arm: Remove coercion from scalar argument to vmin & vmax intrinsics

This patch fixes an issue with vmin* and vmax* intrinsics which accept
a scalar argument. Previously when the scalar was of different width
to the vector elements this would generate __ARM_undef. This change
allows the scalar argument to be implicitly converted to the correct
width. Also tidied up the relevant unit tests, some of which would
have passed even if only one of two or three intrinsic calls had
compiled correctly.

Bootstrapped and tested on arm-none-eabi, gcc and CMSIS_DSP
testsuites are clean. OK for trunk?

Thanks,
Joe

gcc/ChangeLog:

2020-08-10  Joe Ramsay  <joe.ramsay@arm.com>

	* config/arm/arm_mve.h (__arm_vmaxnmavq): Remove coercion of scalar
	argument.
	(__arm_vmaxnmvq): Likewise.
	(__arm_vminnmavq): Likewise.
	(__arm_vminnmvq): Likewise.
	(__arm_vmaxnmavq_p): Likewise.
	(__arm_vmaxnmvq_p): Likewise (and delete duplicate definition).
	(__arm_vminnmavq_p): Likewise.
	(__arm_vminnmvq_p): Likewise.
	(__arm_vmaxavq): Likewise.
	(__arm_vmaxavq_p): Likewise.
	(__arm_vmaxvq): Likewise.
	(__arm_vmaxvq_p): Likewise.
	(__arm_vminavq): Likewise.
	(__arm_vminavq_p): Likewise.
	(__arm_vminvq): Likewise.
	(__arm_vminvq_p): Likewise.

gcc/testsuite/ChangeLog:

	* gcc.target/arm/mve/intrinsics/vmaxavq_p_s16.c: Add test for mismatched
	width of scalar argument.
	* gcc.target/arm/mve/intrinsics/vmaxavq_p_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxavq_p_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxavq_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxavq_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxavq_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmavq_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmavq_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmvq_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmvq_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_u16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_u32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_p_u8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_u16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_u32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vmaxvq_u8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_p_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_p_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_p_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminavq_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmavq_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmavq_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmavq_p_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmavq_p_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmvq_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmvq_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmvq_p_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminnmvq_p_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_u16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_u32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_p_u8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_s8.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_u16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_u32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vminvq_u8.c: Likewise.

53 files changed, 500 insertions, 140 deletions

diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h
index 99cff41..26c83c7 100644
--- a/gcc/config/arm/arm_mve.h
+++ b/gcc/config/arm/arm_mve.h
@@ -41682,16 +41682,16 @@ extern void *__ARM_undef;
 #define __arm_vmaxavq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxavq_s8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, int8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxavq_s16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxavq_s32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, int32x4_t)));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxavq_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxavq_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxavq_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t)));})
 
 #define __arm_vmaxavq_p(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxavq_p_s8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxavq_p_s16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxavq_p_s32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, int32x4_t), p2));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxavq_p_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxavq_p_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxavq_p_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t), p2));})
 
 #define __arm_vmaxq_x(p1,p2,p3) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
@@ -41706,36 +41706,36 @@ extern void *__ARM_undef;
 #define __arm_vmaxvq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxvq_s8 (__ARM_mve_coerce(__p0, int8_t), __ARM_mve_coerce(__p1, int8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxvq_s16 (__ARM_mve_coerce(__p0, int16_t), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxvq_s32 (__ARM_mve_coerce(__p0, int32_t), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vmaxvq_u8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, uint8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vmaxvq_u16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, uint16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vmaxvq_u32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, uint32x4_t)));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxvq_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxvq_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxvq_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vmaxvq_u8 (__p0, __ARM_mve_coerce(__p1, uint8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vmaxvq_u16 (__p0, __ARM_mve_coerce(__p1, uint16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vmaxvq_u32 (__p0,__ARM_mve_coerce(__p1, uint32x4_t)));})
 
 #define __arm_vmaxvq_p(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxvq_p_s8 (__ARM_mve_coerce(__p0, int8_t), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxvq_p_s16 (__ARM_mve_coerce(__p0, int16_t), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxvq_p_s32 (__ARM_mve_coerce(__p0, int32_t), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vmaxvq_p_u8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, uint8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vmaxvq_p_u16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vmaxvq_p_u32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vmaxvq_p_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vmaxvq_p_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vmaxvq_p_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vmaxvq_p_u8 (__p0, __ARM_mve_coerce(__p1, uint8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vmaxvq_p_u16 (__p0, __ARM_mve_coerce(__p1, uint16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vmaxvq_p_u32 (__p0, __ARM_mve_coerce(__p1, uint32x4_t), p2));})
 
