One of the latest initiatives out of the Khronos Group -- the industry consortium that leads various open industry standards like OpenGL, OpenCL, and WebGL -- is SPIR. The goal of Khronos SPIR is to provide a Standard Portable IR for the OpenCL kernel language.
The provisional Khronos SPIR 1.0 specification was released in late August and in the past few days the SPIR discussion has ignited on the LLVM development list, since Khronos is actually basic this OpenCL kernel IR upon the LLVM IR.
SPIR specifies a standard intermediate representation (IR) to be used for the OpenCL kernel language, or simply said it's a standard IR for OpenCL programs so that there's binary compatibility between OpenCL drivers.
The Standard Portable IR is derived from LLVM and its IR, which already is leveraged quite extensively throughout the OpenCL/GPGPU and computing worlds. LLVM was chosen by the Khronos Group since it was deemed to be "highly portable" with its many back-ends and the open-source compiler infrastructure suits the needs of the Khronos stakeholders.
In theory, a SPIR front-end can generate OpenCL binaries in this IR format that can then be directly sent to any OpenCL drivers supporting SPIR. SPIR though doesn't govern the design and implementation of such front-ends. "Today, many of the OpenCL vendors base their technology on LLVM. This makes LLVM IR the de facto OpenCL IR and the immediate candidate to be considered by the Khronos members. An analysis showed that LLVM IR has its limitations but in general provides a very good solution for SPIR."
From Intel's Boaz Ouriel, here's his description from the LLVMdev mailing list thread
why the SPIR work is important:
SPIR offers binary portability between OpenCL implementations, and a stable target for 3rd party compilers without having to go through OpenCL "C".
Binary compatibility simplifies the support burden for developers delivering applications that use OpenCL. The same application can be delivered in fully binary form and work across existing and future OpenCL implementations supporting SPIR. This helps the entire OpenCL ecosystem.
Generally speaking OpenCL is a JIT environment and as such deserves and requires an intermediate representation like other major JIT environments already have.
Also, some developers using OpenCL have requested portability at binary level. Today OpenCL offers portability only at the source level with OpenCL “C”. They are concerned with protecting their IP by meeting “Digital Millennium Copyright Act” requirements. Today, those companies are forced to distribute their OpenCL code using device specific binaries. This leads to many difficulties for SW developers and end users. In addition, the binaries are not guaranteed to be functionally working as new devices and vendors appear in the market. This constraint places OpenCL standard in a disadvantage compared to other standards which already have a portable binary distribution form.
Aside from that original mailing list thread, there's also now a forked thread
The provisional SPIR 1.0 specification is available for viewing from Khronos.org