Compiler Explorer

Source code

// Type your code here, or load an example.
#include <cinttypes>
#if !defined( __aarch64__) && !defined(__arm__)
#include <xmmintrin.h>
#include <tmmintrin.h>
#include <pmmintrin.h>
#include <smmintrin.h>
#else
#include <arm_neon.h>
#endif
#include <array>
#include <limits>
typedef uint8_t  v16ui  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef uint16_t v8ui  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef uint32_t v4ui  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef uint64_t v2ui __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef int8_t  v16si  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef int16_t v8si  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef int32_t v4si  __attribute__((vector_size(16))) __attribute__((aligned(16)));
typedef int64_t v2si __attribute__((vector_size(16))) __attribute__((aligned(16)));

v2ui expand64_u(v16ui a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1]};
        case 1:
        return {a[2], a[3]};
        case 2:
        return {a[4], a[5]};
        case 3:
        return {a[6], a[7]};
        case 4:
        return {a[8], a[9]};
        case 5:
        return {a[10], a[11]};
        case 6:
        return {a[12],a[13]};
        case 7:
        return {a[14],a[15]};
    }
    return {0};
}

v2ui expand64_u(v8ui a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1]};
        case 1:
        return {a[2], a[3]};
        case 2:
        return {a[4], a[5]};
        case 3:
        return {a[6], a[7]};
    }
    return {0};
}

v2si expand64_s(v16si a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1]};
        case 1:
        return {a[2], a[3]};
        case 2:
        return {a[4], a[5]};
        case 3:
        return {a[6], a[7]};
        case 4:
        return {a[8], a[9]};
        case 5:
        return {a[10], a[11]};
        case 6:
        return {a[12],a[13]};
        case 7:
        return {a[14],a[15]};
    }
    return {0};
}

v2si expand64_s(v8si a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1]};
        case 1:
        return {a[2], a[3]};
        case 2:
        return {a[4], a[5]};
        case 3:
        return {a[6], a[7]};
    }
        return {0};
}

v4ui expand32_u(v16ui a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1], a[2], a[3]};
        case 1:
        return {a[4], a[5], a[6], a[7]};
        case 2:
        return {a[8], a[9], a[10], a[11]};
        case 3:
        return {a[12], a[13], a[14], a[15]};
    }
        return {0};
}

v4si expand32_s(v16si a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1], a[2], a[3]};
        case 1:
        return {a[4], a[5], a[6], a[7]};
        case 2:
        return {a[8], a[9], a[10], a[11]};
        case 3:
        return {a[12], a[13], a[14], a[15]};
    }
        return {0};
}
#if defined(__aarch64__)

v4ui expand64_u_insr(v16ui a, uint8_t imm) {
    const v2ui adder{2,2};
    const v2ui tblMask1 = {0xffffffffffffff00,0xffffffffffffff01};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    switch(imm) {
        case 0:
        return vqtbl1q_u8(a,tblMask1);
        case 1:
        return vqtbl1q_u8(a,tblMask2);
        case 2:
        return vqtbl1q_u8(a,tblMask3);
        case 3:
        return vqtbl1q_u8(a,tblMask4);
        case 4:
        return vqtbl1q_u8(a,tblMask5);
        case 5:
        return vqtbl1q_u8(a,tblMask6);
        case 6:
        return vqtbl1q_u8(a,tblMask7);
        case 7:
        return vqtbl1q_u8(a,tblMask8);
    }
        return {0};
}

v4si expand64_s_insr(v16si a, uint8_t imm) {
    const v2ui adder{0x02ffffffffffffff,0x02ffffffffffffff};
    const v2ui tblMask1 = {0x00ffffffffffffff,0x01ffffffffffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    switch(imm) {
        case 0:
        return ((v2si)vqtbl1q_u8(a,tblMask1)) >> 56;
        case 1:
        return ((v2si)vqtbl1q_u8(a,tblMask2)) >> 56;
        case 2:
        return ((v2si)vqtbl1q_u8(a,tblMask3)) >> 56;
        case 3:
        return ((v2si)vqtbl1q_u8(a,tblMask4)) >> 56;
        case 4:
        return ((v2si)vqtbl1q_u8(a,tblMask5)) >> 56;
        case 5:
        return ((v2si)vqtbl1q_u8(a,tblMask6)) >> 56;
        case 6:
        return ((v2si)vqtbl1q_u8(a,tblMask7)) >> 56;
        case 7:
        return ((v2si)vqtbl1q_u8(a,tblMask8)) >> 56;
    }
        return {0};
}

