Re: gcc Digest 21 Mar 2004 05:08:56 -0000 Issue 3435

Re: gcc Digest 21 Mar 2004 05:08:56 -0000 Issue 3435

[R. D. Flowers]

After months of reading this list, I timidly venture my first post.

Dear esteemed,
would not some flag like -fp-pedantic (or however to fit in the namespace) be a good idea (for both decent sets of folks: the fast-and-a-little-dirty and the clean-and-a-little-slow people)?

--
R. D. Flowers, Chattanooga
http://chatta.us/resume.txt

jFP religious wars

[R. D. Flowers]

After months of reading this list, I timidly venture my first post.

Dear esteemed,
would not some flag like -fp-pedantic (or however to fit in the namespace) be a good idea (for both decent sets of folks: the fast-and-a-little-dirty and the clean-and-a-little-slow people)?

--
R. D. Flowers, Chattanooga
http://chatta.us/resume.txt

Re: jFP religious wars

[Zack Weinberg]

"R. D. Flowers" <rd@chatta.us> writes:

> would not some flag like -fp-pedantic (or however to fit in the
> namespace) be a good idea (for both decent sets of folks: the
> fast-and-a-little-dirty and the clean-and-a-little-slow people)?

Maybe.  What would it do?

zw

Re: jFP religious wars

[R. D. Flowers]

Really sort of an inverse of -fast-math -- an insistence on total floating point correctness at the expense of any optimization that would water the correctness down any. Without it, roundings etc. thought to be minor would be allowed to happen in the middle end. I guess.

Zack Weinberg wrote:

> "R. D. Flowers" <rd@chatta.us> writes:
> 
> 
>>would not some flag like -fp-pedantic (or however to fit in the
>>namespace) be a good idea (for both decent sets of folks: the
>>fast-and-a-little-dirty and the clean-and-a-little-slow people)?
> 
> 
> Maybe.  What would it do?
> 
> zw
> 
> 
> .
> 


-- 
--
R. D. Flowers, O. H.
Xinna R. Coloohiy
The Bear At Home
Now a single dad.

I   M i s s   A m e r i c a

http://chalice.us/mylinks
http://chalice.us/pho
http://chalice.us/poe
http://chatta.us/resume.txt
Mailing lists at http://chatta.us

-- "He who cannot protect himself or his nearest and dearest or
their honor by non-violently facing death may and ought to do so
by violently dealing with the oppressor. He who can do neither of
the two is a burden."
-- Mohandes Ghandi

Refer !
Stop Tritium !

Re: jFP religious wars

[Robert Dewar]

R. D. Flowers wrote:

> Really sort of an inverse of -fast-math -- an insistence on total 
> floating point correctness at the expense of any optimization that would 
> water the correctness down any. Without it, roundings etc. thought to be 
> minor would be allowed to happen in the middle end. I guess.

Rounding is not some pedantic issue, proper rounding means that
you can reason much more clearly about floating-point semantics.

Re: jFP religious wars

[Robert Dewar]

Zack Weinberg wrote:
> "R. D. Flowers" <rd@chatta.us> writes:
> 
> 
>>would not some flag like -fp-pedantic (or however to fit in the
>>namespace) be a good idea (for both decent sets of folks: the
>>fast-and-a-little-dirty and the clean-and-a-little-slow people)?
> 
> 
> Maybe.  What would it do?

I don't think this would work. There is nothing "pedantic" in
the carefully formulated rules that make it easier/possible
to write accurate floating-point code. It is far better to
have the existing flags for -ffast-math etc, which clearly
warn that the resulting semantics may be unreliable.

The use of -pedantic to cover little used language features
or rules is one thing, the use of a flag like this to give
incorrect semantics.

Re: jFP religious wars

[Gabriel Dos Reis]

Robert Dewar <dewar@gnat.com> writes:

| Zack Weinberg wrote:
| > "R. D. Flowers" <rd@chatta.us> writes:
| >
| >>would not some flag like -fp-pedantic (or however to fit in the
| >>namespace) be a good idea (for both decent sets of folks: the
| >>fast-and-a-little-dirty and the clean-and-a-little-slow people)?
| > Maybe.  What would it do?
| 
| I don't think this would work. There is nothing "pedantic" in
| the carefully formulated rules that make it easier/possible
| to write accurate floating-point code. It is far better to
| have the existing flags for -ffast-math etc, which clearly
| warn that the resulting semantics may be unreliable.

Fully agreed.

| The use of -pedantic to cover little used language features
| or rules is one thing, the use of a flag like this to give
| incorrect semantics.

In fact, the flag "-pedantic" is misnamed -- at least, for what it does
currently in C++ (I think the same is true for C).  It is there only
for historical accident.  Multiplying that naming by a factor greater
than 1 would be misguided.

-- Gaby

Re: jFP religious wars

[R. D. Flowers]

Gabriel Dos Reis wrote:

> Robert Dewar <dewar@gnat.com> writes:
> 
> | Zack Weinberg wrote:
> | > "R. D. Flowers" <rd@chatta.us> writes:
> | >
> | >>would not some flag like -fp-pedantic (or however to fit in the
> | >>namespace) be a good idea (for both decent sets of folks: the
> | >>fast-and-a-little-dirty and the clean-and-a-little-slow people)?
> | > Maybe.  What would it do?
> | 
> | I don't think this would work. There is nothing "pedantic" in
> | the carefully formulated rules that make it easier/possible
> | to write accurate floating-point code. It is far better to
> | have the existing flags for -ffast-math etc, which clearly
> | warn that the resulting semantics may be unreliable.
> 
> Fully agreed.
> 
> | The use of -pedantic to cover little used language features
> | or rules is one thing, the use of a flag like this to give
> | incorrect semantics.


Welllll, as far as what to do, you folks probably do in fact know best.

But, if some code is broken by -pedantic, isn't it also incorrect in SOME picky litte sense? 

That people in some of their actual situations might not care about? Similar (in some actual situations) to not caring about that final shred of accuracy?


> 
> In fact, the flag "-pedantic" is misnamed -- at least, for what it does
> currently in C++ (I think the same is true for C).  It is there only
> for historical accident.  Multiplying that naming by a factor greater
> than 1 would be misguided.
> 
> -- Gaby
> 
> 
> 


-- 

R. D. Flowers, O. H.
Xinna R. Coloohiy
The Bear At Home
Now a single dad.

I   M i s s   A m e r i c a

http://chalice.us/mylinks
http://chalice.us/pho
http://chalice.us/poe
http://chatta.us/resume.txt
Mailing lists at http://chatta.us

-- "He who cannot protect himself or his nearest and dearest or
their honor by non-violently facing death may and ought to do so
by violently dealing with the oppressor. He who can do neither of
the two is a burden."
-- Mohandes Ghandi

Refer !
Stop Tritium !

Re: jFP religious wars

[Robert Dewar]

R. D. Flowers wrote:

> But, if some code is broken by -pedantic, isn't it also incorrect in 
> SOME picky litte sense?
> That people in some of their actual situations might not care about? 
> Similar (in some actual situations) to not caring about that final shred 
> of accuracy?

Not at all similar, please review what -pedantic does (don't just
reason from the name :-)