laag-libjpeg-turbo 1.5.3.0

data/vendor/github.com/libjpeg-turbo/libjpeg-turbo/release/uninstall.in

@@ -0,0 +1,112 @@
+# Copyright (C)2009-2011, 2013, 2016 D. R. Commander.  All Rights Reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+# - Redistributions of source code must retain the above copyright notice,
+#   this list of conditions and the following disclaimer.
+# - Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+# - Neither the name of the libjpeg-turbo Project nor the names of its
+#   contributors may be used to endorse or promote products derived from this
+#   software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+#!/bin/sh
+
+if [ ! "`id -u`" = "0" ]; then
+	echo "ERROR: This script must be executed as root"
+	exit -1
+fi
+
+PACKAGE=@PKGNAME@
+MACPACKAGE=com.$PACKAGE.$PACKAGE
+RECEIPT=/Library/Receipts/$PACKAGE.pkg
+
+LSBOM=
+if [ -d $RECEIPT ]; then
+	LSBOM='lsbom -s -f -l '$RECEIPT'/Contents/Archive.bom'
+else
+	LSBOM='pkgutil --files '$MACPACKAGE
+fi
+
+mylsbom()
+{
+	$LSBOM || (echo "ERROR: Could not list package contents"; exit -1)
+}
+
+echo Removing package files ...
+EXITSTATUS=0
+pushd /
+mylsbom | while read file; do
+	if [ ! -d "$file" ]; then rm "$file" 2>&1 || EXITSTATUS=-1; fi
+done
+popd
+
+echo Removing package directories ...
+PREFIX=%{__prefix}
+BINDIR=%{__bindir}
+DATADIR=%{__datadir}
+INCLUDEDIR=%{__includedir}
+LIBDIR=%{__libdir}
+MANDIR=%{__mandir}
+
+if [ -d $BINDIR ]; then
+	rmdir $BINDIR 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $LIBDIR/pkgconfig ]; then
+	rmdir $LIBDIR/pkgconfig 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $LIBDIR ]; then
+	rmdir $LIBDIR 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $INCLUDEDIR ]; then
+	rmdir $INCLUDEDIR 2>&1 || EXITSTATUS=-1
+fi
+if [ "$PREFIX" = "/opt/libjpeg-turbo" -a "$LIBDIR" = "/opt/libjpeg-turbo/lib" ]; then
+	if [ -h $LIBDIR\32 ]; then
+		rm $LIBDIR\32 2>&1 || EXITSTATUS=-1
+	fi
+	if [ -h $LIBDIR\64 ]; then
+		rm $LIBDIR\64 2>&1 || EXITSTATUS=-1
+	fi
+fi
+if [ -d $MANDIR/man1 ]; then
+	rmdir $MANDIR/man1 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $MANDIR ]; then
+	rmdir $MANDIR 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $DATADIR/classes ]; then
+	rmdir $DATADIR/classes 2>&1 || EXITSTATUS=-1
+fi
+if [ -d $DATADIR -a "$DATADIR" != "$PREFIX" ]; then
+	rmdir $DATADIR 2>&1 || EXITSTATUS=-1
+fi
+if [ "$PREFIX" = "/opt/libjpeg-turbo" -a -h "$PREFIX/doc" ]; then
+	rm $PREFIX/doc 2>&1 || EXITSTATUS=-1
+fi
+rmdir $PREFIX 2>&1 || EXITSTATUS=-1
+rmdir /Library/Documentation/$PACKAGE 2>&1 || EXITSTATUS=-1
+
+if [ -d $RECEIPT ]; then
+	echo Removing package receipt ...
+	rm -r $RECEIPT 2>&1 || EXITSTATUS=-1
+else
+	echo Forgetting package $MACPACKAGE ...
+	pkgutil --forget $MACPACKAGE
+fi
+
+exit $EXITSTATUS

