How are floating point single and double values stored in big endian mode

[Ben Dooks (bjdooks-ct)]

With the MBE/SBE/UBE bits added, these change the integer endian. Do these also do the same to the floating point numbers or vector? I couldn't see anything in the documents. Currently gcc assumes big endian and qemu is still in little.

[Radim Krčmář (radimkrcmar)]

The spec doesn't seem to make a distinction for integers. Should be big endian for all data.

[Andrew Waterman (aswaterman)]

Radim Krčmář (@radimkrcmar)’s answer is correct; the definition in the spec is deliberately generic and so applies to FP memory accesses, too.

[Ben Dooks (bjdooks-ct)]

Thank you Andrew Waterman (@aswaterman) and Radim Krčmář (@radimkrcmar). For some reason I had gotten into my memory that it was integer only.

I'll update my QEMu patch set and hopefully I can get my big endian project mostly sorted this week. Info is at https://gitlab.com/CodethinkLabs/riscv_bigendian