0N/A/*
2362N/A * Copyright (c) 1998, 2003, Oracle and/or its affiliates. All rights reserved.
0N/A * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
0N/A *
0N/A * This code is free software; you can redistribute it and/or modify it
0N/A * under the terms of the GNU General Public License version 2 only, as
2362N/A * published by the Free Software Foundation. Oracle designates this
0N/A * particular file as subject to the "Classpath" exception as provided
2362N/A * by Oracle in the LICENSE file that accompanied this code.
0N/A *
0N/A * This code is distributed in the hope that it will be useful, but WITHOUT
0N/A * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
0N/A * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
0N/A * version 2 for more details (a copy is included in the LICENSE file that
0N/A * accompanied this code).
0N/A *
0N/A * You should have received a copy of the GNU General Public License version
0N/A * 2 along with this work; if not, write to the Free Software Foundation,
0N/A * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
0N/A *
2362N/A * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
2362N/A * or visit www.oracle.com if you need additional information or have any
2362N/A * questions.
0N/A */
0N/A
0N/A
0N/A
0N/A/*
0N/A * FUNCTIONS
0N/A * mlib_ImageConvCopyEdge - Copy src edges to dst edges
0N/A *
0N/A *
0N/A * SYNOPSIS
0N/A * mlib_status mlib_ImageConvCopyEdge(mlib_image *dst,
0N/A * const mlib_image *src,
0N/A * mlib_s32 dx_l,
0N/A * mlib_s32 dx_r,
0N/A * mlib_s32 dy_t,
0N/A * mlib_s32 dy_b,
0N/A * mlib_s32 cmask)
0N/A *
0N/A * ARGUMENT
0N/A * dst Pointer to an dst image.
0N/A * src Pointer to an src image.
0N/A * dx_l Number of columns on the left side of the
0N/A * image to be copyed.
0N/A * dx_r Number of columns on the right side of the
0N/A * image to be copyed.
0N/A * dy_t Number of rows on the top edge of the
0N/A * image to be copyed.
0N/A * dy_b Number of rows on the top edge of the
0N/A * image to be copyed.
0N/A * cmask Channel mask to indicate the channels to be convolved.
0N/A * Each bit of which represents a channel in the image. The
0N/A * channels corresponded to 1 bits are those to be processed.
0N/A *
0N/A * RESTRICTION
0N/A * The src and the dst must be the same type, same width, same height and have same number
0N/A * of channels (1, 2, 3, or 4). The unselected channels are not
0N/A * overwritten. If both src and dst have just one channel,
0N/A * cmask is ignored.
0N/A *
0N/A * DESCRIPTION
0N/A * Copy src edges to dst edges.
0N/A
0N/A * The unselected channels are not overwritten.
0N/A * If src and dst have just one channel,
0N/A * cmask is ignored.
0N/A */
0N/A
0N/A#include "vis_proto.h"
0N/A#include "mlib_image.h"
0N/A#include "mlib_ImageConvEdge.h"
0N/A
0N/A/***************************************************************/
0N/Astatic void mlib_ImageConvCopyEdge_U8(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_U8_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_S16(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_S16_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_S32(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_S32_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask);
0N/A
0N/Astatic void mlib_ImageConvCopyEdge_S32_4(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask);
0N/A
0N/A/***************************************************************/
0N/A#define VERT_EDGES(chan, type, mask) \
0N/A type *pdst = (type *) mlib_ImageGetData(dst); \
0N/A type *psrc = (type *) mlib_ImageGetData(src); \
0N/A type *pdst_row, *psrc_row, *pdst_row_end; \
0N/A mlib_s32 img_height = mlib_ImageGetHeight(dst); \
