Legacy ImageMagick Discussions Archive

Hi,

I'm building a small web server that crops/resizes gif(note: I'm using JMagick and Java). It works great but I'm having 2 issues.

1. I use coalesce and optimizeLayers method for resizing gif. The issue is converting resized gif(ImagesToBlob) to byte array is extremely slow(takes 3-4 sec for 2MB gif with 60 frames), I believe it's related to 'coalesce' operation. I even noticed it in command line. Is it normal behavior? Can it be optimized?

2. The above issue is much worser in multi-thread environment.
Here is time of resizing a single gif:
Single thread: 1357ms
Multi thread: 96267ms(4 threads)

I have OpenMP disabled and set MAGICK_TMPDIR to a big disk.

My system info:
2.7 GHz Intel Core i7
16 GB 1600 MHz DDR3

Appreciate for any comments or help.

Creating gif files can be slow due to the color quantization. You can try using -tree-depth to make it faster, but less accurate colors. See http://www.imagemagick.org/script/quantize.php and http://www.imagemagick.org/script/comma ... #treedepth.

Coalesce needs to fill out partial frames and that also can be slow. More cores/threads might help, but on some systems it can be slower than limiting the thread to 1.

Thanks for the suggestion.

Is it slow because of reading/writing to disk(since memory isn't big enough)? How much memory does it need not to use disk? Also, what are you referring to as 'some systems'?

What are the physical dimensions of the gif (WxHxframes) and is there transparency. IM will need at least twice the size of
W x H x frames x 32bits/pixel until it quantizes and compresses the output. Certain Linux systems, I am told, may not work well with some IM functions with multiple cores. They run slower. You did not say what platform you are using. I am guessing a Mac. I have not seen that problem before on a Mac.

400x225x60 frames. Uploaded the original image here http://uploadpie.com/eEYDb.

I'm using a Mac. Here's what IM has on my machine.

400x225x60*32*2=345600000 bytes = 345.6 Mbytes (my estimate of RAM needed)