v4ui expand64_u_insr(v8ui a, uint8_t imm) {
    const v2ui adder{0x0202,0x0202};
    const v2ui tblMask1 = {0xffffffffffff0100,0xffffffffffff0302};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;

switch(imm) {
        case 0:
        return vqtbl1q_u8(a,tblMask1);
        case 1:
        return vqtbl1q_u8(a,tblMask2);
        case 2:
        return vqtbl1q_u8(a,tblMask3);
        case 3:
        return vqtbl1q_u8(a,tblMask4);
    }
        return {0};
}

v4si expand64_s_insr(v8si a, uint8_t imm) {
    uint64_t adderval1 = 0x0202l << 48;
    const v2ui adder{adderval1,adderval1};
    const v2ui tblMask1 = {0x0100ffffffffffff,0x0302ffffffffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;

switch(imm) {
        case 0:
        return ((v2si)vqtbl1q_u8(a,tblMask1)) >> 48;
        case 1:
        return ((v2si)vqtbl1q_u8(a,tblMask2)) >> 48;
        case 2:
        return ((v2si)vqtbl1q_u8(a,tblMask3)) >> 48;
        case 3:
        return ((v2si)vqtbl1q_u8(a,tblMask4)) >> 48;
    }
        return {0};
}

v4ui expand32_u_insr(v16ui a, uint8_t imm) {
    const v4ui tblMask1 = {0xffffff00,0xffffff01,0xffffff02, 0xffffff03},
        tblMask2 = {0xffffff04,0xffffff05,0xffffff06, 0xffffff07},
        tblMask3 = {0xffffff08,0xffffff09,0xffffff0a, 0xffffff0b},
        tblMask4 = {0xffffff0c,0xffffff0d,0xffffff0e, 0xffffff0f};

v4si expand32_s_insr(v16si a, uint8_t imm) {
        const v4ui tblMask1 = {0x00ffffff,0x01ffffff,0x02ffffff, 0x03ffffff},
        tblMask2 = {0x04ffffff,0x05ffffff,0x06ffffff, 0x07ffffff},
        tblMask3 = {0x08ffffff,0x09ffffff,0x0affffff, 0x0bffffff},
        tblMask4 = {0x0cffffff,0x0dffffff,0x0effffff, 0x0fffffff};

switch(imm) {
        case 0:
        return ((v4si)vqtbl1q_u8(a,tblMask1))>>24;
        case 1:
        return ((v4si)vqtbl1q_u8(a,tblMask2))>>24;
        case 2:
        return ((v4si)vqtbl1q_u8(a,tblMask3))>>24;
        case 3:
        return ((v4si)vqtbl1q_u8(a,tblMask4))>>24;
    }
        return {0};
}

#elif defined(__x86_64__)
__attribute__((target("ssse3")))
v4ui expand32_u_insr(v16ui a, uint8_t imm) {
    const v4ui tblMask1 = {0xffffff00,0xffffff01,0xffffff02, 0xffffff03},
        tblMask2 = {0xffffff04,0xffffff05,0xffffff06, 0xffffff07},
        tblMask3 = {0xffffff08,0xffffff09,0xffffff0a, 0xffffff0b},
        tblMask4 = {0xffffff0c,0xffffff0d,0xffffff0e, 0xffffff0f};

switch(imm) {
        case 0:
        return _mm_shuffle_epi8(a,tblMask1);
        case 1:
        return _mm_shuffle_epi8(a,tblMask2);
        case 2:
        return _mm_shuffle_epi8(a,tblMask3);
        case 3:
        return _mm_shuffle_epi8(a,tblMask4);
    }
        return {0};
}

