Make sure frexp() returns correct for argument 0.0

[gromacs.git] / src / gromacs / simd / impl_x86_sse2 / impl_x86_sse2_simd_float.h

/*
 * This file is part of the GROMACS molecular simulation package.
 *
 * Copyright (c) 2014,2015,2016,2017,2019,2020, by the GROMACS development team, led by
 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
 * and including many others, as listed in the AUTHORS file in the
 * top-level source directory and at http://www.gromacs.org.
 *
 * GROMACS is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public License
 * as published by the Free Software Foundation; either version 2.1
 * of the License, or (at your option) any later version.
 *
 * GROMACS is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with GROMACS; if not, see
 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
 *
 * If you want to redistribute modifications to GROMACS, please
 * consider that scientific software is very special. Version
 * control is crucial - bugs must be traceable. We will be happy to
 * consider code for inclusion in the official distribution, but
 * derived work must not be called official GROMACS. Details are found
 * in the README & COPYING files - if they are missing, get the
 * official version at http://www.gromacs.org.
 *
 * To help us fund GROMACS development, we humbly ask that you cite
 * the research papers on the package. Check out http://www.gromacs.org.
 */
#ifndef GMX_SIMD_IMPL_X86_SSE2_SIMD_FLOAT_H
#define GMX_SIMD_IMPL_X86_SSE2_SIMD_FLOAT_H

#include "config.h"

#include <cassert>
#include <cstddef>
#include <cstdint>

#include <emmintrin.h>

#include "gromacs/math/utilities.h"

namespace gmx
{

class SimdFloat
{
public:
    SimdFloat() {}

    SimdFloat(float f) : simdInternal_(_mm_set1_ps(f)) {}

    // Internal utility constructor to simplify return statements
    SimdFloat(__m128 simd) : simdInternal_(simd) {}

    __m128 simdInternal_;
};

class SimdFInt32
{
public:
    SimdFInt32() {}

    SimdFInt32(std::int32_t i) : simdInternal_(_mm_set1_epi32(i)) {}

    // Internal utility constructor to simplify return statements
    SimdFInt32(__m128i simd) : simdInternal_(simd) {}

    __m128i simdInternal_;
};

class SimdFBool
{
public:
    SimdFBool() {}

    SimdFBool(bool b) : simdInternal_(_mm_castsi128_ps(_mm_set1_epi32(b ? 0xFFFFFFFF : 0))) {}

    // Internal utility constructor to simplify return statements
    SimdFBool(__m128 simd) : simdInternal_(simd) {}

    __m128 simdInternal_;
};

class SimdFIBool
{
public:
    SimdFIBool() {}

    SimdFIBool(bool b) : simdInternal_(_mm_set1_epi32(b ? 0xFFFFFFFF : 0)) {}

    // Internal utility constructor to simplify return statements
    SimdFIBool(__m128i simd) : simdInternal_(simd) {}

    __m128i simdInternal_;
};

static inline SimdFloat gmx_simdcall simdLoad(const float* m, SimdFloatTag = {})
{
    assert(std::size_t(m) % 16 == 0);
    return { _mm_load_ps(m) };
}

static inline void gmx_simdcall store(float* m, SimdFloat a)
{
    assert(std::size_t(m) % 16 == 0);
    _mm_store_ps(m, a.simdInternal_);
}

static inline SimdFloat gmx_simdcall simdLoadU(const float* m, SimdFloatTag = {})
{
    return { _mm_loadu_ps(m) };
}

static inline void gmx_simdcall storeU(float* m, SimdFloat a)
{
    _mm_storeu_ps(m, a.simdInternal_);
}

static inline SimdFloat gmx_simdcall setZeroF()
{
    return { _mm_setzero_ps() };
}

static inline SimdFInt32 gmx_simdcall simdLoad(const std::int32_t* m, SimdFInt32Tag)
{
    assert(std::size_t(m) % 16 == 0);
    return { _mm_load_si128(reinterpret_cast<const __m128i*>(m)) };
}

