STMCubeIDE: compiler doesn't create code for FPU on STM32G431

[Hansel]

I am using STM32CubeIDE version 1.14.1 on a WIN10 machine.

The following C code

 

asm("VMUL.F32 s0, s0, s0");
volatile float xyz = 3.1415;
 xyz = xyz * 2.3;

 

disassembles to the following:

 

 79 asm("VMUL.F32 s0, s0, s0");
08000df0: vmul.f32 s0, s0, s0
 80 volatile float xyz = 3.1415;
08000df4: ldr r3, [pc, #344] ; (0x8000f50 <main+360>)
08000df6: str r3, [sp, #0]
 81 xyz = xyz * 2.3;
08000df8: ldr r0, [sp, #0]
08000dfa: bl 0x8000598 <__extendsfdf2>
08000dfe: add r3, pc, #328 ; (adr r3, 0x8000f48 <main+352>)
08000e00: ldrd r2, r3, [r3]
08000e04: bl 0x8000648 <__muldf3>
08000e08: bl 0x8000bf8 <__truncdfsf2>

 

The compiler flags are set:

and I see them also used when compiling.

The compiler used is GNU11 (ISO C11 + gnu extensions)(-std=gnu11)

When walking through the code, the debugger stops at the breakpoint for SCB->CPACK:

 

void SystemInit(void)
{
 /* FPU settings ------------------------------------------------------------*/
 #if (__FPU_PRESENT == 1) && (__FPU_USED == 1)
 SCB->CPACR |= ((3UL << (10*2))|(3UL << (11*2))); /* set CP10 and CP11 Full Access */
 #endif

 

Clearly something's wrong but I am out of ideas where to search. Can someone give me a hint what's going on?

[KnarfB]

constant 2.3 is of type double. So the multiplication is for double type operands and the assignment truncates to float.Try 2.3f.

hth

KnarfB

[Hansel]

D'oh! What impact a single character can have. This did the trick. Thank you very much, Frank!

[KnarfB]

Yeah, that's a nuisance, especially when porting existing code. There are some compiler flags for it too, see https://stackoverflow.com/questions/32266864/make-c-floating-point-literals-float-rather-than-double, but they are compiler dependent and don't cure the root cause.

KnarfB