public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report
@ 2004-11-12 2:16 giovannibajo at libero dot it
2004-11-12 2:45 ` [Bug tree-optimization/18441] " pinskia at gcc dot gnu dot org
` (5 more replies)
0 siblings, 6 replies; 7+ messages in thread
From: giovannibajo at libero dot it @ 2004-11-12 2:16 UTC (permalink / raw)
To: gcc-bugs
If an user wants to know if vectorization succeeded or not, he has to use -
fdump-tree-vect-details and then grep for some special pattern to find the
lines that say if loops were vectorized or not.
Such an information can be very useful while developing code, so I think it
should be easily given. ICC has an option /Qvec_report[=n] to setup a
diagnostic (equivalent to our inform()), with different verbosity levels. I
suggest we add something like this to GCC as well.
/Qvec_report[n] control amount of vectorizer diagnostic information:
n=0 no diagnostic information
n=1 indicate vectorized loops (DEFAULT)
n=2 indicate vectorized/non-vectorized loops
n=3 indicate vectorized/non-vectorized loops and prohibiting
data dependence information
n=4 indicate non-vectorized loops
n=5 indicate non-vectorized loops and prohibiting data
dependence information
With such an infrastructure, we could also change the vectorizer testsuite not
to scan the details dump and just do a { dg-inform "VECTORIZED" } (or similar)
on the line with the loop header.
--
Summary: Vectorizer: add a command line for simple vectorizer
report
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Severity: enhancement
Priority: P2
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: giovannibajo at libero dot it
CC: gcc-bugs at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
@ 2004-11-12 2:45 ` pinskia at gcc dot gnu dot org
2004-12-13 20:16 ` dpatel at apple dot com
` (4 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2004-11-12 2:45 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From pinskia at gcc dot gnu dot org 2004-11-12 02:45 -------
Confirmed, We should have done this by now but don't for reasons unknown to me. We have all the
information we need as of right now.
--
What |Removed |Added
----------------------------------------------------------------------------
CC| |pinskia at gcc dot gnu dot
| |org
Status|UNCONFIRMED |NEW
Ever Confirmed| |1
Keywords| |diagnostic
Last reconfirmed|0000-00-00 00:00:00 |2004-11-12 02:45:12
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
2004-11-12 2:45 ` [Bug tree-optimization/18441] " pinskia at gcc dot gnu dot org
@ 2004-12-13 20:16 ` dpatel at apple dot com
2004-12-22 19:00 ` leehod at il dot ibm dot com
` (3 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: dpatel at apple dot com @ 2004-12-13 20:16 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From dpatel at apple dot com 2004-12-13 20:16 -------
Try -fdump-tree-vect-stats. It produces more user friendly output compared to -details.
I am not fan of putting such diagnostics as part of warnings/info/errors. IMO, we should enhance/
update -fdump-tree-vect-stats to meet user's requirements.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
2004-11-12 2:45 ` [Bug tree-optimization/18441] " pinskia at gcc dot gnu dot org
2004-12-13 20:16 ` dpatel at apple dot com
@ 2004-12-22 19:00 ` leehod at il dot ibm dot com
2005-01-19 19:28 ` leehod at il dot ibm dot com
` (2 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: leehod at il dot ibm dot com @ 2004-12-22 19:00 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From leehod at il dot ibm dot com 2004-12-22 18:59 -------
I am working with Dorit on cleaning up the vectorizer dumps. We'd also like to
try to address this PR - here are some preliminary thoughts/questions:
1) The 'inform' utility doesn't look very suitable because it prints the line
information of the function and we want the line info of the loop, so a typical
printout would look like this:
loop at trees.c:499: Success! loop vectorized.
Also, it only prints to stderr, and we want to be able to print the same
information also to a dump-file. Maybe we want a 'vect_inform' that (1) would
not print the function line-number but some other line-number passed as
argument (the loop line-number), and (2) could be made to print either to
stderr or to a dump-file?
