https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128836

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |ubizjak at gmail dot com

--- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
Actually, just a case of missing "&& reload_completed".

--cut here--
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 23a9bc7fa59..461974f7151 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -13896,8 +13896,9 @@ (define_split
        (and:SWI48 (match_dup 0)
                   (const_int -65536)))
    (clobber (reg:CC FLAGS_REG))]
-  "(TARGET_FAST_PREFIX && !TARGET_PARTIAL_REG_STALL)
-    || optimize_function_for_size_p (cfun)"
+  "((TARGET_FAST_PREFIX && !TARGET_PARTIAL_REG_STALL)
+    || optimize_function_for_size_p (cfun))
+   && reload_completed"
   [(set (strict_low_part (match_dup 1)) (const_int 0))]
   "operands[1] = gen_lowpart (HImode, operands[0]);")

--cut here--

However, the compiler generates quite unoptimal code here:

.L5:
        movd    %xmm0, %eax
        xorw    %ax, %ax
        movd    %eax, %xmm0
        ret

clang is better:

        movss   .LCPI0_1(%rip), %xmm1           # xmm1 =
[NaN,0.0E+0,0.0E+0,0.0E+0]
        andps   %xmm1, %xmm0
        retq

Reply via email to