On Sun, Nov 26, 2006 at 06:08:23AM +0000, Deanna Phillips wrote:
> Deanna Phillips writes:
> > The old workarounds weren't needed on i386 or amd64, so I
> > commented them out. I'd appreciate it if anyone could test
> > whether or not they are still necessary on sparc64.
>
> I did manage to test this on sparc64, and the ICE is still
> there, so here's a new diff. The benchmarks finish on amd64,
> i386 and sparc64 (which had the best results by far).
'Make regress' fails for me on i386. (It does terminate, so you might be
aware of this issue; the wording was not sufficiently clear for me to be
sure.)
===> Regression check for chicken-2.5
./csc cscbench.scm -o cscbench -compiler ./chicken
/usr/local/lib/libchicken.so.1.0
/usr/local/lib/libchicken.so.1.0
/usr/local/lib/libchicken.so.1.0
cd benchmarks; ../cscbench
___| | _) |
| __ \ | __| | / _ \ __ \
| | | | | ( < __/ | |
\____|_| |_|_|\___|_|\_\\___|_| _|
Version 2.5 - openbsd-unix-gnu-x86 - [ dload ptables applyhook ]
(c)2000-2006 Felix L. Winkelmann
CC:
Reading specs from /usr/lib/gcc-lib/i386-unknown-openbsd4.0/3.3.5/specs
Configured with:
Thread model: single
gcc version 3.3.5 (propolice)
CFLAGS:
-DHAVE_CHICKEN_CONFIG_H -O2 -pipe -I/usr/local/include -DC_ENABLE_PTABLES
-DC_NO_PIC_NO_DLL
base unsafe max
boyer
0.130
FAILED
FAILED
browse
0.130
FAILED
0.011
conform
0.245
0.107
FAILED
cpstak
0.641
0.544
0.419
ctak
0.252
0.200
0.182
dderiv
0.255
0.198
0.190
deriv
0.185
0.151
0.138
destructive
0.159
0.151
FAILED
div-iter
FAILED
FAILED
FAILED
div-rec
0.138
FAILED
FAILED
dynamic
0.143
0.115
0.107
earley
FAILED
FAILED
FAILED
fft
FAILED
FAILED
fib
0.323
0.266
FAILED
fibc
3.708
1.056
0.961
fprint
0.000
FAILED
FAILED
fread
FAILED
FAILED
FAILED
hanoi
0.815
0.784
0.117
lattice
12.630
10.253
10.294
maze
0.313
0.018
nqueens
0.000
0.000
0.000
puzzle
0.102
FAILED
FAILED
scheme
FAILED
FAILED
FAILED
tak
0.721
0.565
FAILED
takl
0.398
0.101
FAILED
takr
0.922
0.596
0.422
traverse
0.963
0.606
0.588
travinit
FAILED
FAILED
FAILED
triangl
1.468
1.130
0.945
It is worth noting that there is some randomness in the results,
including in which ones do and do not fail. A second run, after `make
clean', resulted in the following differences (the benchmarks are
meaningless, as I was doing quite a few other things at the same time in
both cases):
--- /home/joachim/out Wed Nov 29 00:35:54 2006
+++ /home/joachim/out2 Wed Nov 29 00:35:05 2006
@@ -19,49 +19,49 @@
-DHAVE_CHICKEN_CONFIG_H -O2 -pipe -I/usr/local/include -DC_ENABLE_PTABLES
-DC_NO_PIC_NO_DLL
base unsafe max
boyer
- 0.130
+ 0.141
FAILED
FAILED
browse
- 0.130
- FAILED
- 0.011
+ 0.133
+ 0.104
+ 0.102
conform
- 0.245
- 0.107
- FAILED
+ 0.237
+ FAILED
+ 0.102
cpstak
- 0.641
- 0.544
- 0.419
+ 0.638
+ 0.541
+ 0.416
ctak
- 0.252
- 0.200
- 0.182
-dderiv
0.255
0.198
- 0.190
+ 0.182
+dderiv
+ 0.260
+ 0.188
+ 0.188
deriv
- 0.185
- 0.151
- 0.138
+ 0.190
+ 0.154
+ 0.151
destructive
0.159
- 0.151
+ 0.149
FAILED
div-iter
FAILED
FAILED
FAILED
div-rec
- 0.138
+ 0.135
FAILED
FAILED
dynamic
- 0.143
- 0.115
- 0.107
+ 0.146
+ 0.109
+ 0.011
earley
FAILED
FAILED
@@ -70,13 +70,13 @@
FAILED
FAILED
fib
- 0.323
- 0.266
+ 0.320
+ 0.258
FAILED
fibc
- 3.708
- 1.056
- 0.961
+ 3.690
+ 1.562
+ 0.974
fprint
0.000
FAILED
@@ -87,48 +87,48 @@
FAILED
hanoi
0.815
- 0.784
- 0.117
+ 0.782
+ 0.115
lattice
- 12.630
- 10.253
- 10.294
+ 12.609
+ 10.262
+ 10.234
maze
- 0.313
- 0.018
+ 0.302
+ 0.185
nqueens
0.000
0.000
0.000
puzzle
- 0.102
FAILED
+ FAILED
FAILED
scheme
FAILED
FAILED
FAILED
tak
- 0.721
- 0.565
+ 0.711
+ 0.563
FAILED
takl
- 0.398
- 0.101
+ 0.399
+ FAILED
FAILED
takr
- 0.922
- 0.596
- 0.422
+ 0.093
+ 0.604
+ 0.427
traverse
- 0.963
- 0.606
- 0.588
+ 0.976
+ 0.062
+ 0.595
travinit
FAILED
FAILED
FAILED
triangl
- 1.468
- 1.130
- 0.945
+ 1.492
+ 1.151
+ 0.958
It otherwise builds without errors. csi appears to work, although I know
very little about scheme. I did not test csc (yet).
You might want to add --without-libffi --without-pcre to CONFIGURE_ARGS,
and/or depend on pcre, since configure seems to want to pull that in
(but, at least on my system, fails to compile pcre:
checking pcre.h presence... no
configure: WARNING: pcre.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: pcre.h: proceeding with the compiler's result).
Joachim
