Reduce or improve calls to clamp0to255 in jpg js

[CodingFabian]

The following picture shows a profiling from IE10.

it shows that clamping is quite frequent. And because it is IE, it is super slow.
http://jsperf.com/clamp-functions/12
confirms that IE is way behind all modern browsers.

The question I am asking myself:
From the perspective of the algorithm, is the clamping actually needed?
and if so, is it needed in both directions?
Would a | 0 truncation be sufficient?

Maybe @p01 or @bthorben can give me a hint or help me out here.

[p01]

The clamping is needed, OTOH using Uint8ClampedArrays would work like a charm... except not all browsers support them ( PDF.js aims to work on a very broad range of browsers ) and shiming them using getters and setters would slow the shim'ed code path even more and take more memory.

[CodingFabian]

would it be possible to polyfill Uint8ClampedArray with Canvas­Pixel­Array ? Canvas­Pixel­Array is supported by IE9.

[yurydelendik]

would it be possible to polyfill Uint8ClampedArray with Canvas­Pixel­Array

Yeah, I thought about that. We will need to do that at some point -- it might take time to review entire code base to fix that for all image processing and colorspace parts.

[CodingFabian]

wouldn't the data be automatically clamped when painted to canvas?

[p01]

Unfortunately the HTML5 spec remained fuzzy about clamping vs clipping for a while which lead to different behaviors across browser. Not sure what is the state of affairs in the browsers targeted by PDF.js

[CodingFabian]

here a polyfill
https://gist.github.com/gengkev/2295355