/*
* Copyright (c) 2003, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code 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 General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
#if !defined(JAVA2D_NO_MLIB) || defined(MLIB_ADD_SUFF)
#include <vis_proto.h>
#include "java2d_Mlib.h"
#include "vis_AlphaMacros.h"
/***************************************************************/
mlib_d64 vis_d64_div_tbl[256] = {
0 , 1.0000000000, 0.5000000000, 0.3333333333,
0.2500000000, 0.2000000000, 0.1666666667, 0.1428571429,
0.1250000000, 0.1111111111, 0.1000000000, 0.0909090909,
0.0833333333, 0.0769230769, 0.0714285714, 0.0666666667,
0.0625000000, 0.0588235294, 0.0555555556, 0.0526315789,
0.0500000000, 0.0476190476, 0.0454545455, 0.0434782609,
0.0416666667, 0.0400000000, 0.0384615385, 0.0370370370,
0.0357142857, 0.0344827586, 0.0333333333, 0.0322580645,
0.0312500000, 0.0303030303, 0.0294117647, 0.0285714286,
0.0277777778, 0.0270270270, 0.0263157895, 0.0256410256,
0.0250000000, 0.0243902439, 0.0238095238, 0.0232558140,
0.0227272727, 0.0222222222, 0.0217391304, 0.0212765957,
0.0208333333, 0.0204081633, 0.0200000000, 0.0196078431,
0.0192307692, 0.0188679245, 0.0185185185, 0.0181818182,
0.0178571429, 0.0175438596, 0.0172413793, 0.0169491525,
0.0166666667, 0.0163934426, 0.0161290323, 0.0158730159,
0.0156250000, 0.0153846154, 0.0151515152, 0.0149253731,
0.0147058824, 0.0144927536, 0.0142857143, 0.0140845070,
0.0138888889, 0.0136986301, 0.0135135135, 0.0133333333,
0.0131578947, 0.0129870130, 0.0128205128, 0.0126582278,
0.0125000000, 0.0123456790, 0.0121951220, 0.0120481928,
0.0119047619, 0.0117647059, 0.0116279070, 0.0114942529,
0.0113636364, 0.0112359551, 0.0111111111, 0.0109890110,
0.0108695652, 0.0107526882, 0.0106382979, 0.0105263158,
0.0104166667, 0.0103092784, 0.0102040816, 0.0101010101,
0.0100000000, 0.0099009901, 0.0098039216, 0.0097087379,
0.0096153846, 0.0095238095, 0.0094339623, 0.0093457944,
0.0092592593, 0.0091743119, 0.0090909091, 0.0090090090,
0.0089285714, 0.0088495575, 0.0087719298, 0.0086956522,
0.0086206897, 0.0085470085, 0.0084745763, 0.0084033613,
0.0083333333, 0.0082644628, 0.0081967213, 0.0081300813,
0.0080645161, 0.0080000000, 0.0079365079, 0.0078740157,
0.0078125000, 0.0077519380, 0.0076923077, 0.0076335878,
0.0075757576, 0.0075187970, 0.0074626866, 0.0074074074,
0.0073529412, 0.0072992701, 0.0072463768, 0.0071942446,
0.0071428571, 0.0070921986, 0.0070422535, 0.0069930070,
0.0069444444, 0.0068965517, 0.0068493151, 0.0068027211,
0.0067567568, 0.0067114094, 0.0066666667, 0.0066225166,
0.0065789474, 0.0065359477, 0.0064935065, 0.0064516129,
0.0064102564, 0.0063694268, 0.0063291139, 0.0062893082,
0.0062500000, 0.0062111801, 0.0061728395, 0.0061349693,
0.0060975610, 0.0060606061, 0.0060240964, 0.0059880240,
0.0059523810, 0.0059171598, 0.0058823529, 0.0058479532,
0.0058139535, 0.0057803468, 0.0057471264, 0.0057142857,