0N/A mlib_s32 img_width = mlib_ImageGetWidth(dst); \
0N/A mlib_s32 dst_stride = mlib_ImageGetStride(dst) / sizeof(type); \
0N/A mlib_s32 src_stride = mlib_ImageGetStride(src) / sizeof(type); \
0N/A mlib_s32 i, j, l; \
0N/A mlib_s32 emask, testchan; \
0N/A mlib_s32 img_width_t, img_width_b; \
0N/A mlib_d64 *dpdst, *dpsrc, data0, data1; \
0N/A \
0N/A testchan = 1; \
0N/A for (l = chan - 1; l >= 0; l--) { \
0N/A if ((mask & testchan) == 0) { \
0N/A testchan <<= 1; \
0N/A continue; \
0N/A } \
0N/A testchan <<= 1; \
0N/A for (j = 0; j < dx_l; j++) { \
0N/A for (i = dy_t; i < (img_height - dy_b); i++) { \
0N/A pdst[i*dst_stride + l + j*chan] = \
0N/A psrc[i*src_stride + l + j*chan]; \
0N/A } \
0N/A } \
0N/A for (j = 0; j < dx_r; j++) { \
0N/A for (i = dy_t; i < (img_height - dy_b); i++) { \
0N/A pdst[i*dst_stride + l+(img_width-1 - j)*chan] = \
0N/A psrc[i*src_stride + l+(img_width-1 - j)*chan]; \
0N/A } \
0N/A } \
0N/A } \
0N/A img_width_t = img_width; \
0N/A img_width_b = img_width; \
0N/A if (((img_width * chan) == dst_stride) && \
0N/A ((img_width * chan) == src_stride)) { \
0N/A img_width_t *= dy_t; \
0N/A img_width_b *= dy_b; \
0N/A dst_stride *= (img_height - dy_b); \
0N/A src_stride *= (img_height - dy_b); \
0N/A img_height = 2; \
0N/A dy_t = ((dy_t == 0) ? 0 : 1); \
0N/A dy_b = ((dy_b == 0) ? 0 : 1); \
0N/A }
0N/A
0N/A/***************************************************************/
0N/A#define HORIZ_EDGES(chan, type, mask) { \
0N/A testchan = 1; \
0N/A for (l = chan - 1; l >= 0; l--) { \
0N/A if ((mask & testchan) == 0) { \
0N/A testchan <<= 1; \
0N/A continue; \
0N/A } \
0N/A testchan <<= 1; \
0N/A for (i = 0; i < dy_t; i++) { \
0N/A for (j = 0; j < img_width_t; j++) { \
0N/A pdst[i*dst_stride + l + j*chan] = \
0N/A psrc[i*src_stride + l + j*chan]; \
0N/A } \
0N/A } \
0N/A for (i = 0; i < dy_b; i++) { \
0N/A for (j = 0; j < img_width_b; j++) { \
0N/A pdst[(img_height-1 - i)*dst_stride + l + j*chan] = \
0N/A psrc[(img_height-1 - i)*src_stride + l + j*chan]; \
0N/A } \
0N/A } \
0N/A } \
0N/A return; \
0N/A }
0N/A
0N/A/***************************************************************/
0N/Amlib_status mlib_ImageConvCopyEdge(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask)
0N/A{
0N/A mlib_s32 img_width = mlib_ImageGetWidth(dst);
0N/A mlib_s32 img_height = mlib_ImageGetHeight(dst);
0N/A
0N/A if (dx_l + dx_r > img_width) {
0N/A dx_l = img_width;
0N/A dx_r = 0;
0N/A }
0N/A
0N/A if (dy_t + dy_b > img_height) {
0N/A dy_t = img_height;
0N/A dy_b = 0;
0N/A }
0N/A
0N/A switch (mlib_ImageGetType(dst)) {
0N/A case MLIB_BIT:
0N/A return mlib_ImageConvCopyEdge_Bit(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A
0N/A case MLIB_BYTE:
0N/A switch (mlib_ImageGetChannels(dst)) {
0N/A
0N/A case 1:
0N/A mlib_ImageConvCopyEdge_U8(dst, src, dx_l, dx_r, dy_t, dy_b, 1, 1);
0N/A break;
0N/A
0N/A case 2:
0N/A mlib_ImageConvCopyEdge_U8(dst, src, dx_l, dx_r, dy_t, dy_b, cmask, 2);
0N/A break;
0N/A
0N/A case 3:
0N/A mlib_ImageConvCopyEdge_U8_3(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A break;
0N/A
0N/A case 4:
0N/A mlib_ImageConvCopyEdge_U8(dst, src, dx_l, dx_r, dy_t, dy_b, cmask, 4);
0N/A break;
0N/A
0N/A default:
0N/A return MLIB_FAILURE;
0N/A }
0N/A
0N/A break;
0N/A
0N/A case MLIB_SHORT:
0N/A case MLIB_USHORT:
0N/A switch (mlib_ImageGetChannels(dst)) {
0N/A
0N/A case 1:
0N/A mlib_ImageConvCopyEdge_S16(dst, src, dx_l, dx_r, dy_t, dy_b, 1, 1);
0N/A break;
0N/A
0N/A case 2:
0N/A mlib_ImageConvCopyEdge_S16(dst, src, dx_l, dx_r, dy_t, dy_b, cmask, 2);
0N/A break;
0N/A
0N/A case 3:
0N/A mlib_ImageConvCopyEdge_S16_3(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A break;
0N/A
0N/A case 4:
