quick handling of graphics files

On Thu, 15 Apr 2010, Juan Vuletich wrote:

> This is an updated version. It includes Levente's suggestion, and Yoshiki's
> and Andreas' alternatives as well, although disabled. All three options now
> work for gray level images with other bit depths besides 1 (I only tested 1
> and 8). I also fixed transparent pixels for gray level images (it was
> completely broken).

Nice, I ran the following benchmark with all three (+1) versions:

Smalltalk garbageCollect.
(1 to: 5) collect: [ :runs |
  [ ImageReadWriter formFromFileNamed: 'test.png' ] timeToRun ].

test.png is Ross' test image which is pretty large.
The results:

#Andreas -> #(37 35 36 37 36).
#Juan -> #(100 99 103 100 101).
#Juan2 -> #(75 74 75 75 76).
#Yoshiki -> #(38 38 39 38 38).

#Juan used the original method from your changeset, the others used the
attached version. It includes three enhancements:
- doesn't create a blitter when it's not necessary, this affects all
benchmarks
- holds the bitmap in a variable instead of accessing via #bits
- uses #bitShift: instead of #<<


Levente


>
> Cheers,
> Juan Vuletich
>

Levente Uzonyi wrote:
Cool! Will integrate your enh for Cuis asap!

Cheers,
Juan Vuletich

Andreas Raab wrote:
So, let's play!

<pokerface on>

Mh... Hard challenge. Let's see. I take your technique, but save all
your objects in new instance variables. And I go down from 67msecs down
to 31 msecs. A 110% speed increase over yours!

Who wins now? :)

<pokerface off>

Actually the my smalltalk version is not that bad in many situations.
For example, on 'plogo.png' taken from
http://palmzlib.sourceforge.net/images/dir.html , and evaluating:
            (1 to: 100) collect: [ :i | Smalltalk garbageCollect. [ 1
timesRepeat: [ Form fromFileNamed: 'test.png' ]] timeToRun ]

Yours gives an array with:
    self max -> 117
    self min -> 107
    self average*1.0 -> 108.13

Yoshiki's gives:
    self max -> 118
    self min -> 108
    self average*1.0 -> 110.63

And mine gives:
    self max -> 109
    self min -> 95
    self average*1.0 -> 97.68

So, the bitblt technique only makes sense if we avoid creating the
objects for each scanline, as this is expensive if scanlines are small.
This variant gives:
    self max  ->  103
    self min  ->  91
    self average * 1.0  ->  95.16
The gains is not as big as in test.png, because this one has less
scanlines. But it is still the winner.

Cheers,
Juan Vuletich

On 4/15/10 8:29 AM, Igor Stasenko wrote:
And now things come full circle. I recall reading an ancient article
about someone who used bitblt as a hardware accelerated matrix multiply
(or somesuch) back in the day.

Lawson
>
>

LOL I had been searching for a message by name in my mail app and forgot
to reset the sorting to "by date". That's a rather old thread, isn't it...