<br><br><div class="gmail_quote">On Thu, Jul 25, 2013 at 12:18 PM, Herbert König <span dir="ltr"><<a href="mailto:herbertkoenig@gmx.net" target="_blank">herbertkoenig@gmx.net</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div text="#000000" bgcolor="#FFFFFF">
Hi Eliot,<br>
<br>
under Win7 64 Bit the upgrading from r2701 of March 11th I got a
speedup from 575 to 231 seconds for my "CSV file reading using
symbols to reduce memory consumption and speedup #select" job. <br>
<br>
The average full gc came down from 811 to 450 ms (averaged over 19
GC's)<br>
<br>
So the performance regression may have been in Windows VM's too.<br>
<br>
Thanks a lot!<br>
<br>
What I don't understand:<br>
I do a "purge undo records" then a GC.<br>
In both cases I get around 5MB less used Memory but Squeak uses some
additional 68 MB (old VM) resp 73 MB (new VM). Total Memory after GC
is around 445 MB in both cases.<br>
Is this to be expected and wouldn't happen if the image size were
closer to the 512 MB Limit of Windows?<br></div></blockquote><div><br></div><div>The Cog VM keeps a lot more memory around in reserve. It has a machine code zone, a larger new space and a reserve for flushing the stack zone to the heap as contexts that the standard VM doesn't use. (all of these extras are part of the standard linear heap segment). So the disparity is to be expected.</div>
<div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF">
<br>
Cheers<br>
<br>
Herbert<br></div></blockquote><div><br></div><div>cheers!</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF">
<br>
<div>Am 16.07.2013 02:35, schrieb Eliot
Miranda:<br>
</div><div><div class="h5">
<blockquote type="cite">
<div>are in <a href="http://www.mirandabanda.org/files/Cog/VM/VM.r2749/" target="_blank">http://www.mirandabanda.org/files/Cog/VM/VM.r2749/</a>.</div>
<div><br>
</div>
<div>CogVM binaries as per VMMaker.oscog-eem.302/r2749</div>
<div><br>
</div>
<div>Fix bad performance regression that on certain platforms
(linux) results in all</div>
<div>send misses causing a discarded PIC creation followed by a
slow hash lookup.</div>
<div><br>
</div>
<div>Fix bug when assigning to some context inst vars from
generated methods.</div>
<div>Add accessors to document the context inst var access scheme.</div>
<div>Fix a compiler warning comparing an error code in
cog:selector:</div>
<div><br>
</div>
<div>Update the BitBltPlugin to include the fast Arm ASM option.</div>
<div><br>
</div>
<div>Fix type errors in the Cogit that prevent the Cogit working
when compiled with</div>
<div>clang. Specifically void * pointers are not comparable.
Make sure that</div>
<div>fetchPointer:ofObject: & isIntegerValue: are declared in
cointerp.h.</div>
<div><br>
</div>
<div>Update the Newspeak version of the
VMProfileLinuxSupportPlugin.</div>
<div>Improve robustness of the nscogbuild/unixbuild mvm scripts.</div>
<div><br>
</div>
<div>Add the linux policy change dance to sqUnixVMProfile.c (cf
sqUnixHeartbeat.c).</div>
<div>Fix a bail_out typo.</div>
<div><br>
</div>
<div>Change the VMProfileLinuxSupportPlugin to follow symlinks,</div>
<div>answering pairs of module name, dereferenced symlink or nil.</div>
<div><br>
</div>
<div>Fix 3 (!!) bugs in primitiveDLSymInLibrary.</div>
<div><br>
</div>
<div><br>
</div>
-- <br>
best,
<div>Eliot</div>
<br>
<fieldset></fieldset>
<br>
<pre></pre>
</blockquote>
<br>
</div></div></div>
<br><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br>best,<div>Eliot</div>