RE: [PATCH 08/23] arm: improve tests for vcmlaq*

[Kyrylo Tkachov via Gcc-patches]

> -----Original Message-----
> From: Andrea Corallo <andrea.cora...@arm.com>
> Sent: Friday, January 20, 2023 4:40 PM
> To: gcc-patches@gcc.gnu.org
> Cc: Kyrylo Tkachov <kyrylo.tkac...@arm.com>; Richard Earnshaw
> <richard.earns...@arm.com>; Andrea Corallo <andrea.cora...@arm.com>
> Subject: [PATCH 08/23] arm: improve tests for vcmlaq*
> 
> gcc/testsuite/ChangeLog:
> 
>       * gcc.target/arm/mve/intrinsics/vcmlaq_f16.c: Improve test.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_m_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_m_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f32.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f16.c: Likewise.
>       * gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f32.c: Likewise.

Ok.
Thanks,
Kyrill

> ---
>  .../arm/mve/intrinsics/vcmlaq_f16.c           | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_f32.c           | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_m_f16.c         | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_m_f32.c         | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot180_f16.c    | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot180_f32.c    | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot180_m_f16.c  | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot180_m_f32.c  | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot270_f16.c    | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot270_f32.c    | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot270_m_f16.c  | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot270_m_f32.c  | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot90_f16.c     | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot90_f32.c     | 24 +++++++++++--
>  .../arm/mve/intrinsics/vcmlaq_rot90_m_f16.c   | 34 ++++++++++++++++---
>  .../arm/mve/intrinsics/vcmlaq_rot90_m_f32.c   | 34 ++++++++++++++++---
>  16 files changed, 416 insertions(+), 48 deletions(-)
> 
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f16.c
> index fa7d0c05e8c..bb8a99790a0 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f16.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_f16 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f32.c
> index 166bf421f14..71ec4b8479c 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_f32.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_f32 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f16.c
> index 0929f5a0a89..3db345d0791 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f16.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_m_f16 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f32.c
> index 1f4ba453cbc..dcbd2dccce5 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_m_f32.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_m_f32 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #0(?:        @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f16.c
> index fc6ba30b36a..f76ae2383a2 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f16.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot180_f16 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot180 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f32.c
> index dbe3f26b3b9..c97d0d0d852 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_f32.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot180_f32 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot180 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f16.c
> index 84a3bd81ad9..132cdf9954f 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f16.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot180_m_f16 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot180_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f32.c
> index 61f5716f0b7..99e96ebe3a9 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot180_m_f32.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot180_m_f32 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #180(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot180_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f16.c
> index 1b0bef91f37..fae85105feb 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f16.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot270_f16 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot270 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f32.c
> index 83e15025e44..54a9b662772 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_f32.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot270_f32 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot270 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f16.c
> index 6e033b1d7ab..e34f83165c3 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f16.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot270_m_f16 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot270_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f32.c
> index 4928341076d..cdba91b8e8c 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot270_m_f32.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot270_m_f32 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #270(?:      @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot270_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f16.c
> index 16744a4582c..f767b2b6e6b 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f16.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot90_f16 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f16       q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c)
>  {
>    return vcmlaq_rot90 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f32.c
> index f1f19a87ba7..6c9b24f271d 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_f32.c
> @@ -1,21 +1,41 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot90_f32 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vcmla.f32       q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c)
>  {
>    return vcmlaq_rot90 (a, b, c);
>  }
> 
> -/* { dg-final { scan-assembler "vcmla.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f16.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f16.c
> index 7133ddc7ad5..9141c9e6f90 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f16.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f16.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot90_m_f16 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f16      q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float16x8_t
>  foo1 (float16x8_t a, float16x8_t b, float16x8_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot90_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f16"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> diff --git
> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f32.c
> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f32.c
> index 6022e3be538..f317d411806 100644
> --- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f32.c
> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcmlaq_rot90_m_f32.c
> @@ -1,23 +1,49 @@
>  /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
>  /* { dg-add-options arm_v8_1m_mve_fp } */
>  /* { dg-additional-options "-O2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> 
>  #include "arm_mve.h"
> 
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/*
> +**foo:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot90_m_f32 (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> 
> +/*
> +**foo1:
> +**   ...
> +**   vmsr    p0, (?:ip|fp|r[0-9]+)(?:        @.*|)
> +**   ...
> +**   vpst(?: @.*|)
> +**   ...
> +**   vcmlat.f32      q[0-9]+, q[0-9]+, q[0-9]+, #90(?:       @.*|)
> +**   ...
> +*/
>  float32x4_t
>  foo1 (float32x4_t a, float32x4_t b, float32x4_t c, mve_pred16_t p)
>  {
>    return vcmlaq_rot90_m (a, b, c, p);
>  }
> 
> -/* { dg-final { scan-assembler "vpst" } } */
> -/* { dg-final { scan-assembler "vcmlat.f32"  }  } */
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +/* { dg-final { scan-assembler-not "__ARM_undef" } } */
> \ No newline at end of file
> --
> 2.25.1