| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1292.1 | | QUARK::LIONEL | Free advice is worth every cent | Mon May 12 1997 10:33 | 6 |
| I don't see anything obviously wrong in the program, and the VAX compiler
generates the same answer for /OPT and /NOOPT. I note that the F90 compiler
matches F77's behavior, so it may indeed be an optimizer problem. I'm sure
Dave will have more to say on this later.
Steve
|
| 1292.2 | Yes, a code-generation bug | WIBBIN::NOYCE | Pulling weeds, pickin' stones | Mon May 12 1997 10:42 | 9 |
| Yes, it's a bug. The compiler computed the value of NNNNN, but forgot to
store it to memory across the I/O call. Later it fetched it from memory,
but got random garbage.
Obviously, the compiler doesn't usually make such mistakes. This arises only
for a particular combination of LOGICAL operators.
One "big-hammer" workaround would be to declare NNNNN volatile. Meanwhile,
the compiler fix is underway...
|
| 1292.3 | Fixed | TLE::EKLUND | Always smiling on the inside! | Mon May 12 1997 11:52 | 6 |
| We have fixed the problem. Thanks again for the simple
program!
Cheers!
Dave Eklund
|
| 1292.4 | That was 108 minutes for the fix - could we trim that under 1 hour ;-) | FRSTSC::TLAUER | "I've been designed multi-asking." | Tue May 13 1997 02:53 | 8 |
| Many thanks for that really fast solution!!! Is the fixed compiler available?
I'm sure my customer doesn't want to wait for a fully tested version, but get
hands on an improved version as quick as possible.
Thanks again,
-- Thilo
|
| 1292.5 | | QUARK::LIONEL | Free advice is worth every cent | Tue May 13 1997 09:56 | 3 |
| This fix should make the ECO kit we'll put out next week.
Steve
|
| 1292.6 | | TLE::EKLUND | Always smiling on the inside! | Tue May 13 1997 10:22 | 5 |
| The Fortran team has a keen appreciation for great performance...
Cheers!
Dave Eklund
|
| 1292.7 | unable to locate the promised ECO | FRSTSC::TLAUER | "I've been designed multi-asking." | Fri Jun 06 1997 05:16 | 14 |
| Hello,
today i got the customer's response that the promised ECO is still not
available, and indeed it's not visible within TIMA TOOLS. May i kindly ask for a
special private version of the planned ECO to provide a solution for the
customer? Until a final fix they have to use /NOOPT for all their BIIIG FORTRAN
programs (this is a university), and have to waste expensive cpu cycles...
thanks and regards,
-- Thilo
|
| 1292.8 | also... | CERN::HOBBS | Congrats to the Ignoble Peace Prize winner! (http://www.eecs.harvard.edu/ig_nobel) | Fri Jun 06 1997 06:34 | 7 |
| This program fails on DIGITAL Visual Fortran 5.0.1b1 on Alpha, but is
OK on Intel.
Do you want a standard bug report on DVF, or is the fix in the source
pool already?
-cw
|
| 1292.9 | DVF T5.0.1 Beta 2 will have the fix | TLE::WHITLOCK | Stan Whitlock | Fri Jun 06 1997 09:32 | 12 |
| >>This program fails on DIGITAL Visual Fortran 5.0.1b1 on Alpha, but is
>>OK on Intel.
>>Do you want a standard bug report on DVF, or is the fix in the source
>>pool already?
DVF T5.0.1 Beta 1 was frozen on 16-Apr so it could not contain this fix. The
fix is in the sources so it will appread in DVF T5.0.1 Beta 2.
No extra bug report is necessary.
/Stan
|
| 1292.10 | yes, give them DFA411 | TLE::WHITLOCK | Stan Whitlock | Fri Jun 06 1997 09:35 | 15 |
| >>today i got the customer's response that the promised ECO is still not
>>available, and indeed it's not visible within TIMA TOOLS.
The ECO was submitted yesterday so it isn't yet on TIMA.
>>May i kindly ask for a special private version of the planned ECO to provide
>>a solution for the customer?
The ECO is announced in note 21.last... and says that the kit is in
ftp://format.zko.dec.com/DFA411.tar
Yes, you can give that kit to the customer.
/Stan
|
| 1292.11 | | QUARK::LIONEL | Free advice is worth every cent | Fri Jun 06 1997 10:46 | 4 |
| Unfortunately, Stan, the ECO kit announced is for UNIX - the customer reported
the problem on OpenVMS. We should have the OpenVMS kit ready next week.
Steve
|