__attribute__((target("ssse3")))
v4si expand32_s_insr(v16ui a, uint8_t imm) {
    const v4ui tblMask1 = {0x00ffffff,0x01ffffff,0x02ffffff, 0x03ffffff},
        tblMask2 = {0x04ffffff,0x05ffffff,0x06ffffff, 0x07ffffff},
        tblMask3 = {0x08ffffff,0x09ffffff,0x0affffff, 0x0bffffff},
        tblMask4 = {0x0cffffff,0x0dffffff,0x0effffff, 0x0fffffff};

switch(imm) {
        case 0:
        return ((v4si)_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1))>>24;
        case 1:
        return ((v4si)_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2))>>24;
        case 2:
        return ((v4si)_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3))>>24;
        case 3:
        return ((v4si)_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4))>>24;
    }
    return {0};
}

__attribute__((target("ssse3")))
v4ui expand64_u_insr(v16ui a, uint8_t imm) {
    const v2ui adder{2,2};
    const v2ui tblMask1 = {0xffffffffffffff00,0xffffffffffffff01};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    switch(imm) {
        case 0:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        case 1:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2);
        case 2:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        case 3:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4);
        case 4:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask5);
        case 5:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask6);
        case 6:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask7);
        case 7:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask8);
    }
        return {0};
}
__attribute__((target("ssse3")))
inline v2si expand64_s_insr_old(v16si a, uint8_t imm) {
    const v2ui adder{0x02ffffff02ffffff,0x02ffffff02ffffff};
    const v2ui tblMask1 = {0x01ffffff00ffffff,0x03ffffff02ffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    __m128i shuffleOut;
#if 0
    switch(imm) {
        case 0:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        break;
        case 1:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2));
        break;
        case 2:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        break;
        case 3:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4));
        break;
        case 4:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask5);
        break;
        case 5:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask6));
        break;
        case 6:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask7);
        break;
        case 7:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask8);
        break;
        default:
        return {0};
    }
#else
    switch(imm) {
        case 0:
        case 1:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        break;
        case 2:
        case 3:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        break;
        case 4:
        case 5:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask5);
        break;
        case 6:
        case 7:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask7);
        break;
        default:
        return {0};
    }
#endif 
    auto lowerPart = shuffleOut;
    auto upperPart = lowerPart;
    lowerPart = _mm_srai_epi32(lowerPart,24);
    upperPart = _mm_srai_epi32(upperPart,31);

return (imm % 2 == 1) ? (v2si) _mm_unpackhi_epi32(upperPart,lowerPart) :  (v2si) _mm_unpacklo_epi32(upperPart,lowerPart);
}

__attribute__((target("ssse3")))
inline std::array<v2si,2> expand64_s_insr_old_return2(v16si a, uint8_t imm) {
    const v2ui adder{0x0400000004000000,0x0400000004000000};
    const v2ui tblMask1 = {0x01ffffff00ffffff,0x03ffffff02ffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    __m128i shuffleOut;
#if 0
    switch(imm) {
        case 0:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        break;
        case 1:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2));
        break;
        case 2:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        break;
        case 3:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4));
        break;
        case 4:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask5);
        break;
        case 5:
        shuffleOut = (_mm_shuffle_epi8((__m128i)a,(__m128i)tblMask6));
        break;
        case 6:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask7);
        break;
        case 7:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask8);
        break;
        default:
        return {0};
    }
#else
    switch(imm) {
        case 0:
        case 1:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        break;
        case 2:
        case 3:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2);
        break;
        case 4:
        case 5:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        break;
        case 6:
        case 7:
        shuffleOut = _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4);
        break;
        default:
        return {0};
    }
#endif 
    auto lowerPart = shuffleOut;
    auto upperPart = lowerPart;
    lowerPart = _mm_srai_epi32(lowerPart,24);
    upperPart = _mm_srai_epi32(upperPart,31);

return std::array<v2si,2>{_mm_unpacklo_epi32(upperPart,lowerPart),_mm_unpackhi_epi32(upperPart,lowerPart)};
}

