X-pop3-spooler: POP3MAIL 2.1.0 b 4 980420 -bs- Message-ID: <360F1E5D.611EDBE8@uwyo.edu> Date: Sun, 27 Sep 1998 23:27:57 -0600 From: Katlun Yong X-Mailer: Mozilla 4.5b2 [en] (X11; I; Linux 2.0.35 i686) X-Accept-Language: en MIME-Version: 1.0 To: Marc Lehmann CC: beastium Subject: Re: Snapshot 19980921 References: <360C34DB DOT A97C02AF AT uwyo DOT edu> <19980927233749 DOT B385 AT cerebro DOT laendle> Content-Type: multipart/alternative; boundary="------------FA74AA4E171DC34463C6A2E7" Sender: Marc Lehmann Status: RO Content-Length: 16066 Lines: 295 --------------FA74AA4E171DC34463C6A2E7 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Marc Lehmann wrote: > On Fri, Sep 25, 1998 at 06:27:07AM -0600, Katlun Yong wrote: > > It seems the floaing point performace is really slow in this > > snapshot? About 30% slower then 1.1b > > > Anything's been turn off in this??? > > Not that I'd know. which compiler options did you use? > > -----==- | > ----= > > -- _ > | > > ---==---(_)__ __ ____ __ > Marc Lehmann +-- > > --==---/ / _ \/ // /\ \/ / > pcg AT goof DOT com |e| > > -=====/_/_//_/\_,_/ /_/\_\ > --+ > > The choice of a GNU generation > | > > > | > I am using redhat-5.1. Kernel 2.0.35 compiled with pgcc-1.1b glibc 2.0.7 binutils-2.9.1.0.9 hmmm... possible compile with pgcc-1.0.3. anyway. CFLAGS = -s -static -O6 -fomit-frame-pointer -Wall -mpentiumpro -fforce-addr \ -fforce-mem -malign-loops=4 -malign-functions=4 \ -malign-jumps=4\ -malign-double \ -fthread-jumps \ -funroll-all-loops \ -funroll-loops \ -fstrength-reduce \ Compare the result for the FP Emulation/Fourier/LU Decomposition: Result with pgcc-1.1b: : : Pentium 90* : AMD K6/233* --------------------:------------------:-------------:------------ NUMERIC SORT : 215.04 : 5.51 : 1.81 STRING SORT : 21.099 : 9.43 : 1.46 BITFIELD : 5.9514e+07 : 10.21 : 2.13 FP EMULATION : 18.77 : 9.01 : 2.08 FOURIER : 4743.4 : 5.39 : 3.03 ASSIGNMENT : 2.1476 : 8.17 : 2.12 IDEA : 499.6 : 7.64 : 2.27 HUFFMAN : 204.69 : 5.68 : 1.81 NEURAL NET : 5.3318 : 8.57 : 3.60 LU DECOMPOSITION : 103.32 : 5.35 : 3.87 INTEGER INDEX : 7.760 FLOATING-POINT INDEX: 6.277 Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0 ==============================LINUX DATA BELOW=============================== C compiler : gcc version pgcc-2.91.57 19980901 (egcs-1.1 release) libc : unknown version MEMORY INDEX : 1.875 INTEGER INDEX : 1.984 FLOATING-POINT INDEX : 3.481 Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38 Result with snapshot 19980921 : : Pentium 90* : AMD K6/233* --------------------:------------------:-------------:------------ NUMERIC SORT : 214.44 : 5.50 : 1.81 STRING SORT : 20.533 : 9.17 : 1.42 BITFIELD : 6.0271e+07 : 10.34 : 2.16 FP EMULATION : 17.2 : 8.25 : 1.90 FOURIER : 4388 : 4.99 : 2.80 ASSIGNMENT : 2.1817 : 8.30 : 2.15 IDEA : 530.66 : 8.12 : 2.41 HUFFMAN : 208.73 : 5.79 : 1.85 NEURAL NET : 5.3635 : 8.62 : 3.62 LU DECOMPOSITION : 68.453 : 3.55 : 2.56 INTEGER INDEX : 7.750 FLOATING-POINT INDEX: 5.342 Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0 ==============================LINUX DATA BELOW=============================== C compiler : gcc version pgcc-2.92.11 19980921 (gcc2 ss-980609 experimental) libc : unknown version MEMORY INDEX : 1.876 INTEGER INDEX : 1.978 FLOATING-POINT INDEX : 2.963 Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38 P.S. I'll recompile the binutils with pgcc-1.1b and recompile the snapshot. --------------FA74AA4E171DC34463C6A2E7 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit Marc Lehmann wrote:
On Fri, Sep 25, 1998 at 06:27:07AM -0600, Katlun Yong wrote:
> It seems  the floaing point performace is really slow in this
> snapshot?  About 30% slower then 1.1b

