What is FPU and why does it cause assembler error?

Assembler error: FPU

I built a project on STM32F429 board in PlatformIO, and I got this error:

1
2
3
4
5
6
 
arm-none-eabi-gcc-ar rc .pio/build/disco_f429zi/libFrameworkCMSISDevice.a .pio/build/disco_f429zi/FrameworkCMSISDevice/gcc/startup_stm32f429xx.o .pio/build/disco_f429zi/FrameworkCMSISDevice/system_stm32f4xx.o
arm-none-eabi-gcc-ranlib .pio/build/disco_f429zi/libFrameworkCMSISDevice.a
/var/folders/w7/s2tyt4v54313_kmyn4cy499r0000gn/T//ccYSACNG.s: Assembler messages:
/var/folders/w7/s2tyt4v54313_kmyn4cy499r0000gn/T//ccYSACNG.s:6641: Error: selected processor does not support vstmdbeq r0!,{s16-s31}' in Thumb mode /var/folders/w7/s2tyt4v54313_kmyn4cy499r0000gn/T//ccYSACNG.s:6643: Error: instruction not allowed in IT block -- stmdb r0!,{r4-r11,r14}'
/var/folders/w7/s2tyt4v54313_kmyn4cy499r0000gn/T//ccYSACNG.s:6663: Error: selected processor does not support vldmiaeq r0!,{s16-s31}' in Thumb mode /var/folders/w7/s2tyt4v54313_kmyn4cy499r0000gn/T//ccYSACNG.s:6665: Error: instruction not allowed in IT block -- msr psp,r0'
*** [.pio/build/disco_f429zi/src/freertos_includes.o] Error 1
 

The error message was quite incomprehensible. Turns out it’s the assembler error (vstmdbeq/vldmiaeq/msr psp in IT block). But why did it happen? This was Copilot’s answer:

• The assembler errors you saw (vstmdbeq/vldmiaeq/msr psp in IT block) happen when assembler doesn’t expect VFP/FP instructions or the wrong float-ABI is used. The STM32F429 has an FPU; the ARM_CM4F FreeRTOS port emits VFP instructions and context-switch code that uses VFP registers. The compile flags above tell GCC/assembler to target the FPU and use the hard float ABI, and the __FPU_PRESENT macro enables FPU-specific code paths in the port headers.

There are some new, unheard keywords popping out.

• FPU
• VFP
• ABI

Let’s go over each term.

1. Floating-Point Unit (FPU)

Floating-point unit (FPU) is an actual hardware component; it is specialized in performing mathematical calculations on floating-point numbers.

Floating-point calculations require a lot of resources. If the processor doesn’t have a FPU, all the floating point operations are done by software through the C compiler library (but the performances are very low). On a processor having an FPU, all of the operations are entirely done by hardware in a single cycle, for most of the instructions. The C compiler does not use its own floating-point library but directly generates FPU native instructions.¹

So does my board (STM32F429) support FPU? The answer is Yes.

The STM32F427xx and STM32F429xx devices are based on the high-performance Arm® Cortex®-M4 32-bit RISC core operating at a frequency of up to 180 MHz. The Cortex®-M4 core features a floating-point unit (FPU) single precision, which supports all Arm® single- precision data-processing instructions and data types.²

2. Vector Floating Point (VFP)

A Vector Floating-Point (VFP) instruction is a command for a floating-point coprocessor used in ARM architecture that performs arithmetic on floating-point numbers. VFP can be thought of as FPU or floating point instruction set. ³⁴

3. Application Binary Interface (ABI)

Application Binary Interface (ABI) refers to the specification that defines how the parameters and the results of the floating-point calculations are transferred across function boundaries. ⁵

To make it simple, there are options for ABI and they affect whether the floating-point unit is used. In my case, STM32F429 board has a FPU and that is why I can set ABI option as -mfloat-abi=hard.

Hard ABI: This allows the compiled code to generate codes that directly accesss the FPU and use FPU-specific calling conventions when calling runtime library functions.⁵

VFP insturction in port.c

What did it mean by this?:

The ARM_CM4F FreeRTOS port emits VFP instructions and context-switch code that uses VFP registers.

Let’s look into the actual code. port.c is located under $HOMEPATH/.platformio/packages/framework-stm32cubef4/Middlewares/Third_Party/FreeRTOS/Source/portable/GCC/ARM_CM4F/.

In line 372 of port.c,

1
2
 
/* Ensure the VFP is enabled - it should be anyway. */
vPortEnableVFP();
 

Also in line 431,

1
 
"	tst r14, #0x10						\n" /* Is the task using the FPU context?  If so, push high vfp registers. */
 

Conclusion

These need to be added to build_flags in platformio.ini:

• -D__FPU_PRESENT=1
• -mfloat-abi=hard
• -mfpu=fpv4-sp-d16

References