target-arm: Introduce float64_256 and float64_512 constants.
authorChristophe Lyon <christophe.lyon@st.com>
Mon, 21 Feb 2011 16:38:46 +0000 (17:38 +0100)
committerAurelien Jarno <aurelien@aurel32.net>
Thu, 24 Feb 2011 07:53:36 +0000 (08:53 +0100)
These two constants will be used by helper functions such as recpe_f32
and rsqrte_f32.

Signed-off-by: Christophe Lyon <christophe.lyon@st.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
target-arm/helper.c

index 7f63a280fc80dfd32b7586c2c91ae5ceb85e2b8f..30c18090d25be8fa1ac60669518ab593922c83fc 100644 (file)
@@ -2687,6 +2687,11 @@ float32 HELPER(rsqrts_f32)(float32 a, float32 b, CPUState *env)
 
 /* NEON helpers.  */
 
+/* Constants 256 and 512 are used in some helpers; we avoid relying on
+ * int->float conversions at run-time.  */
+#define float64_256 make_float64(0x4070000000000000LL)
+#define float64_512 make_float64(0x4080000000000000LL)
+
 /* TODO: The architecture specifies the value that the estimate functions
    should return.  We return the exact reciprocal/root instead.  */
 float32 HELPER(recpe_f32)(float32 a, CPUState *env)