data/vendor/github.com/libjpeg-turbo/libjpeg-turbo/sharedlib/CMakeLists.txt

@@ -0,0 +1,73 @@
+# Anything that must be linked against the shared C library on Windows must
+# be built in this subdirectory, because CMake doesn't allow us to override
+# the compiler flags for each build type except at directory scope.  Note
+# to CMake developers:  Add a COMPILE_FLAGS_<CONFIG> target property, or
+# better yet, provide a friendly way of configuring a Windows target to use the
+# static C library.
+
+if(MSVC)
+  # Build all configurations against shared C library
+  foreach(var CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE
+    CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO)
+    if(${var} MATCHES "/MT")
+      string(REGEX REPLACE "/MT" "/MD" ${var} "${${var}}")
+    endif()
+  endforeach()
+endif()
+
+foreach(src ${JPEG_SOURCES})
+  set(JPEG_SRCS ${JPEG_SRCS} ${CMAKE_SOURCE_DIR}/${src})
+endforeach()
+
+if(WITH_SIMD)
+  # This tells CMake that the "source" files haven't been generated yet
+  set_source_files_properties(${SIMD_OBJS} PROPERTIES GENERATED 1)
+endif()
+
+if(WITH_MEM_SRCDST AND NOT WITH_JPEG8)
+  add_library(jpeg SHARED ${JPEG_SRCS} ${SIMD_OBJS}
+    ${CMAKE_SOURCE_DIR}/win/jpeg${DLL_VERSION}-memsrcdst.def)
+else()
+  add_library(jpeg SHARED ${JPEG_SRCS} ${SIMD_OBJS}
+    ${CMAKE_SOURCE_DIR}/win/jpeg${DLL_VERSION}.def)
+endif()
+set_target_properties(jpeg PROPERTIES SOVERSION ${DLL_VERSION}
+  VERSION ${FULLVERSION})
+if(MSVC)
+  set_target_properties(jpeg PROPERTIES SUFFIX ${DLL_VERSION}.dll)
+elseif(MINGW OR CYGWIN)
+  set_target_properties(jpeg PROPERTIES SUFFIX -${DLL_VERSION}.dll)
+endif(MSVC)
+if(WITH_SIMD)
+  add_dependencies(jpeg simd)
+endif()
+
+if(WITH_12BIT)
+  set(COMPILE_FLAGS "-DGIF_SUPPORTED -DPPM_SUPPORTED -DUSE_SETMODE")
+else()
+  set(COMPILE_FLAGS "-DBMP_SUPPORTED -DGIF_SUPPORTED -DPPM_SUPPORTED -DTARGA_SUPPORTED -DUSE_SETMODE")
+	set(CJPEG_BMP_SOURCES ../rdbmp.c ../rdtarga.c)
+	set(DJPEG_BMP_SOURCES ../wrbmp.c ../wrtarga.c)
+endif()
+
+add_executable(cjpeg ../cjpeg.c ../cdjpeg.c ../rdgif.c ../rdppm.c
+  ../rdswitch.c ${CJPEG_BMP_SOURCES})
+set_property(TARGET cjpeg PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+target_link_libraries(cjpeg jpeg)
+
+add_executable(djpeg ../djpeg.c ../cdjpeg.c ../rdcolmap.c ../rdswitch.c
+  ../wrgif.c ../wrppm.c ${DJPEG_BMP_SOURCES})
+set_property(TARGET djpeg PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+target_link_libraries(djpeg jpeg)
+
+add_executable(jpegtran ../jpegtran.c ../cdjpeg.c ../rdswitch.c ../transupp.c)
+target_link_libraries(jpegtran jpeg)
+set_property(TARGET jpegtran PROPERTY COMPILE_FLAGS "-DUSE_SETMODE")
+
+add_executable(jcstest ../jcstest.c)
+target_link_libraries(jcstest jpeg)
+
+install(TARGETS jpeg cjpeg djpeg jpegtran
+  ARCHIVE DESTINATION lib
+  LIBRARY DESTINATION lib
+  RUNTIME DESTINATION bin)

