This patch remove inline assembly and reimplement all mvn/mvnq vector
integer intrinsics through the standard "one_cmpl<mode>2" pattern was
introduced later after the initial implementation of those intrinsics.
that's why inline assembly was used historically.
OK for trunk?
no regression on the exist advsimd-intrinsics/vmvn.c.
2016-05-16 Jiong Wang<jiong.w...@arm.com>
gcc/
* config/aarch64/arm_neon.h (vmvn_s8): Reimplement using C operator.
Remove inline assembly.
(vmvn_s16): Likewise.
(vmvn_s32): Likewise.
(vmvn_u8): Likewise.
(vmvn_u16): Likewise.
(vmvn_u32): Likewise.
(vmvnq_s8): Likewise.
(vmvnq_s16): Likewise.
(vmvnq_s32): Likewise.
(vmvnq_u8): Likewise.
(vmvnq_u16): Likewise.
(vmvnq_u32): Likewise.
(vmvn_p8): Likewise.
(vmvnq_p16): Likewise.
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 432a1fad9a6df6fef844896df5e8ad29cc31f548..ae4c429a87822a8807f2d2ec054d3194b39ef6ac 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -8093,161 +8093,6 @@ vmull_u32 (uint32x2_t a, uint32x2_t b)
return result;
}
-__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
-vmvn_p8 (poly8x8_t a)
-{
- poly8x8_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
-vmvn_s8 (int8x8_t a)
-{
- int8x8_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
-vmvn_s16 (int16x4_t a)
-{
- int16x4_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
-vmvn_s32 (int32x2_t a)
-{
- int32x2_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
-vmvn_u8 (uint8x8_t a)
-{
- uint8x8_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
-vmvn_u16 (uint16x4_t a)
-{
- uint16x4_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
-vmvn_u32 (uint32x2_t a)
-{
- uint32x2_t result;
- __asm__ ("mvn %0.8b,%1.8b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__))
-vmvnq_p8 (poly8x16_t a)
-{
- poly8x16_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int8x16_t __attribute__ ((__always_inline__))
-vmvnq_s8 (int8x16_t a)
-{
- int8x16_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
-vmvnq_s16 (int16x8_t a)
-{
- int16x8_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
-vmvnq_s32 (int32x4_t a)
-{
- int32x4_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__))
-vmvnq_u8 (uint8x16_t a)
-{
- uint8x16_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__))
-vmvnq_u16 (uint16x8_t a)
-{
- uint16x8_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__))
-vmvnq_u32 (uint32x4_t a)
-{
- uint32x4_t result;
- __asm__ ("mvn %0.16b,%1.16b"
- : "=w"(result)
- : "w"(a)
- : /* No clobbers */);
- return result;
-}
-
-
__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
vpadal_s8 (int16x4_t a, int8x8_t b)
{
@@ -18622,6 +18467,92 @@ vmulq_n_u32 (uint32x4_t __a, uint32_t __b)
return __a * __b;
}
+/* vmvn */
+
+__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
+vmvn_p8 (poly8x8_t __a)
+{
+ return (poly8x8_t) ~((int8x8_t) __a);
+}
+
+__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
+vmvn_s8 (int8x8_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vmvn_s16 (int16x4_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vmvn_s32 (int32x2_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
+vmvn_u8 (uint8x8_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
+vmvn_u16 (uint16x4_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
+vmvn_u32 (uint32x2_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__))
+vmvnq_p8 (poly8x16_t __a)
+{
+ return (poly8x16_t) ~((int8x16_t) __a);
+}
+
+__extension__ static __inline int8x16_t __attribute__ ((__always_inline__))
+vmvnq_s8 (int8x16_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vmvnq_s16 (int16x8_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vmvnq_s32 (int32x4_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__))
+vmvnq_u8 (uint8x16_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__))
+vmvnq_u16 (uint16x8_t __a)
+{
+ return ~__a;
+}
+
+__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__))
+vmvnq_u32 (uint32x4_t __a)
+{
+ return ~__a;
+}
+
/* vneg */
__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
