Problem with building "Error: illegal operands `addi sp,sp,-20*(1<<2)-20*FPREGBYTES'"

[Dranzey]

Hello everyone! Firstly i hope this is right place to post my problem. If it's not please redirect me to another section. I'm trying to build a Cmake based project on target: Sipeed Longan Nano, which is GD32VF103CBT6 MCU with RISC-V 32-bit core of GigaDevice. During configuration process i faced a problem with "entry.S" file, which was connected to the toolchain. I thinks it's some startup code. This are Assembler messages:

Error: illegal operands addi sp,sp,-20*(1<<2)-20FPREGBYTES' Error: unrecognized opcode fpstore f0,(20(1<<2)+0FPREGBYTES)(sp)' Error: unrecognized opcode fpstore f1,(20*(1<<2)+1FPREGBYTES)(sp)' Error: unrecognized opcode fpstore f2,(20(1<<2)+2FPREGBYTES)(sp)' Error: unrecognized opcode fpload f0,(20(1<<2)+0FPREGBYTES)(sp)' Error: unrecognized opcode fpload f1,(20(1<<2)+1FPREGBYTES)(sp)'

[GPSBabelDeveloper]

GD32V is an RV32IMAC core. It has integer the integer core (that's the I) and multiply and divide (that's the M) plus Atomics and Compressed. It doesn't have floating point. I would not expect fpload and fpstore to be valid opcodes.

I suspect that startup code that you don't understand doesn't understand this family of chips. I'm pretty sure that Longan Nano just doesn't have floating point. Maybe your configuration process needs a tweak to Not Do That, but if your mystery program really needs FP, you should brace yourself for floating point emulation libraries which can be terribly slow. So a lot depends on how badly your app really assumes the presence of FP.