0.0056818182, 0.0056497175, 0.0056179775, 0.0055865922,
0.0055555556, 0.0055248619, 0.0054945055, 0.0054644809,
0.0054347826, 0.0054054054, 0.0053763441, 0.0053475936,
0.0053191489, 0.0052910053, 0.0052631579, 0.0052356021,
0.0052083333, 0.0051813472, 0.0051546392, 0.0051282051,
0.0051020408, 0.0050761421, 0.0050505051, 0.0050251256,
0.0050000000, 0.0049751244, 0.0049504950, 0.0049261084,
0.0049019608, 0.0048780488, 0.0048543689, 0.0048309179,
0.0048076923, 0.0047846890, 0.0047619048, 0.0047393365,
0.0047169811, 0.0046948357, 0.0046728972, 0.0046511628,
0.0046296296, 0.0046082949, 0.0045871560, 0.0045662100,
0.0045454545, 0.0045248869, 0.0045045045, 0.0044843049,
0.0044642857, 0.0044444444, 0.0044247788, 0.0044052863,
0.0043859649, 0.0043668122, 0.0043478261, 0.0043290043,
0.0043103448, 0.0042918455, 0.0042735043, 0.0042553191,
0.0042372881, 0.0042194093, 0.0042016807, 0.0041841004,
0.0041666667, 0.0041493776, 0.0041322314, 0.0041152263,
0.0040983607, 0.0040816327, 0.0040650407, 0.0040485830,
0.0040322581, 0.0040160643, 0.0040000000, 0.0039840637,
0.0039682540, 0.0039525692, 0.0039370079, 0.0039215686
};
/***************************************************************/
#define D64_FROM_F32x2(ff) \
vis_freg_pair(ff, ff)
/***************************************************************/
#define RGB2GRAY(r, g, b) \
(((77 * (r)) + (150 * (g)) + (29 * (b)) + 128) >> 8)
/***************************************************************/
static void vis_ByteGrayBlendMask(mlib_u8 *rasBase,
mlib_u8 *pMask,
mlib_s32 rasScan,
mlib_s32 maskScan,
mlib_s32 width,
mlib_s32 height,
mlib_s32 *a0_S32,
mlib_s32 srcG)
{
mlib_f32 ff, srcG_f;
mlib_d64 dd, a0, a1;
mlib_d64 d_one = vis_to_double_dup(0x7FFF7FFF);
mlib_d64 d_round = vis_to_double_dup(((1 << 16) | 1) << 6);
mlib_s32 j, pathA;
maskScan -= width;
srcG = (srcG << 8) | srcG;
srcG_f = vis_to_float((srcG << 16) | srcG);
vis_write_gsr((0 << 3) | 6);
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
mlib_u8 *dst_end;
dst_end = dst + width;
while (((mlib_s32)dst & 3) && dst < dst_end) {
dd = vis_ld_u8(dst);
pathA = *pMask++;
a0 = vis_ld_u16(a0_S32 + pathA);
a1 = vis_fpsub16(d_one, a0);
a0 = vis_fmul8x16(vis_read_lo(dd), a0);
a1 = vis_fmul8x16(srcG_f, a1);
a0 = vis_fpadd16(a0, d_round);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
#pragma pipeloop(0)
for (; dst <= (dst_end - 4); dst += 4) {
ff = *(mlib_f32*)dst;
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[3]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[2]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[1]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[0]), a0);
a1 = vis_fpsub16(d_one, a0);
a0 = vis_fmul8x16(ff, a0);
a1 = vis_fmul8x16(srcG_f, a1);
a0 = vis_fpadd16(a0, d_round);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
*(mlib_f32*)dst = ff;
pMask += 4;
}
while (dst < dst_end) {
dd = vis_ld_u8(dst);
