/home/daniel/Projects/java/jdk/src/java.desktop/share/native/libmlib_image/mlib_ImageConv

Bug Summary

File:	jdk/src/java.desktop/share/native/libmlib_image/mlib_ImageConv_8ext.c
Warning:	line 615, column 16 Assigned value is garbage or undefined
Annotated Source Code

Press '?' to see keyboard shortcuts
Show analyzer invocation
clang -cc1 -triple x86_64-pc-linux-gnu -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name mlib_ImageConv_8ext.c -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -mrelocation-model pic -pic-level 2 -mthread-model posix -fno-delete-null-pointer-checks -mframe-pointer=all -relaxed-aliasing -fmath-errno -fno-rounding-math -masm-verbose -mconstructor-aliases -munwind-tables -target-cpu x86-64 -dwarf-column-info -fno-split-dwarf-inlining -debugger-tuning=gdb -resource-dir /usr/lib/llvm-10/lib/clang/10.0.0 -I /home/daniel/Projects/java/jdk/build/linux-x86_64-server-fastdebug/support/modules_include/java.base -I /home/daniel/Projects/java/jdk/build/linux-x86_64-server-fastdebug/support/modules_include/java.base/linux -I /home/daniel/Projects/java/jdk/src/java.base/share/native/libjava -I /home/daniel/Projects/java/jdk/src/java.base/unix/native/libjava -I /home/daniel/Projects/java/jdk/src/hotspot/share/include -I /home/daniel/Projects/java/jdk/src/hotspot/os/posix/include -D LIBC=gnu -D _GNU_SOURCE -D _REENTRANT -D _LARGEFILE64_SOURCE -D LINUX -D DEBUG -D _LITTLE_ENDIAN -D ARCH="amd64" -D amd64 -D _LP64=1 -D __USE_J2D_NAMES -D __MEDIALIB_OLD_NAMES -D MLIB_NO_LIBSUNMATH -D MLIB_OS64BIT -I /home/daniel/Projects/java/jdk/src/java.desktop/share/native/libmlib_image -I /home/daniel/Projects/java/jdk/src/java.desktop/share/native/common/awt/medialib -I /home/daniel/Projects/java/jdk/build/linux-x86_64-server-fastdebug/support/headers/java.desktop -D _FORTIFY_SOURCE=2 -internal-isystem /usr/local/include -internal-isystem /usr/lib/llvm-10/lib/clang/10.0.0/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -Wno-unused-parameter -Wno-unused -Wno-unused-function -std=c99 -fdebug-compilation-dir /home/daniel/Projects/java/jdk/make -ferror-limit 19 -fmessage-length 0 -fvisibility hidden -stack-protector 1 -fgnuc-version=4.2.1 -fobjc-runtime=gcc -fdiagnostics-show-option -vectorize-loops -vectorize-slp -analyzer-output=html -faddrsig -o /home/daniel/Projects/java/scan/2021-12-21-193737-8510-1 -x c /home/daniel/Projects/java/jdk/src/java.desktop/share/native/libmlib_image/mlib_ImageConv_8ext.c
1/*
* Copyright (c) 2003, 2020, 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.
*/


27/*
* FUNCTION
*   Internal functions for mlib_ImageConv* on U8/S16/U16 type and
*   MLIB_EDGE_SRC_EXTEND mask
*/

33#include "mlib_image.h"
34#include "mlib_ImageConv.h"
35#include "mlib_c_ImageConv.h"

37/*
* This define switches between functions of different data types
*/

41#define IMG_TYPE1 1

43/***************************************************************/
44#if IMG_TYPE1 == 1

46#define DTYPEmlib_u8             mlib_u8
47#define CONV_FUNC(KERN)mlib_c_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask)   mlib_c_conv##KERN##ext_u8(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
48#define CONV_FUNC_MxNmlib_c_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)     mlib_c_convMxNext_u8(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
49#define CONV_FUNC_I(KERN)mlib_i_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask) mlib_i_conv##KERN##ext_u8(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
50#define CONV_FUNC_MxN_Imlib_i_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)   mlib_i_convMxNext_u8(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
51#define DSCALE(1 << 24)            (1 << 24)
52#define FROM_S32(x)(((x) >> 24) ^ 128)       (((x) >> 24) ^ 128)
53#define S64TOS32(x)(x)       (x)
54#define SAT_OFF-(1u << 31)           -(1u << 31)

56#elif IMG_TYPE1 == 2

58#define DTYPEmlib_u8             mlib_s16
59#define CONV_FUNC(KERN)mlib_c_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask)   mlib_conv##KERN##ext_s16(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
60#define CONV_FUNC_MxNmlib_c_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)     mlib_convMxNext_s16(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
61#define CONV_FUNC_I(KERN)mlib_i_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask) mlib_i_conv##KERN##ext_s16(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
62#define CONV_FUNC_MxN_Imlib_i_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)   mlib_i_convMxNext_s16(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
63#define DSCALE(1 << 24)            65536.0
64#define FROM_S32(x)(((x) >> 24) ^ 128)       ((x) >> 16)
65#define S64TOS32(x)(x)       ((x) & 0xffffffff)
66#define SAT_OFF-(1u << 31)

68#elif IMG_TYPE1 == 3

70#define DTYPEmlib_u8             mlib_u16
71#define CONV_FUNC(KERN)mlib_c_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask)   mlib_conv##KERN##ext_u16(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
72#define CONV_FUNC_MxNmlib_c_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)     mlib_convMxNext_u16(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
73#define CONV_FUNC_I(KERN)mlib_i_convKERNext_u8(mlib_image *dst, const mlib_image *src,
 mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const
 mlib_s32 *kern, mlib_s32 scalef_expon, mlib_s32 cmask) mlib_i_conv##KERN##ext_u16(PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask)
74#define CONV_FUNC_MxN_Imlib_i_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)   mlib_i_convMxNext_u16(PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask)
75#define DSCALE(1 << 24)            65536.0
76#define FROM_S32(x)(((x) >> 24) ^ 128)       (((x) >> 16) ^ 0x8000)
77#define S64TOS32(x)(x)       (x)
78#define SAT_OFF-(1u << 31)           -(1u << 31)