 #define __arm_vminavq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminavq_s8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, int8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminavq_s16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminavq_s32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, int32x4_t)));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminavq_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminavq_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminavq_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t)));})
 
 #define __arm_vminavq_p(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminavq_p_s8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminavq_p_s16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminavq_p_s32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, int32x4_t), p2));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminavq_p_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminavq_p_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminavq_p_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t), p2));})
 
 #define __arm_vminq_x(p1,p2,p3) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
@@ -41750,22 +41750,22 @@ extern void *__ARM_undef;
 #define __arm_vminvq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminvq_s8 (__ARM_mve_coerce(__p0, int8_t), __ARM_mve_coerce(__p1, int8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminvq_s16 (__ARM_mve_coerce(__p0, int16_t), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminvq_s32 (__ARM_mve_coerce(__p0, int32_t), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vminvq_u8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, uint8x16_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vminvq_u16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, uint16x8_t)), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vminvq_u32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, uint32x4_t)));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminvq_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminvq_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminvq_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vminvq_u8 (__p0, __ARM_mve_coerce(__p1, uint8x16_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vminvq_u16 (__p0, __ARM_mve_coerce(__p1, uint16x8_t)), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vminvq_u32 (__p0, __ARM_mve_coerce(__p1, uint32x4_t)));})
 
 #define __arm_vminvq_p(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminvq_p_s8 (__ARM_mve_coerce(__p0, int8_t), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminvq_p_s16 (__ARM_mve_coerce(__p0, int16_t), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminvq_p_s32 (__ARM_mve_coerce(__p0, int32_t), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vminvq_p_u8 (__ARM_mve_coerce(__p0, uint8_t), __ARM_mve_coerce(__p1, uint8x16_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vminvq_p_u16 (__ARM_mve_coerce(__p0, uint16_t), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vminvq_p_u32 (__ARM_mve_coerce(__p0, uint32_t), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int8x16_t]: __arm_vminvq_p_s8 (__p0, __ARM_mve_coerce(__p1, int8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int16x8_t]: __arm_vminvq_p_s16 (__p0, __ARM_mve_coerce(__p1, int16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_int32x4_t]: __arm_vminvq_p_s32 (__p0, __ARM_mve_coerce(__p1, int32x4_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint8x16_t]: __arm_vminvq_p_u8 (__p0, __ARM_mve_coerce(__p1, uint8x16_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint16x8_t]: __arm_vminvq_p_u16 (__p0, __ARM_mve_coerce(__p1, uint16x8_t), p2), \
+  int (*)[__ARM_mve_type_int_n][__ARM_mve_type_uint32x4_t]: __arm_vminvq_p_u32 (__p0, __ARM_mve_coerce(__p1, uint32x4_t), p2));})
 