data/vendor/github.com/libjpeg-turbo/libjpeg-turbo/simd/CMakeLists.txt

@@ -0,0 +1,81 @@
+if(NOT DEFINED NASM)
+  find_program(NASM NAMES nasm yasm DOC "Path to NASM/YASM executable")
+endif()
+message(STATUS "NASM = ${NASM}")
+
+if(SIMD_X86_64)
+  set(NAFLAGS -fwin64 -DWIN64 -D__x86_64__)
+else()
+  if(BORLAND)
+    set(NAFLAGS -fobj -DOBJ32)
+  else()
+    set(NAFLAGS -fwin32 -DWIN32)
+  endif()
+endif()
+set(NAFLAGS ${NAFLAGS} -I${CMAKE_SOURCE_DIR}/win/ -I${CMAKE_CURRENT_SOURCE_DIR}/)
+
+# This only works if building from the command line.  There is currently no way
+# to set a variable's value based on the build type when using the MSVC IDE.
+if(CMAKE_BUILD_TYPE STREQUAL "Debug"
+  OR CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
+  set(NAFLAGS ${NAFLAGS} -g)
+endif()
+
+if(SIMD_X86_64)
+  set(SIMD_BASENAMES jfdctflt-sse-64 jccolor-sse2-64 jcgray-sse2-64
+    jchuff-sse2-64 jcsample-sse2-64 jdcolor-sse2-64 jdmerge-sse2-64
+    jdsample-sse2-64 jfdctfst-sse2-64 jfdctint-sse2-64 jidctflt-sse2-64
+    jidctfst-sse2-64 jidctint-sse2-64 jidctred-sse2-64 jquantf-sse2-64
+    jquanti-sse2-64)
+  message(STATUS "Building x86_64 SIMD extensions")
+else()
+  set(SIMD_BASENAMES jsimdcpu jfdctflt-3dn jidctflt-3dn jquant-3dn jccolor-mmx
+    jcgray-mmx jcsample-mmx jdcolor-mmx jdmerge-mmx jdsample-mmx jfdctfst-mmx
+    jfdctint-mmx jidctfst-mmx jidctint-mmx jidctred-mmx jquant-mmx jfdctflt-sse
+    jidctflt-sse jquant-sse jccolor-sse2 jcgray-sse2 jchuff-sse2 jcsample-sse2
+    jdcolor-sse2 jdmerge-sse2 jdsample-sse2 jfdctfst-sse2 jfdctint-sse2
+    jidctflt-sse2 jidctfst-sse2 jidctint-sse2 jidctred-sse2 jquantf-sse2
+    jquanti-sse2)
+  message(STATUS "Building i386 SIMD extensions")
+endif()
+
+if(MSVC_IDE)
+  set(OBJDIR "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}")
+else()
+  set(OBJDIR ${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
+file(GLOB INC_FILES *.inc)
+
+foreach(file ${SIMD_BASENAMES})
+  set(DEPFILE "")
+  set(SIMD_SRC ${CMAKE_CURRENT_SOURCE_DIR}/${file}.asm)
+  if(${file} MATCHES jccolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jccolor" "jccolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jcgray)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jcgray" "jcgryext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdcolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdcolor" "jdcolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdmerge)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdmerge" "jdmrgext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  set(SIMD_OBJ ${OBJDIR}/${file}.obj)
+  add_custom_command(OUTPUT ${SIMD_OBJ}
+    DEPENDS ${SIMD_SRC} ${DEPFILE} ${INC_FILES}
+    COMMAND ${NASM} ${NAFLAGS} ${SIMD_SRC} -o${SIMD_OBJ})
+  set(SIMD_OBJS ${SIMD_OBJS} ${SIMD_OBJ})
+endforeach()
+
+set(SIMD_OBJS ${SIMD_OBJS} PARENT_SCOPE)
+add_custom_target(simd DEPENDS ${SIMD_OBJS})

