2009/10/12 mvv <[email protected]>: > > > On Oct 9, 8:45 pm, Erik Corry <[email protected]> wrote: >> Congratulations on getting so far. We would love to see patches if >> you can sign the contributors' agreement. > > Thanks. Once we get it working I believe that this will be the case. > > Now onto the ABI. > > After some testing, disassembling and reading I found that WinCE > clearly uses ARM EABI. So 'USE_ARM_EABI' should be defined. Once we > are using the correct ABI double/double operations work correctly. > This is good. > > However there is still a problem with smi/double and smi/smi > operations. I that case it looks like numbers are not correctly > defined (encoded) in register pairs r0-r1 and r2-r3. > > I also observed that in case of smi/double and smi/smi operations, > results are numbers that have their r1 register set to zero; This > probably leads us to ConvertToDoubleStub class. Do we have endian > problem maybe?
I think the problem is around line 4902 in objects.h. Here we use the __ARM_EABI__ macro instead of USE_ARM_EABI. Probably you haven't defined the former, only the latter. That will lead to confusion about the floating point endianism. The line in objects.h should be fixed. > > While I'm trying to get this work any help is highly appriciated, > since I'm not good at ARM assembler or V8 internals for that matter. > > Thanks, > > -Marko. > > > > -- Erik Corry, Software Engineer Google Denmark ApS. CVR nr. 28 86 69 84 c/o Philip & Partners, 7 Vognmagergade, P.O. Box 2227, DK-1018 Copenhagen K, Denmark. --~--~---------~--~----~------------~-------~--~----~ v8-users mailing list [email protected] http://groups.google.com/group/v8-users -~----------~----~----~----~------~----~------~--~---
