This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH][ARM] Fix broken shift patterns

On 06/10/11 16:01, Andrew Stubbs wrote:
  (define_special_predicate "shift_operator"
    (and (ior (ior (and (match_code "mult")
  		      (match_test "power_of_two_operand (XEXP (op, 1), mode)"))
  		 (and (match_code "rotate")
  		      (match_test "GET_CODE (XEXP (op, 1)) == CONST_INT
  				&&  ((unsigned HOST_WIDE_INT) INTVAL (XEXP (op, 1)))<  32")))
-	    (match_code "ashift,ashiftrt,lshiftrt,rotatert"))
+	    (and (match_code "ashift,ashiftrt,lshiftrt,rotatert")
+		 (ior (match_test "GET_CODE (XEXP (op, 1)) == CONST_INT
+				&&  ((unsigned HOST_WIDE_INT) INTVAL (XEXP (op, 1)))<  32")
+		      (match_test "REG_P (XEXP (op, 1))"))))
         (match_test "mode == GET_MODE (op)")))

Oh, I forgot to say, I don't understand why the "rotate" operator is special cased?

If I understand it correctly, the effect of the (existing) rotate is both to check the constant range, AND to disallow registers as the shift amount. This difference has no effect on Thumb, but might cause ARM mode some troubles?

Is this likely to be deliberate, or an oversight? I can't see any reason in the ARM ARM why this should be the case.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]