__attribute__((target("default")))
v2ui expand64_u_insr(v8ui a, uint8_t imm) {
    return {a[imm*2],a[imm*2+1]};
}

__attribute__((target("ssse3")))
 v2ui expand64_u_insr(v8ui a, uint8_t imm) {
    const v2ui adder{0x0202,0x0202};
    const v2ui tblMask1 = {0xffffffffffff0100,0xffffffffffff0302};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;

switch(imm) {
        case 0:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask1);
        case 1:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask2);
        case 2:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask3);
        case 3:
        return _mm_shuffle_epi8((__m128i)a,(__m128i)tblMask4);
    }
        return {0};
}

inline v8ui genMask64(v8si a, uint8_t imm) {
    return {uint16_t(imm*2),uint16_t(imm*2+1),0xffff,0xffff,uint16_t(imm*2+2),uint16_t(imm*2+3),0xffff,0xffff};
}

inline v16ui genMask64(v16si a, uint8_t imm) {
        return {uint8_t(imm*2),0xff,0xff,0xff,0xff,0xff,0xff,0xff,
        uint8_t(imm*2+1),0xff,0xff,0xff,0xff,0xff,0xff,0xff};
}

__attribute__((target("sse4")))
 v4si expand64_s_insr(v16si const & a, uint8_t const & imm) {
    __m128i *pPtr = (__m128i*)(((int16_t*)&a)+imm);
    return _mm_cvtepi8_epi64(*pPtr);
}
__attribute__((target("sse4")))
 v4si expand64_s_insr(v8si const & a, uint8_t const & imm) {
    __m128i *pPtr = (__m128i*)(((int32_t*)&a)+imm);
    return _mm_cvtepi16_epi64(*pPtr);
}

#endif

#if defined(__arm__) 
v2ui expand64_u_insr(v16ui a, uint8_t imm) {
    const v2ui adder{2,2};
    const v2ui tblMask1 = {0xffffffffffffff00,0xffffffffffffff01};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    switch(imm) {
        case 0:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask1)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask1)));
        case 1:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask2)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask2)));
        case 2:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask3)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask3)));
        case 3:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask4)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask4)));
        case 4:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask5)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask5)));
        case 5:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask6)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask6)));
        case 6:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask7)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask7)));
        case 7:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask8)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask8)));
    }
        return {0};
}

v2si expand64_s_insr(v16si a, uint8_t imm) {
    const v2ui adder{0x02ffffffffffffff,0x02ffffffffffffff};
    const v2ui tblMask1 = {0x00ffffffffffffff,0x01ffffffffffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;
    auto tblMask5 = tblMask4 + adder;
    auto tblMask6 = tblMask5 + adder;
    auto tblMask7 = tblMask6 + adder;
    auto tblMask8 = tblMask7 + adder;
    switch(imm) {
        case 0:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask1)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask1))))>>56;
        case 1:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask2)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask2))))>>56;
        case 2:
           return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask3)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask3))))>>56;
        case 3:
            return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask4)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask4))))>>56;
        case 4:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask5)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask5))))>>56;
        case 5:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask6)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask6))))>>56;
        case 6:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask7)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask7))))>>56;
        case 7:
        return((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask8)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask8))))>>56;
    }
        return {0};
}

v2ui expand64_u_insr(v8ui a, uint8_t imm) {
    const v2ui adder{0x0202,0x0202};
    const v2ui tblMask1 = {0xffffffffffff0100,0xffffffffffff0302};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;

switch(imm) {
        case 0:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask1)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask1)));
        case 1:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask2)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask2)));
        case 2:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask3)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask3)));
        case 3:
        return vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask4)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask4)));
    }
        return {0};
}

