gmane.lisp.steel-bank.devel

Re: How to know when to stop spelling "banana"

Paul Khuong — 2013-05-21T21:29:14

Sometimes, we have > 64 bit constants (SSE), and we might eventually 
have jump table offsets in there.  Do you see a nice way to just 
hex-dump the rest? It'd be even better if references were used to 
determine where to break chunks of data.

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

How to know when to stop spelling "banana"

Douglas Katzman — 2013-05-21T21:16:49

Patch inspired by suggestion of Paul Khuong which he attributed in turn to
Alistair Bridgewater:
dont disassemble memory locations which are part of the RIP-relative
constant area.
(Makes sense now that they're printed as notes on the line using them)
------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may_______________________________________________
Sbcl-devel mailing list
Sbcl-devel< at >lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Re: Smarter EQL -> EQ strength reduction causes consing (was Faster FIND and POSITION in unsafe code)

Paul Khuong — 2013-05-21T20:31:26

 > On Tue, May 21, 2013 at 1:50 AM, Paul Khuong <pvk< at >pvk.ca
 > <mailto:pvk< at >pvk.ca>> wrote:
 >
 >     Douglas Katzman wrote:
 >
 >         Would you believe that this [commit 9ce27ba] EQL transform ruins
 >         my FOO example from the
 >         unboxed >32-bit constants patch?
 >         It's consing now and it wasn't before.  Roll back this patch and
 >         we get
 >         the expected result.
 >
 >
 >     Likely a missing VOP (or bad cost info…). I'll look into it tomorrow
 >     morning, after the SSE stuff.

Fixed on x86 and x86-64 in 09f925d (Implement EQ of unboxed characters 
and small integers on x86oids).

EQ presents the reverse situation of EQL: we try very hard in EQL to 
detect when we can perform an EQ check... For EQ, we need to determine 
when it makes sense to perform unboxed comparisons instead of the 
fallback generic comparison.  I believe it's most important for 
characters and word-sized integers, with and without constants: these 
are the cases that the compiler recognises.

Re: [Sbcl-commits] master: Constant-fold backquote ofconstant expressions

Paul Khuong — 2013-05-21T16:18:28

On my end, this commit causes gc.impure.lisp / BUG-936304 to fail on 
Linux/x86. Running the test by itself works fine, and I'm really not 
sure what it's actually testing for; the scheduling logic for GCs? If 
so, the best way to solve the problem is likely to make sure we can 
force a GC *before* giving up with an OOM condition.

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

SB-GMP

2013-05-21T08:34:54

Hi all,

since we are approaching the end of the month freeze I'd like again to
inquire about thoughts on including SB-GMP [1] as a contrib, what should
be done before this happens, or whether is should continue to exist as an
external SBCL-specific library (then via quicklisp).

If you run the test suite be advised that, depending on your machine, this
will take sometime between 10-30 minutes.

Regards,
Stephan

[1] https://github.com/sfrank/sb-gmp



------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

Re: Smarter EQL -> EQ strength reduction causes consing (was Faster FIND and POSITION in unsafe code)

Douglas Katzman — 2013-05-21T05:52:59

EQL has a boatload of templates:
* (length (fun-info-templates (info :function :info 'eql)))
13

EQ has only one template, in particular nothing for any non-descriptor arg
types. Fascinating.
* (fun-info-templates (info :function :info 'eq))
(#<VOP-INFO
   :NAME IF-EQ
   :ARG-TYPES (* *)
   :RESULT-TYPES (:CONDITIONAL :E)
   :LTN-POLICY :FAST-SAFE
   :COST 3>)




On Tue, May 21, 2013 at 1:50 AM, Paul Khuong <pvk< at >pvk.ca> wrote:

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may_______________________________________________
Sbcl-devel mailing list
Sbcl-devel< at >lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Re: Smarter EQL -> EQ strength reduction causes consing (was Faster FIND and POSITION in unsafe code)

Paul Khuong — 2013-05-21T05:50:35

Likely a missing VOP (or bad cost info…). I'll look into it tomorrow 
morning, after the SSE stuff.


I believe that's what we currently do.

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel< at >lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Re: Faster FIND and POSITION in unsafe code

Douglas Katzman — 2013-05-21T05:40:57

Would you believe that this EQL transform ruins my FOO example from the
unboxed >32-bit constants patch?
It's consing now and it wasn't before.  Roll back this patch and we get the
expected result.

How would one go about writing a unit test that asserts that this doesn't
happen?
Martin suggests disassembling to a string stream and comparing literally
the instruction sequence.
I've already nearly hacked up something to eliminate the address column,
which, incidentally, displays addresses truncated anyway (apropos of
nothing).