 #define __arm_vmladavaq(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p1) __p1 = (p1); \
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s16.c
index 02e0227..74ffad4 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, int16x8_t b, mve_pred16_t p)
   return vmaxavq_p_s16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, int16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, int16x8_t b, mve_pred16_t p)
   return vmaxavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s16"  }  } */
+
+int16_t
+foo2 (uint8_t a, int16x8_t b, mve_pred16_t p)
+{
+  return vmaxavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxavt.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s32.c
index 7ecd94a..40800b0 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, int32x4_t b, mve_pred16_t p)
   return vmaxavq_p_s32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, int32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, int32x4_t b, mve_pred16_t p)
   return vmaxavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s32"  }  } */
+
+int32_t
+foo2 (uint16_t a, int32x4_t b, mve_pred16_t p)
+{
+  return vmaxavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxavt.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s8.c
index 7a21de7..7638737 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_p_s8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, int8x16_t b, mve_pred16_t p)
   return vmaxavq_p_s8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, int8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, int8x16_t b, mve_pred16_t p)
   return vmaxavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxavt.s8"  }  } */
+
+int8_t
+foo2 (uint32_t a, int8x16_t b, mve_pred16_t p)
+{
+  return vmaxavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxavt.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s16.c
index 4621eba..0dca149 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, int16x8_t b)
   return vmaxavq_s16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, int16x8_t b)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, int16x8_t b)
   return vmaxavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s16"  }  } */
+
+int16_t
+foo2 (uint8_t a, int16x8_t b)
+{
+  return vmaxavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxav.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s32.c
index 8813d9d..f419a77 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, int32x4_t b)
   return vmaxavq_s32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, int32x4_t b)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, int32x4_t b)
   return vmaxavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s32"  }  } */
+
+int32_t
+foo2 (uint16_t a, int32x4_t b)
+{
+  return vmaxavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxav.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s8.c
index 961f1d2..214ad88 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxavq_s8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, int8x16_t b)
   return vmaxavq_s8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, int8x16_t b)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, int8x16_t b)
   return vmaxavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxav.s8"  }  } */
+
+int8_t
+foo2 (uint32_t a, int8x16_t b)
+{
+  return vmaxavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxav.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f16.c
index de48ea8..6d8cf19 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b)
   return vmaxnmavq_f16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmav.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b)
   return vmaxnmavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmav.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b)
+{
+  return vmaxnmavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmav.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f32.c
index b4c7f83..ef79030 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b)
   return vmaxnmavq_f32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmav.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b)
   return vmaxnmavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmav.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b)
+{
+  return vmaxnmavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmav.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f16.c
index 9c2eed0..f7f39f5 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b, mve_pred16_t p)
   return vmaxnmavq_p_f16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmavt.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
   return vmaxnmavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmavt.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b, mve_pred16_t p)
+{
+  return vmaxnmavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmavt.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f32.c
index 1cadccb..341f625 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmavq_p_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b, mve_pred16_t p)
   return vmaxnmavq_p_f32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmavt.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
   return vmaxnmavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmavt.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b, mve_pred16_t p)
+{
+  return vmaxnmavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmavt.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f16.c
index 81f4b9b..80bd1d4 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b)
   return vmaxnmvq_f16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmv.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b)
   return vmaxnmvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmv.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b)
+{
+  return vmaxnmvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmv.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f32.c
index ab06c2b..bb2fc46 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b)
   return vmaxnmvq_f32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmv.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b)
   return vmaxnmvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxnmv.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b)
+{
+  return vmaxnmvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmv.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f16.c
index e37c5a1..3efe203 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b, mve_pred16_t p)
   return vmaxnmvq_p_f16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmvt.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
   return vmaxnmvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmvt.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b, mve_pred16_t p)
+{
+  return vmaxnmvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmvt.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f32.c
index 884cd45..6c13247 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxnmvq_p_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b, mve_pred16_t p)
   return vmaxnmvq_p_f32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmvt.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
   return vmaxnmvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxnmvt.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b, mve_pred16_t p)
+{
+  return vmaxnmvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxnmvt.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s16.c
index 79de370..657efc5 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s16.c
@@ -10,7 +10,6 @@ foo (int16_t a, int16x8_t b, mve_pred16_t p)
   return vmaxvq_p_s16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s16"  }  } */
 
 int16_t
 foo1 (int16_t a, int16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int16_t a, int16x8_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s16"  }  } */
+
+int16_t
+foo2 (int8_t a, int16x8_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s32.c
index e526744..5882351 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s32.c
@@ -10,7 +10,6 @@ foo (int32_t a, int32x4_t b, mve_pred16_t p)
   return vmaxvq_p_s32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s32"  }  } */
 
 int32_t
 foo1 (int32_t a, int32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int32_t a, int32x4_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s32"  }  } */
+
+int32_t
+foo2 (int16_t a, int32x4_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s8.c
index d3cedd4..3737ecd 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_s8.c
@@ -10,7 +10,6 @@ foo (int8_t a, int8x16_t b, mve_pred16_t p)
   return vmaxvq_p_s8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s8"  }  } */
 