pathA = *pMask++;
a0 = vis_ld_u16(a0_S32 + pathA);
a1 = vis_fpsub16(d_one, a0);
a0 = vis_fmul8x16(vis_read_lo(dd), a0);
a1 = vis_fmul8x16(srcG_f, a1);
a0 = vis_fpadd16(a0, d_round);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
PTR_ADD(rasBase, rasScan);
PTR_ADD(pMask, maskScan);
}
}
/***************************************************************/
static void vis_ByteGrayBlendMask2(mlib_u8 *rasBase,
mlib_u8 *pMask,
mlib_s32 rasScan,
mlib_s32 maskScan,
mlib_s32 width,
mlib_s32 height,
mlib_s32 *a0_S32,
mlib_s16 *d1_S16)
{
mlib_f32 ff;
mlib_d64 dd, a0, a1;
mlib_s32 j, pathA;
maskScan -= width;
vis_write_gsr((0 << 3) | 6);
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
mlib_u8 *dst_end;
dst_end = dst + width;
while (((mlib_s32)dst & 3) && dst < dst_end) {
dd = vis_ld_u8(dst);
pathA = *pMask++;
a0 = vis_ld_u16(a0_S32 + pathA);
a1 = vis_ld_u16(d1_S16 + pathA);
a0 = vis_fmul8x16(vis_read_lo(dd), a0);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
#pragma pipeloop(0)
for (; dst <= (dst_end - 4); dst += 4) {
ff = *(mlib_f32*)dst;
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[3]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[2]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[1]), a0);
a0 = vis_faligndata(vis_ld_u16(a0_S32 + pMask[0]), a0);
a1 = vis_faligndata(vis_ld_u16(d1_S16 + pMask[3]), a1);
a1 = vis_faligndata(vis_ld_u16(d1_S16 + pMask[2]), a1);
a1 = vis_faligndata(vis_ld_u16(d1_S16 + pMask[1]), a1);
a1 = vis_faligndata(vis_ld_u16(d1_S16 + pMask[0]), a1);
a0 = vis_fmul8x16(ff, a0);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
*(mlib_f32*)dst = ff;
pMask += 4;
}
while (dst < dst_end) {
dd = vis_ld_u8(dst);
pathA = *pMask++;
a0 = vis_ld_u16(a0_S32 + pathA);
a1 = vis_ld_u16(d1_S16 + pathA);
a0 = vis_fmul8x16(vis_read_lo(dd), a0);
a0 = vis_fpadd16(a0, a1);
ff = vis_fpack16(a0);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
PTR_ADD(rasBase, rasScan);
PTR_ADD(pMask, maskScan);
}
}
/***************************************************************/
static void vis_ByteGrayBlend(mlib_u8 *rasBase,
mlib_s32 rasScan,
mlib_s32 width,
mlib_s32 height,
mlib_f32 a0,
mlib_d64 d1)
{
mlib_f32 ff;
mlib_d64 dd;
mlib_s32 j;
vis_write_gsr((0 << 3) | 6);
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
mlib_u8 *dst_end;
dst_end = dst + width;
while (((mlib_s32)dst & 3) && dst < dst_end) {
dd = vis_ld_u8(dst);
dd = vis_fmul8x16al(vis_read_lo(dd), a0);
dd = vis_fpadd16(dd, d1);
ff = vis_fpack16(dd);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
#pragma pipeloop(0)
for (; dst <= (dst_end - 4); dst += 4) {
ff = *(mlib_f32*)dst;
dd = vis_fmul8x16al(ff, a0);
dd = vis_fpadd16(dd, d1);
ff = vis_fpack16(dd);
*(mlib_f32*)dst = ff;
}
while (dst < dst_end) {
dd = vis_ld_u8(dst);
dd = vis_fmul8x16al(vis_read_lo(dd), a0);
dd = vis_fpadd16(dd, d1);
ff = vis_fpack16(dd);
dd = D64_FROM_F32x2(ff);
