Jump to page: 1 2
Thread overview
Passing an in parameter and aliasing
May 23, 2007
renoX
May 23, 2007
Bill Baxter
May 25, 2007
renoX
May 25, 2007
Bill Baxter
May 26, 2007
Dave
May 26, 2007
Bill Baxter
May 26, 2007
Oskar Linde
May 26, 2007
Bill Baxter
May 26, 2007
renoX
May 26, 2007
Dave
May 23, 2007
I'm not sure whether to post this in D.learn or not, but here's my issue:

What happens when you have a function f(in T array[], inout T x)
and that x is in fact an element of the array?

AFAIK, there are three possibilities:
1- the compiler detect the incompatibility and complain, that's nice for the programmer but real hard work for the compiler and it's impossible to catch all the problems for the compiler.

2- the array is modified when x is modified, which can leads to bug depending on the compiler and its optimization level: hopefully this would be a rare bug, but very hard to catch.
And if parameter passing is by default as constants, then maybe the bug won't be so rare.

3- the array is copied which reduce performance is this is done by default.

So which way is-it supposed to be?

Regards,
renoX
May 23, 2007
"renoX" <renosky@free.fr> wrote in message news:f328ns$1f1q$1@digitalmars.com...
> I'm not sure whether to post this in D.learn or not, but here's my issue:
>
> What happens when you have a function f(in T array[], inout T x) and that x is in fact an element of the array?
>
> AFAIK, there are three possibilities:
> 1- the compiler detect the incompatibility and complain, that's nice for
> the programmer but real hard work for the compiler and it's impossible to
> catch all the problems for the compiler.
>
> 2- the array is modified when x is modified, which can leads to bug
> depending on the compiler and its optimization level: hopefully this would
> be a rare bug, but very hard to catch.
> And if parameter passing is by default as constants, then maybe the bug
> won't be so rare.
>
> 3- the array is copied which reduce performance is this is done by default.
>
> So which way is-it supposed to be?
>

It works like 2 right now.  3 is just impractical for most cases.  1 can be put in manually:

void func(int[] a, ref int x)
in
{
    assert(&x < a.ptr || &x >= (a.ptr + a.length), "x can't be a member of
a");
}
body
{
    x = 5;
    writefln(a);
    fflush(stdout); // without this, the assertion confusingly gets printed
first
}

void main()
{
    int[] a = ([1, 2, 3]).dup;
    int x = 0;

    func(a, x);
    func(a, a[0]);
}

You could probably wrap that ugly expression into a mixin or something.


May 23, 2007
Jarrett Billingsley wrote:
> "renoX" <renosky@free.fr> wrote in message news:f328ns$1f1q$1@digitalmars.com...
>> I'm not sure whether to post this in D.learn or not, but here's my issue:
>>
>> What happens when you have a function f(in T array[], inout T x)
>> and that x is in fact an element of the array?
>>
>> AFAIK, there are three possibilities:
>> 1- the compiler detect the incompatibility and complain, that's nice for the programmer but real hard work for the compiler and it's impossible to catch all the problems for the compiler.
>>
>> 2- the array is modified when x is modified, which can leads to bug depending on the compiler and its optimization level: hopefully this would be a rare bug, but very hard to catch.
>> And if parameter passing is by default as constants, then maybe the bug won't be so rare.
>>
>> 3- the array is copied which reduce performance is this is done by default.
>>
>> So which way is-it supposed to be?
>>
> 
> It works like 2 right now.  3 is just impractical for most cases.  1 can be put in manually:
> 
> void func(int[] a, ref int x)
> in
> {
>     assert(&x < a.ptr || &x >= (a.ptr + a.length), "x can't be a member of a");
> }

The thing with this is it may catch the error but it doesn't help the compiler generate more optimized code.  The C99 'restrict' keyword does, though.  I suppose the compiler could be smart enough to deduce 'restrict' semantics are in effect from the assert message above.  That would be pretty neat.