v2si expand64_s_insr(v8si a, uint8_t imm) {
    uint64_t adderval1 = 0x0202l << 48;
    const v2ui adder{adderval1,adderval1};
    const v2ui tblMask1 = {0x0100ffffffffffff,0x0302ffffffffffff};
    auto tblMask2 = tblMask1 + adder;
    auto tblMask3 = tblMask2 + adder;
    auto tblMask4 = tblMask3 + adder;

switch(imm) {
        case 0:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask1)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask1))))>>48;
        case 1:
        return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask2)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask2))))>>48;
        case 2:
           return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask3)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask3))))>>48;
        case 3:
            return ((v2si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask4)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask4))))>>48;
    }
        return {0};
}

switch(imm) {
        case 0:
        return ((v4si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask1)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask1))))>>24;
        case 1:
        return ((v4si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask2)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask2))))>>24;
        case 2:
           return ((v4si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask3)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask3))))>>24;
        case 3:
            return ((v4si)vcombine_u8(vtbl1_u8(vget_low_u8(a),vget_low_u8(tblMask4)),
                vtbl1_u8(vget_high_u8(a),vget_high_u8(tblMask4))))>>24;
    }
        return {0};
}
#endif

v4ui expand32_u(v8ui a, uint8_t imm) {
    switch(imm) {
        case 0:
        return {a[0], a[1], a[2], a[3]};
        case 1:
        return {a[4], a[5], a[6], a[7]};
    }
        return {0};
}

void test100_expand32_u(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand32_u(pIn[i],0);
        pOut[i*4+1] = expand32_u(pIn[i],1); 
        pOut[i*4+2] = expand32_u(pIn[i],2);
        pOut[i*4+3] = expand32_u(pIn[i],3);
    }
}

void test100_expand32_s(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand32_s(pIn[i],0);
        pOut[i*4+1] = expand32_s(pIn[i],1); 
        pOut[i*4+2] = expand32_s(pIn[i],2);
        pOut[i*4+3] = expand32_s(pIn[i],3);
    }
}

void test100_expand32_u_insr(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand32_u_insr(pIn[i],0);
        pOut[i*4+1] = expand32_u_insr(pIn[i],1); 
        pOut[i*4+2] = expand32_u_insr(pIn[i],2);
        pOut[i*4+3] = expand32_u_insr(pIn[i],3);
    }
}

void test100_expand32_s_insr(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand32_s_insr(pIn[i],0);
        pOut[i*4+1] = expand32_s_insr(pIn[i],1); 
        pOut[i*4+2] = expand32_s_insr(pIn[i],2);
        pOut[i*4+3] = expand32_s_insr(pIn[i],3);
    }
}

void test100_expand64_u(v8ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_u(pIn[i],0);
        pOut[i*4+1] = expand64_u(pIn[i],1); 
        pOut[i*4+2] = expand64_u(pIn[i],2);
        pOut[i*4+3] = expand64_u(pIn[i],3);
    }
}

void test100_expand64_u(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_u(pIn[i],0);
        pOut[i*4+1] = expand64_u(pIn[i],1); 
        pOut[i*4+2] = expand64_u(pIn[i],2);
        pOut[i*4+3] = expand64_u(pIn[i],3);
        pOut[i*4+4] = expand64_u(pIn[i],4);
        pOut[i*4+5] = expand64_u(pIn[i],5); 
        pOut[i*4+6] = expand64_u(pIn[i],6);
        pOut[i*4+7] = expand64_u(pIn[i],7);
    }
}

void test100_expand64_s(v8si* pIn, size_t pLength, v4si *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_s(pIn[i],0);
        pOut[i*4+1] = expand64_s(pIn[i],1); 
        pOut[i*4+2] = expand64_s(pIn[i],2);
        pOut[i*4+3] = expand64_s(pIn[i],3);
    }
}