vis_st_u8(dd, dst);
dst++;
}
PTR_ADD(rasBase, rasScan);
}
}
/***************************************************************/
void ADD_SUFF(ByteGraySrcMaskFill)(void *rasBase,
jubyte *pMask,
jint maskOff,
jint maskScan,
jint width,
jint height,
jint fgColor,
SurfaceDataRasInfo *pRasInfo,
NativePrimitive *pPrim,
CompositeInfo *pCompInfo)
{
mlib_s32 rasScan = pRasInfo->scanStride;
mlib_s32 r, g, b, i, j;
mlib_s32 a0_S32[256];
mlib_s32 resA, resG, dstF, pathA, srcA, srcG;
mlib_d64 dscale;
b = (fgColor) & 0xff;
g = (fgColor >> 8) & 0xff;
r = (fgColor >> 16) & 0xff;
srcA = (fgColor >> 24) & 0xff;
srcG = RGB2GRAY(r, g, b);
#ifdef LOOPS_OLD_VERSION
if (srcA == 0) return;
if (pMask == NULL) {
AnyByteSetRect(pRasInfo, 0, 0, width, height, srcG, pPrim, pCompInfo);
return;
}
#else
if (pMask == NULL) {
if (srcA == 0) srcG = 0;
ADD_SUFF(AnyByteSetRect)(pRasInfo,
pRasInfo->bounds.x1, pRasInfo->bounds.y1,
pRasInfo->bounds.x2, pRasInfo->bounds.y2,
srcG, pPrim, pCompInfo);
return;
}
#endif
pMask += maskOff;
if (width < 32) {
srcG = mul8table[srcA][srcG];
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
for (i = 0; i < width; i++) {
pathA = pMask[i];
resG = dst[i];
dstF = 0xff - pathA;
resA = dstF + mul8table[pathA][srcA];
resG = mul8table[dstF][resG] + mul8table[pathA][srcG];
resG = div8table[resA][resG];
dst[i] = resG;
}
PTR_ADD(rasBase, rasScan);
PTR_ADD(pMask, maskScan);
}
return;
}
dscale = (mlib_d64)(1 << 15)*(1 << 16);
a0_S32[0] = dscale - 1;
#pragma pipeloop(0)
for (pathA = 1; pathA < 256; pathA++) {
dstF = 0xff - pathA;
resA = dstF + mul8table[pathA][srcA];
dstF = dscale*dstF*vis_d64_div_tbl[resA];
a0_S32[pathA] = dstF;
}
vis_ByteGrayBlendMask(rasBase, pMask, rasScan, maskScan,
width, height, a0_S32, srcG);
}
/***************************************************************/
void ADD_SUFF(ByteGraySrcOverMaskFill)(void *rasBase,
jubyte *pMask,
jint maskOff,
jint maskScan,
jint width,
jint height,
jint fgColor,
SurfaceDataRasInfo *pRasInfo,
NativePrimitive *pPrim,
CompositeInfo *pCompInfo)
{
mlib_s32 rasScan = pRasInfo->scanStride;
mlib_s32 r, g, b, i, j;
mlib_s32 dstA, pathA, srcA, srcG;
b = (fgColor) & 0xff;
g = (fgColor >> 8) & 0xff;
r = (fgColor >> 16) & 0xff;
srcA = (fgColor >> 24) & 0xff;
srcG = RGB2GRAY(r, g, b);
if (srcA == 0) return;
if (pMask != NULL) pMask += maskOff;
if (width < 16) {
srcG = mul8table[srcA][srcG];
if (pMask != NULL) {
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
for (i = 0; i < width; i++) {
pathA = pMask[i];
dstA = 0xff - mul8table[pathA][srcA];
dst[i] = mul8table[dstA][dst[i]] + mul8table[pathA][srcG];
}
PTR_ADD(rasBase, rasScan);
PTR_ADD(pMask, maskScan);
}
} else {
mlib_u8 *mul8_dstA = mul8table[0xff - srcA];
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
for (i = 0; i < width; i++) {
dst[i] = mul8_dstA[dst[i]] + srcG;
}
PTR_ADD(rasBase, rasScan);
}
}
return;
}
if (pMask != NULL) {
