Intel tests code compression
1 min read
According to Processor Watch, Intel Labs has demonstrated a compile time code compression technology in which a processor unpacks the compressed program at run time. The ultimate goal, says editor Tom Halfhill, is to cut silicon costs by reducing the amount of on chip memory.
The report says that Intel calls the technology 'direct compressed execution', although the compressed program is decompressed on the fly before execution.
Code compression techniques have been developed by other companies in the past in which 32bit instructions are replaced by 16bit opcodes – ARM's Thumb approach is one example.
However, instead of using a modified instruction set, the Intel approach is said to allow programmers to write, compile and link their code as usual, then apply a utility to compress the binary file. The executable file shrinks by as much as 33%.
Because the packed code is decompressed and executed on the fly, not unpacked into memory, the processor needs much less code storage. Halfhill claims this scheme can achieve the same goals as a 16bit instruction set, without further complicating the x86 instruction set architecture.
The report also cites a similar approach called CodePack, developed by IBM in 1998, which failed to catch on.
