[PATCH ARM iWMMXt 2/5] intrinsic head file change

[Ramana Radhakrishnan]

I've only had a brief look at this and point out certain stylistic
issues that I noticed and would like another set of eyes on this and
the next patch.


On 29 May 2012 05:13, Matt Turner <mattst88@gmail.com> wrote:
> From: Xinyu Qi <xyqi@marvell.com>
>
>        gcc/
>        * config/arm/mmintrin.h: Use __IWMMXT__ to enable iWMMXt intrinsics.
>        Use __IWMMXT2__ to enable iWMMXt2 intrinsics.
>        Use C name-mangling for intrinsics.
>        (__v8qi): Redefine.
>        (_mm_cvtsi32_si64, _mm_andnot_si64, _mm_sad_pu8): Revise.
>        (_mm_sad_pu16, _mm_align_si64, _mm_setwcx, _mm_getwcx): Likewise.
>        (_m_from_int): Likewise.
>        (_mm_sada_pu8, _mm_sada_pu16): New intrinsic.
>        (_mm_alignr0_si64, _mm_alignr1_si64, _mm_alignr2_si64): Likewise.
>        (_mm_alignr3_si64, _mm_tandcb, _mm_tandch, _mm_tandcw): Likewise.
>        (_mm_textrcb, _mm_textrch, _mm_textrcw, _mm_torcb): Likewise.
>        (_mm_torch, _mm_torcw, _mm_tbcst_pi8, _mm_tbcst_pi16): Likewise.
>        (_mm_tbcst_pi32): Likewise.
>        (_mm_abs_pi8, _mm_abs_pi16, _mm_abs_pi32): New iWMMXt2 intrinsic.
>        (_mm_addsubhx_pi16, _mm_absdiff_pu8, _mm_absdiff_pu16): Likewise.
>        (_mm_absdiff_pu32, _mm_addc_pu16, _mm_addc_pu32): Likewise.
>        (_mm_avg4_pu8, _mm_avg4r_pu8, _mm_maddx_pi16, _mm_maddx_pu16): Likewise.
>        (_mm_msub_pi16, _mm_msub_pu16, _mm_mulhi_pi32): Likewise.
>        (_mm_mulhi_pu32, _mm_mulhir_pi16, _mm_mulhir_pi32): Likewise.
>        (_mm_mulhir_pu16, _mm_mulhir_pu32, _mm_mullo_pi32): Likewise.
>        (_mm_qmulm_pi16, _mm_qmulm_pi32, _mm_qmulmr_pi16): Likewise.
>        (_mm_qmulmr_pi32, _mm_subaddhx_pi16, _mm_addbhusl_pu8): Likewise.
>        (_mm_addbhusm_pu8, _mm_qmiabb_pi32, _mm_qmiabbn_pi32): Likewise.
>        (_mm_qmiabt_pi32, _mm_qmiabtn_pi32, _mm_qmiatb_pi32): Likewise.
>        (_mm_qmiatbn_pi32, _mm_qmiatt_pi32, _mm_qmiattn_pi32): Likewise.
>        (_mm_wmiabb_si64, _mm_wmiabbn_si64, _mm_wmiabt_si64): Likewise.
>        (_mm_wmiabtn_si64, _mm_wmiatb_si64, _mm_wmiatbn_si64): Likewise.
>        (_mm_wmiatt_si64, _mm_wmiattn_si64, _mm_wmiawbb_si64): Likewise.
>        (_mm_wmiawbbn_si64, _mm_wmiawbt_si64, _mm_wmiawbtn_si64): Likewise.
>        (_mm_wmiawtb_si64, _mm_wmiawtbn_si64, _mm_wmiawtt_si64): Likewise.
>        (_mm_wmiawttn_si64, _mm_merge_si64): Likewise.
>        (_mm_torvscb, _mm_torvsch, _mm_torvscw): Likewise.
>        (_m_to_int): New define.
> ---
>  gcc/config/arm/mmintrin.h |  649 ++++++++++++++++++++++++++++++++++++++++++---
>  1 files changed, 614 insertions(+), 35 deletions(-)
>
> diff --git a/gcc/config/arm/mmintrin.h b/gcc/config/arm/mmintrin.h
> index 2cc500d..0fe551d 100644
> --- a/gcc/config/arm/mmintrin.h
> +++ b/gcc/config/arm/mmintrin.h
> @@ -24,16 +24,30 @@
>  #ifndef _MMINTRIN_H_INCLUDED
>  #define _MMINTRIN_H_INCLUDED
>
> +#ifndef __IWMMXT__
> +#error You must enable WMMX/WMMX2 instructions (e.g. -march=iwmmxt or -march=iwmmxt2) to use iWMMXt/iWMMXt2 intrinsics
> +#else
> +
> +#ifndef __IWMMXT2__
> +#warning You only enable iWMMXt intrinsics. Extended iWMMXt2 intrinsics available only if WMMX2 instructions enabled (e.g. -march=iwmmxt2)
> +#endif
> +