data/vendor/github.com/libjpeg-turbo/libjpeg-turbo/simd/Makefile.am

@@ -0,0 +1,102 @@
+noinst_LTLIBRARIES = libsimd.la
+
+BUILT_SOURCES = jsimdcfg.inc
+
+EXTRA_DIST = nasm_lt.sh CMakeLists.txt \
+	jccolext-mmx.asm   jcgryext-mmx.asm   jdcolext-mmx.asm   jdmrgext-mmx.asm \
+	jccolext-sse2.asm  jcgryext-sse2.asm  jdcolext-sse2.asm  jdmrgext-sse2.asm \
+	jccolext-sse2-64.asm  jcgryext-sse2-64.asm  jdcolext-sse2-64.asm \
+	jdmrgext-sse2-64.asm  jccolext-altivec.c    jcgryext-altivec.c \
+	jdcolext-altivec.c    jdmrgext-altivec.c
+
+if SIMD_X86_64
+
+libsimd_la_SOURCES = jsimd_x86_64.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jfdctflt-sse-64.asm \
+	jccolor-sse2-64.asm   jcgray-sse2-64.asm    jchuff-sse2-64.asm \
+	jcsample-sse2-64.asm  jdcolor-sse2-64.asm   jdmerge-sse2-64.asm \
+	jdsample-sse2-64.asm  jfdctfst-sse2-64.asm  jfdctint-sse2-64.asm \
+	jidctflt-sse2-64.asm  jidctfst-sse2-64.asm  jidctint-sse2-64.asm \
+	jidctred-sse2-64.asm  jquantf-sse2-64.asm   jquanti-sse2-64.asm
+
+jccolor-sse2-64.lo:  jccolext-sse2-64.asm
+jcgray-sse2-64.lo:   jcgryext-sse2-64.asm
+jdcolor-sse2-64.lo:  jdcolext-sse2-64.asm
+jdmerge-sse2-64.lo:  jdmrgext-sse2-64.asm
+
+endif
+
+if SIMD_I386
+
+libsimd_la_SOURCES = jsimd_i386.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jsimdcpu.asm \
+	jfdctflt-3dn.asm   jidctflt-3dn.asm   jquant-3dn.asm \
+	jccolor-mmx.asm    jcgray-mmx.asm     jcsample-mmx.asm \
+	jdcolor-mmx.asm    jdmerge-mmx.asm    jdsample-mmx.asm \
+	jfdctfst-mmx.asm   jfdctint-mmx.asm   jidctfst-mmx.asm \
+	jidctint-mmx.asm   jidctred-mmx.asm   jquant-mmx.asm \
+	jfdctflt-sse.asm   jidctflt-sse.asm   jquant-sse.asm \
+	jccolor-sse2.asm   jcgray-sse2.asm    jchuff-sse2.asm \
+	jcsample-sse2.asm  jdcolor-sse2.asm   jdmerge-sse2.asm \
+	jdsample-sse2.asm  jfdctfst-sse2.asm  jfdctint-sse2.asm \
+	jidctflt-sse2.asm  jidctfst-sse2.asm  jidctint-sse2.asm \
+	jidctred-sse2.asm  jquantf-sse2.asm   jquanti-sse2.asm
+
+jccolor-mmx.lo:   jccolext-mmx.asm
+jcgray.-mmx.lo:   jcgryext-mmx.asm
+jdcolor-mmx.lo:   jdcolext-mmx.asm
+jdmerge-mmx.lo:   jdmrgext-mmx.asm
+jccolor-sse2.lo:  jccolext-sse2.asm
+jcgray-sse2.lo:   jcgryext-sse2.asm
+jdcolor-sse2.lo:  jdcolext-sse2.asm
+jdmerge-sse2.lo:  jdmrgext-sse2.asm
+
+endif
+
+if SIMD_ARM
+
+libsimd_la_SOURCES = jsimd_arm.c jsimd_arm_neon.S
+
+endif
+
+if SIMD_ARM_64
+
+libsimd_la_SOURCES = jsimd_arm64.c jsimd_arm64_neon.S
+
+endif
+
+if SIMD_MIPS
+
+libsimd_la_SOURCES = jsimd_mips.c jsimd_mips_dspr2_asm.h jsimd_mips_dspr2.S
+
+endif
+
+if SIMD_POWERPC
+
+noinst_LTLIBRARIES += libsimd_altivec.la
+
+libsimd_altivec_la_SOURCES = \
+	jccolor-altivec.c     jcgray-altivec.c      jcsample-altivec.c \
+	jdcolor-altivec.c     jdmerge-altivec.c     jdsample-altivec.c \
+	jfdctfst-altivec.c    jfdctint-altivec.c \
+	jidctfst-altivec.c    jidctint-altivec.c \
+	jquanti-altivec.c
+libsimd_altivec_la_CFLAGS = -maltivec
+
+jccolor-altivec.lo:  jccolext-altivec.c
+jcgray-altivec.lo:   jcgryext-altivec.c
+jdcolor-altivec.lo:  jdcolext-altivec.c
+jdmerge-altivec.lo:  jdmrgext-altivec.c
+
+libsimd_la_SOURCES = jsimd_powerpc.c jsimd_altivec.h jcsample.h
+libsimd_la_LIBADD = libsimd_altivec.la
+
+endif
+
+AM_CPPFLAGS = -I$(top_srcdir)
+
+.asm.lo:
+	$(AM_V_GEN) $(LIBTOOL) $(AM_V_lt) --mode=compile --tag NASM $(srcdir)/nasm_lt.sh $(AM_V_lt) $(NASM) $(NAFLAGS) -I$(srcdir) -I. $< -o $@
+
+jsimdcfg.inc: $(srcdir)/jsimdcfg.inc.h ../jpeglib.h ../jconfig.h ../jmorecfg.h
+	$(AM_V_GEN) $(CPP) -I$(top_builddir) -I$(top_builddir)/simd $(srcdir)/jsimdcfg.inc.h | $(EGREP) "^[\;%]|^\ %" | sed 's%_cpp_protection_%%' | sed 's@% define@%define@g' > $@

