mewin/libpng
1
0
Fork 0
mirror of https://git.code.sf.net/p/libpng/code.git synced 2025-07-10 18:04:09 +02:00
Code Issues Projects Releases Wiki Activity

[master] Imported from libpng-1.6.29.tar

Browse Source
This commit is contained in:
Glenn Randers-Pehrson 2017-03-16 07:21:03 -05:00
parent 0a181e1665
commit b363e01e6b
46 changed files with 1487 additions and 477 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
ANNOUNCE
View File

@ -1,4 +1,4 @@
Libpng 1.6.28 - January 5, 2017
Libpng 1.6.29 - March 16, 2017
This is a public release of libpng, intended for use in production codes.
@ -7,28 +7,37 @@ Files available for download:
Source files with LF line endings (for Unix/Linux) and with a
"configure" script
libpng-1.6.28.tar.xz (LZMA-compressed, recommended)
libpng-1.6.28.tar.gz
libpng-1.6.29.tar.xz (LZMA-compressed, recommended)
libpng-1.6.29.tar.gz
Source files with CRLF line endings (for Windows), without the
"configure" script
lpng1628.7z (LZMA-compressed, recommended)
lpng1628.zip
lpng1629.7z (LZMA-compressed, recommended)
lpng1629.zip
Other information:
libpng-1.6.28-README.txt
libpng-1.6.28-LICENSE.txt
libpng-1.6.28-*.asc (armored detached GPG signatures)
libpng-1.6.29-README.txt
libpng-1.6.29-LICENSE.txt
libpng-1.6.29-*.asc (armored detached GPG signatures)
Changes since the last public release (1.6.27):
Fixed arm/aarch64 detection in CMakeLists.txt (Gianfranco Costamagna).
Added option to Cmake build allowing a custom location of zlib to be
specified in a scenario where libpng is being built as a subproject
alongside zlib by another project (Sam Serrels).
Changed png_ptr->options from a png_byte to png_uint_32, to accomodate
up to 16 options.
Changes since the last public release (1.6.28):
Readded "include(GNUInstallDirs)" to CMakeLists.txt (Gianfranco Costamagna).
Moved SSE2 optimization code into the main libpng source directory.
Configure libpng with "configure --enable-intel-sse" or compile
libpng with "-DPNG_INTEL_SSE" in CPPFLAGS to enable it.
Simplified conditional compilation in pngvalid.c, for AIX (Michael Felt).
Avoid conditional directives that break statements in pngrutil.c (Romero
Malaquias)
The contrib/examples/pngtopng.c recovery code was in the wrong "if"
branches; the comments were correct.
Added code for PowerPC VSX optimisation (Vadim Barkov).
Avoid potential overflow of shift operations in png_do_expand() (Aaron Boxer).
Change test ZLIB_VERNUM >= 0x1281 to ZLIB_VERNUM >= 0x1290 in pngrutil.c
because Solaris 11 distributes zlib-1.2.8.f that is older than 1.2.8.1.
Suppress clang warnings about implicit sign changes in png.c
Send comments/corrections/commendations to png-mng-implement at lists.sf.net
(subscription required; visit

23
CHANGES
View File

@ -5794,6 +5794,29 @@ Version 1.6.28rc03 [January 4, 2017]
Version 1.6.28 [January 5, 2017]
No changes.
Version 1.6.29beta01 [January 12, 2017]
Readded "include(GNUInstallDirs)" to CMakeLists.txt (Gianfranco Costamagna).
Moved SSE2 optimization code into the main libpng source directory.
Configure libpng with "configure --enable-intel-sse" or compile
libpng with "-DPNG_INTEL_SSE" in CPPFLAGS to enable it.
Simplified conditional compilation in pngvalid.c, for AIX (Michael Felt).
Version 1.6.29beta02 [February 22, 2017]
Avoid conditional directives that break statements in pngrutil.c (Romero
Malaquias)
The contrib/examples/pngtopng.c recovery code was in the wrong "if"
branches; the comments were correct.
Added code for PowerPC VSX optimisation (Vadim Barkov).
Version 1.6.29beta03 [March 1, 2017]
Avoid potential overflow of shift operations in png_do_expand() (Aaron Boxer).
Change test ZLIB_VERNUM >= 0x1281 to ZLIB_VERNUM >= 0x1290 in pngrutil.c
because Solaris 11 distributes zlib-1.2.8.f that is older than 1.2.8.1.
Suppress clang warnings about implicit sign changes in png.c
Version 1.6.29 [March 16, 2017]
No changes.
Send comments/corrections/commendations to png-mng-implement at lists.sf.net
(subscription required; visit
https://lists.sourceforge.net/lists/listinfo/png-mng-implement

44
CMakeLists.txt
View File

@ -36,13 +36,15 @@ enable_testing()
set(PNGLIB_MAJOR 1)
set(PNGLIB_MINOR 6)
set(PNGLIB_RELEASE 28)
set(PNGLIB_RELEASE 29)
set(PNGLIB_NAME libpng${PNGLIB_MAJOR}${PNGLIB_MINOR})
set(PNGLIB_VERSION ${PNGLIB_MAJOR}.${PNGLIB_MINOR}.${PNGLIB_RELEASE})
include(GNUInstallDirs)
# needed packages
#Allow users to specify location of Zlib,
#Allow users to specify location of Zlib,
# Useful if zlib is being built alongside this as a sub-project
option(PNG_BUILD_ZLIB "Custom zlib Location, else find_package is used" OFF)
@ -96,7 +98,7 @@ if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^arm" OR
arm/arm_init.c
arm/filter_neon.S
arm/filter_neon_intrinsics.c)
if(${PNG_ARM_NEON} STREQUAL "on")
add_definitions(-DPNG_ARM_NEON_OPT=2)
elseif(${PNG_ARM_NEON} STREQUAL "check")
@ -107,6 +109,38 @@ if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^arm" OR
endif()
endif()
# set definitions and sources for powerpc
if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^powerpc*" OR
${CMAKE_SYSTEM_PROCESSOR} MATCHES "^ppc64*" )
set(PNG_POWERPC_VSX_POSSIBLE_VALUES check on off)
set(PNG_POWERPC_VSX "check" CACHE STRING "Enable POWERPC VSX optimizations:
check: (default) use internal checking code;
off: disable the optimizations;
on: turn on unconditionally.")
set_property(CACHE PNG_POWERPC_VSX PROPERTY STRINGS
${PNG_POWERPC_VSX_POSSIBLE_VALUES})
list(FIND PNG_POWERPC_VSX_POSSIBLE_VALUES ${PNG_POWERPC_VSX} index)
if(index EQUAL -1)
message(FATAL_ERROR
" PNG_POWERPC_VSX must be one of [${PNG_POWERPC_VSX_POSSIBLE_VALUES}]")
elseif(NOT ${PNG_POWERPC_VSX} STREQUAL "no")
set(libpng_powerpc_sources
powerpc/powerpc_init.c
powerpc/filter_vsx_intrinsics.c)
if(${PNG_POWERPC_VSX} STREQUAL "on")
add_definitions(-DPNG_POWERPC_VSX_OPT=2)
elseif(${PNG_POWERPC_VSX} STREQUAL "check")
add_definitions(-DPNG_POWERPC_VSX_CHECK_SUPPORTED)
message(WARNING
"[PNG_POWERPC_VSX==check] Please check contrib/powerpc/README file for the list of supported OSes.")
endif()
else()
add_definitions(-DPNG_POWERPC_VSX_OPT=0)
endif()
endif()
# SET LIBNAME
set(PNG_LIB_NAME png${PNGLIB_MAJOR}${PNGLIB_MINOR})
@ -398,6 +432,7 @@ set(libpng_sources
pngwtran.c
pngwutil.c
${libpng_arm_sources}
${libpng_powerpc_sources}
)
set(pngtest_sources
pngtest.c
@ -735,7 +770,7 @@ endif(NOT WIN32 OR CYGWIN OR MINGW)
# SET UP LINKS
if(PNG_SHARED)
set_target_properties(png PROPERTIES
# VERSION 16.${PNGLIB_RELEASE}.1.6.28
# VERSION 16.${PNGLIB_RELEASE}.1.6.29
VERSION 16.${PNGLIB_RELEASE}.0
SOVERSION 16
CLEAN_DIRECT_OUTPUT 1)
@ -840,4 +875,3 @@ endif()
# to create msvc import lib for mingw compiled shared lib
# pexports libpng.dll > libpng.def
# lib /def:libpng.def /machine:x86