 int8_t
 foo1 (int8_t a, int8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int8_t a, int8x16_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.s8"  }  } */
+
+int8_t
+foo2 (int32_t a, int8x16_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u16.c
index 79572f7..348cf39 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, uint16x8_t b, mve_pred16_t p)
   return vmaxvq_p_u16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, uint16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, uint16x8_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u16"  }  } */
+
+uint16_t
+foo2 (uint32_t a, uint16x8_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.u16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u32.c
index e2f7a6f..f2e9762 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, uint32x4_t b, mve_pred16_t p)
   return vmaxvq_p_u32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, uint32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, uint32x4_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u32"  }  } */
+
+uint32_t
+foo2 (uint8_t a, uint32x4_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.u32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u8.c
index f977806..7df5b63 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_p_u8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, uint8x16_t b, mve_pred16_t p)
   return vmaxvq_p_u8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, uint8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, uint8x16_t b, mve_pred16_t p)
   return vmaxvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vmaxvt.u8"  }  } */
+
+uint8_t
+foo2 (uint16_t a, uint8x16_t b, mve_pred16_t p)
+{
+  return vmaxvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxvt.u8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s16.c
index 90f10b5..8412452 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s16.c
@@ -10,7 +10,6 @@ foo (int16_t a, int16x8_t b)
   return vmaxvq_s16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s16"  }  } */
 
 int16_t
 foo1 (int16_t a, int16x8_t b)
@@ -18,4 +17,12 @@ foo1 (int16_t a, int16x8_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s16"  }  } */
+
+int16_t
+foo2 (int8_t a, int16x8_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s32.c
index aa0e88b..09f4909 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s32.c
@@ -10,7 +10,6 @@ foo (int32_t a, int32x4_t b)
   return vmaxvq_s32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s32"  }  } */
 
 int32_t
 foo1 (int32_t a, int32x4_t b)
@@ -18,4 +17,12 @@ foo1 (int32_t a, int32x4_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s32"  }  } */
+
+int32_t
+foo2 (int16_t a, int32x4_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s8.c
index 884b84d..a087bbc 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_s8.c
@@ -10,7 +10,6 @@ foo (int8_t a, int8x16_t b)
   return vmaxvq_s8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s8"  }  } */
 
 int8_t
 foo1 (int8_t a, int8x16_t b)
@@ -18,4 +17,12 @@ foo1 (int8_t a, int8x16_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.s8"  }  } */
+
+int8_t
+foo2 (int32_t a, int8x16_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u16.c
index 2813ebd..47fe0d1 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, uint16x8_t b)
   return vmaxvq_u16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, uint16x8_t b)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, uint16x8_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u16"  }  } */
+
+uint16_t
+foo2 (uint32_t a, uint16x8_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.u16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u32.c
index ab51b1e..aa723da 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, uint32x4_t b)
   return vmaxvq_u32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, uint32x4_t b)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, uint32x4_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u32"  }  } */
+
+uint32_t
+foo2 (uint8_t a, uint32x4_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.u32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u8.c
index 3326cfb..3aae785 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vmaxvq_u8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, uint8x16_t b)
   return vmaxvq_u8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, uint8x16_t b)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, uint8x16_t b)
   return vmaxvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vmaxv.u8"  }  } */
+
+uint8_t
+foo2 (uint16_t a, uint8x16_t b)
+{
+  return vmaxvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vmaxv.u8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s16.c
index 6b87648..9303ae0 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, int16x8_t b, mve_pred16_t p)
   return vminavq_p_s16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, int16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, int16x8_t b, mve_pred16_t p)
   return vminavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s16"  }  } */
+
+int16_t
+foo2 (uint8_t a, int16x8_t b, mve_pred16_t p)
+{
+  return vminavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminavt.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s32.c
index 086ff56..36247f6 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, int32x4_t b, mve_pred16_t p)
   return vminavq_p_s32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, int32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, int32x4_t b, mve_pred16_t p)
   return vminavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s32"  }  } */
+
+int32_t
+foo2 (uint16_t a, int32x4_t b, mve_pred16_t p)
+{
+  return vminavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminavt.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s8.c
index 999c11c..d336161 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_p_s8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, int8x16_t b, mve_pred16_t p)
   return vminavq_p_s8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, int8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, int8x16_t b, mve_pred16_t p)
   return vminavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminavt.s8"  }  } */
+
+int8_t
+foo2 (uint32_t a, int8x16_t b, mve_pred16_t p)
+{
+  return vminavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminavt.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s16.c
index a626e31..17e4edc 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, int16x8_t b)
   return vminavq_s16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, int16x8_t b)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, int16x8_t b)
   return vminavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s16"  }  } */
+
+int16_t
+foo2 (uint8_t a, int16x8_t b)
+{
+  return vminavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminav.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s32.c
index be575cb..032d02b 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, int32x4_t b)
   return vminavq_s32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, int32x4_t b)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, int32x4_t b)
   return vminavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s32"  }  } */
+
+int32_t
+foo2 (uint16_t a, int32x4_t b)
+{
+  return vminavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminav.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s8.c
index c3dfe4b..2a2bb3d6 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminavq_s8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, int8x16_t b)
   return vminavq_s8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, int8x16_t b)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, int8x16_t b)
   return vminavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminav.s8"  }  } */
+
+int8_t
+foo2 (uint32_t a, int8x16_t b)
+{
+  return vminavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminav.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f16.c
index 2111681..fadb23e 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b)
   return vminnmavq_f16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmav.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b)
   return vminnmavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmav.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b)
+{
+  return vminnmavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmav.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f32.c
index bd87b85..84714a9 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b)
   return vminnmavq_f32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmav.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b)
   return vminnmavq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmav.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b)
+{
+  return vminnmavq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmav.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f16.c
index e6d0bb5..c79fa30 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b, mve_pred16_t p)
   return vminnmavq_p_f16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmavt.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
   return vminnmavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmavt.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b, mve_pred16_t p)
+{
+  return vminnmavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmavt.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f32.c
index 6b56b67..bea04c7 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmavq_p_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b, mve_pred16_t p)
   return vminnmavq_p_f32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmavt.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
   return vminnmavq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmavt.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b, mve_pred16_t p)
+{
+  return vminnmavq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmavt.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f16.c
index 4d4caae..0eb3a4a 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b)
   return vminnmvq_f16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmv.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b)
   return vminnmvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmv.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b)
+{
+  return vminnmvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmv.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f32.c
index dab04d9..f318350 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b)
   return vminnmvq_f32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmv.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b)
   return vminnmvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminnmv.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b)
+{
+  return vminnmvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmv.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f16.c
index f5eafb1..16f6ac5 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f16.c
@@ -10,7 +10,6 @@ foo (float16_t a, float16x8_t b, mve_pred16_t p)
   return vminnmvq_p_f16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmvt.f16"  }  } */
 