80#endif /* IMG_TYPE == 1 */

82/***************************************************************/
83#define PARAMmlib_image *dst, const mlib_image *src, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, const mlib_s32 *kern, mlib_s32
 scalef_expon, mlib_s32 cmask                                                   \
mlib_image       *dst,                                        \
const mlib_image *src,                                        \
mlib_s32         dx_l,                                        \
mlib_s32         dx_r,                                        \
mlib_s32         dy_t,                                        \
mlib_s32         dy_b,                                        \
const mlib_s32   *kern,                                       \
mlib_s32         scalef_expon,                                \
mlib_s32         cmask

94/***************************************************************/
95#define PARAM_MxNmlib_image *dst, const mlib_image *src, const mlib_s32 *kernel
, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32 dx_r, mlib_s32
 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32 cmask                                               \
mlib_image       *dst,                                        \
const mlib_image *src,                                        \
const mlib_s32   *kernel,                                     \
mlib_s32         m,                                           \
mlib_s32         n,                                           \
mlib_s32         dx_l,                                        \
mlib_s32         dx_r,                                        \
mlib_s32         dy_t,                                        \
mlib_s32         dy_b,                                        \
mlib_s32         scale,                                       \
mlib_s32         cmask

108/***************************************************************/
109#define FTYPEmlib_d64 mlib_d64

111#ifndef MLIB_USE_FTOI_CLAMPING

113#define CLAMP_S32(x)(((x) <= (-2147483647 -1)) ? (-2147483647 -1) : (((x) >=
 2147483647) ? 2147483647 : (mlib_s32)(x)))                                            \
(((x) <= MLIB_S32_MIN(-2147483647 -1)) ? MLIB_S32_MIN(-2147483647 -1) : (((x) >= MLIB_S32_MAX2147483647) ? MLIB_S32_MAX2147483647 : (mlib_s32)(x)))

116#else

118#define CLAMP_S32(x)(((x) <= (-2147483647 -1)) ? (-2147483647 -1) : (((x) >=
 2147483647) ? 2147483647 : (mlib_s32)(x))) ((mlib_s32)(x))

120#endif /* MLIB_USE_FTOI_CLAMPING */

122/***************************************************************/
123#define D2I(x)((((x) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((x) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((x) -(1u << 31)))) CLAMP_S32((x) SAT_OFF)((((x) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((x) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((x) -(1u << 31))))

125/***************************************************************/
126#ifdef _NO_LONGLONG

128#define LOAD_BUFF(buff)*(mlib_s64*)(buff + i) = (((mlib_s64)sp[chan1]) << 32) |
 ((mlib_s64)sp[0])                                         \
buff[i    ] = sp[0];                                          \
buff[i + 1] = sp[chan1]

132#else /* _NO_LONGLONG */

134#ifdef _LITTLE_ENDIAN1

136#define LOAD_BUFF(buff)*(mlib_s64*)(buff + i) = (((mlib_s64)sp[chan1]) << 32) |
 ((mlib_s64)sp[0])                                         \
*(mlib_s64*)(buff + i) = (((mlib_s64)sp[chan1]) << 32) | S64TOS32((mlib_s64)sp[0])((mlib_s64)sp[0])

139#else /* _LITTLE_ENDIAN */

141#define LOAD_BUFF(buff)*(mlib_s64*)(buff + i) = (((mlib_s64)sp[chan1]) << 32) |
 ((mlib_s64)sp[0])                                         \
*(mlib_s64*)(buff + i) = (((mlib_s64)sp[0]) << 32) | S64TOS32((mlib_s64)sp[chan1])((mlib_s64)sp[chan1])

144#endif /* _LITTLE_ENDIAN */
145#endif /* _NO_LONGLONG */

147/***************************************************************/
148typedef union {
mlib_d64 d64;
struct {
  mlib_s32 i0;
  mlib_s32 i1;
} i32s;
154} d64_2x32;

156/***************************************************************/
157#define GET_SRC_DST_PARAMETERS(type)hgt = mlib_ImageGetHeight(src); wid = mlib_ImageGetWidth(src)
; nchannel = mlib_ImageGetChannels(src); sll = mlib_ImageGetStride
(src) / sizeof(type); dll = mlib_ImageGetStride(dst) / sizeof
(type); adr_src = (type *)mlib_ImageGetData(src); adr_dst = (
type *)mlib_ImageGetData(dst)                            \
hgt = mlib_ImageGetHeight(src);                               \
wid = mlib_ImageGetWidth(src);                                \
nchannel = mlib_ImageGetChannels(src);                        \
sll = mlib_ImageGetStride(src) / sizeof(type);                \
dll = mlib_ImageGetStride(dst) / sizeof(type);                \
adr_src = (type *)mlib_ImageGetData(src);                     \
adr_dst = (type *)mlib_ImageGetData(dst)

166/***************************************************************/
167#if IMG_TYPE1 == 1

169/*
* Test for the presence of any "1" bit in bits
 8 to 31 of val. If present, then val is either
 negative or >255. If over/underflows of 8 bits
 are uncommon, then this technique can be a win,
 since only a single test, rather than two, is
 necessary to determine if clamping is needed.
 On the other hand, if over/underflows are common,
 it adds an extra test.
178*/
179#define CLAMP_STORE(dst, val)if (val & 0xffffff00) { if (val < 0) dst = 0; else dst
 = (127*2 +1); } else { dst = (mlib_u8)val; }                                   \
if (val & 0xffffff00) {                                       \
  if (val < MLIB_U8_MIN0)                                      \
    dst = MLIB_U8_MIN0;                                        \
  else                                                        \
    dst = MLIB_U8_MAX(127*2 +1);                                        \
} else {                                                      \
  dst = (mlib_u8)val;                                         \
}

189#elif IMG_TYPE1 == 2

191#define CLAMP_STORE(dst, val)if (val & 0xffffff00) { if (val < 0) dst = 0; else dst
 = (127*2 +1); } else { dst = (mlib_u8)val; }                                   \
if (val >= MLIB_S16_MAX32767)                                      \
  dst = MLIB_S16_MAX32767;                                         \
else if (val <= MLIB_S16_MIN(-32767 -1))                                 \
  dst = MLIB_S16_MIN(-32767 -1);                                         \
else                                                          \
  dst = (mlib_s16)val