6
LICENSE
View File

@ -10,7 +10,7 @@ this sentence.
This code is released under the libpng license.
libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are
libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are
Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are
derived from libpng-1.0.6, and are distributed according to the same
disclaimer and license as libpng-1.0.6 with the following individuals
@ -23,6 +23,8 @@ added to the list of Contributing Authors:
Gilles Vollant
James Yu
Mandar Sahastrabuddhe
Google Inc.
Vadim Barkov
and with the following additions to the disclaimer:
@ -128,4 +130,4 @@ any encryption software. See the EAR, paragraphs 734.3(b)(3) and
Glenn Randers-Pehrson
glennrp at users.sourceforge.net
January 5, 2017
March 16, 2017

10
Makefile.am
View File

@ -102,6 +102,16 @@ libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += mips/mips_init.c\
mips/filter_msa_intrinsics.c
endif
if PNG_INTEL_SSE
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += intel/intel_init.c\
intel/filter_sse2_intrinsics.c
endif
if PNG_POWERPC_VSX
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += powerpc/powerpc_init.c\
powerpc/filter_vsx_intrinsics.c
endif
nodist_libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES = pnglibconf.h
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_LDFLAGS = -no-undefined -export-dynamic \

8
README
View File

@ -1,4 +1,4 @@
README for libpng version 1.6.28 - January 5, 2017 (shared library 16.0)
README for libpng version 1.6.29 - March 16, 2017 (shared library 16.0)
See the note about version numbers near the top of png.h
See INSTALL for instructions on how to install libpng.
@ -179,14 +179,16 @@ Files in this distribution:
pngwtran.c => Write data transformations
pngwutil.c => Write utility functions
arm => Contains optimized code for the ARM platform
powerpc => Contains optimized code for the PowerPC platform
contrib => Contributions
arm-neon => Optimized code for ARM-NEON platform
powerpc-vsx => Optimized code for POWERPC-VSX platform
examples => Example programs
gregbook => source code for PNG reading and writing, from
Greg Roelofs' "PNG: The Definitive Guide",
O'Reilly, 1999
intel => Optimized code for INTEL-SSE2 platform
libtests => Test programs
mips-msa => Optimized code for MIPS-MSA platform
pngminim => Minimal decoder, encoder, and progressive decoder
programs demonstrating use of pngusr.dfa
pngminus => Simple pnm2png and png2pnm programs
@ -194,6 +196,8 @@ Files in this distribution:
testpngs
tools => Various tools
visupng => Contains a MSVC workspace for VisualPng
intel => Optimized code for INTEL-SSE2 platform
mips => Optimized code for MIPS platform
projects => Contains project files and workspaces for
building a DLL
owatcom => Contains a WATCOM project for building libpng

1
TODO
View File

@ -25,5 +25,6 @@ Use greater precision when changing to linear gamma for compositing against
Investigate pre-incremented loop counters and other loop constructions.
Add interpolated method of handling interlacing.
Extend pngvalid.c to validate more of the libpng transformations.
Refactor preprocessor conditionals to compile entire statements
*/

87
configure.ac
View File

@ -25,7 +25,7 @@ AC_PREREQ([2.68])
dnl Version number stuff here:
AC_INIT([libpng],[1.6.28],[png-mng-implement@lists.sourceforge.net])
AC_INIT([libpng],[1.6.29],[png-mng-implement@lists.sourceforge.net])
AC_CONFIG_MACRO_DIR([scripts])
# libpng does not follow GNU file name conventions (hence 'foreign')
@ -46,7 +46,7 @@ dnl automake, so the following is not necessary (and is not defined anyway):
dnl AM_PREREQ([1.11.2])
dnl stop configure from automagically running automake
PNGLIB_VERSION=1.6.28
PNGLIB_VERSION=1.6.29
PNGLIB_MAJOR=1
PNGLIB_MINOR=6
PNGLIB_RELEASE=28
@ -391,6 +391,89 @@ AM_CONDITIONAL([PNG_MIPS_MSA],
mipsel*|mips64el*) :;;
esac])
# INTEL
# =====
#
# INTEL SSE (SIMD) support.
AC_ARG_ENABLE([intel-sse],
AS_HELP_STRING([[[--enable-intel-sse]]],
[Enable Intel SSE optimizations: =no/off, yes/on:]
[no/off: disable the optimizations;]
[yes/on: enable the optimizations.]
[If not specified: determined by the compiler.]),
[case "$enableval" in
no|off)
# disable the default enabling:
AC_DEFINE([PNG_INTEL_SSE_OPT], [0],
[Disable Intel SSE optimizations])
# Prevent inclusion of the assembler files below:
enable_intel_sse=no;;
yes|on)
AC_DEFINE([PNG_INTEL_SSE_OPT], [1],
[Enable Intel SSE optimizations]);;
*)
AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value])
esac])
# Add Intel specific files to all builds where the host_cpu is Intel ('x86*')
# or where Intel optimizations were explicitly requested (this allows a
# fallback if a future host CPU does not match 'x86*')
AM_CONDITIONAL([PNG_INTEL_SSE],
[test "$enable_intel_sse" != 'no' &&
case "$host_cpu" in
i?86|x86_64) :;;
*) test "$enable_intel_sse" != '';;
esac])
# PowerPC
# ===
#
# PowerPC VSX (SIMD) support.
AC_ARG_ENABLE([powerpc-vsx],
AS_HELP_STRING([[[--enable-powerpc-vsx]]],
[Enable POWERPC VSX optimizations: =no/off, check, api, yes/on:]
[no/off: disable the optimizations; check: use internal checking code]
[api: disable by default, enable by a call to png_set_option]
[yes/on: turn on unconditionally.]
[If not specified: determined by the compiler.]),
[case "$enableval" in
no|off)
# disable the default enabling on __ppc64__ systems:
AC_DEFINE([PNG_POWERPC_VSX_OPT], [0],
[Disable POWERPC VSX optimizations])
# Prevent inclusion of the platform specific files below:
enable_powerpc_vsx=no;;
check)
AC_DEFINE([PNG_POWERPC_VSX_CHECK_SUPPORTED], [],
[Check for POWERPC VSX support at run-time])
AC_MSG_WARN([--enable-powerpc-vsx Please check contrib/powerpc/README file]
[for the list of supported OSes.]);;
api)
AC_DEFINE([PNG_POWERPC_VSX_API_SUPPORTED], [],
[Turn on POWERPC VSX optimizations at run-time]);;
yes|on)
AC_DEFINE([PNG_POWERPC_VSX_OPT], [2],
[Enable POWERPC VSX optimizations])
AC_MSG_WARN([--enable-powerpc-vsx: please specify 'check' or 'api', if]
[you want the optimizations unconditionally pass '-maltivec -mvsx']
[or '-mcpu=power8'to the compiler.]);;
*)
AC_MSG_ERROR([--enable-powerpc-vsx=${enable_powerpc_vsx}: invalid value])
esac])
# Add PowerPC specific files to all builds where the host_cpu is powerpc('powerpc*') or
# where POWERPC optimizations were explicitly requested (this allows a fallback if a
# future host CPU does not match 'powerpc*')
AM_CONDITIONAL([PNG_POWERPC_VSX],
[test "$enable_powerpc_vsx" != 'no' &&
case "$host_cpu" in
powerpc*|ppc64*) :;;
esac])
AC_MSG_NOTICE([[Extra options for compiler: $PNG_COPTS]])
# Config files, substituting as above

