check correctness of byte order, bitfield order

[ThomasWaldmann]

see TODO in tests/opcodes.py.

[ThomasWaldmann]

guess this could be done by assembling some very small sample program with binutils-esp32ulp assembler and looking at the resulting machine code.

[ThomasWaldmann]

BTW, NATIVE byte order was a bad idea. It would work when running on esp32, but we want to do development and run tests on the UNIX port, so we should replace that by the correct, specific byte order esp32 ulp processor expects.

[ThomasWaldmann]

byte order fixed (hopefully): 9982807

TODO: bitfield order

When using C structs, is the first contained bitfield at MSB position and the last mentioned one closer to LSB (or ending at LSB, if all 32bits are used)?

[ThomasWaldmann]

#include <stdio.h>

typedef unsigned int uint32;

union {
    struct {
        uint32 lsb:1;
        uint32 middle:30;
        uint32 msb:1;
    } bits;
    uint32 word;
} U;

int main(){
    printf("%u\n", U.word);
    U.bits.msb = 0;
    U.bits.lsb = 1;
    printf("%u\n", U.word);
    U.bits.msb = 1;
    U.bits.lsb = 0;
    printf("%u\n", U.word);
};

Output:

$ ./test
0
1
2147483648

So, at least a gcc on linux amd64 fills in the bitfields into an uint from lsb to msb.

[ThomasWaldmann]

According to that, opcodes.make_ins_struct_def should be correct as it also starts from lsb.