emv@math.lsa.umich.edu (Edward Vielmetti) (07/29/90)
perl on dec 5000, ultrix 4.0.
fails tests 40-42 in op.s, all having to do with $&.
fails tests 8 and 9 in op.eval, dealing with recursive evals.
I get a warning from the compiler like so for eval.c (2601), toke.c (2089),
teval.c (2618), and yylex (2089), also walk.c from x2p (1510).
uopt: Warning: %s: this procedure not optimized because it
exceeds size threshold; to optimize this procedure, use -Olimit option
with value >= %d.
next step is to try to compile these routines as optimized.
--Ed
Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>thewalt@canuck.ce.berkeley.edu (C. Thewalt) (07/30/90)
In article <EMV.90Jul29143827@urania.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes:
perl on dec 5000, ultrix 4.0.
fails tests 40-42 in op.s, all having to do with $&.
fails tests 8 and 9 in op.eval, dealing with recursive evals.
...
--Ed
Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>
I compiled everything without -O and all tests pass.
Chris
--
Christopher Robin Thewalt These opinions are not necessarily
thewalt@ce.berkeley.edu shared by my employer...
University of California, Berkeleyemv@math.lsa.umich.edu (Edward Vielmetti) (07/30/90)
In article <THEWALT.90Jul29124249@canuck.ce.berkeley.edu> thewalt@canuck.ce.berkeley.edu (C. Thewalt) writes:
Ed said: perl on dec 5000, ultrix 4.0.
fails tests 40-42 in op.s, all having to do with $&.
fails tests 8 and 9 in op.eval, dealing with recursive evals.
Chris said: I compiled everything without -O and all tests pass.
Ed said: thanks Chris and Michael Stryker (strkyer@phyc.ucsf.edu),
compiling with -g (not -O) and -DCRIPPLED_CC and it passes all
tests.
--Ed
Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>tmaeno@cc.titech.ac.jp (Toshinori Maeno) (07/30/90)
In <EMV.90Jul29213617@urania.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes: >In article <THEWALT.90Jul29124249@canuck.ce.berkeley.edu> thewalt@canuck.ce.berkeley.edu (C. Thewalt) writes: > Ed said: perl on dec 5000, ultrix 4.0. > fails tests 40-42 in op.s, all having to do with $&. > fails tests 8 and 9 in op.eval, dealing with recursive evals. > Chris said: I compiled everything without -O and all tests pass. > Ed said: thanks Chris and Michael Stryker (strkyer@phyc.ucsf.edu), > compiling with -g (not -O) and -DCRIPPLED_CC and it passes all > tests. We had the same problem on our MIPS3240 and Sony RISC NEWS. Mr. Ohta found that Mips C compiler generates buggy code for perly.c when we compiled with -O. Maeno Toshinori -- Toshinori Maeno Computer Center, Tokyo Institute of Technology
meissner@osf.org (Michael Meissner) (07/31/90)
In article <5960@titcce.cc.titech.ac.jp> tmaeno@cc.titech.ac.jp (Toshinori Maeno) writes: | In <EMV.90Jul29213617@urania.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes: | | >In article <THEWALT.90Jul29124249@canuck.ce.berkeley.edu> thewalt@canuck.ce.berkeley.edu (C. Thewalt) writes: | | > Ed said: perl on dec 5000, ultrix 4.0. | > fails tests 40-42 in op.s, all having to do with $&. | > fails tests 8 and 9 in op.eval, dealing with recursive evals. | | > Chris said: I compiled everything without -O and all tests pass. | | > Ed said: thanks Chris and Michael Stryker (strkyer@phyc.ucsf.edu), | > compiling with -g (not -O) and -DCRIPPLED_CC and it passes all | > tests. | We had the same problem on our MIPS3240 and Sony RISC NEWS. | Mr. Ohta found that Mips C compiler generates buggy code for perly.c | when we compiled with -O. | Maeno Toshinori If you saved the recent GNU CC patches that I sent to the gnu CC bug list, the compiler produced will compile perl with normal and -O2 optimizations. Note that the patches I submitted do not allow for full debugging, since RMS requested that I delay the debug patches until GCC revision 2.00. -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA Do apple growers tell their kids money doesn't grow on bushes?
emv@math.lsa.umich.edu (Edward Vielmetti) (07/31/90)
Ed said: perl on dec 5000, ultrix 4.0.
fails tests 40-42 in op.s, all having to do with $&.
fails tests 8 and 9 in op.eval, dealing with recursive evals.
Chris said: I compiled everything without -O and all tests pass.
Ed said: thanks Chris and Michael Stryker (strkyer@phyc.ucsf.edu),
compiling with -g (not -O) and -DCRIPPLED_CC and it passes all
tests.
tmaeno@cc.titech.ac.jp (Toshinori Maeno) writes that they
had the same problem on our MIPS3240 and Sony RISC NEWS.
Mr. Ohta found that Mips C compiler generates buggy code for perly.c
when we compiled with -O.
Carl Fongheiser reports the same problem on UWS 2.2 on a 2100,
and he got it to work when he undefined HAS_VOLATILE in config.h.
"Apparently, Larry Wall and the MIPS compiler have different ideas
as to what 'volatile' means." <cmf@unix.cis.pitt.edu>
Ed now tries to compile and optimize, and finds that eval.c took over
14 hours to compile on a DS 5000 (!) as
cc -c -DLANGUAGE_C -O -Olimit 2800 eval.c
and it looks like toke.c is about the same. Time to buy more
memory, it only has 8M in it right now.
I'll let you know day after tomorrow when toke.c, teval.c, and ttoke.c are
done compiling. If Mr. Maeno would try undefining HAS_VOLATILE on the
Sony RISC NEWS C compiler and see if perl works with -O then that would
be useful to know.
--Ed
Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>