
/*
 * Copyright 2006 The Android Open Source Project
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */


#include "SkEmbossMask.h"
#include "SkMath.h"

static inline int nonzero_to_one(int x) {
#if 0
    return x != 0;
#else
    return ((unsigned)(x | -x)) >> 31;
#endif
}

static inline int neq_to_one(int x, int max) {
#if 0
    return x != max;
#else
    SkASSERT(x >= 0 && x <= max);
    return ((unsigned)(x - max)) >> 31;
#endif
}

static inline int neq_to_mask(int x, int max) {
#if 0
    return -(x != max);
#else
    SkASSERT(x >= 0 && x <= max);
    return (x - max) >> 31;
#endif
}

static inline unsigned div255(unsigned x) {
    SkASSERT(x <= (255*255));
    return x * ((1 << 24) / 255) >> 24;
}

#define kDelta  32  // small enough to show off angle differences

#include "SkEmbossMask_Table.h"

#if defined(SK_BUILD_FOR_WIN32) && defined(SK_DEBUG)

#include <stdio.h>

void SkEmbossMask_BuildTable() {
    // build it 0..127 x 0..127, so we use 2^15 - 1 in the numerator for our "fixed" table

    FILE* file = ::fopen("SkEmbossMask_Table.h", "w");
    SkASSERT(file);
    ::fprintf(file, "#include \"SkTypes.h\"\n\n");
    ::fprintf(file, "static const U16 gInvSqrtTable[128 * 128] = {\n");
    for (int dx = 0; dx <= 255/2; dx++) {
        for (int dy = 0; dy <= 255/2; dy++) {
            if ((dy & 15) == 0)
                ::fprintf(file, "\t");

            uint16_t value = SkToU16((1 << 15) / SkSqrt32(dx * dx + dy * dy + kDelta*kDelta/4));

            ::fprintf(file, "0x%04X", value);
            if (dx * 128 + dy < 128*128-1) {
                ::fprintf(file, ", ");
            }
            if ((dy & 15) == 15) {
                ::fprintf(file, "\n");
            }
        }
    }
    ::fprintf(file, "};\n#define kDeltaUsedToBuildTable\t%d\n", kDelta);
    ::fclose(file);
}

#endif

void SkEmbossMask::Emboss(SkMask* mask, const SkEmbossMaskFilter::Light& light) {
    SkASSERT(kDelta == kDeltaUsedToBuildTable);

    SkASSERT(mask->fFormat == SkMask::k3D_Format);

    int     specular = light.fSpecular;
    int     ambient = light.fAmbient;
    SkFixed lx = SkScalarToFixed(light.fDirection[0]);
    SkFixed ly = SkScalarToFixed(light.fDirection[1]);
    SkFixed lz = SkScalarToFixed(light.fDirection[2]);
    SkFixed lz_dot_nz = lz * kDelta;
    int     lz_dot8 = lz >> 8;

    size_t      planeSize = mask->computeImageSize();
    uint8_t*    alpha = mask->fImage;
    uint8_t*    multiply = (uint8_t*)alpha + planeSize;
    uint8_t*    additive = multiply + planeSize;

    int rowBytes = mask->fRowBytes;
    int maxy = mask->fBounds.height() - 1;
    int maxx = mask->fBounds.width() - 1;

    int prev_row = 0;
    for (int y = 0; y <= maxy; y++) {
        int next_row = neq_to_mask(y, maxy) & rowBytes;

        for (int x = 0; x <= maxx; x++) {
            if (alpha[x]) {
                int nx = alpha[x + neq_to_one(x, maxx)] - alpha[x - nonzero_to_one(x)];
                int ny = alpha[x + next_row] - alpha[x - prev_row];

                SkFixed numer = lx * nx + ly * ny + lz_dot_nz;
                int     mul = ambient;
                int     add = 0;

                if (numer > 0) {  // preflight when numer/denom will be <= 0
#if 0
                    int denom = SkSqrt32(nx * nx + ny * ny + kDelta*kDelta);
                    SkFixed dot = numer / denom;
                    dot >>= 8;  // now dot is 2^8 instead of 2^16
#else
                    // can use full numer, but then we need to call SkFixedMul, since
                    // numer is 24 bits, and our table is 12 bits

                    // SkFixed dot = SkFixedMul(numer, gTable[]) >> 8
                    SkFixed dot = (unsigned)(numer >> 4) * gInvSqrtTable[(SkAbs32(nx) >> 1 << 7) | (SkAbs32(ny) >> 1)] >> 20;
#endif
                    mul = SkFastMin32(mul + dot, 255);

                    // now for the reflection

                    //  R = 2 (Light * Normal) Normal - Light
                    //  hilite = R * Eye(0, 0, 1)

                    int hilite = (2 * dot - lz_dot8) * lz_dot8 >> 8;
                    if (hilite > 0) {
                        // pin hilite to 255, since our fast math is also a little sloppy
                        hilite = SkClampMax(hilite, 255);

                        // specular is 4.4
                        // would really like to compute the fractional part of this
                        // and then possibly cache a 256 table for a given specular
                        // value in the light, and just pass that in to this function.
                        add = hilite;
                        for (int i = specular >> 4; i > 0; --i) {
                            add = div255(add * hilite);
                        }
                    }
                }
                multiply[x] = SkToU8(mul);
                additive[x] = SkToU8(add);

            //  multiply[x] = 0xFF;
            //  additive[x] = 0;
            //  ((uint8_t*)alpha)[x] = alpha[x] * multiply[x] >> 8;
            }
        }
        alpha += rowBytes;
        multiply += rowBytes;
        additive += rowBytes;
        prev_row = rowBytes;
    }
}