Re: eliminate discreet arenaroots

[Jim Cromie <jim.cromie@gmail.com>]

I ran latest patch (timestamped-today) thru perlbench.
Since this isnt intended as a 'speed patch', I dont need to explain them ;-)


[jimc@harpo perlbench-0.07]$ mroe bench-out-1
A) perl-5.009003
        path        = ../dev/arena5-b1-bld/perl
        cc          = cc
        optimize    = -O2
        ccflags     = -D_REENTRANT -D_GNU_SOURCE -DTHREADS_HAVE_PIDS 
-fno-strict-aliasing -pipe -Wdeclaration-after-statement 
-I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 
-I/usr/include/gdbm
        usemymalloc = n

B) perl-5.009003
        path        = ../dev/arena5-b1-bldnt/perl
        cc          = cc
        optimize    = -O2
        ccflags     = -fno-strict-aliasing -pipe 
-Wdeclaration-after-statement -I/usr/local/include -D_LARGEFILE_SOURCE 
-D_FILE_OFFSET_BITS=64 -I/usr/include/gdbm
        usemymalloc = n

C) perl-5.009003
        path        = ../bleadperl/perl
        cc          = cc
        optimize    = -O2
        ccflags     = -D_REENTRANT -D_GNU_SOURCE -DTHREADS_HAVE_PIDS 
-fno-strict-aliasing -pipe -Wdeclaration-after-statement 
-I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 
-I/usr/include/gdbm
        usemymalloc = n

D) perl-5.009003
        path        = ../bleadperl-nothread/perl
        cc          = cc
        optimize    = -O2
        ccflags     = -fno-strict-aliasing -pipe 
-Wdeclaration-after-statement -I/usr/local/include -D_LARGEFILE_SOURCE 
-D_FILE_OFFSET_BITS=64 -I/usr/include/gdbm
        usemymalloc = n

                           A       B       C       D
                        ----    ----    ----    ----
arith/mixed              100     108      91     107
arith/trig               100     105      97     109
array/copy               100     106      99     106
array/foreach            100     126      98     104
array/index              100     115      87     105
array/pop                100     110     101     108
array/shift              100     110      99     109
array/sort-num           100     106     100     106
array/sort               100     104     100     104
call/0arg                100     127     100     123
call/1arg                100     126     102     113
call/2arg                100     120      99     115
call/9arg                100     115     101     115
call/empty               100     130      99     136
call/fib                 100     123     102     122
call/method              100     120     101     121
call/wantarray           100     117      99     112
hash/copy                100     107     101     107
hash/each                100     114      98     114
hash/foreach-sort        100     105      99     108
hash/foreach             100     115      99     115
hash/get                 100     136     100     131
hash/set                 100     108      97     101
loop/for-c               100     131     102     126
loop/for-range-const     100     110      90     113
loop/for-range           100     106      99     107
loop/getline             100     109     101     107
loop/while-my            100     123      99     123
loop/while               100     137     112     121
re/const                 100     107     101     107
re/w                     100      93      84      93
startup/fewmod           100     115     101     115
startup/lotsofsub        100     117     100     116
startup/noprog           100     113     100     115
string/base64            100     109     100     105
string/htmlparser        100     110     100     106
string/index-const       100     113     104     112
string/index-var         100     103     100     104
string/ipol              100     107      96     109
string/tr                100     102     102     105

AVERAGE                  100     114      99     112
[jimc@harpo perlbench-0.07]$

Assuming that bigger is faster,
B > A =>   non-threaded is better (as expected)
A > C, B > D  => my patch's simplifications gain something.

if bigger is slower, 1 or 2 points is in the noise  :-)

thanks
jimc

---

References to:

Jim Cromie <jim.cromie@gmail.com>
Nicholas Clark <nick@ccl4.org>

---

[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index][Thread Index][Top&Search][Original]

---