23
contrib/examples/pngtopng.c
View File

@ -1,10 +1,12 @@
/*- pngtopng
*
* COPYRIGHT: Written by John Cunningham Bowler, 2011.
* COPYRIGHT: Written by John Cunningham Bowler, 2011, 2017.
* To the extent possible under law, the author has waived all copyright and
* related or neighboring rights to this work. This work is published from:
* United States.
*
* Last changed in libpng 1.6.29 [%RDATE%]
*
* Read a PNG and write it out in a fixed format, using the 'simplified API'
* that was introduced in libpng-1.6.0.
*
@ -59,26 +61,27 @@ int main(int argc, const char **argv)
else
fprintf(stderr, "pngtopng: write %s: %s\n", argv[2],
image.message);
free(buffer);
}
else
{
fprintf(stderr, "pngtopng: read %s: %s\n", argv[1],
image.message);
/* This is the only place where a 'free' is required; libpng does
* the cleanup on error and success, but in this case we couldn't
* complete the read because of running out of memory.
*/
png_image_free(&image);
}
free(buffer);
}
else
{
fprintf(stderr, "pngtopng: out of memory: %lu bytes\n",
(unsigned long)PNG_IMAGE_SIZE(image));
/* This is the only place where a 'free' is required; libpng does
* the cleanup on error and success, but in this case we couldn't
* complete the read because of running out of memory and so libpng
* has not got to the point where it can do cleanup.
*/
png_image_free(&image);
}
}
else

158
contrib/intel/INSTALL
View File

@ -1,158 +0,0 @@
Enabling SSE support
Copyright (c) 2016 Google, Inc.
Written by Mike Klein, Matt Sarett
This INSTALL file written by Glenn Randers-Pehrson, 2016.
If you have moved intel_init.c and filter_sse2_intrinsics.c to a different
directory, be sure to update the '#include "../../pngpriv.h"' line in both
files if necessary to point to the correct relative location of pngpriv.h
with respect to the new location of those files.
To enable SSE support in libpng, follow the instructions in I, II, or III,
below:
I. Using patched "configure" scripts:
First, apply intel_sse.patch in your build directory.
patch -i contrib/intel/intel_sse.patch -p1
Then, if you are not building in a new GIT clone, e.g., in a tar
distribution, remove any existing pre-built configure scripts:
./configure --enable-maintainer-mode
make maintainer-clean
./autogen.sh --maintainer --clean
Finally, configure libpng with -DPNG_INTEL_SSE in CPPFLAGS:
./autogen.sh --maintainer
CPPFLAGS="-DPNG_INTEL_SSE" ./configure [options]
make CPPFLAGS="-DPNG_INTEL_SSE" [options]
make
II. Using a custom makefile:
If you are using a custom makefile makefile, you will have to update it
manually to include contrib/intel/*.o in the dependencies, and to define
PNG_INTEL_SSE.
III. Using manually updated "configure" scripts:
If you prefer, manually edit pngpriv.h, configure.ac, and Makefile.am,
following the instructions below, then follow the instructions in
section II of INSTALL in the main libpng directory, then configure libpng
with -DPNG_INTEL_SSE in CPPFLAGS.
1. Add the following code to configure.ac under HOST SPECIFIC OPTIONS
directly beneath the section for ARM:
-----------------cut----------------
# INTEL
# =====
#
# INTEL SSE (SIMD) support.
AC_ARG_ENABLE([intel-sse],
AS_HELP_STRING([[[--enable-intel-sse]]],
[Enable Intel SSE optimizations: =no/off, yes/on:]
[no/off: disable the optimizations;]
[yes/on: enable the optimizations.]
[If not specified: determined by the compiler.]),
[case "$enableval" in
no|off)
# disable the default enabling:
AC_DEFINE([PNG_INTEL_SSE_OPT], [0],
[Disable Intel SSE optimizations])
# Prevent inclusion of the assembler files below:
enable_intel_sse=no;;
yes|on)
AC_DEFINE([PNG_INTEL_SSE_OPT], [1],
[Enable Intel SSE optimizations]);;
*)
AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value])
esac])
# Add Intel specific files to all builds where the host_cpu is Intel ('x86*')
# or where Intel optimizations were explicitly requested (this allows a
# fallback if a future host CPU does not match 'x86*')
AM_CONDITIONAL([PNG_INTEL_SSE],
[test "$enable_intel_sse" != 'no' &&
case "$host_cpu" in
i?86|x86_64) :;;
*) test "$enable_intel_sse" != '';;
esac])
-----------------cut----------------
2. Add the following code to Makefile.am under HOST SPECIFIC OPTIONS
directly beneath the "if PNG_ARM_NEON ... endif" statement:
-----------------cut----------------
if PNG_INTEL_SSE
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += contrib/intel/intel_init.c\
contrib/intel/filter_sse2_intrinsics.c
endif
-----------------cut----------------
3. Add the following lines to pngpriv.h, following the PNG_ARM_NEON_OPT
code:
-----------------cut----------------
#ifndef PNG_INTEL_SSE_OPT
# ifdef PNG_INTEL_SSE
/* Only check for SSE if the build configuration has been modified to
* enable SSE optimizations. This means that these optimizations will
* be off by default. See contrib/intel for more details.
*/
# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \
defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
(defined(_M_IX86_FP) && _M_IX86_FP >= 2)
# define PNG_INTEL_SSE_OPT 1
# endif
# endif
#endif
#if PNG_INTEL_SSE_OPT > 0
# ifndef PNG_INTEL_SSE_IMPLEMENTATION
# if defined(__SSE4_1__) || defined(__AVX__)
/* We are not actually using AVX, but checking for AVX is the best
way we can detect SSE4.1 and SSSE3 on MSVC.
*/
# define PNG_INTEL_SSE_IMPLEMENTATION 3
# elif defined(__SSSE3__)
# define PNG_INTEL_SSE_IMPLEMENTATION 2
# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
(defined(_M_IX86_FP) && _M_IX86_FP >= 2)
# define PNG_INTEL_SSE_IMPLEMENTATION 1
# else
# define PNG_INTEL_SSE_IMPLEMENTATION 0
# endif
# endif
# if PNG_INTEL_SSE_IMPLEMENTATION > 0
# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2
# endif
#endif
-----------------cut----------------
4. Add the following lines to pngpriv.h, following the prototype for
png_read_filter_row_paeth4_neon:
-----------------cut----------------
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
-----------------cut----------------