2) How about using a mechanism like -fsched-verbose? If -ftree-vect-verbose=x
is used and a dump-file is available, then the output will be printed into the
dump-file; otherwise, it will be printed into stderr. If -ftree-vect-verbose=x
is used then there will be no difference between -fdump-tree-vect-details and -
fdump-tree-vect-stats (they will both use the same verbosity level x). If -
ftree-vect-verbose is not specified, then -fdump-tree-vect-details could use
the highest verbosity level as default, and -fdump-tree-vect-stats could use
the lowest verbosity level.
3) We could use the different verbosity levels to prune the number of non-
vectorized-loops we report about, not just the amount of detail we report. For
example:
level 0: no diagnostic information.
level 1: report vectorized loops only
level 2: report vectorized loops and non-vectorized inner-most loops that
passed the first analysis phase (vect_analyze_loop_form).
level 3: report vectorized loops and non-vectorized inner-most loops
level 4: report vectorized loops and all non-vectorized loops
level 5: full debug information (vectorized/non-vectorized loops, data-
dependence/scalar-evolution dumps)
4) Is there a mechanism in the testsuite to analyze output to stderr?
thanks,
Leehod
--
What |Removed |Added
----------------------------------------------------------------------------
CC| |dorit at il dot ibm dot com,
| |leehod at il dot ibm dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
` (2 preceding siblings ...)
2004-12-22 19:00 ` leehod at il dot ibm dot com
@ 2005-01-19 19:28 ` leehod at il dot ibm dot com
2005-02-10 0:13 ` dorit at il dot ibm dot com
2005-05-21 12:51 ` pinskia at gcc dot gnu dot org
5 siblings, 0 replies; 7+ messages in thread
From: leehod at il dot ibm dot com @ 2005-01-19 19:28 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From leehod at il dot ibm dot com 2005-01-19 19:28 -------
There is now a patch addressing these issues.
See: http://gcc.gnu.org/ml/gcc-patches/2005-01/msg01247.html
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
` (3 preceding siblings ...)
2005-01-19 19:28 ` leehod at il dot ibm dot com
@ 2005-02-10 0:13 ` dorit at il dot ibm dot com
2005-05-21 12:51 ` pinskia at gcc dot gnu dot org
5 siblings, 0 replies; 7+ messages in thread
From: dorit at il dot ibm dot com @ 2005-02-10 0:13 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From dorit at il dot ibm dot com 2005-02-09 19:30 -------
continued in this thread:
http://gcc.gnu.org/ml/gcc-patches/2005-02/msg00196.html
and cmmitted:
http://gcc.gnu.org/ml/gcc-cvs/2005-02/msg00153.html
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Bug tree-optimization/18441] Vectorizer: add a command line for simple vectorizer report
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
` (4 preceding siblings ...)
2005-02-10 0:13 ` dorit at il dot ibm dot com
@ 2005-05-21 12:51 ` pinskia at gcc dot gnu dot org
5 siblings, 0 replies; 7+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2005-05-21 12:51 UTC (permalink / raw)
To: gcc-bugs
------- Additional Comments From pinskia at gcc dot gnu dot org 2005-05-21 12:51 -------
Fixed for 4.0.0.
--
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution| |FIXED
Target Milestone|--- |4.0.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18441
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2005-05-21 12:51 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-11-12 2:16 [Bug tree-optimization/18441] New: Vectorizer: add a command line for simple vectorizer report giovannibajo at libero dot it
2004-11-12 2:45 ` [Bug tree-optimization/18441] " pinskia at gcc dot gnu dot org
2004-12-13 20:16 ` dpatel at apple dot com
2004-12-22 19:00 ` leehod at il dot ibm dot com
2005-01-19 19:28 ` leehod at il dot ibm dot com
2005-02-10 0:13 ` dorit at il dot ibm dot com
2005-05-21 12:51 ` pinskia at gcc dot gnu dot org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).