199#elif IMG_TYPE1 == 3

201#define CLAMP_STORE(dst, val)if (val & 0xffffff00) { if (val < 0) dst = 0; else dst
 = (127*2 +1); } else { dst = (mlib_u8)val; }                                   \
if (val >= MLIB_U16_MAX(32767 *2 +1))                                      \
  dst = MLIB_U16_MAX(32767 *2 +1);                                         \
else if (val <= MLIB_U16_MIN0)                                 \
  dst = MLIB_U16_MIN0;                                         \
else                                                          \
  dst = (mlib_u16)val

209#endif /* IMG_TYPE == 1 */

211/***************************************************************/
212#define MAX_KER7   7
213#define MAX_N15    15
214#define BUFF_SIZE1600   1600
215#define CACHE_SIZE(64*1024)  (64*1024)

217static mlib_status mlib_ImageConv1xN_ext(mlib_image       *dst,
                                       const mlib_image *src,
                                       const mlib_d64   *k,
                                       mlib_s32         n,
                                       mlib_s32         dy_t,
                                       mlib_s32         dy_b,
                                       mlib_s32         cmask)
224{
DTYPEmlib_u8    *adr_src, *sl;
DTYPEmlib_u8    *adr_dst, *dl, *dp;
FTYPEmlib_d64    buff[BUFF_SIZE1600];
FTYPEmlib_d64    *buffd;
FTYPEmlib_d64    *pbuff = buff;
const FTYPEmlib_d64    *pk;
FTYPEmlib_d64    k0, k1, k2, k3;
FTYPEmlib_d64    p0, p1, p2, p3, p4;
FTYPEmlib_d64    *sbuff;
mlib_s32 l, k_off, off, bsize;
mlib_s32 max_hsize, smax_hsize, shgt, hsize, kh;
mlib_s32 d0, d1, ii;
mlib_s32 wid, hgt, sll, dll;
mlib_s32 nchannel;
mlib_s32 i, j, c;
GET_SRC_DST_PARAMETERS(DTYPE)hgt = mlib_ImageGetHeight(src); wid = mlib_ImageGetWidth(src)
; nchannel = mlib_ImageGetChannels(src); sll = mlib_ImageGetStride
(src) / sizeof(mlib_u8); dll = mlib_ImageGetStride(dst) / sizeof
(mlib_u8); adr_src = (mlib_u8 *)mlib_ImageGetData(src); adr_dst
 = (mlib_u8 *)mlib_ImageGetData(dst);

max_hsize = ((CACHE_SIZE(64*1024)/sizeof(DTYPEmlib_u8))/sll) - (n - 1);

if (max_hsize < 1) max_hsize = 1;
if (max_hsize > hgt) max_hsize = hgt;

shgt = hgt + (n - 1);
smax_hsize = max_hsize + (n - 1);

bsize = 2 * (smax_hsize + 1);

if (bsize > BUFF_SIZE1600) {
  pbuff = mlib_malloc(sizeof(FTYPEmlib_d64)*bsize);

  if (pbuff == NULL((void*)0)) return MLIB_FAILURE;
}

sbuff = pbuff;
buffd = sbuff + smax_hsize;

shgt -= (dy_t + dy_b);
k_off = 0;

for (l = 0; l < hgt; l += hsize) {
  hsize = hgt - l;

  if (hsize > max_hsize) hsize = max_hsize;

  smax_hsize = hsize + (n - 1);

  for (c = 0; c < nchannel; c++) {
    if (!(cmask & (1 << (nchannel - 1 - c)))) continue;

    sl = adr_src + c;
    dl = adr_dst + c;

    for (i = 0; i < hsize; i++) buffd[i] = 0.0;

    for (j = 0; j < wid; j++) {
      FTYPEmlib_d64    *buff = sbuff;

      for (i = k_off, ii = 0; (i < dy_t) && (ii < smax_hsize); i++, ii++) {
        sbuff[i - k_off] = (FTYPEmlib_d64)sl[0];
      }

      for (; (i < shgt + dy_t) && (ii < smax_hsize); i++, ii++) {
        sbuff[i - k_off] = (FTYPEmlib_d64)sl[(i - dy_t)*sll];
      }

      for (; (i < shgt + dy_t + dy_b) && (ii < smax_hsize); i++, ii++) {
        sbuff[i - k_off] = (FTYPEmlib_d64)sl[(shgt - 1)*sll];
      }

      pk = k;

      for (off = 0; off < (n - 4); off += 4) {

        p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
        k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];

        for (i = 0; i < hsize; i += 2) {
          p0 = p2; p1 = p3; p2 = p4;

          p3 = buff[i + 3]; p4 = buff[i + 4];

          buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3;
          buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3;
        }

        pk += 4;
        buff += 4;
      }

      dp = dl;
      kh = n - off;

      if (kh == 4) {
        p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
        k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];

        for (i = 0; i <= (hsize - 2); i += 2) {
          p0 = p2; p1 = p3; p2 = p4;

          p3 = buff[i + 3]; p4 = buff[i + 4];

          d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i ]) -(1u << 31
)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1
*k1 + p2*k2 + p3*k3 + buffd[i ]) -(1u << 31)) >= 2147483647
) ? 2147483647 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd
[i ]) -(1u << 31))));
          d1 = D2I(p1*k0 + p2*k1 + p3*k2 + p4*k3 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + buffd[i + 1]) -(1u <<
 31)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 +
 p2*k1 + p3*k2 + p4*k3 + buffd[i + 1]) -(1u << 31)) >=
 2147483647) ? 2147483647 : (mlib_s32)((p1*k0 + p2*k1 + p3*k2
 + p4*k3 + buffd[i + 1]) -(1u << 31))));

          dp[0  ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
          dp[dll] = FROM_S32(d1)(((d1) >> 24) ^ 128);

          buffd[i    ] = 0.0;
          buffd[i + 1] = 0.0;

          dp += 2*dll;
        }

        if (i < hsize) {
          p0 = p2; p1 = p3; p2 = p4;
          p3 = buff[i + 3];
          d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i]) -(1u << 31
)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1
*k1 + p2*k2 + p3*k3 + buffd[i]) -(1u << 31)) >= 2147483647
) ? 2147483647 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd
[i]) -(1u << 31))));
          dp[0] = FROM_S32(d0)(((d0) >> 24) ^ 128);
          buffd[i] = 0.0;
        }