 float16_t
 foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float16_t a, float16x8_t b, mve_pred16_t p)
   return vminnmvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmvt.f16"  }  } */
+
+float16_t
+foo2 (float32_t a, float16x8_t b, mve_pred16_t p)
+{
+  return vminnmvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmvt.f16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f32.c
index 5ac20bf..a8e4f9f 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminnmvq_p_f32.c
@@ -10,7 +10,6 @@ foo (float32_t a, float32x4_t b, mve_pred16_t p)
   return vminnmvq_p_f32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmvt.f32"  }  } */
 
 float32_t
 foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (float32_t a, float32x4_t b, mve_pred16_t p)
   return vminnmvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminnmvt.f32"  }  } */
+
+float32_t
+foo2 (float16_t a, float32x4_t b, mve_pred16_t p)
+{
+  return vminnmvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminnmvt.f32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s16.c
index c2edb62..91bb63f 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s16.c
@@ -10,7 +10,6 @@ foo (int16_t a, int16x8_t b, mve_pred16_t p)
   return vminvq_p_s16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s16"  }  } */
 
 int16_t
 foo1 (int16_t a, int16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int16_t a, int16x8_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s16"  }  } */
+
+int16_t
+foo2 (int8_t a, int16x8_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s32.c
index ba89217..a846701 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s32.c
@@ -10,7 +10,6 @@ foo (int32_t a, int32x4_t b, mve_pred16_t p)
   return vminvq_p_s32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s32"  }  } */
 
 int32_t
 foo1 (int32_t a, int32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int32_t a, int32x4_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s32"  }  } */
+
+int32_t
+foo2 (int16_t a, int32x4_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s8.c
index 1665c53..716d414 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_s8.c
@@ -10,7 +10,6 @@ foo (int8_t a, int8x16_t b, mve_pred16_t p)
   return vminvq_p_s8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s8"  }  } */
 