void test100_expand64_s(v16si* pIn, size_t pLength, v4si *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_s(pIn[i],0);
        pOut[i*4+1] = expand64_s(pIn[i],1); 
        pOut[i*4+2] = expand64_s(pIn[i],2);
        pOut[i*4+3] = expand64_s(pIn[i],3);
        pOut[i*4+4] = expand64_s(pIn[i],4);
        pOut[i*4+5] = expand64_s(pIn[i],5); 
        pOut[i*4+6] = expand64_s(pIn[i],6);
        pOut[i*4+7] = expand64_s(pIn[i],7);
    }
}

void test100_expand64_u_insr(v8ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_u_insr(pIn[i],0);
        pOut[i*4+1] = expand64_u_insr(pIn[i],1); 
        pOut[i*4+2] = expand64_u_insr(pIn[i],2);
        pOut[i*4+3] = expand64_u_insr(pIn[i],3);
    }
}

void test100_expand64_u_insr(v16ui* pIn, size_t pLength, v4ui *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_u_insr(pIn[i],0);
        pOut[i*4+1] = expand64_u_insr(pIn[i],1); 
        pOut[i*4+2] = expand64_u_insr(pIn[i],2);
        pOut[i*4+3] = expand64_u_insr(pIn[i],3);
        pOut[i*4+4] = expand64_u_insr(pIn[i],4);
        pOut[i*4+5] = expand64_u_insr(pIn[i],5); 
        pOut[i*4+6] = expand64_u_insr(pIn[i],6);
        pOut[i*4+7] = expand64_u_insr(pIn[i],7);
    }
}

void test100_expand64_s_insr(v8si* pIn, size_t pLength, v4si *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_s_insr(pIn[i],0);
        pOut[i*4+1] = expand64_s_insr(pIn[i],1); 
        pOut[i*4+2] = expand64_s_insr(pIn[i],2);
        pOut[i*4+3] = expand64_s_insr(pIn[i],3);
    }
}
void test100_expand64_s_insr(v16si* pIn, size_t pLength, v4si *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_s_insr(pIn[i],0);
        pOut[i*4+1] = expand64_s_insr(pIn[i],1); 
        pOut[i*4+2] = expand64_s_insr(pIn[i],2);
        pOut[i*4+3] = expand64_s_insr(pIn[i],3);
        pOut[i*4+4] = expand64_s_insr(pIn[i],4);
        pOut[i*4+5] = expand64_s_insr(pIn[i],5); 
        pOut[i*4+6] = expand64_s_insr(pIn[i],6);
        pOut[i*4+7] = expand64_s_insr(pIn[i],7);
    }
}

#if defined(__x86_64__)
void test100_expand64_s_insr_old(v16si* pIn, size_t pLength, v4si *pOut) {
    for (size_t i = 0; i < pLength; ++i) {
        pOut[i*4] = expand64_s_insr_old(pIn[i],0);
        pOut[i*4+1] = expand64_s_insr_old(pIn[i],1); 
        pOut[i*4+2] = expand64_s_insr_old(pIn[i],2);
        pOut[i*4+3] = expand64_s_insr_old(pIn[i],3);
        pOut[i*4+4] = expand64_s_insr_old(pIn[i],4);
        pOut[i*4+5] = expand64_s_insr_old(pIn[i],5); 
        pOut[i*4+6] = expand64_s_insr_old(pIn[i],6);
        pOut[i*4+7] = expand64_s_insr_old(pIn[i],7);
    }
}

//expand64_s_insr_old_return2
void test100_expand64_s_insr_old_return2(v16si* pIn, size_t pLength, v4si *pOut) {

for (size_t i = 0; i < pLength; ++i) {
        auto a = expand64_s_insr_old_return2(pIn[i],0);
        pOut[i*4] = a[0];
        pOut[i*4+1] = a[1];
        auto b = expand64_s_insr_old_return2(pIn[i],2);
        pOut[i*4+2] = b[0];
        pOut[i*4+3] = b[1];
        auto c = expand64_s_insr_old_return2(pIn[i],4);
        pOut[i*4+4] =  c[0];
        pOut[i*4+5] = c[1];
        auto d = expand64_s_insr_old_return2(pIn[i],6);
        pOut[i*4+6] = d[0];
        pOut[i*4+7] = d[1];
    }
}
#endif