www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - [Issue 6514] New: dotProduct at compile-time too

reply d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=6514

           Summary: dotProduct at compile-time too
           Product: D
           Version: D2
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Phobos
        AssignedTo: nobody puremagic.com
        ReportedBy: bearophile_hugs eml.cc


--- Comment #0 from bearophile_hugs eml.cc 2011-08-16 14:00:14 PDT ---
Currently (DMD 2.054) I can't run dotProduct() at compile-time, but small
changes allow it too. Not much tested:


/// Ditto
Unqual!(CommonType!(F1, F2))
dotProduct(F1, F2)(in F1[] avector, in F2[] bvector) pure nothrow  system {
    enum size_t big_block_size = 16;
    enum size_t small_block_size = 4;
    //static assert(isPow2(big_block_size)); // isPow2 missing in std.math
    //static assert(isPow2(small_block_size));

    immutable n = avector.length;
    assert(n == bvector.length);
    auto avec = avector.ptr, bvec = bvector.ptr;
    typeof(return) sum0 = 0, sum1 = 0;

    const all_endp = avec + n;
    const smallblock_endp = avec + (n & ~(small_block_size - 1));
    const bigblock_endp = avec + (n & ~(big_block_size - 1));

    for (; avec != bigblock_endp; avec += big_block_size, bvec +=
big_block_size) {
        sum0 += avec[0] * bvec[0];
        sum1 += avec[1] * bvec[1];
        sum0 += avec[2] * bvec[2];
        sum1 += avec[3] * bvec[3];
        sum0 += avec[4] * bvec[4];
        sum1 += avec[5] * bvec[5];
        sum0 += avec[6] * bvec[6];
        sum1 += avec[7] * bvec[7];
        sum0 += avec[8] * bvec[8];
        sum1 += avec[9] * bvec[9];
        sum0 += avec[10] * bvec[10];
        sum1 += avec[11] * bvec[11];
        sum0 += avec[12] * bvec[12];
        sum1 += avec[13] * bvec[13];
        sum0 += avec[14] * bvec[14];
        sum1 += avec[15] * bvec[15];
    }

    for (; avec != smallblock_endp; avec += small_block_size, bvec +=
small_block_size) {
        sum0 += avec[0] * bvec[0];
        sum1 += avec[1] * bvec[1];
        sum0 += avec[2] * bvec[2];
        sum1 += avec[3] * bvec[3];
    }

    sum0 += sum1;

    // Do trailing portion in naive loop.
    if (__ctfe) {
        const size_t left = (n % big_block_size) % small_block_size;
        for (size_t i = n - left; i < n; i++)
            sum0 += avec[i] * bvec[i];
    } else {
        while (avec != all_endp)
            sum0 += (*avec++) * (*bvec++);
    }

    return sum0;
}

----------------------------


Note that this block of code:

sum0 += avec[0] * bvec[0];
sum1 += avec[1] * bvec[1];
sum0 += avec[2] * bvec[2];
sum1 += avec[3] * bvec[3];
sum0 += avec[4] * bvec[4];
sum1 += avec[5] * bvec[5];
sum0 += avec[6] * bvec[6];
sum1 += avec[7] * bvec[7];
sum0 += avec[8] * bvec[8];
sum1 += avec[9] * bvec[9];
sum0 += avec[10] * bvec[10];
sum1 += avec[11] * bvec[11];
sum0 += avec[12] * bvec[12];
sum1 += avec[13] * bvec[13];
sum0 += avec[14] * bvec[14];
sum1 += avec[15] * bvec[15];


Equals to something like:

foreach (i; Iota!(big_block_size / 2)) {
    sum0 += avec[2 * i] * bvec[2 * i];
    sum1 += avec[2 * i + 1] * bvec[2 * i + 1];
}

Where Iota is defined in Issue 4085

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 16 2011
next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=6514


David Simcha <dsimcha yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dsimcha yahoo.com


--- Comment #1 from David Simcha <dsimcha yahoo.com> 2011-08-17 05:04:09 PDT ---
https://github.com/D-Programming-Language/phobos/pull/200

I was curious why dotProduct didn't already work at compile time.  With Don's
recent CTFE improvements that allow pointers, the only reason is bug 6517, so I
put the trivial workaround in dotProduct.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 17 2011
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=6514



--- Comment #2 from bearophile_hugs eml.cc 2011-08-17 05:36:32 PDT ---
(In reply to comment #1)


 With Don's recent CTFE improvements that allow pointers, the only reason
 is bug 6517, so I put the trivial workaround in dotProduct.


Thank you.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 17 2011
prev sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=6514


bearophile_hugs eml.cc changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


--- Comment #3 from bearophile_hugs eml.cc 2011-08-17 11:36:17 PDT ---
Fixed:

https://github.com/D-Programming-Language/phobos/commit/dc7fb2a8470e40929bbb67f74d62f01d61f5bcad

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 17 2011