190
contrib/intel/intel_sse.patch
View File

@ -1,190 +0,0 @@
diff --git a/configure.ac b/configure.ac
--- a/configure.ac 2016-08-29 11:46:27.000000000 -0400
+++ b/configure.ac 2016-08-29 16:57:03.866355018 -0400
@@ -386,16 +386,51 @@ AC_ARG_ENABLE([mips-msa],
# future host CPU does not match 'mips*')
AM_CONDITIONAL([PNG_MIPS_MSA],
[test "$enable_mips_msa" != 'no' &&
case "$host_cpu" in
mipsel*|mips64el*) :;;
esac])
+# INTEL
+# =====
+#
+# INTEL SSE (SIMD) support.
+
+AC_ARG_ENABLE([intel-sse],
+ AS_HELP_STRING([[[--enable-intel-sse]]],
+ [Enable Intel SSE optimizations: =no/off, yes/on:]
+ [no/off: disable the optimizations;]
+ [yes/on: enable the optimizations.]
+ [If not specified: determined by the compiler.]),
+ [case "$enableval" in
+ no|off)
+ # disable the default enabling:
+ AC_DEFINE([PNG_INTEL_SSE_OPT], [0],
+ [Disable Intel SSE optimizations])
+ # Prevent inclusion of the assembler files below:
+ enable_intel_sse=no;;
+ yes|on)
+ AC_DEFINE([PNG_INTEL_SSE_OPT], [1],
+ [Enable Intel SSE optimizations]);;
+ *)
+ AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value])
+ esac])
+
+# Add Intel specific files to all builds where the host_cpu is Intel ('x86*')
+# or where Intel optimizations were explicitly requested (this allows a
+# fallback if a future host CPU does not match 'x86*')
+AM_CONDITIONAL([PNG_INTEL_SSE],
+ [test "$enable_intel_sse" != 'no' &&
+ case "$host_cpu" in
+ i?86|x86_64) :;;
+ *) test "$enable_intel_sse" != '';;
+ esac])
+
AC_MSG_NOTICE([[Extra options for compiler: $PNG_COPTS]])
# Config files, substituting as above
AC_CONFIG_FILES([Makefile libpng.pc:libpng.pc.in])
AC_CONFIG_FILES([libpng-config:libpng-config.in],
[chmod +x libpng-config])
AC_OUTPUT
diff --git a/Makefile.am b/Makefile.am
--- a/Makefile.am 2016-08-29 11:46:27.000000000 -0400
+++ b/Makefile.am 2016-08-29 16:57:45.955528215 -0400
@@ -97,16 +97,21 @@ libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SO
arm/filter_neon.S arm/filter_neon_intrinsics.c
endif
if PNG_MIPS_MSA
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += mips/mips_init.c\
mips/filter_msa_intrinsics.c
endif
+if PNG_INTEL_SSE
+libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += contrib/intel/intel_init.c\
+ contrib/intel/filter_sse2_intrinsics.c
+endif
+
nodist_libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES = pnglibconf.h
libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_LDFLAGS = -no-undefined -export-dynamic \
-version-number @PNGLIB_MAJOR@@PNGLIB_MINOR@:@PNGLIB_RELEASE@:0
if HAVE_LD_VERSION_SCRIPT
# Versioned symbols and restricted exports
if HAVE_SOLARIS_LD
diff --git a/pngpriv.h b/pngpriv.h
--- debug16/pngpriv.h 2016-08-30 10:46:36.000000000 -0400
+++ libpng16/pngpriv.h 2016-08-30 11:57:25.672280202 -0400
@@ -185,16 +185,52 @@
#ifndef PNG_MIPS_MSA_OPT
# if defined(__mips_msa) && (__mips_isa_rev >= 5) && defined(PNG_ALIGNED_MEMORY_SUPPORTED)
# define PNG_MIPS_MSA_OPT 2
# else
# define PNG_MIPS_MSA_OPT 0
# endif
#endif
+#ifndef PNG_INTEL_SSE_OPT
+# ifdef PNG_INTEL_SSE
+ /* Only check for SSE if the build configuration has been modified to
+ * enable SSE optimizations. This means that these optimizations will
+ * be off by default. See contrib/intel for more details.
+ */
+# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \
+ defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
+ (defined(_M_IX86_FP) && _M_IX86_FP >= 2)
+# define PNG_INTEL_SSE_OPT 1
+# endif
+# endif
+#endif
+
+#if PNG_INTEL_SSE_OPT > 0
+# ifndef PNG_INTEL_SSE_IMPLEMENTATION
+# if defined(__SSE4_1__) || defined(__AVX__)
+ /* We are not actually using AVX, but checking for AVX is the best
+ way we can detect SSE4.1 and SSSE3 on MSVC.
+ */
+# define PNG_INTEL_SSE_IMPLEMENTATION 3
+# elif defined(__SSSE3__)
+# define PNG_INTEL_SSE_IMPLEMENTATION 2
+# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
+ (defined(_M_IX86_FP) && _M_IX86_FP >= 2)
+# define PNG_INTEL_SSE_IMPLEMENTATION 1
+# else
+# define PNG_INTEL_SSE_IMPLEMENTATION 0
+# endif
+# endif
+
+# if PNG_INTEL_SSE_IMPLEMENTATION > 0
+# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2
+# endif
+#endif
+
#if PNG_MIPS_MSA_OPT > 0
# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_msa
# ifndef PNG_MIPS_MSA_IMPLEMENTATION
# if defined(__mips_msa)
# if defined(__clang__)
# elif defined(__GNUC__)
# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 7)
# define PNG_MIPS_MSA_IMPLEMENTATION 2
@@ -1251,16 +1287,31 @@ PNG_INTERNAL_FUNCTION(void,png_read_filt
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_msa,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_msa,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_msa,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
#endif
+#if PNG_INTEL_SSE_IMPLEMENTATION > 0
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop
+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
+#endif
+
/* Choose the best filter to use and filter the row data */
PNG_INTERNAL_FUNCTION(void,png_write_find_filter,(png_structrp png_ptr,
png_row_infop row_info),PNG_EMPTY);
#ifdef PNG_SEQUENTIAL_READ_SUPPORTED
PNG_INTERNAL_FUNCTION(void,png_read_IDAT_data,(png_structrp png_ptr,
png_bytep output, png_alloc_size_t avail_out),PNG_EMPTY);
/* Read 'avail_out' bytes of data from the IDAT stream. If the output buffer
@@ -1986,16 +2037,21 @@ PNG_INTERNAL_FUNCTION(void, PNG_FILTER_O
PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_neon,
(png_structp png_ptr, unsigned int bpp), PNG_EMPTY);
#endif
#if PNG_MIPS_MSA_OPT > 0
PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_msa,
(png_structp png_ptr, unsigned int bpp), PNG_EMPTY);
#endif
+
+# if PNG_INTEL_SSE_IMPLEMENTATION > 0
+PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_sse2,
+ (png_structp png_ptr, unsigned int bpp), PNG_EMPTY);
+# endif
#endif
PNG_INTERNAL_FUNCTION(png_uint_32, png_check_keyword, (png_structrp png_ptr,
png_const_charp key, png_bytep new_key), PNG_EMPTY);
/* Maintainer: Put new private prototypes here ^ */
#include "pngdebug.h"

16
contrib/libtests/pngvalid.c
View File

@ -1,8 +1,8 @@
/* pngvalid.c - validate libpng by constructing then reading png files.
*
* Last changed in libpng 1.6.27 [December 29, 2016]
* Copyright (c) 2014-2016 John Cunningham Bowler
* Last changed in libpng 1.6.29 [March 16, 2017]
* Copyright (c) 2014-2017 John Cunningham Bowler
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
@ -7742,13 +7742,11 @@ image_transform_png_set_rgb_to_gray_ini(const image_transform *this,
* NOTE: this number only affects the internal limit check in pngvalid,
* it has no effect on the limits applied to the libpng values.
*/
that->pm->limit += pow(
# if DIGITIZE
2.0
# else
1.0
# endif
/255, data.gamma);
#if DIGITIZE
that->pm->limit += pow( 2.0/255, data.gamma);
#else
that->pm->limit += pow( 1.0/255, data.gamma);
#endif
}
}

