Question, enabling NEON H/W

取消
显示结果 
显示  仅  | 搜索替代 
您的意思是: 
已解决

Question, enabling NEON H/W

跳至解决方案
2,810 次查看
Aemj
Contributor IV

Dear All,

My customer measured the performance of mathematical function, such as Logarithm or Power, on i.MX6S.

He thinks the performance will get better if NEON HW can be enabled.

Now he uses GCC compiler.

Please show me how to enable NOEN H/W for the math functions on i.MX6. (GCC flags, etc.)

Also, please let me know if you have any ideas to improve the math calculation performance.

Best Regards,

Miyamoto

标签 (1)
0 项奖励
回复
1 解答
1,752 次查看
Yuri
NXP Employee
NXP Employee

1.

Default GCC options regarding floating point support are :

-march=armv7-a -mfpu=vfpv3 -mfloat-abi=softfp

To use the NEON :

-march=armv7-a -mfpu=neon -mfloat-abi=softfp

2.

As for Linux :


“How to create ubuntu hardfloat rootfs for imx6d/q”

< https://community.freescale.com/docs/DOC-95185 >

在原帖中查看解决方案

0 项奖励
回复
1 回复
1,753 次查看
Yuri
NXP Employee
NXP Employee

1.

Default GCC options regarding floating point support are :

-march=armv7-a -mfpu=vfpv3 -mfloat-abi=softfp

To use the NEON :

-march=armv7-a -mfpu=neon -mfloat-abi=softfp

2.

As for Linux :


“How to create ubuntu hardfloat rootfs for imx6d/q”

< https://community.freescale.com/docs/DOC-95185 >

0 项奖励
回复