 int8_t
 foo1 (int8_t a, int8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (int8_t a, int8x16_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.s8"  }  } */
+
+int8_t
+foo2 (int32_t a, int8x16_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u16.c
index 5bade0a..cc7f8fe 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, uint16x8_t b, mve_pred16_t p)
   return vminvq_p_u16 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, uint16x8_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, uint16x8_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u16"  }  } */
+
+uint16_t
+foo2 (uint32_t a, uint16x8_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.u16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u32.c
index c4c5748..6bde0be 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, uint32x4_t b, mve_pred16_t p)
   return vminvq_p_u32 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, uint32x4_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint32_t a, uint32x4_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u32"  }  } */
+
+uint32_t
+foo2 (uint8_t a, uint32x4_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.u32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u8.c
index dc890dc..bb89490 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_p_u8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, uint8x16_t b, mve_pred16_t p)
   return vminvq_p_u8 (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, uint8x16_t b, mve_pred16_t p)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, uint8x16_t b, mve_pred16_t p)
   return vminvq_p (a, b, p);
 }
 
-/* { dg-final { scan-assembler "vminvt.u8"  }  } */
+
+uint8_t
+foo2 (uint16_t a, uint8x16_t b, mve_pred16_t p)
+{
+  return vminvq_p (a, b, p);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminvt.u8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s16.c
index f6eed63..6d589aa 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s16.c
@@ -10,7 +10,6 @@ foo (int16_t a, int16x8_t b)
   return vminvq_s16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s16"  }  } */
 
 int16_t
 foo1 (int16_t a, int16x8_t b)
@@ -18,4 +17,11 @@ foo1 (int16_t a, int16x8_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s16"  }  } */
+int16_t
+foo2 (int8_t a, int16x8_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.s16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s32.c
index 4077c32..7c727d6 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s32.c
@@ -10,7 +10,6 @@ foo (int32_t a, int32x4_t b)
   return vminvq_s32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s32"  }  } */
 
 int32_t
 foo1 (int32_t a, int32x4_t b)
@@ -18,4 +17,11 @@ foo1 (int32_t a, int32x4_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s32"  }  } */
+int32_t
+foo2 (int8_t a, int32x4_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.s32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s8.c
index bdf15f4..7630948 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_s8.c
@@ -10,7 +10,6 @@ foo (int8_t a, int8x16_t b)
   return vminvq_s8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s8"  }  } */
 
 int8_t
 foo1 (int8_t a, int8x16_t b)
@@ -18,4 +17,11 @@ foo1 (int8_t a, int8x16_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.s8"  }  } */
+int8_t
+foo2 (int32_t a, int8x16_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.s8" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u16.c
index 5c0935c..698975f 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u16.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u16.c
@@ -10,7 +10,6 @@ foo (uint16_t a, uint16x8_t b)
   return vminvq_u16 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u16"  }  } */
 
 uint16_t
 foo1 (uint16_t a, uint16x8_t b)
@@ -18,4 +17,12 @@ foo1 (uint16_t a, uint16x8_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u16"  }  } */
+
+uint8_t
+foo2 (uint32_t a, uint16x8_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.u16" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u32.c
index 1580c87..7489f81 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u32.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u32.c
@@ -10,7 +10,6 @@ foo (uint32_t a, uint32x4_t b)
   return vminvq_u32 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u32"  }  } */
 
 uint32_t
 foo1 (uint32_t a, uint32x4_t b)
@@ -18,4 +17,11 @@ foo1 (uint32_t a, uint32x4_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u32"  }  } */
+uint32_t
+foo2 (uint16_t a, uint32x4_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.u32" 3 } } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u8.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u8.c
index 95919b4..aa2b986 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u8.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vminvq_u8.c
@@ -10,7 +10,6 @@ foo (uint8_t a, uint8x16_t b)
   return vminvq_u8 (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u8"  }  } */
 
 uint8_t
 foo1 (uint8_t a, uint8x16_t b)
@@ -18,4 +17,12 @@ foo1 (uint8_t a, uint8x16_t b)
   return vminvq (a, b);
 }
 
-/* { dg-final { scan-assembler "vminv.u8"  }  } */
+
+uint16_t
+foo2 (uint32_t a, uint8x16_t b)
+{
+  return vminvq (a, b);
+}
+
+/* { dg-final { scan-assembler-not "__ARM_undef" } } */
+/* { dg-final { scan-assembler-times "vminv.u8" 3 } } */