Hi! On Sun, Oct 22, 2017 at 09:13:11PM +0100, Richard Sandiford wrote: > rtx_equal_for_field_assignment_p had: > > x = adjust_address_nv (x, GET_MODE (y), > -subreg_lowpart_offset (GET_MODE (x), > GET_MODE (y))); > > But subreg_lowpart_offset returns an unsigned int and > adjust_address_nv takes a HWI, so a subreg offset of 4 would > give a memory offset of 0x00000000fffffffffc. > > The SVE series makes this go away by using HWI-based types for > both interfaces, but in this case the fix is also a minor clean-up. > > Tested on aarch64-linux-gnu, powerpc64le-linux-gnu and > x86_64-linux-gnu. OK to install?
Yes please, thanks! Segher > 2017-10-22 Richard Sandiford <richard.sandif...@linaro.org> > > gcc/ > * combine.c (rtx_equal_for_field_assignment_p): Use > byte_lowpart_offset. > > Index: gcc/combine.c > =================================================================== > --- gcc/combine.c 2017-10-22 21:04:50.138830154 +0100 > +++ gcc/combine.c 2017-10-22 21:04:59.000825360 +0100 > @@ -9526,13 +9526,9 @@ rtx_equal_for_field_assignment_p (rtx x, > return 0; > if (BYTES_BIG_ENDIAN != WORDS_BIG_ENDIAN) > return 0; > - /* For big endian, adjust the memory offset. */ > - if (BYTES_BIG_ENDIAN) > - x = adjust_address_nv (x, GET_MODE (y), > - -subreg_lowpart_offset (GET_MODE (x), > - GET_MODE (y))); > - else > - x = adjust_address_nv (x, GET_MODE (y), 0); > + x = adjust_address_nv (x, GET_MODE (y), > + byte_lowpart_offset (GET_MODE (y), > + GET_MODE (x))); > } > > if (x == y || rtx_equal_p (x, y))