mlib_s32 a0_S32[256];
mlib_d64 dscale = (mlib_d64)(1 << 15)*(1 << 16);
a0_S32[0] = dscale - 1;
#pragma pipeloop(0)
for (pathA = 1; pathA < 256; pathA++) {
a0_S32[pathA] = dscale - pathA*srcA*(dscale*(1.0/(255*255)));
}
vis_ByteGrayBlendMask(rasBase, pMask, rasScan, maskScan,
width, height, a0_S32, srcG);
} else {
mlib_s32 a0_int = (1 << 15)*(1.0 - srcA*(1.0/255));
mlib_f32 a0, a1, srcG_f;
mlib_d64 d1;
mlib_d64 d_round = vis_to_double_dup(((1 << 16) | 1) << 6);
srcG = (srcG << 8) | srcG;
srcG_f = vis_to_float((srcG << 16) | srcG);
a0 = vis_to_float(a0_int);
a1 = vis_to_float(0x7FFF - a0_int);
d1 = vis_fmul8x16al(srcG_f, a1);
d1 = vis_fpadd16(d1, d_round);
vis_ByteGrayBlend(rasBase, rasScan, width, height, a0, d1);
}
}
/***************************************************************/
void ADD_SUFF(ByteGrayAlphaMaskFill)(void *rasBase,
jubyte *pMask,
jint maskOff,
jint maskScan,
jint width,
jint height,
jint fgColor,
SurfaceDataRasInfo *pRasInfo,
NativePrimitive *pPrim,
CompositeInfo *pCompInfo)
{
mlib_s32 rasScan = pRasInfo->scanStride;
mlib_s32 pathA, srcA, srcG, dstA, dstFbase, srcFbase;
mlib_s32 SrcOpAnd, SrcOpXor, SrcOpAdd;
mlib_s32 DstOpAnd, DstOpXor, DstOpAdd;
mlib_s32 r, g, b;
mlib_s32 resA, resG, srcF, i, j;
b = (fgColor) & 0xff;
g = (fgColor >> 8) & 0xff;
r = (fgColor >> 16) & 0xff;
srcA = (fgColor >> 24) & 0xff;
srcG = RGB2GRAY(r, g, b);
SrcOpAnd = (AlphaRules[pCompInfo->rule].srcOps).andval;
SrcOpXor = (AlphaRules[pCompInfo->rule].srcOps).xorval;
SrcOpAdd = (AlphaRules[pCompInfo->rule].srcOps).addval - SrcOpXor;
DstOpAnd = (AlphaRules[pCompInfo->rule].dstOps).andval;
DstOpXor = (AlphaRules[pCompInfo->rule].dstOps).xorval;
DstOpAdd = (AlphaRules[pCompInfo->rule].dstOps).addval - DstOpXor;
dstFbase = ((((srcA) & DstOpAnd) ^ DstOpXor) + DstOpAdd);
srcFbase = ((((0xff) & SrcOpAnd) ^ SrcOpXor) + SrcOpAdd);
if (pMask != NULL) pMask += maskOff;
srcG = mul8table[srcA][srcG];
if (width < 100) {
if (pMask != NULL) {
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
for (i = 0; i < width; i++) {
pathA = pMask[i];
srcF = mul8table[pathA][srcFbase];
dstA = 0xff - pathA + mul8table[pathA][dstFbase];
resA = dstA + mul8table[srcF][srcA];
resG = mul8table[dstA][dst[i]] + mul8table[srcF][srcG];
dst[i] = div8table[resA][resG];
}
PTR_ADD(rasBase, rasScan);
PTR_ADD(pMask, maskScan);
}
} else {
mlib_u8 *mul8_dstA;
srcF = srcFbase;
dstA = dstFbase;
resA = dstA + mul8table[srcF][srcA];
srcG = mul8table[srcF][srcG];
mul8_dstA = mul8table[dstA];
for (j = 0; j < height; j++) {
mlib_u8 *dst = rasBase;
for (i = 0; i < width; i++) {
resG = mul8_dstA[dst[i]] + srcG;
dst[i] = div8table[resA][resG];
}
PTR_ADD(rasBase, rasScan);
}
}
return;
}
if (pMask != NULL) {
mlib_s32 a0_S32[256];
mlib_s16 d1_S16[256];
mlib_d64 dscale = (mlib_d64)(1 << 15)*(1 << 16);
a0_S32[0] = dscale - 1;
d1_S16[0] = (1 << 6);
#pragma pipeloop(0)
for (pathA = 1; pathA < 256; pathA++) {
