<br><br><div class="gmail_quote">On Thu, Sep 22, 2011 at 12:29 PM, Henrik Sperre Johansen <span dir="ltr"><<a href="mailto:henrik.s.johansen@veloxit.no">henrik.s.johansen@veloxit.no</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<br>
<div bgcolor="#FFFFFF" text="#000000">
On 22.09.2011 20:20, Eliot Miranda wrote:
<blockquote type="cite">
<pre> </pre>
<br>
<fieldset></fieldset>
<br>
<br>
<br>
<div>On Thu, Sep 22, 2011 at 11:06 AM, Eliot Miranda <span><<a href="mailto:eliot.miranda@gmail.com" target="_blank">eliot.miranda@gmail.com</a>></span>
wrote:<br>
<blockquote>
Hi Igor,<br>
<br>
<div>
<div>
<div>On Thu, Sep 22, 2011 at 10:53 AM, Igor Stasenko <span><<a href="mailto:siguctua@gmail.com" target="_blank">siguctua@gmail.com</a>></span>
wrote:<br>
<blockquote>
<div>
<div>On 22 September 2011 19:16, Eliot Miranda <<a href="mailto:eliot.miranda@gmail.com" target="_blank">eliot.miranda@gmail.com</a>>
wrote:<br>
> (apologies for the duplicate reply; someone
needs to sort out their<br>
> threading for the benefit of the community ;)
)<br>
><br>
> On Thu, Sep 22, 2011 at 2:36 AM, Marcus
Denker <<a href="mailto:marcus.denker@inria.fr" target="_blank">marcus.denker@inria.fr</a>><br>
> wrote:<br>
>><br>
>> Hi,<br>
>><br>
>> There are two changesets waiting for
integrating in 1.4 that have serious<br>
>> consequences:<br>
>><br>
>> - Ephemerons. The VM level changes are in
the Cog VMs build on Jenkins,<br>
>> but have not<br>
>> been integrated in the VMMaker codebase.<br>
>><br>
>> <a href="http://code.google.com/p/pharo/issues/detail?id=4265" target="_blank">http://code.google.com/p/pharo/issues/detail?id=4265</a><br>
><br>
> I would *really* like to back out these
changes. The Ephemeron<br>
> implementation is very much a prototype,
requiring a hack to determine<br>
> whether an object is an ephemeron (the
presence of a marker class in the<br>
> first inst var) that I'm not at all happy
with. There is a neater<br>
> implementation available via using an unused
instSpec which IMO has<br>
> significant advantages (much simpler &
faster, instSpec is valid at all<br>
> times, including during compaction, less
overhead, doesn't require a marker<br>
> class), and is the route I'm taking with the
new GC/object-representation<br>
> I'm working on now. Note that other than
determining whether an object is<br>
> an ephemeron (instSpec/format vs inst var
test) the rest of Igor's code<br>
> remains the same. I'd like to avoid too much
VM forking. Would you all<br>
> consider putting these changes on hold for
now?<br>
> If so, I'll make the effort to produce
prototype changes (in the area of<br>
> ClassBuilder and class definition; no VM code
necessary as yet) to allow<br>
> defining Ephemerons via the int spec route by
next week at the latest.<br>
><br>
<br>
</div>
</div>
i agree that in my implementation this is a weak
point. But its hard<br>
to do anything without<br>
making changes to object format to identify these
special objects.<br>
<br>
The main story behind this is can we afford to change
the internals of<br>
VM without being beaten hard<br>
by "backwards compatibility" party? :)<br>
</blockquote>
<div><br>
</div>
</div>
</div>
<div>I don't think we get stuck in this at all. The
instSpec/format field has an unused value (5 i believe)
and this can easily be used for Ephemerons. All that is
needed is a little image work on these methods:</div>
<div><br>
</div>
<div> Behavior>>typeOfClass</div>
<div> needs to answer e.g. #ephemeron for ephemeron
classes</div>
<div><br>
</div>
<div>
ClassBuilder>>computeFormat:instSize:forSuper:ccIndex:</div>
<div>
needs to accept e.g. #ephemeron for type and pass
variable: false and weak: true for ephemerons
to format:variable:words:pointers:weak:.</div>
<div><br>
</div>
<div>
ClassBuilder>>format:variable:words:pointers:weak:</div>
<div> needs to respond to variable: false and weak:
true by computing the ephemeron instSpec.</div>
<div><br>
</div>
<div>
Class>>weakSubclass:instanceVariableNames:classVariableNames:poolDictionaries:category:</div>
<div>
ClassBuilder>>superclass:weakSubclass:instanceVariableNames:classVariableNames:poolDictionaries:category:</div>
<div> need siblings, e.g.</div>
<div>
ephemeronSubclass:instanceVariableNames:classVariableNames:poolDictionaries:category </div>
<div>
superclass:ephemeronSubclass:instanceVariableNames:classVariableNames:poolDictionaries:category:</div>
<div><br>
</div>
<div>Right? This is easy. Then in the VM there are a few
places where pointer indexability (formats 3 and 4) need
to be firmed up to exclude 5, but nothing difficult. We
talked about this in email last week.</div>
</div>
</blockquote>
<div><br>
</div>
<div>Here's the format field (Behavior>instSpec at the image
level) as currently populated:</div>
<div> </div>
<div>
<div> 0 = 0 sized objects (UndefinedObject True False et al)</div>
<div> 1 = non-indexable objects with inst vars (Point et al)</div>
<div> 2 = indexable objects with no inst vars (Array et al)</div>
<div> 3 = indexable objects with inst vars (MethodContext
AdditionalMethodState et al)</div>
<div> 4 = weak indexable objects with inst vars (WeakArray et
al)</div>
</div>
<div> 6 = 32-bit indexable objects (Float, Bitmap ert al)</div>
<div> 8 = 8-bit indexable objects (ByteString, ByteArray et al)</div>
<div>12 = CompiledMethod</div>
<div><br>
</div>
<div>N.B. in the VM the least two bits of the format/instSpec
for byte objects (formats 8 and 12) is used to encode the
number of odd bytes in the object, so that a 1 character
ByteString has a format of 11, = 8 + 3, size = 1 word - 3
bytes.</div>
<div><br>
</div>
<div><br>
</div>
<div>For the future (i.e. the new GC/object representation,
/not/ for the first implementation of ephemerons which we can
do now, for Pharo 1.4 or 1.5) we need to extend
format/instSpec to support 64 bits. I think format needs to
be made a 5 bit field with room for 4 bits of odd bytes for
64-bit images. [For VMers, the Size4Bit is a horrible hack).
