This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: State of m68k float emulation in lb1sf68.asm (muldf, divdf)
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: tm_gccmail at kloo dot net
- Cc: Jim Wilson <wilson at specifixinc dot com>, Pavel Pisa <pisa at cmp dot felk dot cvut dot cz>, gcc at gcc dot gnu dot org
- Date: Mon, 17 Nov 2003 15:27:23 -0500
- Subject: Re: State of m68k float emulation in lb1sf68.asm (muldf, divdf)
- References: <Pine.LNX.4.21.0311171149230.26515-100000@mail.kloo.net>
>>>>> Toshi writes:
Toshi> The gcc SH port uses fp-bit.c, and I have many users bitterly complaining
Toshi> that the gcc SH floating point is slow compared to the Renesas compiler.
Toshi> On Java benchmarks, the gcc sh-elf compiled code runs *3* times slower
Toshi> then the Renesas compiled one.
Toshi> That's 3 times slower even including all the JVM overhead, so I
Toshi> suspect the actual ratio for FP only is probably something like 10x
Toshi> slower.
Swox contributed another floating point emulation library to GCC
that is suppose to be faster. I believe that Aldy was experimenting with
it, but I do not know the final outcome of the analysis. There may have
been some technical limitation in the alternate library.
David