Linky (top google hit for 'C99 restrict keyword'): http://www.cellperformance.com/mike_acton/2006/05/demystifying_the_restrict_keyw.html


--bb
May 25, 2007
Bill Baxter a écrit :
> Jarrett Billingsley wrote:
>> "renoX" <renosky@free.fr> wrote in message news:f328ns$1f1q$1@digitalmars.com...
>>> I'm not sure whether to post this in D.learn or not, but here's my issue:
>>>
>>> What happens when you have a function f(in T array[], inout T x)
>>> and that x is in fact an element of the array?
>>>
>>> AFAIK, there are three possibilities:
>>> 1- the compiler detect the incompatibility and complain, that's nice for the programmer but real hard work for the compiler and it's impossible to catch all the problems for the compiler.
>>>
>>> 2- the array is modified when x is modified, which can leads to bug depending on the compiler and its optimization level: hopefully this would be a rare bug, but very hard to catch.
>>> And if parameter passing is by default as constants, then maybe the bug won't be so rare.
>>>
>>> 3- the array is copied which reduce performance is this is done by default.
>>>
>>> So which way is-it supposed to be?
>>>
>>
>> It works like 2 right now.  3 is just impractical for most cases.  1 can be put in manually:
>>
>> void func(int[] a, ref int x)
>> in
>> {
>>     assert(&x < a.ptr || &x >= (a.ptr + a.length), "x can't be a member of a");
>> }
> 
> The thing with this is it may catch the error but it doesn't help the compiler generate more optimized code.  The C99 'restrict' keyword does, though.  I suppose the compiler could be smart enough to deduce 'restrict' semantics are in effect from the assert message above.  That would be pretty neat.

Well, if I understood correctly Jarrett Billingsley's answer, the semantic on 'in' is the same as 'restrict': the fastest as the compiler will be free to optimize but the most fragile from programmers point of view because if you violate the restrict restriction the result is unpredictable (no bug, a bug depending on optimization flags, fun!)

Sure, it'd be nice if the compiler was able to do this kind of optimization without the in|restrict hint but this is hard.

renoX

May 25, 2007
renoX wrote:
> Bill Baxter a écrit :

>> The thing with this is it may catch the error but it doesn't help the compiler generate more optimized code.  The C99 'restrict' keyword does, though.  I suppose the compiler could be smart enough to deduce 'restrict' semantics are in effect from the assert message above.  That would be pretty neat.
> 
> Well, if I understood correctly Jarrett Billingsley's answer, the semantic on 'in' is the same as 'restrict': the fastest as the compiler will be free to optimize but the most fragile from programmers point of view because if you violate the restrict restriction the result is unpredictable (no bug, a bug depending on optimization flags, fun!)
> 
> Sure, it'd be nice if the compiler was able to do this kind of optimization without the in|restrict hint but this is hard.

The 'in' parameter type does nothing currently.  Leaving it off is the same as putting it in.  So I don't believe it acts like 'restrict'.  I certainly don't remember reading anywhere in the spec that D assumes all 'in' array parameters are unaliased.  That would be a big break from C/C++ if it were so, so I think the spec would go out of its way to point out the difference.

--bb
May 26, 2007
Bill Baxter wrote:
> renoX wrote:
>> Bill Baxter a écrit :
> 
>>> The thing with this is it may catch the error but it doesn't help the compiler generate more optimized code.  The C99 'restrict' keyword does, though.  I suppose the compiler could be smart enough to deduce 'restrict' semantics are in effect from the assert message above.  That would be pretty neat.
>>
>> Well, if I understood correctly Jarrett Billingsley's answer, the semantic on 'in' is the same as 'restrict': the fastest as the compiler will be free to optimize but the most fragile from programmers point of view because if you violate the restrict restriction the result is unpredictable (no bug, a bug depending on optimization flags, fun!)
>>
>> Sure, it'd be nice if the compiler was able to do this kind of optimization without the in|restrict hint but this is hard.
> 
> The 'in' parameter type does nothing currently.  Leaving it off is the same as putting it in.  So I don't believe it acts like 'restrict'.  I 

That's right (verifiable through the asm).