0N/A mlib_ImageConvCopyEdge_S16(dst, src, dx_l, dx_r, dy_t, dy_b, cmask, 4);
0N/A break;
0N/A
0N/A default:
0N/A return MLIB_FAILURE;
0N/A }
0N/A
0N/A break;
0N/A
0N/A case MLIB_INT:
0N/A case MLIB_FLOAT:
0N/A switch (mlib_ImageGetChannels(dst)) {
0N/A
0N/A case 1:
0N/A mlib_ImageConvCopyEdge_S32(dst, src, dx_l, dx_r, dy_t, dy_b, 1, 1);
0N/A break;
0N/A
0N/A case 2:
0N/A mlib_ImageConvCopyEdge_S32(dst, src, dx_l, dx_r, dy_t, dy_b, cmask, 2);
0N/A break;
0N/A
0N/A case 3:
0N/A mlib_ImageConvCopyEdge_S32_3(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A break;
0N/A
0N/A case 4:
0N/A mlib_ImageConvCopyEdge_S32_4(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A break;
0N/A
0N/A default:
0N/A return MLIB_FAILURE;
0N/A }
0N/A
0N/A break;
0N/A
0N/A case MLIB_DOUBLE:
0N/A return mlib_ImageConvCopyEdge_Fp(dst, src, dx_l, dx_r, dy_t, dy_b, cmask);
0N/A
0N/A default:
0N/A return MLIB_FAILURE;
0N/A }
0N/A
0N/A return MLIB_SUCCESS;
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_U8(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan)
0N/A{
0N/A mlib_s32 tmask = cmask & ((1 << nchan) - 1), mask1, offset;
0N/A VERT_EDGES(nchan, mlib_u8, cmask);
0N/A
0N/A if (img_width < 16 / nchan)
0N/A HORIZ_EDGES(nchan, mlib_u8, cmask);
0N/A
0N/A if (nchan == 1)
0N/A tmask = 0xFFFF;
0N/A else if (nchan == 2) {
0N/A tmask |= (tmask << 2);
0N/A tmask |= (tmask << 4);
0N/A tmask |= (tmask << 8);
0N/A }
0N/A else if (nchan == 4) {
0N/A tmask |= (tmask << 4);
0N/A tmask |= (tmask << 8);
0N/A }
0N/A
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_u8 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -offset);
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_u8 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_t * nchan - 8); j += 8) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_u8 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -offset);
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_u8 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_b * nchan - 8); j += 8) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_U8_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask)
0N/A{
0N/A mlib_s32 tmask = cmask & 7, mask0, mask1, mask2, offset;
0N/A
0N/A VERT_EDGES(3, mlib_u8, cmask);
0N/A
0N/A if (img_width < 16)
0N/A HORIZ_EDGES(3, mlib_u8, cmask);
0N/A
0N/A tmask |= (tmask << 3);
0N/A tmask |= (tmask << 6);
0N/A tmask |= (tmask << 12);
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * 3 - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_u8 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -offset);
0N/A mask2 = (tmask >> (offset + 1));
0N/A mask0 = mask2 >> 1;
0N/A mask1 = mask0 >> 1;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_u8 *) dpdst - pdst_row);
0N/A for (; j < (img_width_t * 3 - 24); j += 24) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_t * 3 - 8)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_t * 3 - 16)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * 3 - 1;
0N/A
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_u8 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -offset);
0N/A mask2 = (tmask >> (offset + 1));
0N/A mask0 = mask2 >> 1;
0N/A mask1 = mask0 >> 1;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_u8 *) dpdst - pdst_row);
0N/A for (; j < (img_width_b * 3 - 24); j += 24) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_b * 3 - 8)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_b * 3 - 16)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge8(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_8(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_S16(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan)