Extra newline.

> +
> +#if defined __cplusplus
> +extern "C" { /* Begin "C" */
> +/* Intrinsics use C name-mangling.  */
> +#endif /* __cplusplus */
> +
>  /* The data type intended for user use.  */
>  typedef unsigned long long __m64, __int64;
>
>  /* Internal data types for implementing the intrinsics.  */
>  typedef int __v2si __attribute__ ((vector_size (8)));
>  typedef short __v4hi __attribute__ ((vector_size (8)));
> -typedef char __v8qi __attribute__ ((vector_size (8)));
> +typedef signed char __v8qi __attribute__ ((vector_size (8)));
>
>  /* "Convert" __m64 and __int64 into each other.  */
> -static __inline __m64
> +static __inline __m64
>  _mm_cvtsi64_m64 (__int64 __i)
>  {
>   return __i;
> @@ -54,7 +68,7 @@ _mm_cvtsi64_si32 (__int64 __i)
>  static __inline __int64
>  _mm_cvtsi32_si64 (int __i)
>  {
> -  return __i;
> +  return (__i & 0xffffffff);
>  }
>
>  /* Pack the four 16-bit values from M1 into the lower four 8-bit values of
> @@ -603,7 +617,7 @@ _mm_and_si64 (__m64 __m1, __m64 __m2)
>  static __inline __m64
>  _mm_andnot_si64 (__m64 __m1, __m64 __m2)
>  {
> -  return __builtin_arm_wandn (__m1, __m2);
> +  return __builtin_arm_wandn (__m2, __m1);
>  }
>
>  /* Bit-wise inclusive OR the 64-bit values in M1 and M2.  */
> @@ -935,7 +949,13 @@ _mm_avg2_pu16 (__m64 __A, __m64 __B)
>  static __inline __m64
>  _mm_sad_pu8 (__m64 __A, __m64 __B)
>  {
> -  return (__m64) __builtin_arm_wsadb ((__v8qi)__A, (__v8qi)__B);
> +  return (__m64) __builtin_arm_wsadbz ((__v8qi)__A, (__v8qi)__B);
> +}
> +
> +static __inline __m64
> +_mm_sada_pu8 (__m64 __A, __m64 __B, __m64 __C)
> +{
> +  return (__m64) __builtin_arm_wsadb ((__v2si)__A, (__v8qi)__B, (__v8qi)__C);
>  }
>
>  /* Compute the sum of the absolute differences of the unsigned 16-bit
> @@ -944,9 +964,16 @@ _mm_sad_pu8 (__m64 __A, __m64 __B)
>  static __inline __m64
>  _mm_sad_pu16 (__m64 __A, __m64 __B)
>  {
> -  return (__m64) __builtin_arm_wsadh ((__v4hi)__A, (__v4hi)__B);
> +  return (__m64) __builtin_arm_wsadhz ((__v4hi)__A, (__v4hi)__B);
>  }
>
> +static __inline __m64
> +_mm_sada_pu16 (__m64 __A, __m64 __B, __m64 __C)
> +{
> +  return (__m64) __builtin_arm_wsadh ((__v2si)__A, (__v4hi)__B, (__v4hi)__C);
> +}
> +
> +
>  /* Compute the sum of the absolute differences of the unsigned 8-bit
>    values in A and B.  Return the value in the lower 16-bit word; the
>    upper words are cleared.  */
> @@ -965,11 +992,8 @@ _mm_sadz_pu16 (__m64 __A, __m64 __B)
>   return (__m64) __builtin_arm_wsadhz ((__v4hi)__A, (__v4hi)__B);
>  }
>
> -static __inline __m64
> -_mm_align_si64 (__m64 __A, __m64 __B, int __C)
> -{
> -  return (__m64) __builtin_arm_walign ((__v8qi)__A, (__v8qi)__B, __C);
> -}
> +#define _mm_align_si64(__A,__B, N) \
> +  (__m64) __builtin_arm_walign ((__v8qi) (__A),(__v8qi) (__B), (N))
>
>  /* Creates a 64-bit zero.  */
>  static __inline __m64
> @@ -987,42 +1011,76 @@ _mm_setwcx (const int __value, const int __regno)
>  {
>   switch (__regno)
>     {
> -    case 0:  __builtin_arm_setwcx (__value, 0); break;
> -    case 1:  __builtin_arm_setwcx (__value, 1); break;
> -    case 2:  __builtin_arm_setwcx (__value, 2); break;
> -    case 3:  __builtin_arm_setwcx (__value, 3); break;
