|
From: John R. <joh...@cr...> - 2004-03-16 18:38:49
|
Hi all,
I just downloaded and tried Valgrind 2.1.1
on my Redhat Enterprise Linux 3 system
using gcc 3.2.3.
(1) First, I encountered a build problem when
I tried to build Valgrind. After I did
my "configure --prefix=path", then tried
make, I got the following error:
make[2]: Entering directory `/export/jroberts/valgrind/build/coregrind'
make[2]: *** No rule to make target
`../../valgrind-2.1.1/coregrind/Makefile', needed by `vg_toolint.c'. Stop.
So I went to valgrind-2.1.1/coregrind
and made an empty file called "Makefile".
Then I returned to the build directory
and the make succeeded.
(2) I tried Massif and its very cool!
I have one question/comment. I don't
understand the concept of "spacetime".
When I ran Massif, it made a very cool
postscript chart. The top band for
method1 (most memory used) peaked out
at about a 4 megabyte thickness.
The third band, lets call this method3,
looks to be about a 2 meg thickness.
When I look at the textfile output
though, method3 is listed at 16.7%
and method1 at 15.7%.
So the percentages are very unintuitive
to me. I think they would be better
reflecting just plain space utilization.
What is the motive behind "spacetime"?
(3) Massif gave a cryptic label/title
to my postscript chart:
3,240,776,-39,-221 bytes xms
Did this overflow some value?
I don't understand this either.
My printout shows about a 10 meg peak
and the max X-axis time is 350000.0 ms.
Anyway it looks _very_ cool and is
potentially extremely useful!
I think you're too timid versioning this
at 0.0.3. It should be more like 0.3.0,
unless you don't want to reach 1.0.0
in your lifetime. :)
John Roberts
Staff Software Engineer
5975 N.W. Pinefarm Place
Credence Systems Corporation
Hillsboro, Oregon 97124
email: joh...@cr...
tel: (503) 466-8056
|
|
From: Nicholas N. <nj...@ca...> - 2004-03-16 20:28:46
|
On Tue, 16 Mar 2004, John Roberts wrote: > (1) First, I encountered a build problem when I tried to build Valgrind. > After I did my "configure --prefix=path", then tried make, I got the > following error: > > make[2]: Entering directory `/export/jroberts/valgrind/build/coregrind' > make[2]: *** No rule to make target > `../../valgrind-2.1.1/coregrind/Makefile', needed by `vg_toolint.c'. Stop. > > So I went to valgrind-2.1.1/coregrind and made an empty file called > "Makefile". > > Then I returned to the build directory and the make succeeded. That's strange... and the fact that it worked when you added an empty Makefile is even stranger... is 'build/' the name of Valgrind's top-level directory, from which the install is being done? > (2) I tried Massif and its very cool! I have one question/comment. I > don't understand the concept of "spacetime". > [...] > So the percentages are very unintuitive to me. I think they would be > better reflecting just plain space utilization. > > What is the motive behind "spacetime"? Spacetime = space x time :) ie. the area under the graph. I used spacetime because that's what existing space profilers for Haskell that I based Massif on used; in particular that's what hp2ps, the graph-drawing program, uses. And I find it a more general measurement than, say, peak memory use. > (3) Massif gave a cryptic label/title > to my postscript chart: > > 3,240,776,-39,-221 bytes xms > > Did this overflow some value? > I don't understand this either. > My printout shows about a 10 meg peak > and the max X-axis time is 350000.0 ms. Yes, it's an overflow in hp2ps, code I grabbed wholesale from somewhere else, and I don't know how it works. I'll have a look; in the meantime you can ignore the value in the title (but you can trust the one that Massif prints out in its summary). > I think you're too timid versioning this at 0.0.3. It should be more > like 0.3.0, unless you don't want to reach 1.0.0 in your lifetime. :) Oh, that version number's irrelevant now that Massif is in the main distro. I'll change it. Thanks for the feedback. N |