This patch makes use of the subreg_offset_from_lsb function added by an earlier patch in the SVE series.
Tested on aarch64-linux-gnu and x86_64-linux-gnu, and by making sure that there were no differences in testsuite assembly output for one target per CPU. OK to install? Richard 2017-08-23 Richard Sandiford <richard.sandif...@linaro.org> Alan Hayward <alan.hayw...@arm.com> David Sherwood <david.sherw...@arm.com> gcc/ * combine.c (make_extraction): Use subreg_offset_from_lsb. Index: gcc/combine.c =================================================================== --- gcc/combine.c 2017-08-23 10:44:56.293484623 +0100 +++ gcc/combine.c 2017-08-23 10:45:34.493718192 +0100 @@ -7482,26 +7482,15 @@ make_extraction (machine_mode mode, rtx return a new hard register. */ if (pos || in_dest) { - HOST_WIDE_INT final_word = pos / BITS_PER_WORD; - - if (WORDS_BIG_ENDIAN - && GET_MODE_SIZE (inner_mode) > UNITS_PER_WORD) - final_word = ((GET_MODE_SIZE (inner_mode) - - GET_MODE_SIZE (tmode)) - / UNITS_PER_WORD) - final_word; - - final_word *= UNITS_PER_WORD; - if (BYTES_BIG_ENDIAN && - GET_MODE_SIZE (inner_mode) > GET_MODE_SIZE (tmode)) - final_word += (GET_MODE_SIZE (inner_mode) - - GET_MODE_SIZE (tmode)) % UNITS_PER_WORD; + unsigned int offset + = subreg_offset_from_lsb (tmode, inner_mode, pos); /* Avoid creating invalid subregs, for example when simplifying (x>>32)&255. */ - if (!validate_subreg (tmode, inner_mode, inner, final_word)) + if (!validate_subreg (tmode, inner_mode, inner, offset)) return NULL_RTX; - new_rtx = gen_rtx_SUBREG (tmode, inner, final_word); + new_rtx = gen_rtx_SUBREG (tmode, inner, offset); } else new_rtx = gen_lowpart (tmode, inner);