(disassemble '(lambda (x)
  (declare (fixnum x) (optimize speed (safety 0)))
  (zerop (logxor (ldb (byte 64 0) x) #xffffffff80aabbcc))))

; in: LAMBDA (X)
;     (ZEROP (LOGXOR (LDB (BYTE 64 0) SB-C::X) 18446744071573257164))
; --> = IF = EQL IF EQL EQ IF
; ==>
;   (EQ SB-C::X SB-C::Y)
;
; note: doing unsigned word to integer coercion (cost 20), for:
;       the first argument of IF-EQ
;
; compilation unit finished
;   printed 1 note

; disassembly for (LAMBDA (X))
; Size: 64 bytes
; 03

Re: Improved signed=>unsigned conversion on x86-64

Paul Khuong — 2013-05-21T03:21:19

[...]

Both patches were committed tonight: aae8dd3 (POPCNT instruction on 
x86-64) and 39117fb (Cleverer handling of medium (32 < bit width <= 64) 
constants on x86-64). For the latter, I also yanked out all the load-if 
logic: we always want to the remaining operand and the result in 
registers, as the constant operand is either an immediate or RIP-rel.

Thank you,

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

Re: Faster FIND and POSITION in unsafe code

Paul Khuong — 2013-05-21T03:17:59

That part was committed in 9ce27ba (Exploit specialised VOPs for EQL of 
anything/constant fixnum).

Thank you,

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

Re: Patch for review: move-to-word/integer and fixnump hack

Paul Khuong — 2013-05-21T03:17:07

Committed in 6794373 (More efficient integer=>word conversion and 
fixnump tests on x86-64)

Thank you,

Paul Khuong

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

Re: [Sbcl-commits] master: Make some instances of IF/IF conversion more direct

Paul Khuong — 2013-05-21T03:16:20

I tried to extend the logic to predicates, but only managed to get it to 
fire on (eq (if ...) constant), and with fairly suspect code at that.

Still, patch attached if someone can find the inspiration and energy to 
make it work.

Paul Khuong
diff --git a/src/compiler/ir1opt.lisp b/src/compiler/ir1opt.lisp
index 9c01bc1..b8b7c9a 100644
--- a/src/compiler/ir1opt.lisp
+++ b/src/compiler/ir1opt.lisp
< at >< at > -682,6 +682,7 < at >< at >
       (when victim
         (kill-if-branch-1 node test block victim)
         (return-from ir1-optimize-if (values))))
+    (mini-trick-1 node test block)
     (tension-if-if-1 node test block)
     (duplicate-if-if-1 node test block)
     (values)))
< at >< at > -696,6 +697,63 < at >< at >
   (setf (component-reanalyze (node-component node)) t)
   (unlink-node node))
 
+;; When the predicate is foldable, and all but one argument is constant,
+;; duplicate the whole TEST/CIF combo in predecessors in which the
+;; remaining variable argument is constant: we constant fold that into
+;; a jump to the right branch.

Re: GSoC 2013 Mentoring - Unicode

Christophe Rhodes — 2013-05-18T21:01:50


I have not come up with a viciously clever indexing scheme for primary
composition; I've contented myself with a fairly simple scheme, and also
not implementing any of the optimizations (obvious or otherwise) in the
normalization area -- not even the various quick-check properties for
the various normalization forms.  I dare say I'll get to them
eventually, but in the absence of substantial impending time I have gone
ahead and merged the normalization work into master.  Still plenty of
low-hanging fruit...

Cheers,

Christophe

------------------------------------------------------------------------------
AlienVault Unified Security Management (USM) platform delivers complete
security visibility with the essential security capabilities. Easily and
efficiently configure, manage, and operate all of your security controls
from a single console and one unified framework. Download a free trial.
http://p.sf.net/sfu/alienvault_d2d

Patch for review: move-to-word/integer and fixnump hack

Douglas Katzman — 2013-05-16T18:33:14

With single-bit fixnum tags, move-to-word/integer can be done in 4 or 5
instructions (including the initial move) versus 7 instructions.
fixnump/signed-byte-64 can be done without use of a constant, sometimes
just 1 instruction, plus the conditional jump or move.
Also fixnump/unsigned-byte-64 could emit 'mov rdx,rdx' or similar because
it used (inst mov tmp value) where it meant (move tmp value).
------------------------------------------------------------------------------
AlienVault Unified Security Management (USM) platform delivers complete
security visibility with the essential security capabilities. Easily and
efficiently configure, manage, and operate all of your security controls
from a single console and one unified framework. Download a free trial.
http://p.sf.net/sfu/alienvault_d2d_______________________________________________
Sbcl-devel mailing list
Sbcl-devel< at >lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Re: Lisp Interface to GC

Alastair Bridgewater — 2013-05-15T23:54:59

If you can guarantee that your memory areas will always have sufficient
space for all of the allocation that you need to do, are willing to run
WITHOUT-INTERRUPTS while allocating into that space, and the space will
contain no external references other than to static space, then I have a
nasty, nasty hack in mind that will run on a stock system (hint: there are
only two words used for the basic allocation and overflow check, and the
region is in a known global location on single-thread systems and in the
thread structure on multi-thread systems, easily poked at via SAP functions
either way).