      } else if (kh == 3) {

        p2 = buff[0]; p3 = buff[1];
        k0 = pk[0]; k1 = pk[1]; k2 = pk[2];

        for (i = 0; i <= (hsize - 2); i += 2) {
          p0 = p2; p1 = p3;

          p2 = buff[i + 2]; p3 = buff[i + 3];

          d0 = D2I(p0*k0 + p1*k1 + p2*k2 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + buffd[i ]) -(1u << 31)) <=
 (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + p2
*k2 + buffd[i ]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + buffd[i ]) -(1u <<
 31))));
          d1 = D2I(p1*k0 + p2*k1 + p3*k2 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + buffd[i + 1]) -(1u << 31)) <=
 (-2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 + p2*k1 + p3
*k2 + buffd[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p1*k0 + p2*k1 + p3*k2 + buffd[i + 1]) -(1u <<
 31))));

          dp[0  ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
          dp[dll] = FROM_S32(d1)(((d1) >> 24) ^ 128);

          buffd[i    ] = 0.0;
          buffd[i + 1] = 0.0;

          dp += 2*dll;
        }

        if (i < hsize) {
          p0 = p2; p1 = p3;
          p2 = buff[i + 2];
          d0 = D2I(p0*k0 + p1*k1 + p2*k2 + buffd[i])((((p0*k0 + p1*k1 + p2*k2 + buffd[i]) -(1u << 31)) <=
 (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + p2
*k2 + buffd[i]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + buffd[i]) -(1u <<
 31))));
          dp[0] = FROM_S32(d0)(((d0) >> 24) ^ 128);

          buffd[i] = 0.0;
        }

      } else if (kh == 2) {

        p2 = buff[0];
        k0 = pk[0]; k1 = pk[1];

        for (i = 0; i <= (hsize - 2); i += 2) {
          p0 = p2;

          p1 = buff[i + 1]; p2 = buff[i + 2];

          d0 = D2I(p0*k0 + p1*k1 + buffd[i    ])((((p0*k0 + p1*k1 + buffd[i ]) -(1u << 31)) <= (-2147483647
 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + buffd[i ]) -(1u
 << 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p0
*k0 + p1*k1 + buffd[i ]) -(1u << 31))));
          d1 = D2I(p1*k0 + p2*k1 + buffd[i + 1])((((p1*k0 + p2*k1 + buffd[i + 1]) -(1u << 31)) <= (-
2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 + p2*k1 + buffd
[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647 : (
mlib_s32)((p1*k0 + p2*k1 + buffd[i + 1]) -(1u << 31))));

          dp[0  ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
          dp[dll] = FROM_S32(d1)(((d1) >> 24) ^ 128);

          buffd[i    ] = 0.0;
          buffd[i + 1] = 0.0;

          dp += 2*dll;
        }

        if (i < hsize) {
          p0 = p2;
          p1 = buff[i + 1];
          d0 = D2I(p0*k0 + p1*k1 + buffd[i])((((p0*k0 + p1*k1 + buffd[i]) -(1u << 31)) <= (-2147483647
 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + buffd[i]) -(1u
 << 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p0
*k0 + p1*k1 + buffd[i]) -(1u << 31))));
          dp[0] = FROM_S32(d0)(((d0) >> 24) ^ 128);

          buffd[i] = 0.0;
        }

      } else /* kh == 1 */{

        k0 = pk[0];

        for (i = 0; i <= (hsize - 2); i += 2) {
          p0 = buff[i]; p1 = buff[i + 1];

          d0 = D2I(p0*k0 + buffd[i    ])((((p0*k0 + buffd[i ]) -(1u << 31)) <= (-2147483647 -
1)) ? (-2147483647 -1) : ((((p0*k0 + buffd[i ]) -(1u <<
 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p0*k0 + buffd
[i ]) -(1u << 31))));
          d1 = D2I(p1*k0 + buffd[i + 1])((((p1*k0 + buffd[i + 1]) -(1u << 31)) <= (-2147483647
 -1)) ? (-2147483647 -1) : ((((p1*k0 + buffd[i + 1]) -(1u <<
 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p1*k0 + buffd
[i + 1]) -(1u << 31))));

          dp[0  ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
          dp[dll] = FROM_S32(d1)(((d1) >> 24) ^ 128);

          buffd[i    ] = 0.0;
          buffd[i + 1] = 0.0;

          dp += 2*dll;
        }

        if (i < hsize) {
          p0 = buff[i];
          d0 = D2I(p0*k0 + buffd[i])((((p0*k0 + buffd[i]) -(1u << 31)) <= (-2147483647 -
1)) ? (-2147483647 -1) : ((((p0*k0 + buffd[i]) -(1u << 31
)) >= 2147483647) ? 2147483647 : (mlib_s32)((p0*k0 + buffd
[i]) -(1u << 31))));
          dp[0] = FROM_S32(d0)(((d0) >> 24) ^ 128);

          buffd[i] = 0.0;
        }
      }

      /* next line */
      sl += nchannel;
      dl += nchannel;
    }
  }

  k_off += max_hsize;
  adr_dst += max_hsize*dll;
}

if (pbuff != buff) mlib_free(pbuff);

return MLIB_SUCCESS;
449}

451/***************************************************************/
452mlib_status CONV_FUNC_MxNmlib_c_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)
453{
DTYPEmlib_u8    *adr_src, *sl, *sp = NULL((void*)0);
DTYPEmlib_u8    *adr_dst, *dl, *dp = NULL((void*)0);
FTYPEmlib_d64    buff[BUFF_SIZE1600], *buffs_arr[2*(MAX_N15 + 1)];
FTYPEmlib_d64    **buffs = buffs_arr, *buffd;
FTYPEmlib_d64    akernel[256], *k = akernel, fscale = DSCALE(1 << 24);
FTYPEmlib_d64    *pbuff = buff;
FTYPEmlib_d64    k0, k1, k2, k3, k4, k5, k6;
FTYPEmlib_d64    p0, p1, p2, p3, p4, p5, p6, p7;
mlib_s32 *buffi;
mlib_s32 mn, l, off, kw, bsize, buff_ind;
mlib_s32 d0, d1;
mlib_s32 wid, hgt, sll, dll;
mlib_s32 nchannel, chan1, chan2;
mlib_s32 i, j, c, swid;
d64_2x32 dd;
mlib_status status = MLIB_SUCCESS;

