simd_conv_r3()

static void fl::gfx::blur_detail::simd_conv_r3 ( const u8 * p0, const u8 * p1, const u8 * p2, const u8 * p3, const u8 * p4, const u8 * p5, const u8 * p6, u8 * out, int nbytes )

static

Definition at line 674 of file blur.cpp.hpp.

                                                             {
    namespace fsimd = fl::simd; // ok bare using
    const auto w6w = fsimd::set1_u16_16(6), w15w = fsimd::set1_u16_16(15), w20w = fsimd::set1_u16_16(20);
    const auto w6 = fsimd::set1_u16_8(6), w15 = fsimd::set1_u16_8(15), w20 = fsimd::set1_u16_8(20);
    int i = 0;
    for (; i + 31 < nbytes; i += 32) {
        auto v0 = fsimd::load_u8_32(p0+i), v1 = fsimd::load_u8_32(p1+i), v2 = fsimd::load_u8_32(p2+i);
        auto v3 = fsimd::load_u8_32(p3+i), v4 = fsimd::load_u8_32(p4+i), v5 = fsimd::load_u8_32(p5+i);
        auto v6v = fsimd::load_u8_32(p6+i);
        auto s06 = fsimd::add_u16_16(fsimd::widen_lo_u8x32_to_u16(v0), fsimd::widen_lo_u8x32_to_u16(v6v));
        auto s15 = fsimd::add_u16_16(fsimd::widen_lo_u8x32_to_u16(v1), fsimd::widen_lo_u8x32_to_u16(v5));
        auto s24 = fsimd::add_u16_16(fsimd::widen_lo_u8x32_to_u16(v2), fsimd::widen_lo_u8x32_to_u16(v4));
        auto lo = fsimd::add_u16_16(s06, fsimd::add_u16_16(fsimd::mullo_u16_16(s15, w6w),
                   fsimd::add_u16_16(fsimd::mullo_u16_16(s24, w15w), fsimd::mullo_u16_16(fsimd::widen_lo_u8x32_to_u16(v3), w20w))));
        lo = fsimd::srli_u16_16(lo, 6);
        auto s06h = fsimd::add_u16_16(fsimd::widen_hi_u8x32_to_u16(v0), fsimd::widen_hi_u8x32_to_u16(v6v));
        auto s15h = fsimd::add_u16_16(fsimd::widen_hi_u8x32_to_u16(v1), fsimd::widen_hi_u8x32_to_u16(v5));
        auto s24h = fsimd::add_u16_16(fsimd::widen_hi_u8x32_to_u16(v2), fsimd::widen_hi_u8x32_to_u16(v4));
        auto hi = fsimd::add_u16_16(s06h, fsimd::add_u16_16(fsimd::mullo_u16_16(s15h, w6w),
                   fsimd::add_u16_16(fsimd::mullo_u16_16(s24h, w15w), fsimd::mullo_u16_16(fsimd::widen_hi_u8x32_to_u16(v3), w20w))));
        hi = fsimd::srli_u16_16(hi, 6);
        fsimd::store_u8_32(out+i, fsimd::narrow_u16x16_to_u8(lo, hi));
    }
    for (; i + 15 < nbytes; i += 16) {
        auto v0 = fsimd::load_u8_16(p0+i), v1 = fsimd::load_u8_16(p1+i), v2 = fsimd::load_u8_16(p2+i);
        auto v3 = fsimd::load_u8_16(p3+i), v4 = fsimd::load_u8_16(p4+i), v5 = fsimd::load_u8_16(p5+i);
        auto v6v = fsimd::load_u8_16(p6+i);
        auto s06 = fsimd::add_u16_8(fsimd::widen_lo_u8_to_u16(v0), fsimd::widen_lo_u8_to_u16(v6v));
        auto s15 = fsimd::add_u16_8(fsimd::widen_lo_u8_to_u16(v1), fsimd::widen_lo_u8_to_u16(v5));
        auto s24 = fsimd::add_u16_8(fsimd::widen_lo_u8_to_u16(v2), fsimd::widen_lo_u8_to_u16(v4));
        auto lo = fsimd::add_u16_8(s06, fsimd::add_u16_8(fsimd::mullo_u16_8(s15, w6),
                   fsimd::add_u16_8(fsimd::mullo_u16_8(s24, w15), fsimd::mullo_u16_8(fsimd::widen_lo_u8_to_u16(v3), w20))));
        lo = fsimd::srli_u16_8(lo, 6);
        auto s06h = fsimd::add_u16_8(fsimd::widen_hi_u8_to_u16(v0), fsimd::widen_hi_u8_to_u16(v6v));
        auto s15h = fsimd::add_u16_8(fsimd::widen_hi_u8_to_u16(v1), fsimd::widen_hi_u8_to_u16(v5));
        auto s24h = fsimd::add_u16_8(fsimd::widen_hi_u8_to_u16(v2), fsimd::widen_hi_u8_to_u16(v4));
        auto hi = fsimd::add_u16_8(s06h, fsimd::add_u16_8(fsimd::mullo_u16_8(s15h, w6),
                   fsimd::add_u16_8(fsimd::mullo_u16_8(s24h, w15), fsimd::mullo_u16_8(fsimd::widen_hi_u8_to_u16(v3), w20))));
        hi = fsimd::srli_u16_8(hi, 6);
        fsimd::store_u8_16(out+i, fsimd::narrow_u16_to_u8(lo, hi));
    }
    for (; i < nbytes; ++i) {
        u16 s06=(u16)p0[i]+(u16)p6[i], s15=(u16)p1[i]+(u16)p5[i], s24=(u16)p2[i]+(u16)p4[i];
        out[i]=(u8)((s06+s15*6+s24*15+(u16)p3[i]*20)>>6);
    }
}

Referenced by fl::gfx::blur_detail::simd_hconv_dispatch< 3 >::apply(), and fl::gfx::blur_detail::simd_vconv_dispatch< 3 >::apply().

Here is the caller graph for this function: