What I 'think' I know.... FDLibM most likely clashes with newer MiNTLib because it uses 68881 asm for the 68020-60 target. It would need to be rebuilt using soft-float, like 68000, to fix, I think. I am not sure though as I have not been able to test. Mainly, lack of time, and for that matter, CPU time. Need a second Falcon :).
I have been working with GMP / MPFR / MPC, which are the complex math libs that are the backbone of GCC 4.5. With CVS MiNTLib and PML, they are 100% for the 68020-60 target, at least according to the tests included with them. 68000 has a ton of problems with PML just with GMP, did not mess with MPFR and MPC for it. Can not yet test 5475. I have not had time to run full tests with using FDLibM as the math backend.
Right now I am working on a GCC, hopefully, targeted for 68020-60, using these working complex math libs. It is built (RPMs), just getting ready to move them over to my Falcon and install.
Have not really read thru the last couple weeks of messages, need to get to that still.... :)
Keith On Mon, 28 Feb 2011, m0n0 wrote:
On 28.02.2011 20:05, Peter Slegg wrote:On Sun, 27 Feb 2011 19:36:59 , Vincent Rivière<vincent.riviere@freesbee.fr> wrote:Is a tool, to test the installed libs, needed to make sure that all have beenAlso, be sure that *all* the libs you are using have been compiled with t he same compiler, because of the change about the register to return floats.updated ?I'm not trusting this stuff anymore ! Before I will try this a second time, I will make several tests with small executables I wrote myself, and if that works ok, i will try to link bigger stuff, and of course, I will recompile all the packages. AND... I will make an backup before trying any of this stuff. I would love to have a m68kmint cross compile cluster system for that task ;) Greets, m