GET_SRC_DST_PARAMETERS(DTYPE)hgt = mlib_ImageGetHeight(src); wid = mlib_ImageGetWidth(src)
; nchannel = mlib_ImageGetChannels(src); sll = mlib_ImageGetStride
(src) / sizeof(mlib_u8); dll = mlib_ImageGetStride(dst) / sizeof
(mlib_u8); adr_src = (mlib_u8 *)mlib_ImageGetData(src); adr_dst
 = (mlib_u8 *)mlib_ImageGetData(dst);

if (scale > 30) {
1
Assuming 'scale' is <= 30→
2
←
Taking false branch→
  fscale *= 1.0/(1 << 30);
  scale -= 30;
}

fscale /= (1 << scale);

mn = m*n;

if (mn > 256) {
3
←
Assuming 'mn' is <= 256→
4
←
Taking false branch→
  k = mlib_malloc(mn*sizeof(mlib_d64));

  if (k == NULL((void*)0)) return MLIB_FAILURE;
}

for (i = 0; i < mn; i++) {
5
←
Assuming 'i' is >= 'mn'→
6
←
Loop condition is false. Execution continues on line 492→
  k[i] = kernel[i]*fscale;
}

if (m == 1) {
7
←
Assuming 'm' is not equal to 1→
8
←
Taking false branch→
  status = mlib_ImageConv1xN_ext(dst, src, k, n, dy_t, dy_b, cmask);
  FREE_AND_RETURN_STATUSif (pbuff != buff) mlib_free(pbuff); if (k != akernel) mlib_free
(k); return status;
}

swid = wid + (m - 1);

bsize = (n + 3)*swid;

if ((bsize > BUFF_SIZE1600) || (n > MAX_N15)) {
9
←
Assuming 'bsize' is <= BUFF_SIZE→
10
←
Assuming 'n' is <= MAX_N→
11
←
Taking false branch→
  pbuff = mlib_malloc(sizeof(FTYPEmlib_d64)*bsize + sizeof(FTYPEmlib_d64 *)*2*(n + 1));

  if (pbuff == NULL((void*)0)) {
    status = MLIB_FAILURE;
    FREE_AND_RETURN_STATUSif (pbuff != buff) mlib_free(pbuff); if (k != akernel) mlib_free
(k); return status;
  }
  buffs = (FTYPEmlib_d64   **)(pbuff + bsize);
}

for (l = 0; l < (n + 1); l++) buffs[l] = pbuff + l*swid;
12
←
Assuming the condition is true→
13
←
Loop condition is true.  Entering loop body→
14
←
Assuming the condition is true→
15
←
Loop condition is true.  Entering loop body→
16
←
Assuming the condition is false→
17
←
Loop condition is false. Execution continues on line 512→
for (l = 0; l < (n + 1); l++) buffs[l + (n + 1)] = buffs[l];
18
←
Loop condition is true.  Entering loop body→
19
←
Loop condition is true.  Entering loop body→
20
←
Loop condition is false. Execution continues on line 513→
buffd = buffs[n] + swid;
buffi = (mlib_s32*)(buffd + swid);

chan1 = nchannel;
chan2 = chan1 + chan1;

swid -= (dx_l + dx_r);

for (c = 0; c < nchannel; c++) {
21
←
Assuming 'c' is < 'nchannel'→
22
←
Loop condition is true.  Entering loop body→
  if (!(cmask & (1 << (chan1 - 1 - c)))) continue;
23
←
Assuming the condition is false→
24
←
Taking false branch→

  sl = adr_src + c;
  dl = adr_dst + c;

  for (l = 0; l < n; l++) {
25
←
Loop condition is true.  Entering loop body→
33
←
Loop condition is false. Execution continues on line 545→
    FTYPEmlib_d64    *buff = buffs[l];

    for (i = 0; i < dx_l; i++) {
26
←
Assuming 'i' is >= 'dx_l'→
27
←
Loop condition is false. Execution continues on line 534→
      buff[i] = (FTYPEmlib_d64)sl[0];
    }

    for (i = 0; i < swid; i++) {
28
←
Assuming 'i' is >= 'swid'→
29
←
Loop condition is false. Execution continues on line 538→
      buff[i + dx_l] = (FTYPEmlib_d64)sl[i*chan1];
    }

    for (i = 0; i < dx_r; i++) {
30
←
Assuming 'i' is >= 'dx_r'→
31
←
Loop condition is false. Execution continues on line 542→
      buff[swid + dx_l + i] = buff[swid + dx_l - 1];
    }

    if ((l >= dy_t) && (l < hgt + n - dy_b - 2)) sl += sll;
32
←
Assuming 'l' is < 'dy_t'→
  }

  buff_ind = 0;

  for (i = 0; i < wid; i++) buffd[i] = 0.0;
34
←
Assuming 'i' is >= 'wid'→
35
←
Loop condition is false. Execution continues on line 549→

  for (j = 0; j < hgt; j++) {
36
←
Assuming 'j' is < 'hgt'→
37
←
Loop condition is true.  Entering loop body→
    FTYPEmlib_d64    **buffc = buffs + buff_ind;
    FTYPEmlib_d64    *buffn = buffc[n];
    FTYPEmlib_d64    *pk = k;

    for (l = 0; l < n; l++) {
38
←
Loop condition is true.  Entering loop body→
      FTYPEmlib_d64    *buff_l = buffc[l];

      for (off = 0; off < m;) {
39
←
Assuming 'off' is < 'm'→
40
←
Loop condition is true.  Entering loop body→
        FTYPEmlib_d64    *buff = buff_l + off;

        kw = m - off;