So then</div>
<div><br>
</div>
<div>0 = 0 sized objects (UndefinedObject True False et al)</div>
<div>1 = non-indexable objects with inst vars (Point et al)</div>
<div>2 = indexable objects with no inst vars (Array et al)</div>
<div>3 = indexable objects with inst vars (MethodContext
AdditionalMethodState et al)</div>
<div>4 = weak indexable objects with inst vars (WeakArray et al)</div>
<div>5 = weak non-indexable objects with inst vars (ephemerons)
(Ephemeron)</div>
<div><br>
</div>
<div>and we need 8 CompiledMethod values, 8 byte values, 4
16-bit values, 2 32-bit values and a 64-bit value, = 23
values, 23 + 5 = 30, so there is room, e.g.</div>
<div><br>
</div>
<div>9 (?) 64-bit indexable</div>
<div>10 - 11 32-bit indexable</div>
<div>12 - 15 16-bit indexable</div>
<div>16 - 23 byte indexable</div>
<div>24 - 31 compiled method</div>
<div><br>
</div>
<div>In 32-bit images only the least significant 2 bits would be
used for formats 16 & 24, and the least significant bit
for format 12.<br>
</div>
</div>
</blockquote>
If we are changing the format for 64bit images anyways, why not
simplify it/ be more consistent by spending a full byte?<br>
<br>
Bit: 8 7 6 5
4 3 2 1 <br>
| 64bit | 32bit |16bit | 8bit |compiled | weak | indexable |
instVars | <br>
(Odd number encoded in remaining indexable bit fields)<br></div></blockquote><div><br></div><div>I used to prefer this approach but I've realised that the format/instSpec approach (I think Dan came up with) makes better use of bits because so many of the bit combinations are mutually exclusive. For example, pointers excludes all the byte/short/32-bit/64-bit indexability combinations. Also, see below...</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><div bgcolor="#FFFFFF" text="#000000">
<br>
Could get away with 7 if you put f.ex. the unused indexable weak
combination (6) as compiled method/8bit<br>
<br>
Or is the header space in your new 64bit format already quite
filled, so this is a bad idea?<br></div></blockquote><div><br></div><div>Yes, ish. But they're scarce, and very useful for experiments etc. Right now I have </div><div><br></div><div><div>typedef struct {</div>
<div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned short<span class="Apple-tab-span" style="white-space:pre">        </span>classIndex;</div><div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>unused0 : 6;</div>
<div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>isPinned : 1;</div><div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>isImmutable : 1;</div>
<div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>format : 5; /* on a byte boundary */</div><div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>isMarked : 1;</div>
<div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>isGrey : 1;</div><div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>isRemembered : 1;</div>
<div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned<span class="Apple-tab-span" style="white-space:pre">                </span>objHash : 24; /* on a 32-bit word boundary */</div><div><span class="Apple-tab-span" style="white-space:pre">        </span>unsigned char<span class="Apple-tab-span" style="white-space:pre">        </span>slotSize; /* on a byte boundary */</div>
<div> } CogObjectHeader;</div></div><div><br></div><div>Where classIndex is 16-bits simply for efficiency and will grow to 20 or 22 bits as needed. So one could steal one or two bits from unused0 and two bits from objHash, and give these to format, but it would be a waste. Better keep these back for other uses.</div>
<div><br></div><div>Also, can I ask the assembled company exactly how many bits you'd spend on the objHash (identityHash)? Think forward to 64-bits. Is 24 bits about all we can afford or still too generous? Anybody have any data to contribute?</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><div bgcolor="#FFFFFF" text="#000000">Cheers,<br>
Henry</div></blockquote></div><br>-- <br>best,<div>Eliot</div><br>