Re: Clozure CL 1.12.2 release

Tim Bradshaw <[email protected]> Thu, 10 Aug 2023 17:20:26 +0100
Newsgroups gmane.lisp.openmcl.devel
Message-ID <[email protected]>
I may be wrong about how long it took!  My notes say 'very slow' but not how slow.  I could try it again when I'm back home.

Three things which might matter: I'm using a CCL built (from the current tree at the time) by me, and I can't remember now if I (tried?) to rebuild (obviously only the runtime) on M1.  M1 not newer Apple Silicon.  And finally my experience of CCL on the M1 is that it's OK ... except for the compiler, which is glacial.  I have scripts which find and kill all the fasls on the system and then everything is set up to recompile things if need be.  With SBCL or LW that means there's a second or two waiting for my inits & Quicklisp to rebuild, with CCL it took really a while (not hours but probably minutes).

I'll try to remember to try it again when I'm back.

--tim

> On 10 Aug 2023, at 16:21, Gary Palter <[email protected]> wrote:
> 
> Actually, CCL under Rosetta 2 can be very slow if executing code that it just loaded or compiled as Rosetta won’t be able to pre-translate the code. 
> 
> For example, running the ANSI test suite on an M2 Pro takes around a half hour vs. under 5 minutes on Intel. 
> 
> Sent from my iPhone
> 
>> On Aug 10, 2023, at 10:55 AM, Ron Garret <[email protected]> wrote:
>> 
>> That's actually very interesting.  I've never been able to get CCL to run on my M1 for more than about 10-15 minutes at a time under load.
>> 
>> Your report is also weird on another level: I bootstrapped SBCL on my M1 using ECL and the whole process took substantially less than an hour.  (I didn't time it either, but I sat through the whole thing and I tend towards impatience.)  CCL under Rosetta is flaky, but doesn't seem particularly slow.  So why would it have taken you so much longer than it took me?
>> 
>> rg
>> 
>> 
>>>> On Aug 10, 2023, at 5:10 AM, Tim Bradshaw <[email protected]> wrote:
>>> 
>>> What Ron says is entirely correct.  But it's worth noting that I used CCL to bootstrap SBCL on an M1 mac quite recently.  So it does run well enough for that.  It did take a very long time!  I didn't write down how long it took, but I think it was 'stop the machine from sleeping and let it run overnight' levels of slow.
>>> 
>>> --tim
>>> 
>>>>> On 10 Aug 2023, at 08:35, Ron Garret <[email protected]> wrote:
>>>> 
>>>> and it is generally not reliable in that environment.
>>>> 
>>>>> On Aug 10, 2023, at 12:19 AM, Gary Palter <[email protected]> wrote:
>>>>> 
>>>>> Please note that CCL is not yet native on Apple silicon. Instead, it runs under Rosetta 2. 
>>>>> 
>>>>> Sent from my iPhone
>>>>> 
>>>>>>> On Aug 10, 2023, at 2:46 AM, James Anderson <[email protected]> wrote:
>>>>>> 
>>>>>> good morning;
>>>>>> 
>>>>>>> On 10. Aug 2023, at 01:14, Laughing Water <[email protected]> wrote:
>>>>>>> 
>>>>>>> Sorry, I’m nowhere up to speed with this. Could someone tell me in simple terms whether this release will run on an M1 MacBook Pro running Ventura 13.4.1 and if so, what the steps are? In particular, does it require a build?
>>>>>> 
>>>>>> it does run on ventura 13.0.1.
>>>>>> there is a README.
>>>>>> it includes a link to a document at https://github.com/Clozure/ccl/releases/tag/v1.12.2 .
>>>>>> follow its instructions.
>>>>>> once you have an executable, in order to get the ide, follow the instructions in the README in the cocoa-ide directory.
>>>>>> 
>>>>>> best regards, from berlin,
>>>>>> 
>>>>>>> 
>>>>>>> Thanks!
>>>>>>> 
>>>>>>> LW
>>>>>>> 
>>>>>>>>> On Aug 8, 2023, at 11:43 PM, R. Matthew Emerson <[email protected]> wrote:
>>>>>>>> 
>>>>>>>> Clozure CL 1.12.2 is now available.
>>>>>>>> 
>>>>>>>> https://github.com/Clozure/ccl/releases/tag/v1.12.2
>>>>>>>> 
>>>>>>>> In addition to fixing a number of bugs, this release will, I hope, establish a fresh baseline that future work can build on.
>>>>>>>> 
>>>>>>>> 
>>>>>>> 
>>>>>> 
>>>>>> ---
>>>>>> james anderson | [email protected] | https://dydra.com
>>>>>> 
>>>>>> 
>>>>> 
>>>> 
>> 
>