        if (kw > 2*MAX_KER7) kw = MAX_KER7; else
41
←
Assuming the condition is false→
42
←
Taking false branch→
          if (kw > MAX_KER7) kw = kw/2;
43
←
Assuming 'kw' is <= MAX_KER→
44
←
Taking false branch→
        off += kw;

        sp = sl;
        dp = dl;

        if (kw == 7) {
45
←
Assuming 'kw' is not equal to 7→
46
←
Taking false branch→

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
          p5 = buff[3]; p6 = buff[4]; p7 = buff[5];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4]; k5 = pk[5]; k6 = pk[6];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6; p5 = p7;

              p6 = buff[i + 6]; p7 = buff[i + 7];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6*k6;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7*k6;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6; p5 = p7;

              p6 = buff[i + 6]; p7 = buff[i + 7];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6*k6 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6*k6 + buffd
[i ]) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6
*k6 + buffd[i ]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 +
 p6*k6 + buffd[i ]) -(1u << 31))));
              d1 = D2I(p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7*k6 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7*k6 + buffd
[i + 1]) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7
*k6 + buffd[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 +
 p7*k6 + buffd[i + 1]) -(1u << 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 6) {
47
←
Assuming 'kw' is equal to 6→
48
←
Taking true branch→

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
49
←
Assigned value is garbage or undefined
          p5 = buff[3]; p6 = buff[4];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4]; k5 = pk[5];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6;

              p5 = buff[i + 5]; p6 = buff[i + 6];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6;

              p5 = buff[i + 5]; p6 = buff[i + 6];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + buffd[i ]
) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647 -
1) : ((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + buffd
[i ]) -(1u << 31)) >= 2147483647) ? 2147483647 : (mlib_s32
)((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + buffd[i ])
 -(1u << 31))));
              d1 = D2I(p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + buffd[i +
 1]) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + buffd
[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647 : (
mlib_s32)((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + buffd
[i + 1]) -(1u << 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 5) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
          p5 = buff[3];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5;

              p4 = buff[i + 4]; p5 = buff[i + 5];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5;

              p4 = buff[i + 4]; p5 = buff[i + 5];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + buffd[i ]) -(1u <<
 31)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 +
 p1*k1 + p2*k2 + p3*k3 + p4*k4 + buffd[i ]) -(1u << 31)
) >= 2147483647) ? 2147483647 : (mlib_s32)((p0*k0 + p1*k1 +
 p2*k2 + p3*k3 + p4*k4 + buffd[i ]) -(1u << 31))));
              d1 = D2I(p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + buffd[i + 1]) -(1u
 << 31)) <= (-2147483647 -1)) ? (-2147483647 -1) : (
(((p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + buffd[i + 1]) -(1u
 << 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p1
*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + buffd[i + 1]) -(1u <<
 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 4) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4;

              p3 = buff[i + 3]; p4 = buff[i + 4];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4;

              p3 = buff[i + 3]; p4 = buff[i + 4];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i ]) -(1u << 31
)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1
*k1 + p2*k2 + p3*k3 + buffd[i ]) -(1u << 31)) >= 2147483647
) ? 2147483647 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd
[i ]) -(1u << 31))));
              d1 = D2I(p1*k0 + p2*k1 + p3*k2 + p4*k3 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + p4*k3 + buffd[i + 1]) -(1u <<
 31)) <= (-2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 +
 p2*k1 + p3*k2 + p4*k3 + buffd[i + 1]) -(1u << 31)) >=
 2147483647) ? 2147483647 : (mlib_s32)((p1*k0 + p2*k1 + p3*k2
 + p4*k3 + buffd[i + 1]) -(1u << 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 3) {

          p2 = buff[0]; p3 = buff[1];
          k0 = pk[0]; k1 = pk[1]; k2 = pk[2];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3;

              p2 = buff[i + 2]; p3 = buff[i + 3];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3;

              p2 = buff[i + 2]; p3 = buff[i + 3];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + p2*k2 + buffd[i    ])((((p0*k0 + p1*k1 + p2*k2 + buffd[i ]) -(1u << 31)) <=
 (-2147483647 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + p2
*k2 + buffd[i ]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p0*k0 + p1*k1 + p2*k2 + buffd[i ]) -(1u <<
 31))));
              d1 = D2I(p1*k0 + p2*k1 + p3*k2 + buffd[i + 1])((((p1*k0 + p2*k1 + p3*k2 + buffd[i + 1]) -(1u << 31)) <=
 (-2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 + p2*k1 + p3
*k2 + buffd[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((p1*k0 + p2*k1 + p3*k2 + buffd[i + 1]) -(1u <<
 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else /* if (kw == 2) */ {

          p2 = buff[0];
          k0 = pk[0]; k1 = pk[1];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2;

              p1 = buff[i + 1]; p2 = buff[i + 2];

              buffd[i    ] += p0*k0 + p1*k1;
              buffd[i + 1] += p1*k0 + p2*k1;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2;

              p1 = buff[i + 1]; p2 = buff[i + 2];

              LOAD_BUFF(buffi)*(mlib_s64*)(buffi + i) = (((mlib_s64)sp[chan1]) << 32)
 | ((mlib_s64)sp[0]);

              dd.d64 = *(FTYPEmlib_d64   *)(buffi + i);
              buffn[i + dx_l    ] = (FTYPEmlib_d64)dd.i32s.i0;
              buffn[i + dx_l + 1] = (FTYPEmlib_d64)dd.i32s.i1;

              d0 = D2I(p0*k0 + p1*k1 + buffd[i    ])((((p0*k0 + p1*k1 + buffd[i ]) -(1u << 31)) <= (-2147483647
 -1)) ? (-2147483647 -1) : ((((p0*k0 + p1*k1 + buffd[i ]) -(1u
 << 31)) >= 2147483647) ? 2147483647 : (mlib_s32)((p0
*k0 + p1*k1 + buffd[i ]) -(1u << 31))));
              d1 = D2I(p1*k0 + p2*k1 + buffd[i + 1])((((p1*k0 + p2*k1 + buffd[i + 1]) -(1u << 31)) <= (-
2147483647 -1)) ? (-2147483647 -1) : ((((p1*k0 + p2*k1 + buffd
[i + 1]) -(1u << 31)) >= 2147483647) ? 2147483647 : (
mlib_s32)((p1*k0 + p2*k1 + buffd[i + 1]) -(1u << 31))));