static inline void gmx_simdcall store(std::int32_t* m, SimdFInt32 a)
{
    assert(std::size_t(m) % 16 == 0);
    _mm_store_si128(reinterpret_cast<__m128i*>(m), a.simdInternal_);
}

static inline SimdFInt32 gmx_simdcall simdLoadU(const std::int32_t* m, SimdFInt32Tag)
{
    return { _mm_loadu_si128(reinterpret_cast<const __m128i*>(m)) };
}

static inline void gmx_simdcall storeU(std::int32_t* m, SimdFInt32 a)
{
    _mm_storeu_si128(reinterpret_cast<__m128i*>(m), a.simdInternal_);
}

static inline SimdFInt32 gmx_simdcall setZeroFI()
{
    return { _mm_setzero_si128() };
}


// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
template<int index>
static inline std::int32_t gmx_simdcall extract(SimdFInt32 a)
{
    return _mm_cvtsi128_si32(_mm_srli_si128(a.simdInternal_, 4 * index));
}
#endif

static inline SimdFloat gmx_simdcall operator&(SimdFloat a, SimdFloat b)
{
    return { _mm_and_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall andNot(SimdFloat a, SimdFloat b)
{
    return { _mm_andnot_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall operator|(SimdFloat a, SimdFloat b)
{
    return { _mm_or_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall operator^(SimdFloat a, SimdFloat b)
{
    return { _mm_xor_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall operator+(SimdFloat a, SimdFloat b)
{
    return { _mm_add_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall operator-(SimdFloat a, SimdFloat b)
{
    return { _mm_sub_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall operator-(SimdFloat x)
{
    return { _mm_xor_ps(x.simdInternal_, _mm_set1_ps(GMX_FLOAT_NEGZERO)) };
}

static inline SimdFloat gmx_simdcall operator*(SimdFloat a, SimdFloat b)
{
    return { _mm_mul_ps(a.simdInternal_, b.simdInternal_) };
}

// Override for AVX-128-FMA and higher
#if GMX_SIMD_X86_SSE2 || GMX_SIMD_X86_SSE4_1
static inline SimdFloat gmx_simdcall fma(SimdFloat a, SimdFloat b, SimdFloat c)
{
    return { _mm_add_ps(_mm_mul_ps(a.simdInternal_, b.simdInternal_), c.simdInternal_) };
}

static inline SimdFloat gmx_simdcall fms(SimdFloat a, SimdFloat b, SimdFloat c)
{
    return { _mm_sub_ps(_mm_mul_ps(a.simdInternal_, b.simdInternal_), c.simdInternal_) };
}

static inline SimdFloat gmx_simdcall fnma(SimdFloat a, SimdFloat b, SimdFloat c)
{
    return { _mm_sub_ps(c.simdInternal_, _mm_mul_ps(a.simdInternal_, b.simdInternal_)) };
}

static inline SimdFloat gmx_simdcall fnms(SimdFloat a, SimdFloat b, SimdFloat c)
{
    return { _mm_sub_ps(_mm_setzero_ps(),
                        _mm_add_ps(_mm_mul_ps(a.simdInternal_, b.simdInternal_), c.simdInternal_)) };
}
#endif

static inline SimdFloat gmx_simdcall rsqrt(SimdFloat x)
{
    return { _mm_rsqrt_ps(x.simdInternal_) };
}

static inline SimdFloat gmx_simdcall rcp(SimdFloat x)
{
    return { _mm_rcp_ps(x.simdInternal_) };
}

static inline SimdFloat gmx_simdcall maskAdd(SimdFloat a, SimdFloat b, SimdFBool m)
{
    return { _mm_add_ps(a.simdInternal_, _mm_and_ps(b.simdInternal_, m.simdInternal_)) };
}

static inline SimdFloat gmx_simdcall maskzMul(SimdFloat a, SimdFloat b, SimdFBool m)
{
    return { _mm_and_ps(_mm_mul_ps(a.simdInternal_, b.simdInternal_), m.simdInternal_) };
}

static inline SimdFloat gmx_simdcall maskzFma(SimdFloat a, SimdFloat b, SimdFloat c, SimdFBool m)
{
    return { _mm_and_ps(_mm_add_ps(_mm_mul_ps(a.simdInternal_, b.simdInternal_), c.simdInternal_),
                        m.simdInternal_) };
}

// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
static inline SimdFloat gmx_simdcall maskzRsqrt(SimdFloat x, SimdFBool m)
{
#    ifndef NDEBUG
    x.simdInternal_ = _mm_or_ps(_mm_andnot_ps(m.simdInternal_, _mm_set1_ps(1.0F)),
                                _mm_and_ps(m.simdInternal_, x.simdInternal_));
#    endif
    return { _mm_and_ps(_mm_rsqrt_ps(x.simdInternal_), m.simdInternal_) };
}

static inline SimdFloat gmx_simdcall maskzRcp(SimdFloat x, SimdFBool m)
{
#    ifndef NDEBUG
    x.simdInternal_ = _mm_or_ps(_mm_andnot_ps(m.simdInternal_, _mm_set1_ps(1.0F)),
                                _mm_and_ps(m.simdInternal_, x.simdInternal_));
#    endif
    return { _mm_and_ps(_mm_rcp_ps(x.simdInternal_), m.simdInternal_) };
}
#endif

static inline SimdFloat gmx_simdcall abs(SimdFloat x)
{
    return { _mm_andnot_ps(_mm_set1_ps(GMX_FLOAT_NEGZERO), x.simdInternal_) };
}

static inline SimdFloat gmx_simdcall max(SimdFloat a, SimdFloat b)
{
    return { _mm_max_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFloat gmx_simdcall min(SimdFloat a, SimdFloat b)
{
    return { _mm_min_ps(a.simdInternal_, b.simdInternal_) };
}

// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
static inline SimdFloat gmx_simdcall round(SimdFloat x)
{
    return { _mm_cvtepi32_ps(_mm_cvtps_epi32(x.simdInternal_)) };
}

static inline SimdFloat gmx_simdcall trunc(SimdFloat x)
{
    return { _mm_cvtepi32_ps(_mm_cvttps_epi32(x.simdInternal_)) };
}

#endif

template<MathOptimization opt = MathOptimization::Safe>
static inline SimdFloat gmx_simdcall frexp(SimdFloat value, SimdFInt32* exponent)
{
    const __m128 exponentMask = _mm_castsi128_ps(_mm_set1_epi32(0x7F800000));
    const __m128 mantissaMask = _mm_castsi128_ps(_mm_set1_epi32(0x807FFFFF));
    const __m128i exponentBias = _mm_set1_epi32(126); // add 1 to make our definition identical to frexp()
    const __m128 half = _mm_set1_ps(0.5F);

    __m128i iExponent = _mm_castps_si128(_mm_and_ps(value.simdInternal_, exponentMask));
    iExponent         = _mm_sub_epi32(_mm_srli_epi32(iExponent, 23), exponentBias);

    // Combine mantissa and exponent for result
    __m128 result = _mm_or_ps(_mm_and_ps(value.simdInternal_, mantissaMask), half);

    if (opt == MathOptimization::Safe)
    {
        __m128 valueIsZero = _mm_cmpeq_ps(_mm_setzero_ps(), value.simdInternal_);
        // If value was non-zero, use the exponent we calculated, otherwise set return-value exponent to zero.
        iExponent = _mm_andnot_si128(_mm_castps_si128(valueIsZero), iExponent);
        // set the fraction result to zero for all elements where the input value was zero.
        result = _mm_or_ps(_mm_andnot_ps(valueIsZero, result),
                           _mm_and_ps(valueIsZero, value.simdInternal_));
    }