> Anything's been turn off in this???

Not that I'd know. which compiler options did you use?

      -----==-                                              |
      ----=

-- _                                           
|

      ---==---(_)__  __ ____  __      
Marc Lehmann       +--

      --==---/ / _ \/ // /\ \/ /      
pcg AT goof DOT com       |e|

      -=====/_/_//_/\_,_/ /_/\_\                         
--+

    The choice of a GNU generation                       
|

                                                         
|
I am using redhat-5.1. Kernel 2.0.35 compiled with pgcc-1.1b
glibc 2.0.7
binutils-2.9.1.0.9 hmmm... possible compile with pgcc-1.0.3. anyway.
 

CFLAGS = -s -static -O6 -fomit-frame-pointer -Wall -mpentiumpro -fforce-addr \
          -fforce-mem -malign-loops=4 -malign-functions=4 \
          -malign-jumps=4\
          -malign-double \
          -fthread-jumps \
          -funroll-all-loops \
          -funroll-loops \
          -fstrength-reduce \

Compare the result for the FP Emulation/Fourier/LU Decomposition:
 
 

Result with pgcc-1.1b:

                    :                  :                         Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          215.04  :       5.51  :       1.81
STRING SORT            :          21.099  :       9.43  :       1.46
BITFIELD                     :      5.9514e+07  :      10.21  :       2.13
FP EMULATION        :           18.77  :       9.01                                      :       2.08
FOURIER                      :          4743.4  :       5.39                                     :       3.03
ASSIGNMENT            :          2.1476  :       8.17  :       2.12
IDEA                               :           499.6  :       7.64  :       2.27
HUFFMAN                    :          204.69  :       5.68  :       1.81
NEURAL NET               :          5.3318  :       8.57  :       3.60
LU DECOMPOSITION    :          103.32  :       5.35                              :       3.87
INTEGER INDEX       : 7.760
FLOATING-POINT INDEX: 6.277
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
C compiler          : gcc version pgcc-2.91.57 19980901 (egcs-1.1 release)
libc                : unknown version
MEMORY INDEX        : 1.875
INTEGER INDEX       : 1.984
FLOATING-POINT INDEX             : 3.481
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
 

Result with snapshot 19980921

                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          214.44  :       5.50  :       1.81
STRING SORT         :          20.533  :       9.17  :       1.42
BITFIELD            :      6.0271e+07  :      10.34  :       2.16
FP EMULATION        :            17.2  :       8.25                                  :       1.90
FOURIER             :            4388  :       4.99  :       2.80
ASSIGNMENT          :          2.1817  :       8.30                                 :       2.15
IDEA                :          530.66  :       8.12  :       2.41
HUFFMAN             :          208.73  :       5.79  :       1.85
NEURAL NET          :          5.3635  :       8.62  :       3.62
LU DECOMPOSITION    :          68.453  :       3.55                        :       2.56
INTEGER INDEX       : 7.750
FLOATING-POINT INDEX: 5.342
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
C compiler          : gcc version pgcc-2.92.11 19980921 (gcc2 ss-980609 experimental)
libc                        : unknown version
MEMORY INDEX        : 1.876
INTEGER INDEX         : 1.978
FLOATING-POINT INDEX            : 2.963
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
 
 

P.S. I'll recompile the binutils with pgcc-1.1b and recompile the snapshot. --------------FA74AA4E171DC34463C6A2E7--