0N/A{
0N/A mlib_s32 tmask = cmask & ((1 << nchan) - 1), mask1, offset;
0N/A VERT_EDGES(nchan, mlib_s16, cmask);
0N/A
0N/A if (img_width < 16 / nchan)
0N/A HORIZ_EDGES(nchan, mlib_s16, cmask);
0N/A
0N/A if (nchan == 1)
0N/A tmask = 0xFFFF;
0N/A else if (nchan == 2) {
0N/A tmask |= (tmask << 2);
0N/A tmask |= (tmask << 4);
0N/A }
0N/A else if (nchan == 4)
0N/A tmask |= (tmask << 4);
0N/A
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s16 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 1));
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_s16 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_t * nchan - 4); j += 4) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s16 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 1));
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_s16 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_b * nchan - 4); j += 4) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_S16_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask)
0N/A{
0N/A mlib_s32 tmask = cmask & 7, mask0, mask1, mask2, offset;
0N/A
0N/A VERT_EDGES(3, mlib_s16, cmask);
0N/A
0N/A if (img_width < 16)
0N/A HORIZ_EDGES(3, mlib_s16, cmask);
0N/A
0N/A tmask |= (tmask << 3);
0N/A tmask |= (tmask << 6);
0N/A tmask |= (tmask << 12);
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * 3 - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s16 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 1));
0N/A mask2 = (tmask >> (offset + 2));
0N/A mask0 = mask2 >> 2;
0N/A mask1 = mask0 >> 2;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s16 *) dpdst - pdst_row);
0N/A for (; j < (img_width_t * 3 - 12); j += 12) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_t * 3 - 4)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_t * 3 - 8)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * 3 - 1;
0N/A
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s16 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 1));
0N/A mask2 = (tmask >> (offset + 2));
0N/A mask0 = mask2 >> 2;
0N/A mask1 = mask0 >> 2;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s16 *) dpdst - pdst_row);
0N/A for (; j < (img_width_b * 3 - 12); j += 12) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_b * 3 - 4)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_b * 3 - 8)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge16(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_16(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_S32(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask,
0N/A mlib_s32 nchan)
0N/A{
0N/A mlib_s32 tmask = cmask & ((1 << nchan) - 1), mask1, offset;
0N/A VERT_EDGES(nchan, mlib_s32, cmask);
0N/A
0N/A if (img_width < 16 / nchan)
0N/A HORIZ_EDGES(nchan, mlib_s32, cmask);
0N/A
0N/A if (nchan == 1)
0N/A tmask = 0xFFFF;
0N/A else if (nchan == 2) {
0N/A tmask |= (tmask << 2);
0N/A tmask |= (tmask << 4);
0N/A }
0N/A
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_t * nchan - 2); j += 2) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * nchan - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask1 = (tmask >> offset);
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A data0 = data1;
0N/A for (; j < (img_width_b * nchan - 2); j += 2) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_S32_3(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask)
0N/A{
0N/A mlib_s32 tmask = cmask & 7, mask0, mask1, mask2, offset;
0N/A
0N/A VERT_EDGES(3, mlib_s32, cmask);
0N/A
0N/A if (img_width < 16)
0N/A HORIZ_EDGES(3, mlib_s32, cmask);
0N/A
0N/A tmask |= (tmask << 3);
0N/A tmask |= (tmask << 6);
0N/A tmask |= (tmask << 12);