> certainly don't remember reading anywhere in the spec that D assumes all 'in' array parameters are unaliased.  That would be a big break from C/C++ if it were so, so I think the spec would go out of its way to point out the difference.
> 

I believe that's part of the overall justification for the new for 2.0 'in', which will mean 'scope const final' for any type of reference parameter (correct me if I'm wrong Walter). Then the optimizer will be free to do its thing with those references, w/o any complicated pointer analysis or runtime checking.

IIUC, the compiler front-end will do it's best to enforce it (unlike C's 'restrict') and I believe it won't be as easy to cast away like C++'s 'const&'.

If so that'd be pretty cool, because it would make D competitive with Fortran in this regard.

Toward the end of the thread on this over in d.D.announce, it looks like 'in' will be the default parameter (so 'foo(int[] array){}' would be the same as 'foo(in int[] array){}'). I'm hoping that's the case -- at least given a try first for 2.0 anyway.

> --bb
May 26, 2007
Dave wrote:
> Bill Baxter wrote:
>> renoX wrote:
>>> Bill Baxter a écrit :
>>
>>>> The thing with this is it may catch the error but it doesn't help the compiler generate more optimized code.  The C99 'restrict' keyword does, though.  I suppose the compiler could be smart enough to deduce 'restrict' semantics are in effect from the assert message above.  That would be pretty neat.
>>>
>>> Well, if I understood correctly Jarrett Billingsley's answer, the semantic on 'in' is the same as 'restrict': the fastest as the compiler will be free to optimize but the most fragile from programmers point of view because if you violate the restrict restriction the result is unpredictable (no bug, a bug depending on optimization flags, fun!)
>>>
>>> Sure, it'd be nice if the compiler was able to do this kind of optimization without the in|restrict hint but this is hard.
>>
>> The 'in' parameter type does nothing currently.  Leaving it off is the same as putting it in.  So I don't believe it acts like 'restrict'.  I 
> 
> That's right (verifiable through the asm).
> 
>> certainly don't remember reading anywhere in the spec that D assumes all 'in' array parameters are unaliased.  That would be a big break from C/C++ if it were so, so I think the spec would go out of its way to point out the difference.
>>
> 
> I believe that's part of the overall justification for the new for 2.0 'in', which will mean 'scope const final' for any type of reference parameter (correct me if I'm wrong Walter). Then the optimizer will be free to do its thing with those references, w/o any complicated pointer analysis or runtime checking.
> 
> IIUC, the compiler front-end will do it's best to enforce it (unlike C's 'restrict') and I believe it won't be as easy to cast away like C++'s 'const&'.
> 
> If so that'd be pretty cool, because it would make D competitive with Fortran in this regard.

I don't think Walter has said that 'in' as 'scope const final' is going to imply no aliasing.  I think it's kind of an orthogonal concern.

Say you have a vector difference function:

    void sub(mutable float[] ret, in float[] a, in float[] b) {
        for(size_t i=1; i<ret.length; i++) {
           ret[i] = a[i] - b[i];
        }
    }

Should you expect this to fail?
    float[10] c;
    for(i=0; i<=10; i++) c[i]=i;
    sub(c[0..9],c[1..$],c[0..9]);

Looking at the loop it should be fine.  It should calc
        c[0] = c[1] - c[0];
        c[1] = c[2] - c[1];
        c[2] = c[3] - c[2];
        ...
and leave you with c[0..9] containing all 1's.
But a smarty pants compiler might decide for some reason that it's faster to rewrite the loop as:
       memcpy(ret,a,float.sizeof*a.length);
       for (i=0;i<N;i++) { ret[i] -= b[i]; }

That's a fine transformation if b and ret are distinct.  But it isn't fine when b and ret are aliased.

That's the reason why C doesn't have restrict as default. The thinking is that whatever the compiler does under the hood, the result should be output that looks exactly like if you really did run the C code on a C virtual machine in the exact order written.  Even if users get tricky with inputs and do things that are guaranteed to give incorrect results, it should still give the "exact" incorrect result.  It really is a separate issue from whether the parameter can be modified or not.

