[Bug target/29776] result of ffs/clz/ctz/popcount/parity are already sign-extended

["steven at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29776

Steven Bosscher <steven at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2011-12-24 00:00:00         |2012-07-29 0:00
                 CC|                            |uros at gcc dot gnu.org

--- Comment #3 from Steven Bosscher <steven at gcc dot gnu.org> 2012-07-29 13:23:30 UTC ---
Tested with trunk r189929 on x86-64 (generic).

#define MODE signed // or unsigned
unsigned mask[8];

unsigned foo(unsigned y, unsigned char x)
{
  MODE int c = (MODE) __builtin_ctz(x);
  unsigned int m = mask[c];
  return y & m;
}


With MODE==signed:
        movzbl  %sil, %esi
        movl    %edi, %eax
        rep; bsfl       %esi, %esi
        movslq  %esi, %rsi
        andl    mask(,%rsi,4), %eax
        ret

With MODE==unsigned:
        movzbl  %sil, %esi
        movl    %edi, %eax
        rep; bsfl       %esi, %esi
        movl    %esi, %esi
        andl    mask(,%rsi,4), %eax
        ret

The "movl %esi, %esi" is a zeroextendsidi:
#(insn:TI 9 2 10 2 (parallel [
#            (set (reg:SI 4 si [orig:60 D.1716 ] [60])
#                (ctz:SI (reg:SI 4 si [orig:68 D.1715 ] [68])))
#            (clobber (reg:CC 17 flags))
#        ]) t.c:6 666 {ctzsi2}
#     (expr_list:REG_UNUSED (reg:CC 17 flags)
#        (nil)))
        rep; bsfl       %esi, %esi      # 9     ctzsi2  [length = 4]
#(insn:TI 10 9 12 2 (set (reg:DI 4 si [orig:70 D.1716 ] [70])
#        (zero_extend:DI (reg:SI 4 si [orig:60 D.1716 ] [60]))) t.c:7 113
{*zero_extendsidi2_rex64}
#     (nil))
        movl    %esi, %esi      # 10    *zero_extendsidi2_rex64/1       [length
= 2]

Perhaps REE can be taught about ctz giving a non-negative result.