On Wed, 2008-01-16 at 19:36 +0200, Asko Kauppi wrote:
Though most people would take this as self-evident, please check out
the LNUM patch to allow faster processing (about 33% but varies
greatly) on ARM7 platform.
I'm surprised it's such a small advantage: but then most of my ARM
experience comes from machines that have a floating-point instruction
emulator rather than using softfloat, so FP operations are extremely
expensive.
I should try firing your patch up on a RISC OS box and do some
timings.
B.