> -    case 8:  __builtin_arm_setwcx (__value, 8); break;
> -    case 9:  __builtin_arm_setwcx (__value, 9); break;
> -    case 10: __builtin_arm_setwcx (__value, 10); break;
> -    case 11: __builtin_arm_setwcx (__value, 11); break;
> -    default: break;
> +    case 0:
> +      __asm __volatile ("tmcr wcid, %0" :: "r"(__value));
> +      break;
> +    case 1:
> +      __asm __volatile ("tmcr wcon, %0" :: "r"(__value));
> +      break;
> +    case 2:
> +      __asm __volatile ("tmcr wcssf, %0" :: "r"(__value));
> +      break;
> +    case 3:
> +      __asm __volatile ("tmcr wcasf, %0" :: "r"(__value));
> +      break;
> +    case 8:
> +      __builtin_arm_setwcgr0 (__value);
> +      break;
> +    case 9:
> +      __builtin_arm_setwcgr1 (__value);
> +      break;
> +    case 10:
> +      __builtin_arm_setwcgr2 (__value);
> +      break;
> +    case 11:
> +      __builtin_arm_setwcgr3 (__value);
> +      break;
> +    default:
> +      break;
>     }
>  }
>
>  static __inline int
>  _mm_getwcx (const int __regno)
>  {
> +  int __value;
>   switch (__regno)
>     {
> -    case 0:  return __builtin_arm_getwcx (0);
> -    case 1:  return __builtin_arm_getwcx (1);
> -    case 2:  return __builtin_arm_getwcx (2);
> -    case 3:  return __builtin_arm_getwcx (3);
> -    case 8:  return __builtin_arm_getwcx (8);
> -    case 9:  return __builtin_arm_getwcx (9);
> -    case 10: return __builtin_arm_getwcx (10);
> -    case 11: return __builtin_arm_getwcx (11);
> -    default: return 0;
> +    case 0:
> +      __asm __volatile ("tmrc %0, wcid" : "=r"(__value));
> +      break;
> +    case 1:
> +      __asm __volatile ("tmrc %0, wcon" : "=r"(__value));
> +      break;
> +    case 2:
> +      __asm __volatile ("tmrc %0, wcssf" : "=r"(__value));
> +      break;
> +    case 3:
> +      __asm __volatile ("tmrc %0, wcasf" : "=r"(__value));
> +      break;
> +    case 8:
> +      return __builtin_arm_getwcgr0 ();
> +    case 9:
> +      return __builtin_arm_getwcgr1 ();
> +    case 10:
> +      return __builtin_arm_getwcgr2 ();
> +    case 11:
> +      return __builtin_arm_getwcgr3 ();
> +    default:
> +      break;
>     }
> +  return __value;
>  }
>
>  /* Creates a vector of two 32-bit values; I0 is least significant.  */
>  static __inline __m64
>  _mm_set_pi32 (int __i1, int __i0)
>  {
> -  union {
> +  union
> +  {
>     __m64 __q;
> -    struct {
> +    struct
> +    {
>       unsigned int __i0;
>       unsigned int __i1;
>     } __s;
> @@ -1041,7 +1099,7 @@ _mm_set_pi16 (short __w3, short __w2, short __w1, short __w0)
>   unsigned int __i1 = (unsigned short)__w3 << 16 | (unsigned short)__w2;
>   unsigned int __i0 = (unsigned short)__w1 << 16 | (unsigned short)__w0;
>   return _mm_set_pi32 (__i1, __i0);
> -
> +
Extra newline again here.
>  }
>
>  /* Creates a vector of eight 8-bit values; B0 is least significant.  */
> @@ -1108,11 +1166,526 @@ _mm_set1_pi8 (char __b)
>   return _mm_set1_pi32 (__i);
>  }
>
> -/* Convert an integer to a __m64 object.  */
> +#ifdef __IWMMXT2__
> +static __inline __m64
> +_mm_abs_pi8 (__m64 m1)
> +{
> +  return (__m64) __builtin_arm_wabsb ((__v8qi)m1);
> +}
> +
> +static __inline __m64
> +_mm_abs_pi16 (__m64 m1)
> +{
> +  return (__m64) __builtin_arm_wabsh ((__v4hi)m1);
> +

