On some FSL BookE cores (such as e500mc, e5500 and e6500) there does have
a FPU. But some floating point instructions are not supported by the FPU.
Introduce a kernel option so we can make the kernel only support to
emulate these unimplemented instructions. This will definitely reduce
the footprint of the kernel. And it also trim down the kernel build time
a little as predicted by Scott.
Before the patch:
Name Size
.text 00618948
After the patch:
Name Size
.text 00605328
The kernel build time(I have reboot the build server before each kernel build)
Before the patch:
real 2m29.653s
user 10m21.776s
sys 0m32.832s
After the patch:
real 2m27.767s
user 10m11.808s
sys 0m32.717s
Kevin Hao (2):
powerpc: split the math emulation into two parts
powerpc/mpc85xx: only emulate the unimplemented FP instructions on
corenet64
arch/powerpc/Kconfig | 20 ++++++++++++++++++++
arch/powerpc/configs/corenet64_smp_defconfig | 1 +
arch/powerpc/math-emu/Makefile | 24 ++++++++++++------------
arch/powerpc/math-emu/math.c | 20 ++++++++++++++------
4 files changed, 47 insertions(+), 18 deletions(-)
--
1.8.1.4
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
