new-ra code sizes on Hitachi H8

tm_gccmail@mail.kloo.net tm_gccmail@mail.kloo.net
Thu Feb 13 20:47:00 GMT 2003


7On Thu, 13 Feb 2003 tm_gccmail@mail.kloo.net wrote:

> Just for fun, here's the data I generated for the SH last year while
> tracking down code bloat issues:
> 

Updated with new data, but a little hard to read in 80 columns.

                                                         -nrb
-nrb
                 SH4    SH4-CVS  SH4CVS SH4CVS  SH4CVS   SH4    SH4
SH4
file            3.0.2   7/17/2   8/8/2  9/6/2   10/15/2 2002r1 2/13/3
2/13/3
----------------------------------------------------------------------------
* layer3.i       3cb4    4b64     4900   4b20    4b20    4bc0    4c80
4c80
* l3bitstream.i  1894    1914     1740   18a0    18a0    18a0    1980
1780
* navion_aero.i   5e0     5a0      600    5c0     5c0     5a0     5c0
600
@ advdomestic.i  1d2c    1ee0     1a20   1ea0    1e80    1e80    1ea0
1b20
@ aiunit.i       4374    4394     3e60   4320    43e0    43a0    44a0
3fa0
@ chanserv.i     de7c    e12c     c580   de20    dfc0    dfa0    e7a0
c660
@ melee2.i       60f4    5334     4fc0   5460    54a0    5480    56e0
50a0
@ wizard1.i      46f4    4474     4100   4700    4740    4700    4580
4060
+ adler32.i       130     114      120    120     120     120     120
140
+ blowfish.i     11f4    1188     1120   1160    1160    1160    1200
1140
+ imdecode.i    1ef38   1d754    1b0c0  1d7a0   1d880   1d8c0   1e640
1b540
+ map_fog.i      6bbc    8f20     72a0   73c0    73c0    7360    74c0
73e0
+ scanline.i      9c8     8f0      900    900     900     900     900
900
+ tif_fax3.i     1dec    1e08     1c00   1e00    1e00    1e00    1ea0
1c60
. quantize.i     247c    23b8     2160   2360    2360    2360    2500
2140
. tif_packbits.i  4fc     4d8      4a0    480     4a0     4a0     4a0
4c0
. s_serv.i       666c    6148     5c00   6060    6060    6060    63e0
5d20

* = heavily floating-point testcase
@ = branchy testcase
+ = loopy testcase
. = mixed testcase

-nrb = -fno-reorder-blocks

Note that BBRO increases code size about 7-15% on many of the 
testcases. The most extreme example is chanserv.i which is from Andy
Church's ircservices package. This code is almost entirely control logic
and suffers about 8k of code size increase on a 49k .text section.

Toshi




More information about the Gcc mailing list