Quote:
Originally Posted by Mathesar
Just checked and yes it does unfortunately...
However, the 68020-68040 version does run now on my 68LC060 (REALLY GREAT JOB!) but very, very slow.
Looking at the makefile it seems like an easy job to make the 060 version use soft float too. I guess most of the encrytion code is integer anyway so the speed impact is probably minimal?
I know I'm in the minority here (using an 68LC060) but it would be very welcome to have an FPU-less 060 version.
|
I thought it would be better to let GCC use the FPU for the 060, but I kinda forgot there was a FPU-less 060! All the FPU crash reports that I saw were from 68020/030 users.
On m68k, yes, floating point use is minimal (on PPC, for example, it is used in asm versions of some algorithms to boost performance). When I disassembled the 68060 AmiSSL build, I think I spotted a couple of uses of fsin or fcos, for example, but mainly I think GCC just makes use of the FPU registers as extra storage space, to reduce the need to make standard register saves to the stack.
I'll do some checking to make sure soft float doesn't affect performance the 68060, but unfortunately you'll have to wait until AmiSSL 4.5 for the fix. But, it will not take 12 months for an update this time
. I have a few more things I want to check in AmiSSL, such as possible memory leaks in OpenSSL and also to see if I can figure out why OpenSSL does not work correctly when using a higher GCC optimisation level on m68k.