IBM Corp’s PPC 403GA embedded PowerPC processor stands out in one way – it is the only processor with the PowerPC label that lacks a floating point unit. This can make things difficult for developers, since some graphical devices, or indeed laser printers, can make good use of floating point functions. Until now developers’ choices have been limited: they could either avoid floating point altogether, or carry on regardless and rely on the chip to trap floating point instructions and jump to an emulation routine. Unfortunately the latter method tends to be slow. But the Swedish Diab Data AB Compagnie des Machines Bull SA subsidiary says it will launch the next release of its compilers in the next couple of weeks, and this will include inline support for floating point functions on the 403GA. So when the compiler comes across a floating point call, and when it knows that the target processor is the 403GA, it will insert a routine to emulate the necessary instruction directly into the binary. IBM is also said to be working on a similar facility for its compiler. Diab’s US base is in Foster City, California.
