使用gnueabihf为ARMv6构建 [英] build for ARMv6 with gnueabihf

问题描述

我尝试为ARMv6构建应用程序，但是失败.

I try to build an app for ARMv6 but, fail.

我想问题出在工具链上，它支持Hard Float，但是ARMv6不支持.

I guess the problem is the Toolchain which supports Hard Float but, ARMv6 does not.

好吧，首先我设置-march = armv6，编译失败.

Well, first i set -march=armv6, where compilation fails.

/opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/arm-linux-gnueabihf/libc/usr/include/bits/byteswap.h: In function ‘__bswap_32’: /opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/arm-linux-gnueabihf/libc/usr/include/bits/byteswap.h:45:1: sorry, unimplemented: Thumb-1 hard-float VFP ABI

/opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/arm-linux-gnueabihf/libc/usr/include/bits/byteswap.h: In function ‘__bswap_32’: /opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/arm-linux-gnueabihf/libc/usr/include/bits/byteswap.h:45:1: sorry, unimplemented: Thumb-1 hard-float VFP ABI

此外，我设置了-mfloat-abi = soft，这使链接程序抱怨:

Additionally, i set -mfloat-abi=soft which make the Linker complain:

/opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/bin/../lib/gcc/arm-linux-gnueabihf/4.8.3/../../../../arm-linux-gnueabihf/bin/ld: error: prog uses VFP register arguments, prog.o does not /opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/bin/../lib/gcc/arm-linux-gnueabihf/4.8.3/../../../../arm-linux-gnueabihf/bin/ld: failed to merge target specific data of file prog.o

/opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/bin/../lib/gcc/arm-linux-gnueabihf/4.8.3/../../../../arm-linux-gnueabihf/bin/ld: error: prog uses VFP register arguments, prog.o does not /opt/gcc-linaro-arm-linux-gnueabihf-4.8-2014.03_linux/bin/../lib/gcc/arm-linux-gnueabihf/4.8.3/../../../../arm-linux-gnueabihf/bin/ld: failed to merge target specific data of file prog.o

那么，我实际上如何为ARMv6进行构建? 还是工具链中存在错误?

Well, how do i actually build for ARMv6? Or is there a bug in the Toolchain?

推荐答案

默认情况下，您的工具链配置为发出Thumb代码.这样做的问题在于，浮点指令的唯一Thumb编码是32位Thumb-2编码，就所有目的和目的而言，它们仅存在于ARMv7或更高版本的 ^* 上，并且肯定不是在普通的ARMv6上可用.

Your toolchain is configured to emit Thumb code by default. The issue with that is that the only Thumb encodings of floating-point instructions are 32-bit Thumb-2 ones, which to all intents and purposes only exist on ARMv7 or later^*, and certainly aren't available on plain ARMv6.

要针对v7之前的体系结构，最好不要使用-marm进行编译以生成ARM代码，而不要使用相对有限的16位Thumb指令集.

To target pre-v7 architectures, you're best off compiling with -marm to generate ARM code instead of the relatively limited set of 16-bit Thumb instructions.

_{*我怀疑大多数人永远不会有意识地遇到ARM1156.}

这篇关于使用gnueabihf为ARMv6构建的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！