OpenCL SPIR (Standard Portable Intermediate Representation) is an intermediate representation for OpenCL-code, comparable to LLVM IL and HSAIL. It is a search for what would be a good representation, such that parallel software runs well on all kinds of accelerators. LLVM IL is too general, but SPIR is a subset of it. I’ll discuss HSAIL, on where it differs from SPIR – I thought SPIR was a better way to start introducing these. In my next article I’d like to give you an overview of the whole ecosphere around OpenCL (including SPIR and HSAIL), to give you an understanding what it all means and where we’re going to, and why.Know that the new SPIR-V is something completely different in implementation, and we are only discussing the old SPIR here.
Contributors for the SPIR specifications are: Intel, AMD, Altera, ARM, Apple, Broadcom, Codeplay, Nvidia, Qualcomm and Xilinx. Boaz Ouriel of Intel is the pen-holder of the specifications and to no surprise Intel has had the first SPIR-compiler. I am happy to see Nvidia is in the committee too, and hope they don’t just take ideas for CUDA from this collaboration but finally join. Broadcom and Xilinx are new, so we can expect stuff from them.
For now, just see what SPIR is – as it can help us understand how the compiler work and write better OpenCL code. I used Intel’s offline OpenCL compiler for compiling the below kernel to SPIR can be done on the command line with: ioc64 -cmd=build -input=sum.cl -llvm-spir32=sum.ll (you need an Intel CPU to use the compiler).
[raw]
__kernel void sum(const int size, __global float * vec1, __global float * vec2){
int ii = get_global_id(0);
if(ii < size) vec2[ii] += vec1[ii];
}
[/raw]
There are two variations for generating SPIR-code: binary SPIR, LLVM-SPIR (both in 32 and 64 bit versions). As you might understand, the binary form is not really readable, but SPIR described in the LLVM IL language luckily is. Run ioc64 without parameters to see more options (Assembly, pure LLVM, Intermediate Binary).
See the difference?
Let the competition on large memory GPUs begin!
The past year you might not have heard much from OpenCL-on-ARM, besides the Arndale developer-board. You have heard just a small portion of what has been going on.
There is an interesting book coming up: “Numerical Computations with GPUs” – a book explaining various numerical algorithms with code in CUDA or OpenCL.