81
contrib/powerpc-vsx/README Normal file
View File

@ -0,0 +1,81 @@
OPERATING SYSTEM SPECIFIC POWERPC DETECTION
--------------------------------------------
Detection of the ability to execute POWERPC on processor requires
operating system support. (The information is not available in user mode.)
Currently only this feature is supported only for linux platform.
HOW TO USE THIS
---------------
This directory contains C code fragments that can be included in powerpc/powerpc_init.c
by setting the macro PNG_POWERPC_VSX_FILE to the file name in "" or <> at build
time. This setting is not recorded in pnglibconf.h and can be changed simply by
rebuilding arm/arm_init.o with the required macro definition.
For any of this code to be used the POWERPC code must be enabled and run time
checks must be supported. I.e.:
#if PNG_POWERPC_VSX_OPT > 0
#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED
This is done in a 'configure' build by passing configure the argument:
--enable-powerpc-vsx=check
FILE FORMAT
-----------
Each file documents its testing status as of the last time it was tested (which
may have been a long time ago):
STATUS: one of:
SUPPORTED: This indicates that the file is included in the regularly
performed test builds and bugs are fixed when discovered.
COMPILED: This indicates that the code did compile at least once. See the
more detailed description for the extent to which the result was
successful.
TESTED: This means the code was fully compiled into the libpng test programs
and these were run at least once.
BUG REPORTS: an email address to which to send reports of problems
The file is a fragment of C code. It should not define any 'extern' symbols;
everything should be static. It must define the function:
static int png_have_vsx(png_structp png_ptr);
That function must return 1 if ARM NEON instructions are supported, 0 if not.
It must not execute png_error unless it detects a bug. A png_error will prevent
the reading of the PNG and in the future, writing too.
BUG REPORTS
-----------
If you mail a bug report for any file that is not SUPPORTED there may only be
limited response. Consider fixing it and sending a patch to fix the problem -
this is more likely to result in action.
CONTRIBUTIONS
-------------
You may send contributions of new implementations to
png-mng-implement@sourceforge.net. Please write code in strict C90 C where
possible. Obviously OS dependencies are to be expected. If you submit code you
must have the authors permission and it must have a license that is acceptable
to the current maintainer; in particular that license must permit modification
and redistribution.
Please try to make the contribution a single file and give the file a clear and
unambiguous name that identifies the target OS. If multiple files really are
required put them all in a sub-directory.
You must also be prepared to handle bug reports from users of the code, either
by joining the png-mng-implement mailing list or by providing an email for the
"BUG REPORTS" entry or both. Please make sure that the header of the file
contains the STATUS and BUG REPORTS fields as above.
Please list the OS requirements as precisely as possible. Ideally you should
also list the environment in which the code has been tested and certainly list
any environments where you suspect it might not work.

57
contrib/powerpc-vsx/linux.c Normal file
View File

@ -0,0 +1,57 @@
/* contrib/powerpc-vsx/linux.c
*
* Copyright (c) 2017 Glenn Randers-Pehrson
* Written by Vadim Barkov, 2017.
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*
* STATUS: TESTED
* BUG REPORTS: png-mng-implement@sourceforge.net
*
* png_have_vsx implemented for Linux by reading the widely available
* pseudo-file /proc/cpuinfo.
*
* This code is strict ANSI-C and is probably moderately portable; it does
* however use <stdio.h> and it assumes that /proc/cpuinfo is never localized.
*/
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include "png.h"
#ifndef MAXLINE
# define MAXLINE 1024
#endif
static int
png_have_vsx(png_structp png_ptr)
{
FILE *f;
const char *string = "altivec supported";
char input[MAXLINE];
char *token = NULL;
PNG_UNUSED(png_ptr)
f = fopen("/proc/cpuinfo", "r");
if (f != NULL)
{
memset(input,0,MAXLINE);
while(fgets(input,MAXLINE,f) != NULL)
{
token = strstr(input,string);
if(token != NULL)
return 1;
}
}
#ifdef PNG_WARNINGS_SUPPORTED
else
png_warning(png_ptr, "/proc/cpuinfo open failed");
#endif
return 0;
}

36
contrib/powerpc-vsx/linux_aux.c Normal file
View File

@ -0,0 +1,36 @@
/* contrib/powerpc-vsx/linux_aux.c
*
* Copyright (c) 2017 Glenn Randers-Pehrson
* Written by Vadim Barkov, 2017.
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*
* STATUS: TESTED
* BUG REPORTS: png-mng-implement@sourceforge.net
*
* png_have_vsx implemented for Linux by using the auxiliary vector mechanism.
*
* This code is strict ANSI-C and is probably moderately portable; it does
* however use <stdio.h> and it assumes that /proc/cpuinfo is never localized.
*/
#include "sys/auxv.h"
#include "png.h"
static int
png_have_vsx(png_structp png_ptr)
{
const unsigned long auxv = getauxval( AT_HWCAP );
PNG_UNUSED(png_ptr)
if(auxv & (PPC_FEATURE_HAS_ALTIVEC|PPC_FEATURE_HAS_VSX ))
return 1;
else
return 0;
}

9
contrib/intel/filter_sse2_intrinsics.c → intel/filter_sse2_intrinsics.c
View File

@ -1,19 +1,18 @@
/* filter_sse2_intrinsics.c - SSE2 optimized filter functions
*
* Copyright (c) 2016 Google, Inc.
* Copyright (c) 2016-2017 Glenn Randers-Pehrson
* Written by Mike Klein and Matt Sarett
* Derived from arm/filter_neon_intrinsics.c, which was
* Copyright (c) 2014,2016 Glenn Randers-Pehrson
* Derived from arm/filter_neon_intrinsics.c
*
* Last changed in libpng 1.6.24 [August 4, 2016]
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
#include "../../pngpriv.h"
#include "../pngpriv.h"
#ifdef PNG_READ_SUPPORTED

11
contrib/intel/intel_init.c → intel/intel_init.c
View File

@ -1,19 +1,18 @@
/* intel_init.c - SSE2 optimized filter functions
*
* Copyright (c) 2016 Google, Inc.
* Written by Mike Klein and Matt Sarett
* Derived from arm/arm_init.c, which was
* Copyright (c) 2014,2016 Glenn Randers-Pehrson
* Copyright (c) 2016-2017 Glenn Randers-Pehrson
* Written by Mike Klein and Matt Sarett, Google, Inc.
* Derived from arm/arm_init.c
*
* Last changed in libpng 1.6.22 [May 26, 2016]
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
#include "../../pngpriv.h"
#include "../pngpriv.h"
#ifdef PNG_READ_SUPPORTED
#if PNG_INTEL_SSE_IMPLEMENTATION > 0

6
libpng-manual.txt
View File

@ -1,6 +1,6 @@
libpng-manual.txt - A description on how to use and modify libpng
libpng version 1.6.28 - January 5, 2017
libpng version 1.6.29 - March 16, 2017
Updated and distributed by Glenn Randers-Pehrson
<glennrp at users.sourceforge.net>
Copyright (c) 1998-2016 Glenn Randers-Pehrson
@ -11,7 +11,7 @@ libpng-manual.txt - A description on how to use and modify libpng
Based on:
libpng versions 0.97, January 1998, through 1.6.28 - January 5, 2017
libpng versions 0.97, January 1998, through 1.6.29 - March 16, 2017
Updated and distributed by Glenn Randers-Pehrson
Copyright (c) 1998-2016 Glenn Randers-Pehrson
@ -5355,7 +5355,7 @@ Since the PNG Development group is an ad-hoc body, we can't make
an official declaration.
This is your unofficial assurance that libpng from version 0.71 and
upward through 1.6.28 are Y2K compliant. It is my belief that earlier
upward through 1.6.29 are Y2K compliant. It is my belief that earlier
versions were also Y2K compliant.
Libpng only has two year fields. One is a 2-byte unsigned integer

