4.9.2. Vector Floating-Point support in the ARM Compiler

VFP is fully supported by the ARM compiler. However, you can configure some builds to assume no VFP support by default, in which case floating-point calculations use library code.

To enable VFP support, use the following option:

To specify support for a specific VFP implementation on an ARM Cortex-A series processor, use the following options:

You can use these options for code that only runs on these VFP implementations, and does not require backward compatibility with older VFP implementations. Use --fpu=list to see the full list of FPUs supported.

Use the following options for linkage support:

Hardware floating-point linkage uses the FPU registers to pass the arguments and return values.

Software floating-point linkage means that the parameters and return value for a function are passed using the ARM integer registers R0 to R3 and the stack. --apcs=/hardfp and --apcs=/softfp interact with or override the explicit or implicit use of --fpu.

To achieve better application performance, recompile source files for a target with an FPU and hardware floating-point linkage. If it is impossible to recompile existing code, use options, such as --fpu=softvfp+vfpv4, to maintain compatibility with existing software.

Copyright © 2014 ARM. All rights reserved.ARM DAI0425