              dp[0    ] = FROM_S32(d0)(((d0) >> 24) ^ 128);
              dp[chan1] = FROM_S32(d1)(((d1) >> 24) ^ 128);

              buffd[i    ] = 0.0;
              buffd[i + 1] = 0.0;

              sp += chan2;
              dp += chan2;
            }
          }
        }

        pk += kw;
      }
    }

    /* last pixels */
    for (; i < wid; i++) {
      FTYPEmlib_d64    *pk = k, s = 0;
      mlib_s32 x, d0;

      for (l = 0; l < n; l++) {
        FTYPEmlib_d64    *buff = buffc[l] + i;

        for (x = 0; x < m; x++) s += buff[x] * (*pk++);
      }

      d0 = D2I(s)((((s) -(1u << 31)) <= (-2147483647 -1)) ? (-2147483647
 -1) : ((((s) -(1u << 31)) >= 2147483647) ? 2147483647
 : (mlib_s32)((s) -(1u << 31))));
      dp[0] = FROM_S32(d0)(((d0) >> 24) ^ 128);

      buffn[i + dx_l] = (FTYPEmlib_d64)sp[0];

      sp += chan1;
      dp += chan1;
    }

    for (; i < swid; i++) {
      buffn[i + dx_l] = (FTYPEmlib_d64)sp[0];
      sp += chan1;
    }

    for (i = 0; i < dx_l; i++) buffn[i] = buffn[dx_l];
    for (i = 0; i < dx_r; i++) buffn[swid + dx_l + i] = buffn[swid + dx_l - 1];

    /* next line */

    if (j < hgt - dy_b - 2) sl += sll;
    dl += dll;

    buff_ind++;

    if (buff_ind >= n + 1) buff_ind = 0;
  }
}

FREE_AND_RETURN_STATUSif (pbuff != buff) mlib_free(pbuff); if (k != akernel) mlib_free
(k); return status;
870}

872/***************************************************************/
873/* for x86, using integer multiplies is faster */

875#define STORE_RES(res, x)x >>= shift2; if (x & 0xffffff00) { if (x < 0) res
 = 0; else res = (127*2 +1); } else { res = (mlib_u8)x; }                                       \
x >>= shift2;                                                 \
CLAMP_STORE(res, x)if (x & 0xffffff00) { if (x < 0) res = 0; else res = (
127*2 +1); } else { res = (mlib_u8)x; }

879mlib_status CONV_FUNC_MxN_Imlib_i_convMxNext_u8(mlib_image *dst, const mlib_image *src, const
 mlib_s32 *kernel, mlib_s32 m, mlib_s32 n, mlib_s32 dx_l, mlib_s32
 dx_r, mlib_s32 dy_t, mlib_s32 dy_b, mlib_s32 scale, mlib_s32
 cmask)
880{
DTYPEmlib_u8    *adr_src, *sl, *sp = NULL((void*)0);
DTYPEmlib_u8    *adr_dst, *dl, *dp = NULL((void*)0);
mlib_s32 buff[BUFF_SIZE1600], *buffs_arr[2*(MAX_N15 + 1)];
mlib_s32 *pbuff = buff;
mlib_s32 **buffs = buffs_arr, *buffd;
mlib_s32 l, off, kw, bsize, buff_ind;
mlib_s32 d0, d1, shift1, shift2;
mlib_s32 k0, k1, k2, k3, k4, k5, k6;
mlib_s32 p0, p1, p2, p3, p4, p5, p6, p7;
mlib_s32 wid, hgt, sll, dll;
mlib_s32 nchannel, chan1;
mlib_s32 i, j, c, swid;
mlib_s32 chan2;
mlib_s32 k_locl[MAX_N15*MAX_N15], *k = k_locl;
GET_SRC_DST_PARAMETERS(DTYPE)hgt = mlib_ImageGetHeight(src); wid = mlib_ImageGetWidth(src)
; nchannel = mlib_ImageGetChannels(src); sll = mlib_ImageGetStride
(src) / sizeof(mlib_u8); dll = mlib_ImageGetStride(dst) / sizeof
(mlib_u8); adr_src = (mlib_u8 *)mlib_ImageGetData(src); adr_dst
 = (mlib_u8 *)mlib_ImageGetData(dst);

897#if IMG_TYPE1 != 1
shift1 = 16;
899#else
shift1 = 8;
901#endif /* IMG_TYPE != 1 */
shift2 = scale - shift1;

chan1 = nchannel;
chan2 = chan1 + chan1;

swid = wid + (m - 1);

bsize = (n + 2)*swid;

if ((bsize > BUFF_SIZE1600) || (n > MAX_N15)) {
  pbuff = mlib_malloc(sizeof(mlib_s32)*bsize + sizeof(mlib_s32 *)*2*(n + 1));

  if (pbuff == NULL((void*)0)) return MLIB_FAILURE;
  buffs = (mlib_s32 **)(pbuff + bsize);
}

for (l = 0; l < (n + 1); l++) buffs[l] = pbuff + l*swid;
for (l = 0; l < (n + 1); l++) buffs[l + (n + 1)] = buffs[l];
buffd = buffs[n] + swid;

if (m*n > MAX_N15*MAX_N15) {
  k = mlib_malloc(sizeof(mlib_s32)*(m*n));

  if (k == NULL((void*)0)) {
    if (pbuff != buff) mlib_free(pbuff);
    return MLIB_FAILURE;
  }
}

for (i = 0; i < m*n; i++) {
  k[i] = kernel[i] >> shift1;
}

swid -= (dx_l + dx_r);

for (c = 0; c < nchannel; c++) {
  if (!(cmask & (1 << (nchannel - 1 - c)))) continue;

  sl = adr_src + c;
  dl = adr_dst + c;

  for (l = 0; l < n; l++) {
    mlib_s32  *buff = buffs[l];

    for (i = 0; i < dx_l; i++) {
      buff[i] = (mlib_s32)sl[0];
    }

    for (i = 0; i < swid; i++) {
      buff[i + dx_l] = (mlib_s32)sl[i*chan1];
    }

    for (i = 0; i < dx_r; i++) {
      buff[swid + dx_l + i] = buff[swid + dx_l - 1];
    }

