

$ ./nbench

BYTEmark* Native Mode Benchmark ver. 2 (10/95)

Index-split by Andrew D. Balsa (11/97)

Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index

                    :                  : Pentium 90* : AMD K6/233*

--------------------:------------------:-------------:------------

NUMERIC SORT        :          13.928  :       0.36  :       0.12

STRING SORT         :         0.33333  :       0.15  :       0.02

BITFIELD            :       3.687e+06  :       0.63  :       0.13

FP EMULATION        :          1.1346  :       0.54  :       0.13

FOURIER             :          97.295  :       0.11  :       0.06

ASSIGNMENT          :         0.12933  :       0.49  :       0.13

IDEA                :          32.134  :       0.49  :       0.15

HUFFMAN             :          20.272  :       0.56  :       0.18

NEURAL NET          :        0.089446  :       0.14  :       0.06

LU DECOMPOSITION    :          3.7595  :       0.19  :       0.14

==========================ORIGINAL BYTEMARK RESULTS==========================

INTEGER INDEX       : 0.425

FLOATING-POINT INDEX: 0.146

Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0

==============================LINUX DATA BELOW===============================

CPU                 : 50MHz microSPARC

L2 Cache            :

OS                  : SunOS 5.5.1

C compiler          : gcc 2.95.2

libc                : /usr/lib/libdl.so.1

MEMORY INDEX        : 0.073

INTEGER INDEX       : 0.140

FLOATING-POINT INDEX: 0.081

Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38

* Trademarks are property of their respective holder.

Blue · 06.11.2011 14:13

fast as hell wut

**oreissig** · 06.11.2011 21:39

Blue schrieb: fast as hell wut

lol nee, der microSPARC is ne lahme krücke :fresse:

leider gibts nichts schnelleres in dem Formfaktor
[Bild: s_p_22647_1__72022_zoom.jpg]

EDIT: der microSPARC ist pro takt praktisch genau so schnell, wie ein alter SPARC V7
die einzigen CPUs, die ich mal gebenchmarkt habe und die pro takt noch langsamer waren, waren embedded Sachen in PDAs (StrongARM und SH-3 in HP Jornadas), allerdings warn die höher getaktet

**Xaar** · 06.11.2011 23:11

Mini-Würfel

~~chiaki~~ · 29.11.2011 19:45

dec 3000/400s unter tru64
133MHZ ALPHA
64mb RAM

Code:

Number of digits of pi to calculate?

1000000

initializing...

nfft= 262144

radix= 10000

error_margin= 0.00611182

calculating 1048576 digits of PI...

AGM iteration

precision= 48: 13.10 sec

precision= 80: 13.07 sec

precision= 176: 13.07 sec

precision= 352: 13.10 sec

precision= 688: 13.08 sec

precision= 1392: 13.00 sec

precision= 2784: 12.98 sec

precision= 5584: 12.98 sec

precision= 11168: 12.98 sec

precision= 22336: 12.98 sec

precision= 44688: 12.98 sec

precision= 89408: 13.00 sec

precision= 178816: 12.97 sec

precision= 357648: 13.00 sec

precision= 715312: 13.00 sec

precision= 1430640: 13.00 sec

writing pi1048576.txt...

244.22 sec. (real time)