code review 1021041: pidigits: ~10% performance win by using adds instead of...

pidigits: ~10% performance win by using adds instead of shifts

user time for pidigits -s -n=10000:
6.466s w/ adds
7.138s w/ shifts

[gri, 2010-04-27 21:00:41]

Hello rsc (cc: golang-dev@googlegroups.com),

I'd like you to review this change.

[rsc, 2010-04-27 21:02:44]

LGTM

TODO belongs in package big

On Tue, Apr 27, 2010 at 14:00,  <gri@golang.org> wrote:
> Reviewers: rsc,
>
> Message:
> Hello rsc (cc: golang-dev@googlegroups.com),
>
> I'd like you to review this change.
>
>
> Description:
> pidigits: ~10% performance win by using adds instead of shifts
>
> user time for pidigits -s -n=10000:
> 6.466s w/ adds
> 7.138s w/ shifts
>
> Please review this at http://codereview.appspot.com/1021041/show
>
> Affected files:
>  M test/bench/pidigits.go
>
>
> Index: test/bench/pidigits.go
> ===================================================================
> --- a/test/bench/pidigits.go
> +++ b/test/bench/pidigits.go
> @@ -63,7 +63,7 @@
>        }
>
>        // Compute (numer * 3 + accum) / denom
> -       tmp1.Lsh(numer, 1)
> +       tmp1.Add(numer, numer) // tmp1.Lsh(numer, 1)  TODO(gri) make Lsh
> faster
>        tmp1.Add(tmp1, numer)
>        tmp1.Add(tmp1, accum)
>        tmp1.DivMod(tmp1, denom, tmp2)
> @@ -84,7 +84,7 @@
>        y2.New(k*2 + 1)
>        bigk.New(k)
>
> -       tmp1.Lsh(numer, 1)
> +       tmp1.Add(numer, numer)  // tmp1.Lsh(numer, 1)  TODO(gri) make Lsh
> faster
>        accum.Add(accum, tmp1)
>        accum.Mul(accum, y2)
>        numer.Mul(numer, bigk)
>
>
>

[gri, 2010-04-27 21:06:56]

*** Submitted as http://code.google.com/p/go/source/detail?r=4623445fd946 ***

pidigits: ~10% performance win by using adds instead of shifts

user time for pidigits -s -n=10000:
6.466s w/ adds
7.138s w/ shifts

R=rsc
CC=golang-dev
http://codereview.appspot.com/1021041