Probably I am missing something important, I have tried this code with the 1.034 (that compiles my large d libs fine), but I have found many problems:

import std.stdio: putr = writefln;

void main() {
    int[] a1 = [1, 2, 3];
    int[] a2 = [2, 4, 6];

    //putr(a1[] + a2[]); // test.d(6): Error: Array operations not implemented

    auto a3 = a1[] + 4;
    putr(a3); // [1,2,3,0,0,0,0]

    int[] a4 = a1[] + a2[]; // test.d(12): Error: Array operations not implemented

    int[] a5 = [3, 5, 7, 9];
    int[] a6 = a1 + a5; // test.d(16): Error: Array operations not implemented

    int[] a7;
    a7[] = a1[] + a2[];
    putr(a7); // prints: []

    auto a8 = a1 + a2; // test.d(21): Error: Array operations not implemented
    putr(a8);
}


Few more questions/notes:
- I like a syntax as a+b and a[]+4 instead of a[]+b[] and a[]+4, I am used to that from PyLab, etc.
- How does it works (or not works) with jagged/square matrices?
- When possible I'm do few benchmarks compared to normal D code, C code compiled normally with GCC and C code automatically vectorized by GCC.
- Is it able to compute a+b+c with a single loop (as all Fortran compilers do)? I presume the answer is negative.
- Hopefully in the future they may support the SSE3/SSSE3 too that my CPU supports.

Bye, and good work,
bearophile

Walter Bright Wrote:

> This one has (finally) got array operations implemented. For those who want to show off their leet assembler skills, the initial assembler implementation code is in phobos/internal/array*.d. Burton Radons wrote the assembler. Can you make it faster?
> 
> http://www.digitalmars.com/d/1.0/changelog.html http://ftp.digitalmars.com/dmd.1.034.zip
> 
> http://www.digitalmars.com/d/2.0/changelog.html http://ftp.digitalmars.com/dmd.2.018.zip


This will probably make me sound like an idiot, but what are these array operations everyone's so stoked about? I've only been learning D for a week and a half, fill me in!
BTW, nice update!

bearophile:
> - I like a syntax as a+b and a[]+4 instead of a[]+b[] and a[]+4,

I meant:
a + b
a + 4

instead of:
a[] + b[]
a[] + 4

Michael P.:
> This will probably make me sound like an idiot, but what are these array operations everyone's so stoked about? I've only been learning D for a week and a half, fill me in!

If Walter gives you one link, you have to follow it before asking :-) http://www.digitalmars.com/d/1.0/arrays.html#array-operations

Bye,
bearophile

bearophile Wrote:

> Michael P.:
> > This will probably make me sound like an idiot, but what are these array operations everyone's so stoked about? I've only been learning D for a week and a half, fill me in!
> 
> If Walter gives you one link, you have to follow it before asking :-) http://www.digitalmars.com/d/1.0/arrays.html#array-operations
> 
> Bye,
> bearophile

Who knows where that could have led to... I was just playing it safe. :D

"Lars Ivar Igesund" <larsivar@igesund.net> wrote in message news:g7ias2$2kbo$1@digitalmars.com...
> Walter Bright wrote:
>
>> This one has (finally) got array operations implemented. For those who want to show off their leet assembler skills, the initial assembler implementation code is in phobos/internal/array*.d. Burton Radons wrote the assembler. Can you make it faster?
>>
>> http://www.digitalmars.com/d/1.0/changelog.html http://ftp.digitalmars.com/dmd.1.034.zip
>
> The array op docs aren't actually on the 1.0 array page. But great! I remember trying to use these 3 years ago :D

Too bad Tango doesn't support them yet.  :C

On Sat, Aug 9, 2008 at 7:30 AM, Walter Bright <newshound1@digitalmars.com> wrote:
> Bill Baxter wrote:
>>
>> That is pretty neat.
>>
>> So does this mean you've reconsidered your position on adding new
>> features to D1.x?
>> Because q{} strings,  1..10 literals, and that enhancement to IFTI
>> used by std.algorithm, all sure would be nice.  I'd take those over
>> fancy array ops any day.
>
> Array ops were always supposed to be there.

Ok, I thought the charter for D1 was no new language features, period.

FWIW, I always thought D1 IFTI was supposed to be smarter, so to me
adding array ops seems to be similar to porting the fix for 493 to D1.
 (http://d.puremagic.com/issues/show_bug.cgi?id=493)

--bb

bearophile wrote:
> Probably I am missing something important, I have tried this code
> with the 1.034 (that compiles my large d libs fine), but I have found
> many problems:
> 
> import std.stdio: putr = writefln;
> 
> void main() { int[] a1 = [1, 2, 3]; int[] a2 = [2, 4, 6];
> 
> //putr(a1[] + a2[]); // test.d(6): Error: Array operations not
> implemented

It only works if the top level is an assignment operation.

> 
> auto a3 = a1[] + 4; putr(a3); // [1,2,3,0,0,0,0]
> 
> int[] a4 = a1[] + a2[]; // test.d(12): Error: Array operations not
> implemented

Doesn't work for initializers.

> 
> int[] a5 = [3, 5, 7, 9]; int[] a6 = a1 + a5; // test.d(16): Error:
> Array operations not implemented

Doesn't work for initializers.

> 
> int[] a7; a7[] = a1[] + a2[]; putr(a7); // prints: []

I don't know what putr is.

> 
> auto a8 = a1 + a2; // test.d(21): Error: Array operations not
> implemented

Have to use slice [] operator.

> putr(a8); }
> 
> 
> Few more questions/notes: - I like a syntax as a+b and a[]+4 instead
> of a[]+b[] and a[]+4, I am used to that from PyLab, etc. - How does
> it works (or not works) with jagged/square matrices?

It doesn't.

 - When possible
> I'm do few benchmarks compared to normal D code, C code compiled
> normally with GCC and C code automatically vectorized by GCC. - Is it
> able to compute a+b+c with a single loop (as all Fortran compilers
> do)?

Yes.

 I presume the answer is negative. - Hopefully in the future they
> may support the SSE3/SSSE3 too that my CPU supports.
> 
> Bye, and good work, bearophile

bearophile wrote:
> bearophile:
>> - I like a syntax as a+b and a[]+4 instead of a[]+b[] and a[]+4,
> 
> I meant:
> a + b
> a + 4
> 
> instead of:
> a[] + b[]
> a[] + 4

D already distinguishes operations on the array handle, a, from operations on the contents of a, a[]. I think this is a good distinction.

Walter Bright wrote:
> This one has (finally) got array operations implemented. For those who want to show off their leet assembler skills, the initial assembler implementation code is in phobos/internal/array*.d. Burton Radons wrote the assembler. Can you make it faster?
> 
> http://www.digitalmars.com/d/1.0/changelog.html
> http://ftp.digitalmars.com/dmd.1.034.zip
> 
> http://www.digitalmars.com/d/2.0/changelog.html
> http://ftp.digitalmars.com/dmd.2.018.zip

Sweet!  I love the way you put this forth as a challenge.  Maybe D will have the worlds fastest array operations :)

-Joel