[FYI wayland] tests: add noop to fixed-benchmark

[Kristian Høgsberg]

On Tue, May 15, 2012 at 10:29:12AM +0300, Pekka Paalanen wrote:
> I was just curious of how much the looping takes time without
> conversion, so I added this.
> 
> My results on Intel(R) Core(TM) i5-2410M CPU @ 2.30GHz:
> 
> benchmarked noop:	1.876349827s
> benchmarked magic:	2.245844470s
> benchmarked div:	12.709085309s
> benchmarked mul:	7.504838141s
> 
> Mul seems to take 15x the time magic does, cool!

Hmm, I get

   8:   48 89 44 24 f8          mov    %rax,-0x8(%rsp)
   d:   48 83 c0 01             add    $0x1,%rax
  11:   f2 0f 10 44 24 f8       movsd  -0x8(%rsp),%xmm0
  17:   48 3d ff ff ff 7f       cmp    $0x7fffffff,%rax
  1d:   f2 0f 11 05 00 00 00    movsd  %xmm0,0x0(%rip)        # 25 <noop_conversion+0x25>
  24:   00 
                        21: R_X86_64_PC32       global_d-0x4
  25:   75 e1                   jne    8 <noop_conversion+0x8>

and

benchmarked noop:	1.624387315s
benchmarked magic:	1.387375406s
benchmarked div:	8.840397131s
benchmarked mul:	5.192666376s

so it's a little wonky... just doing

        for (f = 0; f < INT32_MAX; f++)
                global_d = 1.0;

gives me

   8:   48 ba 00 00 00 00 00    movabs $0x3ff0000000000000,%rdx
   f:   00 f0 3f 
  12:   83 e8 01                sub    $0x1,%eax
  15:   48 89 15 00 00 00 00    mov    %rdx,0x0(%rip)        # 1c <noop_conversion+0x1c>
                        18: R_X86_64_PC32       global_d-0x4
  1c:   75 ea                   jne    8 <noop_conversion+0x8>

and

benchmarked noop:	0.652343463s
benchmarked magic:	1.528544144s
benchmarked div:	8.675264684s
benchmarked mul:	5.123568706s

which is probably a more accurate estimate of just the loop overhead.
Either way, it's clear that the integer magic is faster by a good
margin, which is what I was trying to establish.

Kristian

> Btw. the simple default cast of int32_t to double is slower than magic
> for me, hence the use of union.
> 
> Signed-off-by: Pekka Paalanen <ppaalanen at gmail.com>
> ---
>  tests/fixed-benchmark.c |   16 ++++++++++++++++
>  1 files changed, 16 insertions(+), 0 deletions(-)
> 
> diff --git a/tests/fixed-benchmark.c b/tests/fixed-benchmark.c
> index 3f7aae3..0d7abd0 100644
> --- a/tests/fixed-benchmark.c
> +++ b/tests/fixed-benchmark.c
> @@ -30,6 +30,21 @@
>  volatile double global_d;
>  
>  static void
> +noop_conversion(void)
> +{
> +	wl_fixed_t f;
> +	union {
> +		int64_t i;
> +		double d;
> +	} u;
> +
> +	for (f = 0; f < INT32_MAX; f++) {
> +		u.i = f;
> +		global_d = u.d;
> +	}
> +}
> +
> +static void
>  magic_conversion(void)
>  {
>  	wl_fixed_t f;
> @@ -80,6 +95,7 @@ benchmark(const char *s, void (*f)(void))
>  
>  int main(int argc, char *argv[])
>  {
> +	benchmark("noop", noop_conversion);
>  	benchmark("magic", magic_conversion);
>  	benchmark("div", div_conversion);
>  	benchmark("mul", mul_conversion);
> -- 
> 1.7.3.4
> 
> _______________________________________________
> wayland-devel mailing list
> wayland-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/wayland-devel