data/vendor/github.com/libjpeg-turbo/libjpeg-turbo/simd/jccolext-altivec.c

@@ -0,0 +1,267 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2014, Jay Foad.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-altivec.c */
+
+
+void jsimd_rgb_ycc_convert_altivec (JDIMENSION img_width, JSAMPARRAY input_buf,
+                                    JSAMPIMAGE output_buf,
+                                    JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr0, outptr1, outptr2;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1 = {0}, rgb2 = {0},
+    rgbg0, rgbg1, rgbg2, rgbg3, y, cb, cr;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = {0};
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = {0};
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh, crl, crh, cbl, cbh;
+  __vector int y0, y1, y2, y3, cr0, cr1, cr2, cr3, cb0, cb1, cb2, cb3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) },
+    pw_mf016_mf033 = { __4X2(-F_0_168, -F_0_331) },
+    pw_mf008_mf041 = { __4X2(-F_0_081, -F_0_418) };
+  __vector unsigned short pw_f050_f000 = { __4X2(F_0_500, 0) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) },
+    pd_onehalfm1_cj = { __4X(ONE_HALF - 1 + (CENTERJSAMPLE << SCALEBITS)) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr0 += 16, outptr1 += 16, outptr2 += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = VEC_LD(0, tmpbuf);
+          rgb1 = VEC_LD(16, tmpbuf);
+          rgb2 = VEC_LD(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = VEC_LD(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = VEC_LD(0, inptr);
+          if (num_cols > 16)
+            rgb1 = VEC_LD(16, inptr);
+          if (num_cols > 32)
+            rgb2 = VEC_LD(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = VEC_LD(48, inptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr0);
+
+      /* Calculate Cb values */
+      cb0 = vec_msums(rg0, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb1 = vec_msums(rg1, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb2 = vec_msums(rg2, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb3 = vec_msums(rg3, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb0 = (__vector int)vec_msum((__vector unsigned short)bg0, pw_f050_f000,
+                                   (__vector unsigned int)cb0);
+      cb1 = (__vector int)vec_msum((__vector unsigned short)bg1, pw_f050_f000,
+                                   (__vector unsigned int)cb1);
+      cb2 = (__vector int)vec_msum((__vector unsigned short)bg2, pw_f050_f000,
+                                   (__vector unsigned int)cb2);
+      cb3 = (__vector int)vec_msum((__vector unsigned short)bg3, pw_f050_f000,
+                                   (__vector unsigned int)cb3);
+      cbl = vec_perm((__vector unsigned short)cb0,
+                     (__vector unsigned short)cb1, shift_pack_index);
+      cbh = vec_perm((__vector unsigned short)cb2,
+                     (__vector unsigned short)cb3, shift_pack_index);
+      cb = vec_pack(cbl, cbh);
+      vec_st(cb, 0, outptr1);
+
+      /* Calculate Cr values */
+      cr0 = vec_msums(bg0, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr1 = vec_msums(bg1, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr2 = vec_msums(bg2, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr3 = vec_msums(bg3, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr0 = (__vector int)vec_msum((__vector unsigned short)rg0, pw_f050_f000,
+                                   (__vector unsigned int)cr0);
+      cr1 = (__vector int)vec_msum((__vector unsigned short)rg1, pw_f050_f000,
+                                   (__vector unsigned int)cr1);
+      cr2 = (__vector int)vec_msum((__vector unsigned short)rg2, pw_f050_f000,
+                                   (__vector unsigned int)cr2);
+      cr3 = (__vector int)vec_msum((__vector unsigned short)rg3, pw_f050_f000,
+                                   (__vector unsigned int)cr3);
+      crl = vec_perm((__vector unsigned short)cr0,
+                     (__vector unsigned short)cr1, shift_pack_index);
+      crh = vec_perm((__vector unsigned short)cr2,
+                     (__vector unsigned short)cr3, shift_pack_index);
+      cr = vec_pack(crl, crh);
+      vec_st(cr, 0, outptr2);
+    }
+  }
+}