About BitBlt for non graphical problems

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

About BitBlt for non graphical problems

Rafael Luque
Hi all,

I remember to read in the Blue Book about a Game Of Life solution based on the BitBlt operation. I wondered if it was only a smart and elegant demonstration or it may be a useful tool in the case you need parallelism, even in non-graphical contexts. I've never seen BitBlt used to solve non-graphical problems, but I don't have experience with BitBlt in my everyday work, so maybe I'm biased.

I've found a paper by Leo J. Guibas and Jorge Stolfi (https://www.cs.tufts.edu/~nr/cs257/archive/leo-guibas/language-bitmap.pdf) about a "calculus of bitmap operations" including an specific language called MUMBLE, but I suspect it is not a common use case.

Anyway, I've experimented with this option to solve the Advent of Code's Day 3 challenge. The puzzle consists of counting the number of trees you will encounter traversing a given map with open squares and trees, starting at the top-left corner and following a certain slope ((https://adventofcode.com/2020/day/3).

After implementing the straight-forward and iterative solution, I tried using Form instances representing the map and the slope and a BitBlt operation for ANDing both forms in order to leave only bits where you encounter a tree (see attached image). In this alternative solution I like how the iteration is replaced by only one "BitBlt>>copyBits" operation. Additionally, I understand that BitBlt operations are native primitives that could be hardware optimized.

However, I didn't know how to count the resulting bits without recurring again to the iteration of the Form's bits. This is my current implementation:

AoCBitBltForestMap>>countTreesOnPath
    | mapForm |
    mapForm := self form.
    (BitBlt toForm: mapForm)
        copyForm: self slopeMaskForm
        to: 0 @ 0
        rule: Form and.

    ^ mapForm bits count: [ :bit |
           (Color colorFromPixelValue: bit depth: 32) green
           closeTo: 1
           precision: 0.1 ]

You can find all the sources at the following repository: https://github.com/luque/AdventOfCode2020

Any comments about BitBlt use cases or this specific puzzle will be of interest for me.

Thank you.

 



Captura de pantalla 2020-12-21 a las 10.50.26.png (492K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: About BitBlt for non graphical problems

Stéphane Ducasse
I do not have your answer but I see that you are having fun… :)
Continue. 

I’m playing with VM implementation for the exact same reason :)

On 21 Dec 2020, at 10:56, Rafael Luque <[hidden email]> wrote:

Hi all,

I remember to read in the Blue Book about a Game Of Life solution based on the BitBlt operation. I wondered if it was only a smart and elegant demonstration or it may be a useful tool in the case you need parallelism, even in non-graphical contexts. I've never seen BitBlt used to solve non-graphical problems, but I don't have experience with BitBlt in my everyday work, so maybe I'm biased.

I've found a paper by Leo J. Guibas and Jorge Stolfi (https://www.cs.tufts.edu/~nr/cs257/archive/leo-guibas/language-bitmap.pdf) about a "calculus of bitmap operations" including an specific language called MUMBLE, but I suspect it is not a common use case.

Anyway, I've experimented with this option to solve the Advent of Code's Day 3 challenge. The puzzle consists of counting the number of trees you will encounter traversing a given map with open squares and trees, starting at the top-left corner and following a certain slope ((https://adventofcode.com/2020/day/3).

After implementing the straight-forward and iterative solution, I tried using Form instances representing the map and the slope and a BitBlt operation for ANDing both forms in order to leave only bits where you encounter a tree (see attached image). In this alternative solution I like how the iteration is replaced by only one "BitBlt>>copyBits" operation. Additionally, I understand that BitBlt operations are native primitives that could be hardware optimized.

However, I didn't know how to count the resulting bits without recurring again to the iteration of the Form's bits. This is my current implementation:

AoCBitBltForestMap>>countTreesOnPath
    | mapForm |
    mapForm := self form.
    (BitBlt toForm: mapForm)
        copyForm: self slopeMaskForm
        to: 0 @ 0
        rule: Form and.

    ^ mapForm bits count: [ :bit |
           (Color colorFromPixelValue: bit depth: 32) green
           closeTo: 1
           precision: 0.1 ]

You can find all the sources at the following repository: https://github.com/luque/AdventOfCode2020

Any comments about BitBlt use cases or this specific puzzle will be of interest for me.

Thank you.

 


<Captura de pantalla 2020-12-21 a las 10.50.26.png>

--------------------------------------------
Stéphane Ducasse
03 59 35 87 52
Assistant: Aurore Dalle 
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley, 
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France