simd_sse.h

/* -----------------------------------------------------------------------------

    Copyright (c) 2006 Simon Brown                          si@sjbrown.co.uk

    Permission is hereby granted, free of charge, to any person obtaining
    a copy of this software and associated documentation files (the
    "Software"), to deal in the Software without restriction, including
    without limitation the rights to use, copy, modify, merge, publish,
    distribute, sublicense, and/or sell copies of the Software, and to
    permit persons to whom the Software is furnished to do so, subject to
    the following conditions:

    The above copyright notice and this permission notice shall be included
    in all copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
    OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
    MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
    IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
    CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
    TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
    SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

   -------------------------------------------------------------------------- */

#ifndef SQUISH_SIMD_SSE_H
#define SQUISH_SIMD_SSE_H

#include <xmmintrin.h>
#if ( SQUISH_USE_SSE > 1 )
#include <emmintrin.h>
#endif

#define SQUISH_SSE_SPLAT( a )                                        \
    ( ( a ) | ( ( a ) << 2 ) | ( ( a ) << 4 ) | ( ( a ) << 6 ) )

#define SQUISH_SSE_SHUF( x, y, z, w )                                \
    ( ( x ) | ( ( y ) << 2 ) | ( ( z ) << 4 ) | ( ( w ) << 6 ) )

namespace squish {

#define VEC4_CONST( X ) Vec4( X )

class Vec4
{
public:
    typedef Vec4 const& Arg;

    Vec4() {}

    explicit Vec4( __m128 v ) : m_v( v ) {}

    Vec4( Vec4 const& arg ) : m_v( arg.m_v ) {}

    Vec4& operator=( Vec4 const& arg )
    {
        m_v = arg.m_v;
        return *this;
    }

    explicit Vec4( float s ) : m_v( _mm_set1_ps( s ) ) {}

    Vec4( float x, float y, float z, float w ) : m_v( _mm_setr_ps( x, y, z, w ) ) {}

    Vec3 GetVec3() const
    {
#ifdef __GNUC__
        __attribute__ ((__aligned__ (16))) float c[4];
#else
        __declspec(align(16)) float c[4];
#endif
        _mm_store_ps( c, m_v );
        return Vec3( c[0], c[1], c[2] );
    }

    Vec4 SplatX() const { return Vec4( _mm_shuffle_ps( m_v, m_v, SQUISH_SSE_SPLAT( 0 ) ) ); }
    Vec4 SplatY() const { return Vec4( _mm_shuffle_ps( m_v, m_v, SQUISH_SSE_SPLAT( 1 ) ) ); }
    Vec4 SplatZ() const { return Vec4( _mm_shuffle_ps( m_v, m_v, SQUISH_SSE_SPLAT( 2 ) ) ); }
    Vec4 SplatW() const { return Vec4( _mm_shuffle_ps( m_v, m_v, SQUISH_SSE_SPLAT( 3 ) ) ); }

    Vec4& operator+=( Arg v )
    {
        m_v = _mm_add_ps( m_v, v.m_v );
        return *this;
    }

    Vec4& operator-=( Arg v )
    {
        m_v = _mm_sub_ps( m_v, v.m_v );
        return *this;
    }

    Vec4& operator*=( Arg v )
    {
        m_v = _mm_mul_ps( m_v, v.m_v );
        return *this;
    }

    friend Vec4 operator+( Vec4::Arg left, Vec4::Arg right  )
    {
        return Vec4( _mm_add_ps( left.m_v, right.m_v ) );
    }

    friend Vec4 operator-( Vec4::Arg left, Vec4::Arg right  )
    {
        return Vec4( _mm_sub_ps( left.m_v, right.m_v ) );
    }

    friend Vec4 operator*( Vec4::Arg left, Vec4::Arg right  )
    {
        return Vec4( _mm_mul_ps( left.m_v, right.m_v ) );
    }

    //! Returns a*b + c
    friend Vec4 MultiplyAdd( Vec4::Arg a, Vec4::Arg b, Vec4::Arg c )
    {
        return Vec4( _mm_add_ps( _mm_mul_ps( a.m_v, b.m_v ), c.m_v ) );
    }

    //! Returns -( a*b - c )
    friend Vec4 NegativeMultiplySubtract( Vec4::Arg a, Vec4::Arg b, Vec4::Arg c )
    {
        return Vec4( _mm_sub_ps( c.m_v, _mm_mul_ps( a.m_v, b.m_v ) ) );
    }

    friend Vec4 Reciprocal( Vec4::Arg v )
    {
        // get the reciprocal estimate
        __m128 estimate = _mm_rcp_ps( v.m_v );

        // one round of Newton-Rhaphson refinement
        __m128 diff = _mm_sub_ps( _mm_set1_ps( 1.0f ), _mm_mul_ps( estimate, v.m_v ) );
        return Vec4( _mm_add_ps( _mm_mul_ps( diff, estimate ), estimate ) );
    }

    friend Vec4 Min( Vec4::Arg left, Vec4::Arg right )
    {
        return Vec4( _mm_min_ps( left.m_v, right.m_v ) );
    }

    friend Vec4 Max( Vec4::Arg left, Vec4::Arg right )
    {
        return Vec4( _mm_max_ps( left.m_v, right.m_v ) );
    }

    friend Vec4 Truncate( Vec4::Arg v )
    {
#if ( SQUISH_USE_SSE == 1 )
        // convert to ints
        __m128 input = v.m_v;
        __m64 lo = _mm_cvttps_pi32( input );
        __m64 hi = _mm_cvttps_pi32( _mm_movehl_ps( input, input ) );

        // convert to floats
        __m128 part = _mm_movelh_ps( input, _mm_cvtpi32_ps( input, hi ) );
        __m128 truncated = _mm_cvtpi32_ps( part, lo );

        // clear out the MMX multimedia state to allow FP calls later
        _mm_empty();
        return Vec4( truncated );
#else
        // use SSE2 instructions
        return Vec4( _mm_cvtepi32_ps( _mm_cvttps_epi32( v.m_v ) ) );
#endif
    }

    friend bool CompareAnyLessThan( Vec4::Arg left, Vec4::Arg right )
    {
        __m128 bits = _mm_cmplt_ps( left.m_v, right.m_v );
        int value = _mm_movemask_ps( bits );
        return value != 0;
    }

private:
    __m128 m_v;
};

} // namespace squish

#endif // ndef SQUISH_SIMD_SSE_H