20
libpng.3
View File

@ -1,6 +1,6 @@
.TH LIBPNG 3 "January 5, 2017"
.TH LIBPNG 3 "March 16, 2017"
.SH NAME
libpng \- Portable Network Graphics (PNG) Reference Library 1.6.28
libpng \- Portable Network Graphics (PNG) Reference Library 1.6.29
.SH SYNOPSIS
\fB
#include <png.h>\fP
@ -510,7 +510,7 @@ Following is a copy of the libpng-manual.txt file that accompanies libpng.
.SH LIBPNG.TXT
libpng-manual.txt - A description on how to use and modify libpng
libpng version 1.6.28 - January 5, 2017
libpng version 1.6.29 - March 16, 2017
Updated and distributed by Glenn Randers-Pehrson
<glennrp at users.sourceforge.net>
Copyright (c) 1998-2016 Glenn Randers-Pehrson
@ -521,7 +521,7 @@ libpng-manual.txt - A description on how to use and modify libpng
Based on:
libpng versions 0.97, January 1998, through 1.6.28 - January 5, 2017
libpng versions 0.97, January 1998, through 1.6.29 - March 16, 2017
Updated and distributed by Glenn Randers-Pehrson
Copyright (c) 1998-2016 Glenn Randers-Pehrson
@ -5865,7 +5865,7 @@ Since the PNG Development group is an ad-hoc body, we can't make
an official declaration.
This is your unofficial assurance that libpng from version 0.71 and
upward through 1.6.28 are Y2K compliant. It is my belief that earlier
upward through 1.6.29 are Y2K compliant. It is my belief that earlier
versions were also Y2K compliant.
Libpng only has two year fields. One is a 2-byte unsigned integer
@ -5967,7 +5967,7 @@ the first widely used release:
...
1.5.28 15 10528 15.so.15.28[.0]
...
1.6.28 16 10628 16.so.16.28[.0]
1.6.29 16 10629 16.so.16.29[.0]
Henceforth the source version will match the shared-library minor
and patch numbers; the shared-library major version number will be
@ -6023,7 +6023,7 @@ possible without all of you.
Thanks to Frank J. T. Wojcik for helping with the documentation.
Libpng version 1.6.28 - January 5, 2017:
Libpng version 1.6.29 - March 16, 2017:
Initially created in 1995 by Guy Eric Schalnat, then of Group 42, Inc.
Currently maintained by Glenn Randers-Pehrson (glennrp at users.sourceforge.net).
@ -6048,7 +6048,7 @@ this sentence.
This code is released under the libpng license.
libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are
libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are
Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are
derived from libpng-1.0.6, and are distributed according to the same
disclaimer and license as libpng-1.0.6 with the following individuals
@ -6061,6 +6061,8 @@ added to the list of Contributing Authors:
Gilles Vollant
James Yu
Mandar Sahastrabuddhe
Google Inc.
Vadim Barkov
and with the following additions to the disclaimer:
@ -6174,7 +6176,7 @@ files "pngbar.png" and "pngbar.jpg (88x31) and "pngnow.png" (98x31).
Glenn Randers-Pehrson
glennrp at users.sourceforge.net
January 5, 2017
March 16, 2017
.\" end of man page

4
libpngpf.3
View File

@ -1,6 +1,6 @@
.TH LIBPNGPF 3 "January 5, 2017"
.TH LIBPNGPF 3 "March 16, 2017"
.SH NAME
libpng \- Portable Network Graphics (PNG) Reference Library 1.6.28
libpng \- Portable Network Graphics (PNG) Reference Library 1.6.29
(private functions)
.SH SYNOPSIS
\fB#include \fI"pngpriv.h"

2
png.5
View File

@ -1,4 +1,4 @@
.TH PNG 5 "January 5, 2017"
.TH PNG 5 "March 16, 2017"
.SH NAME
png \- Portable Network Graphics (PNG) format
.SH DESCRIPTION

14
png.c
View File

@ -1,7 +1,7 @@
/* png.c - location for general purpose libpng functions
*
* Last changed in libpng 1.6.28 [January 5, 2017]
* Last changed in libpng 1.6.29 [March 16, 2017]
* Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
* (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
@ -14,7 +14,7 @@
#include "pngpriv.h"
/* Generate a compiler error if there is an old png.h in the search path. */
typedef png_libpng_version_1_6_28 Your_png_h_is_not_version_1_6_28;
typedef png_libpng_version_1_6_29 Your_png_h_is_not_version_1_6_29;
/* Tells libpng that we have already handled the first "num_bytes" bytes
* of the PNG file signature. If the PNG data is embedded into another
@ -776,14 +776,14 @@ png_get_copyright(png_const_structrp png_ptr)
#else
# ifdef __STDC__
return PNG_STRING_NEWLINE \
"libpng version 1.6.28 - January 5, 2017" PNG_STRING_NEWLINE \
"libpng version 1.6.29 - March 16, 2017" PNG_STRING_NEWLINE \
"Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson" \
PNG_STRING_NEWLINE \
"Copyright (c) 1996-1997 Andreas Dilger" PNG_STRING_NEWLINE \
"Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc." \
PNG_STRING_NEWLINE;
# else
return "libpng version 1.6.28 - January 5, 2017\
return "libpng version 1.6.29 - March 16, 2017\
Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson\
Copyright (c) 1996-1997 Andreas Dilger\
Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc.";
@ -4260,13 +4260,13 @@ png_set_option(png_structrp png_ptr, int option, int onoff)
if (png_ptr != NULL && option >= 0 && option < PNG_OPTION_NEXT &&
(option & 1) == 0)
{
png_uint_32 mask = 3 << option;
png_uint_32 setting = (2 + (onoff != 0)) << option;
png_uint_32 mask = 3U << option;
png_uint_32 setting = (2U + (onoff != 0)) << option;
png_uint_32 current = png_ptr->options;
png_ptr->options = (png_uint_32)(((current & ~mask) | setting) & 0xff);
return (current & mask) >> option;
return (int)(current & mask) >> option;
}
return PNG_OPTION_INVALID;

