Re: [Monetdb-developers] [Monetdb-checkins] MonetDB5/src/modules/kernel unix.mx, , 1.24, 1.25
Why do you want to be able to flush all changed files to disk? MonetDB already does flushes when they are needed for transaction support. I don't think we need to tell the kernel to synchronize any other files with the disk. This function doesn't do anything with memory management within the process.
Furthermore, this function is not available on Windows.
Martin Kersten wrote:
Update of /cvsroot/monetdb/MonetDB5/src/modules/kernel In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv20459
Modified Files: unix.mx Log Message: We want to experiment with complete flushing memory in out of memory queries.
U unix.mx Index: unix.mx =================================================================== RCS file: /cvsroot/monetdb/MonetDB5/src/modules/kernel/unix.mx,v retrieving revision 1.24 retrieving revision 1.25 diff -u -d -r1.24 -r1.25 --- unix.mx 7 Jan 2009 14:15:12 -0000 1.24 +++ unix.mx 27 Jul 2009 06:29:44 -0000 1.25 @@ -35,6 +35,10 @@ address UNIXsetenv comment "Set the environment variable string.";
+command sync() :void +address UNIXsync +comment "Flush memory, expensive but sometimes needed";
@- Implementation The remainder is a straight forward mapping to the underlying facilities @@ -113,4 +117,10 @@ return MAL_SUCCEED; }
+unix_export str UNIXsync(int *res); +str UNIXsync(int *res){
- (void) res;
- sync();
- return MAL_SUCCEED;
+} @}
Monetdb-checkins mailing list Monetdb-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-checkins
Sjoerd Mullender wrote:
Why do you want to be able to flush all changed files to disk? MonetDB
Hint, we see very bad behavior in the second query, while it is identical to the first. One idea is to 'somehow' kick the system into a state as it was started. The assumption was that dirty pages may be polluting the memory, so a forceful sync could do that.
" According to the standard specification (e.g., POSIX.1-2001), sync() schedules the writes, but may return before the actual writing is done. However, since version 1.3.20 Linux does actually wait. (This still does not guarantee data integrity: modern disks have large caches.) " I was betting on the last remark
already does flushes when they are needed for transaction support. I don't think we need to tell the kernel to synchronize any other files with the disk. This function doesn't do anything with memory management within the process.
Furthermore, this function is not available on Windows.
was to experiment with the blkjoin issues, it can be dropped again.
Martin Kersten wrote:
Update of /cvsroot/monetdb/MonetDB5/src/modules/kernel In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv20459
Modified Files: unix.mx Log Message: We want to experiment with complete flushing memory in out of memory queries.
U unix.mx Index: unix.mx =================================================================== RCS file: /cvsroot/monetdb/MonetDB5/src/modules/kernel/unix.mx,v retrieving revision 1.24 retrieving revision 1.25 diff -u -d -r1.24 -r1.25 --- unix.mx 7 Jan 2009 14:15:12 -0000 1.24 +++ unix.mx 27 Jul 2009 06:29:44 -0000 1.25 @@ -35,6 +35,10 @@ address UNIXsetenv comment "Set the environment variable string.";
+command sync() :void +address UNIXsync +comment "Flush memory, expensive but sometimes needed";
@- Implementation The remainder is a straight forward mapping to the underlying facilities @@ -113,4 +117,10 @@ return MAL_SUCCEED; }
+unix_export str UNIXsync(int *res); +str UNIXsync(int *res){
- (void) res;
- sync();
- return MAL_SUCCEED;
+} @}
Monetdb-checkins mailing list Monetdb-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-checkins
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july
Monetdb-checkins mailing list Monetdb-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-checkins
participants (2)
-
Martin Kersten -
Sjoerd Mullender