    exponent->simdInternal_ = iExponent;
    return { result };
}

// Override for SSE4.1
#if GMX_SIMD_X86_SSE2
template<MathOptimization opt = MathOptimization::Safe>
static inline SimdFloat gmx_simdcall ldexp(SimdFloat value, SimdFInt32 exponent)
{
    const __m128i exponentBias = _mm_set1_epi32(127);
    __m128i       iExponent;

    iExponent = _mm_add_epi32(exponent.simdInternal_, exponentBias);

    if (opt == MathOptimization::Safe)
    {
        // Make sure biased argument is not negative
        iExponent = _mm_and_si128(iExponent, _mm_cmpgt_epi32(iExponent, _mm_setzero_si128()));
    }

    iExponent = _mm_slli_epi32(iExponent, 23);

    return { _mm_mul_ps(value.simdInternal_, _mm_castsi128_ps(iExponent)) };
}
#endif

// Override for AVX-128-FMA and higher
#if GMX_SIMD_X86_SSE2 || GMX_SIMD_X86_SSE4_1
static inline float gmx_simdcall reduce(SimdFloat a)
{
    // Shuffle has latency 1/throughput 1, followed by add with latency 3, t-put 1.
    // This is likely faster than using _mm_hadd_ps, which has latency 5, t-put 2.
    a.simdInternal_ = _mm_add_ps(
            a.simdInternal_, _mm_shuffle_ps(a.simdInternal_, a.simdInternal_, _MM_SHUFFLE(1, 0, 3, 2)));
    a.simdInternal_ = _mm_add_ss(
            a.simdInternal_, _mm_shuffle_ps(a.simdInternal_, a.simdInternal_, _MM_SHUFFLE(0, 3, 2, 1)));
    return *reinterpret_cast<float*>(&a);
}
#endif

static inline SimdFBool gmx_simdcall operator==(SimdFloat a, SimdFloat b)
{
    return { _mm_cmpeq_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFBool gmx_simdcall operator!=(SimdFloat a, SimdFloat b)
{
    return { _mm_cmpneq_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFBool gmx_simdcall operator<(SimdFloat a, SimdFloat b)
{
    return { _mm_cmplt_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFBool gmx_simdcall operator<=(SimdFloat a, SimdFloat b)
{
    return { _mm_cmple_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFBool gmx_simdcall testBits(SimdFloat a)
{
    __m128i ia = _mm_castps_si128(a.simdInternal_);
    __m128i res = _mm_andnot_si128(_mm_cmpeq_epi32(ia, _mm_setzero_si128()), _mm_cmpeq_epi32(ia, ia));

    return { _mm_castsi128_ps(res) };
}

static inline SimdFBool gmx_simdcall operator&&(SimdFBool a, SimdFBool b)
{
    return { _mm_and_ps(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFBool gmx_simdcall operator||(SimdFBool a, SimdFBool b)
{
    return { _mm_or_ps(a.simdInternal_, b.simdInternal_) };
}

static inline bool gmx_simdcall anyTrue(SimdFBool a)
{
    return _mm_movemask_ps(a.simdInternal_) != 0;
}

static inline SimdFloat gmx_simdcall selectByMask(SimdFloat a, SimdFBool mask)
{
    return { _mm_and_ps(a.simdInternal_, mask.simdInternal_) };
}

static inline SimdFloat gmx_simdcall selectByNotMask(SimdFloat a, SimdFBool mask)
{
    return { _mm_andnot_ps(mask.simdInternal_, a.simdInternal_) };
}

// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
static inline SimdFloat gmx_simdcall blend(SimdFloat a, SimdFloat b, SimdFBool sel)
{
    return { _mm_or_ps(_mm_andnot_ps(sel.simdInternal_, a.simdInternal_),
                       _mm_and_ps(sel.simdInternal_, b.simdInternal_)) };
}
#endif

static inline SimdFInt32 gmx_simdcall operator&(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_and_si128(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall andNot(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_andnot_si128(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall operator|(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_or_si128(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall operator^(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_xor_si128(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall operator+(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_add_epi32(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall operator-(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_sub_epi32(a.simdInternal_, b.simdInternal_) };
}

// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
static inline SimdFInt32 gmx_simdcall operator*(SimdFInt32 a, SimdFInt32 b)
{
    __m128i a1 = _mm_srli_si128(a.simdInternal_, 4); // - a[3] a[2] a[1]
    __m128i b1 = _mm_srli_si128(b.simdInternal_, 4); // - b[3] b[2] b[1]
    __m128i c  = _mm_mul_epu32(a.simdInternal_, b.simdInternal_);
    __m128i c1 = _mm_mul_epu32(a1, b1);

    c  = _mm_shuffle_epi32(c, _MM_SHUFFLE(3, 1, 2, 0));  // - - a[2]*b[2] a[0]*b[0]
    c1 = _mm_shuffle_epi32(c1, _MM_SHUFFLE(3, 1, 2, 0)); // - - a[3]*b[3] a[1]*b[1]

    return { _mm_unpacklo_epi32(c, c1) };
}
#endif

static inline SimdFIBool gmx_simdcall operator==(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_cmpeq_epi32(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFIBool gmx_simdcall testBits(SimdFInt32 a)
{
    __m128i x   = a.simdInternal_;
    __m128i res = _mm_andnot_si128(_mm_cmpeq_epi32(x, _mm_setzero_si128()), _mm_cmpeq_epi32(x, x));

    return { res };
}

static inline SimdFIBool gmx_simdcall operator<(SimdFInt32 a, SimdFInt32 b)
{
    return { _mm_cmplt_epi32(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFIBool gmx_simdcall operator&&(SimdFIBool a, SimdFIBool b)
{
    return { _mm_and_si128(a.simdInternal_, b.simdInternal_) };
}

static inline SimdFIBool gmx_simdcall operator||(SimdFIBool a, SimdFIBool b)
{
    return { _mm_or_si128(a.simdInternal_, b.simdInternal_) };
}

static inline bool gmx_simdcall anyTrue(SimdFIBool a)
{
    return _mm_movemask_epi8(a.simdInternal_) != 0;
}

static inline SimdFInt32 gmx_simdcall selectByMask(SimdFInt32 a, SimdFIBool mask)
{
    return { _mm_and_si128(a.simdInternal_, mask.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall selectByNotMask(SimdFInt32 a, SimdFIBool mask)
{
    return { _mm_andnot_si128(mask.simdInternal_, a.simdInternal_) };
}

// Override for SSE4.1 and higher
#if GMX_SIMD_X86_SSE2
static inline SimdFInt32 gmx_simdcall blend(SimdFInt32 a, SimdFInt32 b, SimdFIBool sel)
{
    return { _mm_or_si128(_mm_andnot_si128(sel.simdInternal_, a.simdInternal_),
                          _mm_and_si128(sel.simdInternal_, b.simdInternal_)) };
}
#endif

static inline SimdFInt32 gmx_simdcall cvtR2I(SimdFloat a)
{
    return { _mm_cvtps_epi32(a.simdInternal_) };
}

static inline SimdFInt32 gmx_simdcall cvttR2I(SimdFloat a)
{
    return { _mm_cvttps_epi32(a.simdInternal_) };
}

static inline SimdFloat gmx_simdcall cvtI2R(SimdFInt32 a)
{
    return { _mm_cvtepi32_ps(a.simdInternal_) };
}

static inline SimdFIBool gmx_simdcall cvtB2IB(SimdFBool a)
{
    return { _mm_castps_si128(a.simdInternal_) };
}

static inline SimdFBool gmx_simdcall cvtIB2B(SimdFIBool a)
{
    return { _mm_castsi128_ps(a.simdInternal_) };
}

} // namespace gmx

#endif // GMX_SIMD_IMPL_X86_SSE2_SIMD_FLOAT_H