    if ((l >= dy_t) && (l < hgt + n - dy_b - 2)) sl += sll;
  }

  buff_ind = 0;

  for (i = 0; i < wid; i++) buffd[i] = 0;

  for (j = 0; j < hgt; j++) {
    mlib_s32 **buffc = buffs + buff_ind;
    mlib_s32 *buffn = buffc[n];
    mlib_s32 *pk = k;

    for (l = 0; l < n; l++) {
      mlib_s32  *buff_l = buffc[l];

      for (off = 0; off < m;) {
        mlib_s32 *buff = buff_l + off;

        sp = sl;
        dp = dl;

        kw = m - off;

        if (kw > 2*MAX_KER7) kw = MAX_KER7; else
          if (kw > MAX_KER7) kw = kw/2;
        off += kw;

        if (kw == 7) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
          p5 = buff[3]; p6 = buff[4]; p7 = buff[5];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4]; k5 = pk[5]; k6 = pk[6];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6; p5 = p7;

              p6 = buff[i + 6]; p7 = buff[i + 7];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6*k6;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7*k6;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6; p5 = p7;

              p6 = buff[i + 6]; p7 = buff[i + 7];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + p6*k6 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + p7*k6 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 6) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
          p5 = buff[3]; p6 = buff[4];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4]; k5 = pk[5];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6;

              p5 = buff[i + 5]; p6 = buff[i + 6];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5; p4 = p6;

              p5 = buff[i + 5]; p6 = buff[i + 6];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + p5*k5 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + p6*k5 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 5) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];
          p5 = buff[3];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];
          k4 = pk[4];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5;

              p4 = buff[i + 4]; p5 = buff[i + 5];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4; p3 = p5;

              p4 = buff[i + 4]; p5 = buff[i + 5];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + p2*k2 + p3*k3 + p4*k4 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + p3*k2 + p4*k3 + p5*k4 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 4) {

          p2 = buff[0]; p3 = buff[1]; p4 = buff[2];

          k0 = pk[0]; k1 = pk[1]; k2 = pk[2]; k3 = pk[3];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4;

              p3 = buff[i + 3]; p4 = buff[i + 4];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2 + p3*k3;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2 + p4*k3;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3; p2 = p4;

              p3 = buff[i + 3]; p4 = buff[i + 4];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + p2*k2 + p3*k3 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + p3*k2 + p4*k3 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 3) {

          p2 = buff[0]; p3 = buff[1];
          k0 = pk[0]; k1 = pk[1]; k2 = pk[2];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3;

              p2 = buff[i + 2]; p3 = buff[i + 3];

              buffd[i    ] += p0*k0 + p1*k1 + p2*k2;
              buffd[i + 1] += p1*k0 + p2*k1 + p3*k2;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2; p1 = p3;

              p2 = buff[i + 2]; p3 = buff[i + 3];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + p2*k2 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + p3*k2 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else if (kw == 2) {

          p2 = buff[0];
          k0 = pk[0]; k1 = pk[1];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2;

              p1 = buff[i + 1]; p2 = buff[i + 2];

              buffd[i    ] += p0*k0 + p1*k1;
              buffd[i + 1] += p1*k0 + p2*k1;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = p2;

              p1 = buff[i + 1]; p2 = buff[i + 2];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + p1*k1 + buffd[i    ]);
              d1 = (p1*k0 + p2*k1 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }

        } else /* kw == 1 */{

          k0 = pk[0];

          if (l < (n - 1) || off < m) {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = buff[i]; p1 = buff[i + 1];

              buffd[i    ] += p0*k0;
              buffd[i + 1] += p1*k0;
            }

          } else {
            for (i = 0; i <= (wid - 2); i += 2) {
              p0 = buff[i]; p1 = buff[i + 1];

              buffn[i + dx_l    ] = (mlib_s32)sp[0];
              buffn[i + dx_l + 1] = (mlib_s32)sp[chan1];

              d0 = (p0*k0 + buffd[i    ]);
              d1 = (p1*k0 + buffd[i + 1]);

              STORE_RES(dp[0    ], d0)d0 >>= shift2; if (d0 & 0xffffff00) { if (d0 < 0
) dp[0 ] = 0; else dp[0 ] = (127*2 +1); } else { dp[0 ] = (mlib_u8
)d0; };
              STORE_RES(dp[chan1], d1)d1 >>= shift2; if (d1 & 0xffffff00) { if (d1 < 0
) dp[chan1] = 0; else dp[chan1] = (127*2 +1); } else { dp[chan1
] = (mlib_u8)d1; };

              buffd[i    ] = 0;
              buffd[i + 1] = 0;

              sp += chan2;
              dp += chan2;
            }
          }
        }

        pk += kw;
      }
    }

    /* last pixels */
    for (; i < wid; i++) {
      mlib_s32 *pk = k, x, s = 0;

      for (l = 0; l < n; l++) {
        mlib_s32 *buff = buffc[l] + i;

        for (x = 0; x < m; x++) s += buff[x] * (*pk++);
      }

      STORE_RES(dp[0], s)s >>= shift2; if (s & 0xffffff00) { if (s < 0) dp
[0] = 0; else dp[0] = (127*2 +1); } else { dp[0] = (mlib_u8)s
; };

      buffn[i + dx_l] = (mlib_s32)sp[0];

      sp += chan1;
      dp += chan1;
    }

    for (; i < swid; i++) {
      buffn[i + dx_l] = (mlib_s32)sp[0];
      sp += chan1;
    }

    for (i = 0; i < dx_l; i++) buffn[i] = buffn[dx_l];
    for (i = 0; i < dx_r; i++) buffn[swid + dx_l + i] = buffn[swid + dx_l - 1];

    /* next line */

    if (j < hgt - dy_b - 2) sl += sll;
    dl += dll;

    buff_ind++;

    if (buff_ind >= n + 1) buff_ind = 0;
  }
}

if (pbuff != buff) mlib_free(pbuff);
if (k != k_locl) mlib_free(k);

return MLIB_SUCCESS;
1302}

1304/***************************************************************/