29
png.h
View File

@ -1,7 +1,7 @@
/* png.h - header file for PNG reference library
*
* libpng version 1.6.28, January 5, 2017
* libpng version 1.6.29, March 16, 2017
*
* Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
@ -12,7 +12,7 @@
* Authors and maintainers:
* libpng versions 0.71, May 1995, through 0.88, January 1996: Guy Schalnat
* libpng versions 0.89, June 1996, through 0.96, May 1997: Andreas Dilger
* libpng versions 0.97, January 1998, through 1.6.28, January 5, 2017:
* libpng versions 0.97, January 1998, through 1.6.29, March 16, 2017:
* Glenn Randers-Pehrson.
* See also "Contributing Authors", below.
*/
@ -25,7 +25,7 @@
*
* This code is released under the libpng license.
*
* libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are
* libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are
* Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are
* derived from libpng-1.0.6, and are distributed according to the same
* disclaimer and license as libpng-1.0.6 with the following individuals
@ -38,6 +38,8 @@
* Gilles Vollant
* James Yu
* Mandar Sahastrabuddhe
* Google Inc.
* Vadim Barkov
*
* and with the following additions to the disclaimer:
*
@ -211,7 +213,7 @@
* ...
* 1.5.28 15 10527 15.so.15.28[.0]
* ...
* 1.6.28 16 10628 16.so.16.28[.0]
* 1.6.29 16 10629 16.so.16.29[.0]
*
* Henceforth the source version will match the shared-library major
* and minor numbers; the shared-library major version number will be
@ -239,13 +241,13 @@
* Y2K compliance in libpng:
* =========================
*
* January 5, 2017
* March 16, 2017
*
* Since the PNG Development group is an ad-hoc body, we can't make
* an official declaration.
*
* This is your unofficial assurance that libpng from version 0.71 and
* upward through 1.6.28 are Y2K compliant. It is my belief that
* upward through 1.6.29 are Y2K compliant. It is my belief that
* earlier versions were also Y2K compliant.
*
* Libpng only has two year fields. One is a 2-byte unsigned integer
@ -307,8 +309,8 @@
*/
/* Version information for png.h - this should match the version in png.c */
#define PNG_LIBPNG_VER_STRING "1.6.28"
#define PNG_HEADER_VERSION_STRING " libpng version 1.6.28 - January 5, 2017\n"
#define PNG_LIBPNG_VER_STRING "1.6.29"
#define PNG_HEADER_VERSION_STRING " libpng version 1.6.29 - March 16, 2017\n"
#define PNG_LIBPNG_VER_SONUM 16
#define PNG_LIBPNG_VER_DLLNUM 16
@ -316,7 +318,7 @@
/* These should match the first 3 components of PNG_LIBPNG_VER_STRING: */
#define PNG_LIBPNG_VER_MAJOR 1
#define PNG_LIBPNG_VER_MINOR 6
#define PNG_LIBPNG_VER_RELEASE 28
#define PNG_LIBPNG_VER_RELEASE 29
/* This should match the numeric part of the final component of
* PNG_LIBPNG_VER_STRING, omitting any leading zero:
@ -347,7 +349,7 @@
* version 1.0.0 was mis-numbered 100 instead of 10000). From
* version 1.0.1 it's xxyyzz, where x=major, y=minor, z=release
*/
#define PNG_LIBPNG_VER 10628 /* 1.6.28 */
#define PNG_LIBPNG_VER 10629 /* 1.6.29 */
/* Library configuration: these options cannot be changed after
* the library has been built.
@ -457,7 +459,7 @@ extern "C" {
/* This triggers a compiler error in png.c, if png.c and png.h
* do not agree upon the version number.
*/
typedef char* png_libpng_version_1_6_28;
typedef char* png_libpng_version_1_6_29;
/* Basic control structions. Read libpng-manual.txt or libpng.3 for more info.
*
@ -3224,7 +3226,10 @@ PNG_EXPORT(245, int, png_image_write_to_memory, (png_imagep image, void *memory,
# define PNG_MIPS_MSA 6 /* HARDWARE: MIPS Msa SIMD instructions supported */
#endif
#define PNG_IGNORE_ADLER32 8
#define PNG_OPTION_NEXT 10 /* Next option - numbers must be even */
#ifdef PNG_POWERPC_VSX_API_SUPPORTED
# define PNG_POWERPC_VSX 10 /* HARDWARE: PowerPC VSX SIMD instructions supported */
#endif
#define PNG_OPTION_NEXT 12 /* Next option - numbers must be even */
/* Return values: NOTE: there are four values and 'off' is *not* zero */
#define PNG_OPTION_UNSET 0 /* Unset - defaults to off */

2
pngconf.h
View File

@ -1,7 +1,7 @@
/* pngconf.h - machine configurable file for libpng
*
* libpng version 1.6.28, January 5, 2017
* libpng version 1.6.29, March 16, 2017
*
* Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)

90
pngpriv.h
View File

@ -1,8 +1,8 @@
/* pngpriv.h - private declarations for use inside libpng
*
* Last changed in libpng 1.6.26 [October 20, 2016]
* Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson
* Last changed in libpng 1.6.29 [March 16, 2017]
* Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
* (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
*
@ -190,6 +190,50 @@
# endif
#endif
#ifndef PNG_POWERPC_VSX_OPT
# if defined(__PPC64__) && defined(__ALTIVEC__) && defined(__VSX__)
# define PNG_POWERPC_VSX_OPT 2
# else
# define PNG_POWERPC_VSX_OPT 0
# endif
#endif
#ifndef PNG_INTEL_SSE_OPT
# ifdef PNG_INTEL_SSE
/* Only check for SSE if the build configuration has been modified to
* enable SSE optimizations. This means that these optimizations will
* be off by default. See contrib/intel for more details.
*/
# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \
defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
(defined(_M_IX86_FP) && _M_IX86_FP >= 2)
# define PNG_INTEL_SSE_OPT 1
# endif
# endif
#endif
#if PNG_INTEL_SSE_OPT > 0
# ifndef PNG_INTEL_SSE_IMPLEMENTATION
# if defined(__SSE4_1__) || defined(__AVX__)
/* We are not actually using AVX, but checking for AVX is the best
way we can detect SSE4.1 and SSSE3 on MSVC.
*/
# define PNG_INTEL_SSE_IMPLEMENTATION 3
# elif defined(__SSSE3__)
# define PNG_INTEL_SSE_IMPLEMENTATION 2
# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \
(defined(_M_IX86_FP) && _M_IX86_FP >= 2)
# define PNG_INTEL_SSE_IMPLEMENTATION 1
# else
# define PNG_INTEL_SSE_IMPLEMENTATION 0
# endif
# endif
# if PNG_INTEL_SSE_IMPLEMENTATION > 0
# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2
# endif
#endif
#if PNG_MIPS_MSA_OPT > 0
# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_msa
# ifndef PNG_MIPS_MSA_IMPLEMENTATION
@ -210,6 +254,11 @@
# endif
#endif /* PNG_MIPS_MSA_OPT > 0 */
#if PNG_POWERPC_VSX_OPT > 0
# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_vsx
# define PNG_POWERPC_VSX_IMPLEMENTATION 1
#endif
/* Is this a build of a DLL where compilation of the object modules requires
* different preprocessor settings to those required for a simple library? If
@ -1256,6 +1305,38 @@ PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_msa,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
#endif
#if PNG_POWERPC_VSX_OPT > 0
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_up_vsx,(png_row_infop row_info,
png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_vsx,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
#endif
#if PNG_INTEL_SSE_IMPLEMENTATION > 0
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop
row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY);
#endif
/* Choose the best filter to use and filter the row data */
PNG_INTERNAL_FUNCTION(void,png_write_find_filter,(png_structrp png_ptr,
png_row_infop row_info),PNG_EMPTY);
@ -1991,6 +2072,11 @@ PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_neon,
PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_msa,
(png_structp png_ptr, unsigned int bpp), PNG_EMPTY);
#endif
# if PNG_INTEL_SSE_IMPLEMENTATION > 0
PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_sse2,
(png_structp png_ptr, unsigned int bpp), PNG_EMPTY);
# endif
#endif
PNG_INTERNAL_FUNCTION(png_uint_32, png_check_keyword, (png_structrp png_ptr,

