flattexzc.h

Go to the documentation of this file.

inline void LeRasterizer::fillFlatTexZC(int y, float x1, float x2, float w1, float w2, float u1, float u2, float v1, float v2)
{
    float d = x2 - x1;
    if (d == 0.0f) return;

    float id = 1.0f / d;
    float au = (u2 - u1) * id;
    float av = (v2 - v1) * id;
    float aw = (w2 - w1) * id;

    __m128 u_4 = _mm_set_ps(u1 + 3.0f * au, u1 + 2.0f * au, u1 + au, u1);
    __m128 v_4 = _mm_set_ps(v1 + 3.0f * av, v1 + 2.0f * av, v1 + av, v1);
    __m128 w_4 = _mm_set_ps(w1 + 3.0f * aw, w1 + 2.0f * aw, w1 + aw, w1);

    __m128 au_4 = _mm_set1_ps(au * 4.0f);
    __m128 av_4 = _mm_set1_ps(av * 4.0f);
    __m128 aw_4 = _mm_set1_ps(aw * 4.0f);

    int xb = (int)(x1);
    int xe = (int)(x2 + 1.9999f);
    if (xe > frame.tx) xe = frame.tx;

    LeColor * p = xb + ((int) y) * frame.tx + pixels;
    int b = (xe - xb) >> 2;
    int r = (xe - xb) & 0x3;

    for (int x = 0; x < b; x ++) {
        __m128 z_4 = _mm_rcp_ps(w_4);

        __m128 mu_4, mv_4;
        mu_4 = _mm_mul_ps(u_4, z_4);
        mv_4 = _mm_mul_ps(v_4, z_4);
        mv_4 = _mm_mul_ps(mv_4, texScale_4);

        __m128i mui_4, mvi_4;
        mui_4 = _mm_cvtps_epi32(mu_4);
        mvi_4 = _mm_cvtps_epi32(mv_4);
        mui_4 = _mm_and_si128(mui_4, texMaskU_4);
        mvi_4 = _mm_and_si128(mvi_4, texMaskV_4);
        mui_4 = _mm_add_epi32(mui_4, mvi_4);

        __m128i zv = _mm_set1_epi32(0);
        __m128i tp, tq, t1, t2;
        tp = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *) &mui_4)[0]]);
        tq = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *) &mui_4)[1]]);
        t1 = _mm_unpacklo_epi32(tp, tq);
        t1 = _mm_unpacklo_epi8(t1, zv);
        t1 = _mm_mullo_epi16(t1, color_4);
        t1 = _mm_srli_epi16(t1, 8);

        tp = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *)&mui_4)[2]]);
        tq = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *)&mui_4)[3]]);
        t2 = _mm_unpacklo_epi32(tp, tq);
        t2 = _mm_unpacklo_epi8(t2, zv);
        t2 = _mm_mullo_epi16(t2, color_4);
        t2 = _mm_srli_epi16(t2, 8);

        tp = _mm_packus_epi16(t1, t2);
        _mm_storeu_si128((__m128i *) p,  tp);
        p += 4;

        w_4 = _mm_add_ps(w_4, aw_4);
        u_4 = _mm_add_ps(u_4, au_4);
        v_4 = _mm_add_ps(v_4, av_4);
    }

    if (r == 0) return;
    __m128 z_4 = _mm_rcp_ps(w_4);

    __m128 mu_4, mv_4;
    mu_4 = _mm_mul_ps(u_4, z_4);
    mv_4 = _mm_mul_ps(v_4, z_4);
    mv_4 = _mm_mul_ps(mv_4, texScale_4);

    __m128i mui_4, mvi_4;
    mui_4 = _mm_cvtps_epi32(mu_4);
    mvi_4 = _mm_cvtps_epi32(mv_4);
    mui_4 = _mm_and_si128(mui_4, texMaskU_4);
    mvi_4 = _mm_and_si128(mvi_4, texMaskV_4);
    mui_4 = _mm_add_epi32(mui_4, mvi_4);

    __m128i zv = _mm_set1_epi32(0);
    __m128i tp;
    tp = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *)&mui_4)[0]]);
    tp = _mm_unpacklo_epi8(tp, zv);
    tp = _mm_mullo_epi16(tp, color_4);
    tp = _mm_srli_epi16(tp, 8);
    tp = _mm_packus_epi16(tp, zv);
    *p++ = _mm_cvtsi128_si32(tp);

    if (r == 1) return;
    tp = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *)&mui_4)[1]]);
    tp = _mm_unpacklo_epi8(tp, zv);
    tp = _mm_mullo_epi16(tp, color_4);
    tp = _mm_srli_epi16(tp, 8);
    tp = _mm_packus_epi16(tp, zv);
    *p++ = _mm_cvtsi128_si32(tp);

    if (r == 2) return;
    tp = _mm_loadl_epi64((__m128i *) &texDiffusePixels[((uint32_t *)&mui_4)[2]]);
    tp = _mm_unpacklo_epi8(tp, zv);
    tp = _mm_mullo_epi16(tp, color_4);
    tp = _mm_srli_epi16(tp, 8);
    tp = _mm_packus_epi16(tp, zv);
    *p++ = _mm_cvtsi128_si32(tp);

}