"Dave N6NZ" wrote: > Ron Kreymborg wrote: >>> -----Original Message----- >>> From: Weddington, Eric [mailto:ewedding...@cso.atmel.com] >>> Sent: Sunday, 21 December 2008 6:00 AM >>> To: Ron Kreymborg; avr-libc-dev@nongnu.org >>> Subject: RE: [avr-libc-dev] 64-bit doubles >>> >>> >>> >>>> -----Original Message----- >>>> From: >>>> avr-libc-dev-bounces+eweddington=cso.atmel....@nongnu.org >>>> [mailto:avr-libc-dev-bounces+eweddington=cso.atmel....@nongnu. >>>> org] On Behalf Of Ron Kreymborg >>>> Sent: Saturday, December 20, 2008 5:03 AM >>>> To: avr-libc-dev@nongnu.org >>>> Subject: [avr-libc-dev] 64-bit doubles >>>> >>>> A quick look at AVR Libc for 20081205 still lists doubles as >>>> only 32-bit. Is >>>> this so? If yes does anyone know of a 64-bit double library? >>>> I am pointing >>>> heliostats and 32-bits are just not enough. >>>> >>>> Ron >>> Hi Ron, >>> >>> Yes, that is correct. Floats and doubles are 32-bits. No, I don't know >>> of a library. We (the avr-libc developers) have been advertising for >>> volunteers to help put together 64-bit double support for quite some >>> time. Nobody has volunteered yet. >>> >>> Eric Weddington >> >> Ok. For this project my good sense is saying go to a mcf52xx or M16C and a >> commercial compiler, but I have written fp code for now ancient processors >> and used the CORDIC transformations for transcendentals, so I'll have a look >> at AVRs over the xmas break. No promises... >> >> Ron >> > Is support for 64 bit double limited only by availability of a math > library? I would expect that 8-byte variables, register allocations, > etc, would also be an issue, since avr-gcc doesn't have anything bigger > than a 32-bit long or 32-bit float right now. > > Isn't there some compiler infrastructure that would have to happen also?
avr-gcc does support 64-bit integers (signed long long int, and unsigned long long int), and there is nothing to stop someone from writing a library that uses their own 64-bit float type without making changes to the compiler. You would not be able to use the built in double type for 64-bit floats without making changes to the compiler itself. -Preston _______________________________________________ AVR-libc-dev mailing list AVR-libc-dev@nongnu.org http://lists.nongnu.org/mailman/listinfo/avr-libc-dev
