Re: [arm-gnu] benchmark: arm-none-linux-gnueabi-2009q3_67 much slower in floating point operation than arm-none-linux-gnueabi-gcc 2009q1_203?
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [arm-gnu] benchmark: arm-none-linux-gnueabi-2009q3_67 much slower in floating point operation than arm-none-linux-gnueabi-gcc 2009q1_203?
- To: Hui Chen <usa.chen@xxxxxxxxx>
- Subject: Re: [arm-gnu] benchmark: arm-none-linux-gnueabi-2009q3_67 much slower in floating point operation than arm-none-linux-gnueabi-gcc 2009q1_203?
- From: Mark Mitchell <mark@xxxxxxxxxxxxxxxx>
- Date: Fri, 23 Oct 2009 09:38:19 -0700
Hui Chen wrote:
> CFLAGS:=-std=gnu99 -O3 -mfpu=neon -mfloat-abi=softfp -ffast-math
> -mtune=cortex-a8 -mcpu=cortex-a8
Those command-line options look reasonable, though you might try -O2, or
-fno-tree-vectorize. Sometimes -O3 (which is quite aggressive) can
result in a lot of code growth and there are situations where
vectorization might not help.
Overall, our benchmarking shows significant benefits in 2009q3; for
example, on Cortex-A8 we saw 8% improvement on one suite of popular
embedded benchmarks. But, in every release, there are going to be some
applications that run slower, even if many run faster.
--
Mark Mitchell
CodeSourcery
mark@xxxxxxxxxxxxxxxx
(650) 331-3385 x713