ARM9 with VFP : how is double precision math implemented
posted this on June 23, 2011 08:15 PM
Working with Crossworks ARM version 2.01. Target is an LPC3250. Have properties set for ARM FPU Type = "VFP9". And ARM FP ABI Type to "SoftFP". From the map file I see library "libm_v5te_vfp_a_le.a
And am wondering if double precision maths use double precision VFP instructions? Or are all math operations - both single and double precision - implemented with single precision VFP instructions? Or maybe it is the other way around. All math operations are implemented with double precision VFP instructions and the answer converted to single at the end if needed?
I'd like to know and cannot tell the answer when looking at the assembly. Not familiar enough with such.