[ ghc-Bugs-604889 ] Optimizer introduces non-termination
Bugs item #604889, was opened at 2002-09-05 04:54
You can respond by visiting:
https://sourceforge.net/tracker/?func=detailatid=108032aid=604889group_id=8032
Category: Compiler
Group: 5.04
Status: Closed
Resolution: Fixed
Priority: 5
Submitted By: Henrik Nilsson (nhn)
Assigned to: Simon Marlow (simonmar)
Summary: Optimizer introduces non-termination
Initial Comment:
When compiled with GHC 5.04 and optimization turned on
(-O), AFRP's regression tests fail to terminate. When
compiled with -O0, everything works, except that
somewhat dubious C code seems to be generated (see
below).
I've included a version of the regression tests with
as much code as possible commented out, but you may
need the entire AFRP distribution to make progress on
this one. You can get it as part of the AFrob/AFRP
bundle at www.haskell.org/afrp.
In the regression tests, the following logical
conjunction is used to check if all tests succeeded:
allGood = {- arr_tr
comp_tr
-} first_tr
second_tr
laws_tr
loop_tr {-
looplaws_tr
basicsf_tr
evsrc_tr
coc_tr
switch_tr
kswitch_tr
rswitch_tr
pswitch_tr
rpswitch_tr
wfg_tr
accum_tr
delay_tr
der_tr
loopPre_tr
loopIntegral_tr
react_tr
embed_tr
utils_tr
task_tr
-}
With only the code for the tests first, second,
laws, and loop left, and consequently only the
corresponding four boolean test results left in the
above conjunction, the regression tests fail to
terminate when compiled with -O.
However, if I comment out ANY ONE (or more) of the
four boolean result flags from the conjunction above,
the remaining tests do terminate. Thus there does
not appear to be one single test that fails, but
somehow it is the combination of these four tests
that for some reason results in invalid optimizations.
Additionally, when I compile the main module
(testAFRPMain.hs) with no optimization (-O0), I get
the follwing warning:
ghc -c -fglasgow-exts -O -package lang -package
concurrent -i/usr/local/lib/ghc/imports/arrow:../src
-O0 -o testAFRPMain.o testAFRPMain.hs
/tmp/ghc17120.hc:1150: warning: initialization discards
qualifiers from pointer target type
Again, I'm a bit surprised that C code is generated
here? I thought I was using the native code
generator? This is on a Linux x86 platform.
/Henrik
--
Henrik Nilsson
Yale University
[EMAIL PROTECTED]
--
Comment By: Simon Marlow (simonmar)
Date: 2002-09-06 12:53
Message:
Logged In: YES
user_id=48280
This was caused by a bug in the garbage collector, which
has now been fixed. Thanks for the report.
--
You can respond by visiting:
https://sourceforge.net/tracker/?func=detailatid=108032aid=604889group_id=8032
___
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
[ ghc-Bugs-604889 ] Optimizer introduces non-termination
Bugs item #604889, was opened at 2002-09-05 04:54
You can respond by visiting:
https://sourceforge.net/tracker/?func=detailatid=108032aid=604889group_id=8032
Category: Compiler
Group: 5.04
Status: Open
Resolution: None
Priority: 5
Submitted By: Henrik Nilsson (nhn)
Assigned to: Nobody/Anonymous (nobody)
Summary: Optimizer introduces non-termination
Initial Comment:
When compiled with GHC 5.04 and optimization turned on
(-O), AFRP's regression tests fail to terminate. When
compiled with -O0, everything works, except that
somewhat dubious C code seems to be generated (see
below).
I've included a version of the regression tests with
as much code as possible commented out, but you may
need the entire AFRP distribution to make progress on
this one. You can get it as part of the AFrob/AFRP
bundle at www.haskell.org/afrp.
In the regression tests, the following logical
conjunction is used to check if all tests succeeded:
allGood = {- arr_tr
comp_tr
-} first_tr
second_tr
laws_tr
loop_tr {-
looplaws_tr
basicsf_tr
evsrc_tr
coc_tr
switch_tr
kswitch_tr
rswitch_tr
pswitch_tr
rpswitch_tr
wfg_tr
accum_tr
delay_tr
der_tr
loopPre_tr
loopIntegral_tr
react_tr
embed_tr
utils_tr
task_tr
-}
With only the code for the tests first, second,
laws, and loop left, and consequently only the
corresponding four boolean test results left in the
above conjunction, the regression tests fail to
terminate when compiled with -O.
However, if I comment out ANY ONE (or more) of the
four boolean result flags from the conjunction above,
the remaining tests do terminate. Thus there does
not appear to be one single test that fails, but
somehow it is the combination of these four tests
that for some reason results in invalid optimizations.
Additionally, when I compile the main module
(testAFRPMain.hs) with no optimization (-O0), I get
the follwing warning:
ghc -c -fglasgow-exts -O -package lang -package
concurrent -i/usr/local/lib/ghc/imports/arrow:../src
-O0 -o testAFRPMain.o testAFRPMain.hs
/tmp/ghc17120.hc:1150: warning: initialization discards
qualifiers from pointer target type
Again, I'm a bit surprised that C code is generated
here? I thought I was using the native code
generator? This is on a Linux x86 platform.
/Henrik
--
Henrik Nilsson
Yale University
[EMAIL PROTECTED]
--
You can respond by visiting:
https://sourceforge.net/tracker/?func=detailatid=108032aid=604889group_id=8032
___
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