srcF = mul8table[pathA][srcFbase];
dstA = 0xff - pathA + mul8table[pathA][dstFbase];
resA = dstA + mul8table[srcF][srcA];
a0_S32[pathA] = dscale*dstA*vis_d64_div_tbl[resA] + (1 << 15);
d1_S16[pathA] = (1 << 7)*srcG*srcF*vis_d64_div_tbl[resA] + (1 << 6);
}
vis_ByteGrayBlendMask2(rasBase, pMask, rasScan, maskScan,
width, height, a0_S32, d1_S16);
} else {
mlib_d64 dscale = (mlib_d64)(1 << 15)*(1 << 16);
mlib_s32 _a0, _d1;
mlib_f32 a0;
mlib_d64 d1;
srcF = srcFbase;
dstA = dstFbase;
resA = dstA + mul8table[srcF][srcA];
_a0 = dscale*dstA*vis_d64_div_tbl[resA] + (1 << 15);
_d1 = (1 << 7)*vis_d64_div_tbl[resA]*srcF*srcG + (1 << 6);
a0 = vis_to_float(_a0 >> 16);
d1 = vis_to_double_dup((_d1 << 16) | _d1);
vis_ByteGrayBlend(rasBase, rasScan, width, height, a0, d1);
}
}
/***************************************************************/
#define TBL_MUL ((mlib_s16*)vis_mul8s_tbl + 1)
void ADD_SUFF(ByteGrayDrawGlyphListAA)(GLYPH_LIST_PARAMS)
{
mlib_s32 glyphCounter;
mlib_s32 scan = pRasInfo->scanStride;
mlib_u8 *pPix;
mlib_s32 srcG;
int i, j, r, g, b;
mlib_d64 mix0, mix1, dd, d0, d1, e0, e1, fgpixel_d;
mlib_d64 done, d_half;
mlib_s32 pix, mask0, mask1;
mlib_f32 fgpixel_f, srcG_f;
b = (argbcolor) & 0xff;
g = (argbcolor >> 8) & 0xff;
r = (argbcolor >> 16) & 0xff;
srcG = RGB2GRAY(r, g, b);
if (clipRight - clipLeft >= 16) {
done = vis_to_double_dup(0x7fff7fff);
d_half = vis_to_double_dup((1 << (16 + 6)) | (1 << 6));
fgpixel &= 0xff;
fgpixel_f = F32_FROM_U8x4(fgpixel, fgpixel, fgpixel, fgpixel);
fgpixel_d = vis_freg_pair(fgpixel_f, fgpixel_f);
srcG_f = F32_FROM_U8x4(srcG, srcG, srcG, srcG);
vis_write_gsr((0 << 3) | 6);
}
for (glyphCounter = 0; glyphCounter < totalGlyphs; glyphCounter++) {
const jubyte *pixels;
unsigned int rowBytes;
int left, top;
int width, height;
int right, bottom;
pixels = (const jubyte *) glyphs[glyphCounter].pixels;
if (!pixels) continue;
left = glyphs[glyphCounter].x;
top = glyphs[glyphCounter].y;
width = glyphs[glyphCounter].width;
height = glyphs[glyphCounter].height;
rowBytes = width;
right = left + width;
bottom = top + height;
if (left < clipLeft) {
pixels += clipLeft - left;
left = clipLeft;
}
if (top < clipTop) {
pixels += (clipTop - top) * rowBytes;
top = clipTop;
}
if (right > clipRight) {
right = clipRight;
}
if (bottom > clipBottom) {
bottom = clipBottom;
}
if (right <= left || bottom <= top) {
continue;
}
width = right - left;
height = bottom - top;
pPix = pRasInfo->rasBase;
PTR_ADD(pPix, top * scan + left);
if (width < 16) {
for (j = 0; j < height; j++) {
for (i = 0; i < width; i++) {
jint dstG;
jint mixValSrc = pixels[i];
if (mixValSrc) {
if (mixValSrc < 255) {
jint mixValDst = 255 - mixValSrc;
dstG = pPix[i];
dstG =
mul8table[mixValDst][dstG] +
mul8table[mixValSrc][srcG];
pPix[i] = dstG;
} else {
pPix[i] = fgpixel;
}
}
}
PTR_ADD(pPix, scan);
pixels += rowBytes;
}
/**代码未完, 请加载全部代码(NowJava.com).**/