[Bug target/103383] Microblaze bswaphi2 can cause issues with delay slots

[rkujoth at motorolasolutions dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103383

--- Comment #8 from Rich <rkujoth at motorolasolutions dot com> ---
(In reply to Michael Eager from comment #7)
> Do you have a test case which shows the problem?

It's not difficult to get it to show up if you're using the builtin swap16 and
conditionals. Here's a simple one (the optimizations make writing a simple test
case weird):

main.c:

#include <stdint.h>
#include <stdlib.h>

int main()
{
  uint16_t src1 = (uint16_t)rand();
  uint16_t src2 = (uint16_t)rand();
  if(__builtin_bswap16(src1) != rand())
  {
    return -1;
  }
  return 0;
}

Then using the Vivado 2017.1 toolchain...
mb-gcc -O2 -mcpu=v10.0 -mlittle-endian -xl-barrel-shirt main.c

Then I see the following sequence in a.out, which shouldn't happen:
brlid
swapb
swaph