And here.

> +}
> +
> +static __inline __m64
> +_mm_abs_pi32 (__m64 m1)
> +{
> +  return (__m64) __builtin_arm_wabsw ((__v2si)m1);
> +
and here.

<large part snipped.>

> +
> +#define _mm_qmiabb_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiabb ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiabbn_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiabbn ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiabt_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiabt ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiabtn_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc=acc;\
> +   __m64 _m1=m1;\
> +   __m64 _m2=m2;\
> +   _acc = (__m64) __builtin_arm_wqmiabtn ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiatb_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiatb ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiatbn_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiatbn ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiatt_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiatt ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_qmiattn_pi32(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wqmiattn ((__v2si)_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiabb_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiabb (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiabbn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiabbn (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiabt_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiabt (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiabtn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiabtn (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiatb_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiatb (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiatbn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiatbn (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiatt_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiatt (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiattn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiattn (_acc, (__v4hi)_m1, (__v4hi)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawbb_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawbb (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawbbn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawbbn (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawbt_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawbt (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawbtn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawbtn (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawtb_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawtb (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawtbn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawtbn (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawtt_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawtt (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })
> +
> +#define _mm_wmiawttn_si64(acc, m1, m2) \
> +  ({\
> +   __m64 _acc = acc;\
> +   __m64 _m1 = m1;\
> +   __m64 _m2 = m2;\
> +   _acc = (__m64) __builtin_arm_wmiawttn (_acc, (__v2si)_m1, (__v2si)_m2);\
> +   _acc;\
> +   })

I assume someone knows why these are macros and not inline functions
like the others ?


> +
> +/* The third arguments should be an immediate.  */

s/arguments/argument

> +#define _mm_merge_si64(a, b, n) \
> +  ({\
> +   __m64 result;\
> +   result = (__m64) __builtin_arm_wmerge ((__m64) (a), (__m64) (b), (n));\
> +   result;\
> +   })
> +#endif  /* __IWMMXT2__ */
> +