OpenCL Kernel Code Decorator [Updated Sep 04 2011]

[Petr Schreiber]

OpenCL is powerful technology, but it has one problem - it takes many lines of code to just setup the computation. To target this obstacle, thinBasic OpenCL headers (http://www.thinbasic.com/community/showthread.php?10159-OpenCL-Headers) offer some high level functionality and as a bonus, here comes the OpenCL Kernel Code Decorator tool.
Now remastered and OpenCL 1.1 headers ready!

How to use?

Start the program
Choose .cl file with your OpenCL C kernels
Program will detect their names
Choose for which kernel you wish to generate code
Hit generate
Program will generate for you code template, with prepaired bindings to kernel parameters and other automagically detected features
Generated code is 100% of code needed to run your computation, but you need of course to add few lines of code to specify the data passed

I wish it serves you well!,
Petr

[Charles Pegge]

Thanks Petr,

I think it will take about four years before most desktops have the hardware to do generic parallel computing on the GPU. I think double precision maths is necessary for this to come about, and only the latest generation of video cards have this ability.

It would be great if this parallelism could be tightly integrated in to basic. It could be done almost transparently if the interpreter/compiler could identify parallel situations - iteratively processing large arrays for instance.

Charles

[kryton9]

You are a brave guy Petr taking on OpenCL.