[Pixman] [PATCH 3/3] MIPS: DSPr2: Added more fast-paths for OVER operation:

[Lukic, Nemanja]

Hi Soren,

You are right. This is certainly a performance decrease.
I made a fix for this. Actually case when srca == 0xff was not handled at all in these two routines, but only on over_n_8888 this showed up, since MIPS fast paths are using OVER combiner for 32bit destination.
I'll push another commit for these two fast-paths.

Current results look like this:

    Referent (before):
            over_n_0565 =  L1:  14.48  L2:  21.36  M: 17.57 ( 23.30%)  HT:  6.95  VT:  6.44  R:  6.39  RT:  2.16 (  22Kops/s)
            over_n_8888 =  L1:  92.60  L2:  86.13  M: 24.41 ( 64.74%)  HT:  8.94  VT:  8.06  R:  8.00  RT:  2.53 (  25Kops/s)

    Optimized:
            over_n_0565 =  L1:  27.65  L2: 189.22  M: 58.19 ( 77.12%)  HT: 52.80  VT: 49.88  R: 47.53  RT: 23.67 (  72Kops/s)
            over_n_8888 =  L1: 235.99  L2: 230.86  M: 29.09 ( 77.11%)  HT: 27.95  VT: 27.24  R: 26.58  RT: 18.10 (  67Kops/s)

Thanks,
Nemanja Lukic

-----Original Message-----
From: Søren Sandmann [mailto:sandmann at cs.au.dk] 
Sent: Monday, November 05, 2012 7:39 PM
To: Lukic, Nemanja
Cc: pixman at lists.freedesktop.org; nemanja.lukic at rt-rk.com
Subject: Re: [Pixman] [PATCH 3/3] MIPS: DSPr2: Added more fast-paths for OVER operation:

Nemanja Lukic <nlukic at mips.com> writes:

> From: Nemanja Lukic <nemanja.lukic at rt-rk.com>
>
> Performance numbers before/after on MIPS-74kc @ 1GHz:
>
> lowlevel-blt-bench results
>
> Referent (before):
>         over_n_0565 =  L1:  12.04  L2:  21.45  M: 18.50 ( 24.55%)  HT:  6.93  VT:  6.45  R:  6.38  RT:  2.16 (  22Kops/s)

This one:

>         over_n_8888 =  L1:  93.76  L2:  85.96  M: 24.41 ( 64.78%)  HT:  8.93  VT:  8.08  R:  7.99  RT:  2.54 (  25Kops/s)

>         over_n_8888 =  L1:  55.31  L2:  49.07  M: 28.60 ( 75.93%)  HT: 23.99  VT: 22.95  R: 22.34  RT: 12.85 (  61Kops/s)

looks like a performance regression for L1 and L2?


Søren