That's C's philosophy anyway.  And I haven't heard anything from Walter that sounded to me like D was taking a different approach.  I think a 'restrict' type keyword probably is the best way to handle it in a C-like language.  Most code doesn't need the performance boost as much as it needs predictability and reliability.  Aliasing bugs can also be very tricky to debug because things might work great on your computer but only fail on some other machine when compiled with SSE3 instructions and -03 optimizations.  So it makes sense for performance critical code to declare in a visible way that it's placing some restrictions on the kind of inputs that will work.

So I think D could still use a separate restrict-like keyword.  But I think it could be made a little more user friendly than C's by allowing it to apply to a whole function rather than having to say 'restrict' 'restrict' 'restrict' for every single parameter.

--bb
May 26, 2007
Bill Baxter skrev:
> Dave wrote:
>> [restrict]
> That's C's philosophy anyway.  And I haven't heard anything from Walter that sounded to me like D was taking a different approach.  I think a 'restrict' type keyword probably is the best way to handle it in a C-like language.  Most code doesn't need the performance boost as much as it needs predictability and reliability.  Aliasing bugs can also be very tricky to debug because things might work great on your computer but only fail on some other machine when compiled with SSE3 instructions and -03 optimizations.  So it makes sense for performance critical code to declare in a visible way that it's placing some restrictions on the kind of inputs that will work.
> 
> So I think D could still use a separate restrict-like keyword.  But I think it could be made a little more user friendly than C's by allowing it to apply to a whole function rather than having to say 'restrict' 'restrict' 'restrict' for every single parameter.

Here is an old thread about "noalias, restrict, and Fortran optimizations":

http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28215&header

/Oskar
May 26, 2007
Oskar Linde wrote:
> Bill Baxter skrev:
>> Dave wrote:
>>> [restrict]
>> That's C's philosophy anyway.  And I haven't heard anything from Walter that sounded to me like D was taking a different approach.  I think a 'restrict' type keyword probably is the best way to handle it in a C-like language.  Most code doesn't need the performance boost as much as it needs predictability and reliability.  Aliasing bugs can also be very tricky to debug because things might work great on your computer but only fail on some other machine when compiled with SSE3 instructions and -03 optimizations.  So it makes sense for performance critical code to declare in a visible way that it's placing some restrictions on the kind of inputs that will work.
>>
>> So I think D could still use a separate restrict-like keyword.  But I think it could be made a little more user friendly than C's by allowing it to apply to a whole function rather than having to say 'restrict' 'restrict' 'restrict' for every single parameter.
> 
> Here is an old thread about "noalias, restrict, and Fortran optimizations":
> 
> http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28215&header 

Thanks for that.  Here's a few follow-up posts I found ... that news interfaces doesn't seem to have a "threaded" view, so finding the rest of the thread isn't easy.

http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28236
http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28372
http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28377
http://www.digitalmars.com/pnews/read.php?server=news.digitalmars.com&group=D&artnum=28379

Last thing I see from Walter is musing about making everything noalias with a way to turn that off.  My problem with that is most code doesn't need it, and it would be confusing for users not expecting it to find out that a[i] = b[i] - c[i]; doesn't always execute exactly as a[i]=b[i]-c[i].    Walter's right that the vast majority of code doesn't rely on C-like non-restrict behavior.  But I'd counter that the vast majority of code also doesn't need it.  Better to put something potentially dangerous as an option usable by people who know what they're doing rather than making it the default.

Anyway, generating optimized code is probably at least 2 years away from being anywhere near a priority for Walter.  :-)  [he says hoping he turns out to be wrong]

--bb
May 26, 2007
Thinking about it, you're right, but the hard thing is as always when you have func(in int[]array, int* p) that the programmer must not assume that array is unchanged because of the 'in' mode, it can be modified by the pointer reference (to prevent this the variable declaration must be change, the function parameter passing is not enough).

int f[1];
int* p=&f[0];
func(f,p)

func(in int[]array, int* p)
{
	*p = 1;
	/* here array[0] is 1 not 0 in this case */
}


renoX
PS: I couldn't read the thread linked by oskar linde as the web interface doesn't thread correctly the posts.
« First   ‹ Prev
1 2