This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/21485] [4.2/4.3/4.4 Regression] codegen regression due to PRE increasing register pressure (missing load PRE really)
- From: "rguenth at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 3 Oct 2008 23:18:33 -0000
- Subject: [Bug tree-optimization/21485] [4.2/4.3/4.4 Regression] codegen regression due to PRE increasing register pressure (missing load PRE really)
- References: <bug-21485-10607@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #28 from rguenth at gcc dot gnu dot org 2008-10-03 23:18 -------
nbench 2.2.3 numeric sort test executes 40% less iterations per second
when compiled with 4.4 snapshot than with 3.4.6
iterations/s - version
2439 - 3.4.6
1530 - 4.4.0 20080926 (experimental)
1526 - 4.3.2
1580 - 4.2.4
CFLAGS = -s -static -Wall -O3 -g0 -march=nocona -fomit-frame-pointer
-funroll-loops -ffast-math
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 : 2439.1 : 62.55 : 20.54
STRING SORT : 373.28 : 166.79 : 25.82
BITFIELD : 5.9879e+08 : 102.71 : 21.45
FP EMULATION : 267.84 : 128.52 : 29.66
FOURIER : 43702 : 49.70 : 27.92
ASSIGNMENT : 56.657 : 215.59 : 55.92
IDEA : 5407.7 : 82.71 : 24.56
HUFFMAN : 3204.3 : 88.86 : 28.37
NEURAL NET : 57.485 : 92.35 : 38.84
LU DECOMPOSITION : 2363.5 : 122.44 : 88.41
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 111.792
FLOATING-POINT INDEX: 82.519
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU : Dual GenuineIntel Intel(R) Core(TM)2 Duo CPU E6750 @
2.66GHz 3200MHz
L2 Cache : 4096 KB
OS : Linux 2.6.26.5
C compiler : gcc version 3.4.6 (Gentoo 3.4.6-r2 p1.5, ssp-3.4.6-1.0,
pie-8.7.10)
libc : libc-2.8.so
MEMORY INDEX : 31.404
INTEGER INDEX : 25.525
FLOATING-POINT INDEX: 45.768
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.
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 : 1530.9 : 39.26 : 12.89
STRING SORT : 372.64 : 166.51 : 25.77
BITFIELD : 6.0348e+08 : 103.52 : 21.62
FP EMULATION : 310.4 : 148.94 : 34.37
FOURIER : 31760 : 36.12 : 20.29
ASSIGNMENT : 48.361 : 184.02 : 47.73
IDEA : 9204 : 140.77 : 41.80
HUFFMAN : 3554.3 : 98.56 : 31.47
NEURAL NET : 73.882 : 118.69 : 49.92
LU DECOMPOSITION : 2322.2 : 120.30 : 86.87
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 114.457
FLOATING-POINT INDEX: 80.190
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU : Dual GenuineIntel Intel(R) Core(TM)2 Duo CPU E6750 @
2.66GHz 3200MHz
L2 Cache : 4096 KB
OS : Linux 2.6.26.5
C compiler : gcc version 4.4.0 20080926 (experimental) (GCC)
libc : libc-2.8.so
MEMORY INDEX : 29.851
INTEGER INDEX : 27.632
FLOATING-POINT INDEX: 44.477
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.
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 : 1526.2 : 39.14 : 12.85
STRING SORT : 374.28 : 167.24 : 25.89
BITFIELD : 6.0262e+08 : 103.37 : 21.59
FP EMULATION : 320.64 : 153.86 : 35.50
FOURIER : 33352 : 37.93 : 21.30
ASSIGNMENT : 57.834 : 220.07 : 57.08
IDEA : 9288 : 142.06 : 42.18
HUFFMAN : 3211 : 89.04 : 28.43
NEURAL NET : 74.69 : 119.98 : 50.47
LU DECOMPOSITION : 2655.7 : 137.58 : 99.35
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 116.415
FLOATING-POINT INDEX: 85.548
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU : Dual GenuineIntel Intel(R) Core(TM)2 Duo CPU E6750 @
2.66GHz 3200MHz
L2 Cache : 4096 KB
OS : Linux 2.6.26.5
C compiler : gcc-4.3.2
libc : libc-2.8.so
MEMORY INDEX : 31.716
INTEGER INDEX : 27.199
FLOATING-POINT INDEX: 47.448
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.
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 : 1580.2 : 40.52 : 13.31
STRING SORT : 359.88 : 160.80 : 24.89
BITFIELD : 5.9462e+08 : 102.00 : 21.30
FP EMULATION : 291.28 : 139.77 : 32.25
FOURIER : 32567 : 37.04 : 20.80
ASSIGNMENT : 57.097 : 217.26 : 56.35
IDEA : 7876.8 : 120.47 : 35.77
HUFFMAN : 3433.1 : 95.20 : 30.40
NEURAL NET : 72.091 : 115.81 : 48.71
LU DECOMPOSITION : 2664.4 : 138.03 : 99.67
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 112.739
FLOATING-POINT INDEX: 83.966
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU : Dual GenuineIntel Intel(R) Core(TM)2 Duo CPU E6750 @
2.66GHz 3200MHz
L2 Cache : 4096 KB
OS : Linux 2.6.26.5
C compiler : gcc-4.2.4
libc : libc-2.8.so
MEMORY INDEX : 31.032
INTEGER INDEX : 26.138
FLOATING-POINT INDEX: 46.571
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.
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Known to fail|4.0.0 4.0.4 4.1.0 |4.0.0 4.0.4 4.1.0 4.2.4
| |4.3.2 4.4.0
Summary|[4.2 Regression] codegen |[4.2/4.3/4.4 Regression]
|regression due to PRE |codegen regression due to
|increasing register pressure|PRE increasing register
|(missing load PRE really) |pressure (missing load PRE
| |really)
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21485