The requirement to always have sufficient space is so that you don't trip
an allocation overflow "trap" (though they are actually straight calls on
x86oids, but on PPC they are a TWI instruction that the trap handler can
easily recognize).  To alleviate this, we would need to define a way to
hook the overflow trap on a per-region basis to point to the "correct" GC.
Note that it should be plausible to run the trap handl

Re: Lisp Interface to GC

Craig Lanning — 2013-05-15T18:41:54

I can give a generic description of what I want to do.  I just can't
tell you the specific reason why the application needs to do this.

Basically, the application has a certain object organization that it
needs to be able to keep localized.  If we need to drop one of our main
objects so that we can load another one, it would be advantageous to
know that all allocation related to that object is contained within a
known memory block.  Then we just declare the block empty and zero it.

I started my Lisp career working on Symbolics Lisp Machines.  One thing
that they had was a concept of areas.  An area was an allocation block
that could be constrained somehow.  For the LispM's some areas held only
cons cells, others held symbol names.  Areas could also be excluded from
collection.  The aspect I'm interested in is that the user could create
an area, designate whether it should be handled by the GC, what objects
it would hold, and when the contents of the area are no longer needed,
just flush the contents witho

Re: Lisp Interface to GC

Alastair Bridgewater — 2013-05-15T17:15:20

Are you able to share any details about the sort of changes that you have
in mind?  Or about the use case, even in general terms?


Some years ago (late 2008, maybe), I did some preliminary investigation
into writing parts of the GC in Lisp, although I was mostly focusing on how
to implement things like the dispatch involved in scavenging an object.

If you're planning to implement a GC in Lisp, one of the things to make
sure of is that the code and data that implements the GC is available while
you're running the GC, which is something that seemed very difficult in my
original context, but for what you're doing simply arranging for any data
tables required to be in static space or otherwise pinned and for code
objects to not move would cover the worst of it.  Well, and there are the
FDEFINITION objects to consider, but putting them into static space as well
should work, if you can arrange that (I can think of a couple of approaches
here).

Another aspect to consider is if the GC code itself should be allow

Re: Lisp Interface to GC

Attila Lendvai — 2013-05-15T06:01:27


well, it's easy to do it in a system where all that you have is random
access to the underying huge binary number we call memory, and as such
you must manage the object layout yourself.

it's an entirely different story to introduce first class heaps into a
system that also promises transparent memory management for you (GC).

Re: Lisp Interface to GC

Philipp Marek — 2013-05-15T05:32:56

APR has something similar: nested memory pools, and destroying one destroys all nested,
too.

   http://apr.apache.org/docs/apr/1.4/group__apr__pools.html



------------------------------------------------------------------------------
AlienVault Unified Security Management (USM) platform delivers complete
security visibility with the essential security capabilities. Easily and
efficiently configure, manage, and operate all of your security controls
from a single console and one unified framework. Download a free trial.
http://p.sf.net/sfu/alienvault_d2d

ASDF 3.0.0

Faré — 2013-05-15T05:16:09

Dear SBCL hackers,

I'm proud to announce the release of release ASDF 3.0.0.
Please test it, and include it in SBCL and/or send me patches.
At the request of some #sbcl hackers,
I notably nulled out the *uninteresting-conditions* by default.

I'll have to produce some document explaining the innovations since
ASDF 2.26, 2.000 and/or 1.369, but for now here are just the changes
since 2.33.

As you can see, it's very minor stuff, and ASDF has been mostly stable
these last two months, which is a good sign and the reason why I'm
making an official 3.0.0 release.

I'm also inserting a new digit in the version, so releases will have
three digits:
3.0.0, 3.0.1, etc. will be minor continuations of 3.0. 3.1.0 will be
the next "major"
milestone in a series that preserves backward compatibility, and 4.0 (if ever)
will be the next major release that doesn't.
But I probably won't be there to see it, because I'm moving away from
my Common Lisp job in Cambridge MA
to some new as yet undetermined opportunities in NYC that
a

Re: Lisp Interface to GC

Faré — 2013-05-15T04:26:22

On Tue, May 14, 2013 at 10:53 PM, Attila Lendvai
<attila.lendvai< at >gmail.com> wrote:
The ML-Kit with Region bootstrapped its compiler entirely using such
nested "dynamic extents". Matthew Fluet wrote a nice thesis on the
general topic that generalizes the notion allowing for first-class
such regions.

I'd like to look into it and write a "linear lisp" in the style of
Henry Baker that has some such notion of first-class region. We'll see
after ELS2013. (Admittedly, I said I'd do it after ILC2012, and
instead I wrote ASDF3.)

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
There cannot be Ethics without Models of possible behaviors, and Imagination
to explore them. [Corollary: there is no Ethics for an all-knowing God,
but there are Ethics for mostly-ignorant but nevertheless thinking humans]

------------------------------------------------------------------------------
AlienVault Unified Security Management (USM) platform delivers complete
security visibility with t