Re: -msoft-float for mipsel-linux?

["Kai Ruottu" <karuottu AT freenet.hut.fi>]

> i think i have to  #undef US_SOFTWARE_GOFAST
> in file t-linux

 Yes, there seemed to be the GOFAST-symbol-names instead
of GNU-ones in fp-bit/dp-bit...

> now df compilation passed but output result is strange
> still checking   (probably endian problem )

 I checked the little-endian support in my Win32-hosted mips-elf32
cross-tools and they all seemed to work nicely with '-EL -msoft-float'
(the default is '-EB')...

E:\usr\local\samples>objdump -p tst_mipsel-elf.gx

tst_mipsel-elf.gx:     file format elf32-littlemips

Program Header:
0x70000000 off    0x000000000000b510 vaddr 0x00000000a002a510 paddr 0x00000000a0
02a510 align 2**2
         filesz 0x0000000000000018 memsz 0x0000000000000018 flags r--
    LOAD off    0x0000000000001000 vaddr 0x00000000a0020000 paddr 0x00000000a002
0000 align 2**12
         filesz 0x000000000000b224 memsz 0x000000000000b3a0 flags rwx
private flags = 20000001: [no abi set] [mips3] [not 32bitmode]

How your 'output result is strange' ?  Can you tell more about this ?

Ok, the built-in soft-float in GCC may be ok in the embedded case, but what
about Linux ?  The x86-version has a 'x87'-emulator, a FP-exception triggers
it, so all programs are compiled just as there was a math-coprocessor...

But does the MIPS-processors without a FPU have something equal, i.e. a FP-
instruction without a FPU causes a trap or something ?  And Linux has a FPU-
emulator for this...

If there isn't a FPU-emulator in Linux/MIPS, using the basic soft-float 
routines in GCC sources, or something like the commercial GOFAST-library,
is the solution...

Cheers, Kai

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sourceware.cygnus.com