Path: ...!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: BGB <cr88192@gmail.com>
Newsgroups: comp.lang.c
Subject: Re: transpiling to low level C
Date: Wed, 25 Dec 2024 00:51:37 -0600
Organization: A noiseless patient Spider
Lines: 114
Message-ID: <vkga20$29lmo$1@dont-email.me>
References: <vjlh19$8j4k$1@dont-email.me>
 <vjn9g5$n0vl$1@raubtier-asyl.eternal-september.org>
 <vjnhsq$oh1f$1@dont-email.me> <vjnq5s$pubt$1@dont-email.me>
 <vjpn29$17jub$1@dont-email.me> <86ikrdg6yq.fsf@linuxsc.com>
 <vk78it$77aa$1@dont-email.me> <20241222002032.0000104c@yahoo.com>
 <vk7lik$9iga$1@dont-email.me> <20241222021851.0000059f@yahoo.com>
 <vk7n4l$9okb$1@dont-email.me> <20241222030451.00005565@yahoo.com>
 <vk7s80$ajak$1@dont-email.me> <86o713dvo7.fsf@linuxsc.com>
 <vkaik4$ttg2$1@paganini.bofh.team> <vkb4ab$13mrr$1@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Wed, 25 Dec 2024 07:51:44 +0100 (CET)
Injection-Info: dont-email.me; posting-host="d0b6fc13f6ed5769626df891f2167fc1";
	logging-data="2414296"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18E/dRWZmh5MhXhoCGwis/IAM0TIJgAKqI="
User-Agent: Mozilla Thunderbird
Cancel-Lock: sha1:IWwjUhoujTnwcBEWCBx5LI6IG6g=
In-Reply-To: <vkb4ab$13mrr$1@dont-email.me>
Content-Language: en-US
Bytes: 6730

On 12/23/2024 1:43 AM, David Brown wrote:
> On 23/12/2024 03:41, Waldek Hebisch wrote:
>> Tim Rentsch <tr.17687@z991.linuxsc.com> wrote:
>>>
>>> The comments I made here, in two responses to postings of yours,
>>> were not statements of opinion but statements of fact.
>>
>> They are opinions _about facts_, or if you prefer, opinion
>> about truth value of some statements.
> 
> You can program in C without the "normal" conditional statements or 
> expressions.  You can make an array of two (or more) function pointers 
> and select between them using your controlling expression, and that 
> should be sufficient for conditionals.  (There may be other methods too.)
> 
> So as far as I can see, Tim gave statements of fact, not opinion.
> 
Jumping back in:
That one can do this seems obvious enough;
Downside, as I see it, is that there is no current or likely processor 
hardware where this is likely to be performance competitive with the 
more traditional if-goto mechanism (and if the backend is expected to 
optimize it away, not obvious what would be gained).


Sort of like with "continuation passing style":
Yes, you can do this, but the performance overhead relative to 
conventional call-frames is severe.

But, CPS does at least have use-cases which can justify this overhead.



Though, FWIW, doing control flow via a combination of CPS and plugging 
things together with function pointers is fairly useful in implementing 
things like fast interpreters (where calling through function pointers 
can be faster than going through big if/else trees or "switch()" blocks).

Where, early on in writing interpreters, I had often ran into a limit 
that the interpreter would become bottle-necked by how quickly it could 
spin in a loop and feed instructions through a big "switch()" block. 
Using function pointers can theoretically sidestep this limit (then one 
is more limited by how quickly they can walk the trace graph and call 
the relevant function pointers).

But, can get within 10x of native code in some cases, which is pretty 
fast by interpreter standards (to get much faster usually requires a JIT).


Well, except in my current emulator, where in trying to be 
cycle-accurate, the much bigger overhead is in trying to mimic behavior 
and cycle costs of the cache hierarchy and similar.


> You can say that Tim's posts were patronising, arrogant, and irritating. 
>   /That/ would be an opinion - a /justified/ opinion because it is 
> backed up in the evidence of these posts and corroborating evidence from 
> previous posts and discussions from Tim.  But without some kind of 
> precise definition of the terms involved and a robust and repeatable 
> method of classification, it could not be called "fact".
> 
> You could say that Tim's posts were intended to be annoying, or you 
> could say that he has refused to give an answer to how C can be used 
> without the "normal" conditionals because he realises he was wrong in 
> his posts and won't admit it.  That would be /unjustified/ opinion - or 
> "speculation" - because we have no way of knowing his motives or 
> anything more than what he wrote in his posts.
> 
> 
> You could, quite fairly, characterise Tim's posts as unjustified 
> statements of fact - because he has stated his claim as fact, but has 
> given no justification or reasoning, and it is not something that is 
> obvious or well-known to people.
> 
>>
>>>   They are
>>> no more statements of opinion than a statement about whether the
>>> Riemann Hypothesis is true is a statement of opinion.  Someone
>>> might wonder whether an assertion "The Riemann Hypothesis is
>>> true" is true or false, but it is still a matter of fact, not a
>>> matter of opinion.
>>
>> It is reasobable to assume that you do not know if Riemann Hypothesis
>> is true or false.
> 
> I think if anyone knew the truth of falsity of the Riemann Hypothesis - 
> i.e., they had a proof one way or the other - we'd have heard about it!
> 
>> So if you say "Riemann Hypothesis is true",
>> this is just your opinion.
> 
> No, that would not be an opinion.  It would be an unjustified claim.  
> "I /believe/ the Riemann Hypothesis is true" is an opinion.
> 
>>  I am not a native English speaker
>> but I believed that "statements of opinion" means just that:
>> person does not know the truth, but makes a statement.
>>
> 
> No, an opinion is a personal preference or judgement.  That's very 
> different from not knowing about something factual.  If I say "the 
> number 17 will turn up in next week's lottery numbers", that's not an 
> opinion, it's a claim about facts.  It's an unjustified claim, since I 
> don't know if it is true or not, but it's not an opinion.
> 
> It is not always clear when something is a fact or not, and whether a 
> statement is a justified statement of fact, an unjustified statement of 
> fact (i.e., it might happen to be true, but you have not presented 
> evidence of it), a justified opinion, or an unjustified opinion.  I'm 
> sure there's a philosophy group on Usenet somewhere, but I doubt if 
> cross-posting there would lead to any clarification!
> 
>