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * 3 - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask2 = (tmask >> (offset + 1));
0N/A mask0 = mask2 >> 1;
0N/A mask1 = mask0 >> 1;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A for (; j < (img_width_t * 3 - 6); j += 6) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_t * 3 - 2)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_t * 3 - 4)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * 3 - 1;
0N/A
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask2 = (tmask >> (offset + 1));
0N/A mask0 = mask2 >> 1;
0N/A mask1 = mask0 >> 1;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask2;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A for (; j < (img_width_b * 3 - 6); j += 6) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst + 1, mask1);
0N/A data0 = data1;
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst + 2, mask2);
0N/A data0 = data1;
0N/A dpdst += 3;
0N/A }
0N/A
0N/A if (j < (img_width_b * 3 - 2)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A
0N/A if (j < (img_width_b * 3 - 4)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask1);
0N/A data0 = data1;
0N/A mask0 = mask2;
0N/A }
0N/A else {
0N/A mask0 = mask1;
0N/A }
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/
0N/Avoid mlib_ImageConvCopyEdge_S32_4(mlib_image *dst,
0N/A const mlib_image *src,
0N/A mlib_s32 dx_l,
0N/A mlib_s32 dx_r,
0N/A mlib_s32 dy_t,
0N/A mlib_s32 dy_b,
0N/A mlib_s32 cmask)
0N/A{
0N/A mlib_s32 tmask = cmask & 15, mask0, mask1, offset;
0N/A
0N/A VERT_EDGES(4, mlib_s32, cmask);
0N/A
0N/A if (img_width < 16)
0N/A HORIZ_EDGES(4, mlib_s32, cmask);
0N/A
0N/A tmask |= (tmask << 4);
0N/A tmask |= (tmask << 8);
0N/A for (i = 0; i < dy_t; i++) {
0N/A pdst_row = pdst + i * dst_stride,
0N/A psrc_row = psrc + i * src_stride, pdst_row_end = pdst_row + img_width_t * 4 - 1;
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask1 = (tmask >> (offset + 2));
0N/A mask0 = mask1 >> 2;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A for (; j < (img_width_t * 4 - 4); j += 4) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data1, data0), dpdst + 1, mask1);
0N/A dpdst += 2;
0N/A }
0N/A
0N/A if (j < (img_width_t * 4 - 2)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A mask0 = mask1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A
0N/A for (i = 0; i < dy_b; i++) {
0N/A pdst_row = pdst + (img_height - 1 - i) * dst_stride;
0N/A psrc_row = psrc + (img_height - 1 - i) * src_stride;
0N/A pdst_row_end = pdst_row + img_width_b * 4 - 1;
0N/A
0N/A dpdst = (mlib_d64 *) ((mlib_addr) pdst_row & ~7);
0N/A offset = pdst_row - (mlib_s32 *) dpdst;
0N/A dpsrc = (mlib_d64 *) vis_alignaddr(psrc_row, -(offset << 2));
0N/A mask1 = (tmask >> (offset + 2));
0N/A mask0 = mask1 >> 2;
0N/A data0 = *dpsrc++;
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(pdst_row, pdst_row_end) & mask1;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, emask);
0N/A data0 = data1;
0N/A j = (mlib_s32) ((mlib_s32 *) dpdst - pdst_row);
0N/A for (; j < (img_width_b * 4 - 4); j += 4) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, mask0);
0N/A data0 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data1, data0), dpdst + 1, mask1);
0N/A dpdst += 2;
0N/A }
0N/A
0N/A if (j < (img_width_b * 4 - 2)) {
0N/A data1 = *dpsrc++;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst++, mask0);
0N/A data0 = data1;
0N/A mask0 = mask1;
0N/A }
0N/A
0N/A data1 = *dpsrc++;
0N/A emask = vis_edge32(dpdst, pdst_row_end) & mask0;
0N/A vis_pst_32(vis_faligndata(data0, data1), dpdst, emask);
0N/A }
0N/A}
0N/A
0N/A/***************************************************************/