12
pngrtran.c
View File

@ -1,8 +1,8 @@
/* pngrtran.c - transforms the data in a row for PNG readers
*
* Last changed in libpng 1.6.24 [August 4, 2016]
* Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson
* Last changed in libpng 1.6.29 [March 16, 2017]
* Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
* (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
*
@ -4302,7 +4302,7 @@ png_do_expand_palette(png_row_infop row_info, png_bytep row,
if (num_trans > 0)
{
sp = row + (png_size_t)row_width - 1;
dp = row + (png_size_t)(row_width << 2) - 1;
dp = row + ((png_size_t)row_width << 2) - 1;
for (i = 0; i < row_width; i++)
{
@ -4463,7 +4463,7 @@ png_do_expand(png_row_infop row_info, png_bytep row,
{
gray = gray & 0xff;
sp = row + (png_size_t)row_width - 1;
dp = row + (png_size_t)(row_width << 1) - 1;
dp = row + ((png_size_t)row_width << 1) - 1;
for (i = 0; i < row_width; i++)
{
@ -4519,7 +4519,7 @@ png_do_expand(png_row_infop row_info, png_bytep row,
png_byte green = (png_byte)(trans_color->green & 0xff);
png_byte blue = (png_byte)(trans_color->blue & 0xff);
sp = row + (png_size_t)row_info->rowbytes - 1;
dp = row + (png_size_t)(row_width << 2) - 1;
dp = row + ((png_size_t)row_width << 2) - 1;
for (i = 0; i < row_width; i++)
{
if (*(sp - 2) == red && *(sp - 1) == green && *(sp) == blue)
@ -4542,7 +4542,7 @@ png_do_expand(png_row_infop row_info, png_bytep row,
png_byte green_low = (png_byte)(trans_color->green & 0xff);
png_byte blue_low = (png_byte)(trans_color->blue & 0xff);
sp = row + row_info->rowbytes - 1;
dp = row + (png_size_t)(row_width << 3) - 1;
dp = row + ((png_size_t)row_width << 3) - 1;
for (i = 0; i < row_width; i++)
{
if (*(sp - 5) == red_high &&

6
pngrutil.c
View File

@ -1,8 +1,8 @@
/* pngrutil.c - utilities to read a PNG file
*
* Last changed in libpng 1.6.27 [January 5, 2017]
* Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson
* Last changed in libpng 1.6.29 [March 16, 2017]
* Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson
* (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
* (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
*
@ -418,7 +418,7 @@ png_inflate_claim(png_structrp png_ptr, png_uint_32 owner)
png_ptr->flags |= PNG_FLAG_ZSTREAM_INITIALIZED;
}
#if ZLIB_VERNUM >= 0x1281 && \
#if ZLIB_VERNUM >= 0x1290 && \
defined(PNG_SET_OPTION_SUPPORTED) && defined(PNG_IGNORE_ADLER32)
if (((png_ptr->options >> PNG_IGNORE_ADLER32) & 3) == PNG_OPTION_ON)
/* Turn off validation of the ADLER32 checksum in IDAT chunks */

2
pngtest.c
View File

@ -2088,4 +2088,4 @@ main(void)
#endif
/* Generate a compiler error if there is an old png.h in the search path. */
typedef png_libpng_version_1_6_28 Your_png_h_is_not_version_1_6_28;
typedef png_libpng_version_1_6_29 Your_png_h_is_not_version_1_6_29;

19
pngwutil.c
View File

@ -675,6 +675,7 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height,
int interlace_type)
{
png_byte buf[13]; /* Buffer to store the IHDR info */
int is_invalid_depth;
png_debug(1, "in png_write_IHDR");
@ -700,11 +701,11 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height,
break;
case PNG_COLOR_TYPE_RGB:
is_invalid_depth = (bit_depth != 8);
#ifdef PNG_WRITE_16BIT_SUPPORTED
if (bit_depth != 8 && bit_depth != 16)
#else
if (bit_depth != 8)
is_invalid_depth = (is_invalid_depth && bit_depth != 16);
#endif
if (is_invalid_depth)
png_error(png_ptr, "Invalid bit depth for RGB image");
png_ptr->channels = 3;
@ -726,18 +727,22 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height,
break;
case PNG_COLOR_TYPE_GRAY_ALPHA:
if (bit_depth != 8 && bit_depth != 16)
is_invalid_depth = (bit_depth != 8);
#ifdef PNG_WRITE_16BIT_SUPPORTED
is_invalid_depth = (is_invalid_depth && bit_depth != 16);
#endif
if (is_invalid_depth)
png_error(png_ptr, "Invalid bit depth for grayscale+alpha image");
png_ptr->channels = 2;
break;
case PNG_COLOR_TYPE_RGB_ALPHA:
is_invalid_depth = (bit_depth != 8);
#ifdef PNG_WRITE_16BIT_SUPPORTED
if (bit_depth != 8 && bit_depth != 16)
#else
if (bit_depth != 8)
is_invalid_depth = (is_invalid_depth && bit_depth != 16);
#endif
if (is_invalid_depth)
png_error(png_ptr, "Invalid bit depth for RGBA image");
png_ptr->channels = 4;

767
powerpc/filter_vsx_intrinsics.c Normal file
View File

@ -0,0 +1,767 @@
/* filter_vsx_intrinsics.c - PowerPC optimised filter functions
*
* Copyright (c) 2017 Glenn Randers-Pehrson
* Written by Vadim Barkov, 2017.
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
#include <stdio.h>
#include <stdint.h>
#include "../pngpriv.h"
#ifdef PNG_READ_SUPPORTED
/* This code requires -maltivec and -mvsx on the command line: */
#if PNG_POWERPC_VSX_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */
#include <altivec.h>
#if PNG_POWERPC_VSX_OPT > 0
#ifndef __VSX__
# error "This code requires VSX support (POWER7 and later). Please provide -mvsx compiler flag."
#endif
#define vec_ld_unaligned(vec,data) vec = vec_vsx_ld(0,data)
#define vec_st_unaligned(vec,data) vec_vsx_st(vec,0,data)
/* Functions in this file look at most 3 pixels (a,b,c) to predict the 4th (d).
* They're positioned like this:
* prev: c b
* row: a d
* The Sub filter predicts d=a, Avg d=(a+b)/2, and Paeth predicts d to be
* whichever of a, b, or c is closest to p=a+b-c.
* ( this is taken from ../intel/filter_sse2_intrinsics.c )
*/
#define vsx_declare_common_vars(row_info,row,prev_row,offset) \
png_byte i;\
png_bytep rp = row + offset;\
png_const_bytep pp = prev_row;\
png_size_t unaligned_top = 16 - (((png_size_t)rp % 16));\
png_size_t istop;\
if(unaligned_top == 16)\
unaligned_top = 0;\
istop = row_info->rowbytes;\
if((unaligned_top < istop))\
istop -= unaligned_top;\
else{\
unaligned_top = istop;\
istop = 0;\
}
void png_read_filter_row_up_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
vector unsigned char rp_vec;
vector unsigned char pp_vec;
vsx_declare_common_vars(row_info,row,prev_row,0)
/* Altivec operations require 16-byte aligned data
* but input can be unaligned. So we calculate
* unaligned part as usual.
*/
for (i = 0; i < unaligned_top; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff);
rp++;
}
/* Using SIMD while we can */
while( istop >= 16 )
{
rp_vec = vec_ld(0,rp);
vec_ld_unaligned(pp_vec,pp);
rp_vec = vec_add(rp_vec,pp_vec);
vec_st(rp_vec,0,rp);
pp += 16;
rp += 16;
istop -= 16;
}
if(istop > 0)
{
/* If byte count of row is not divisible by 16
* we will process remaining part as usual
*/
for (i = 0; i < istop; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff);
rp++;
}
}
}
static const vector unsigned char VSX_LEFTSHIFTED1_4 = {16,16,16,16, 0, 1, 2, 3,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_LEFTSHIFTED2_4 = {16,16,16,16,16,16,16,16, 4, 5, 6, 7,16,16,16,16};
static const vector unsigned char VSX_LEFTSHIFTED3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 8, 9,10,11};
static const vector unsigned char VSX_LEFTSHIFTED1_3 = {16,16,16, 0, 1, 2,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_LEFTSHIFTED2_3 = {16,16,16,16,16,16, 3, 4, 5,16,16,16,16,16,16,16};
static const vector unsigned char VSX_LEFTSHIFTED3_3 = {16,16,16,16,16,16,16,16,16, 6, 7, 8,16,16,16,16};
static const vector unsigned char VSX_LEFTSHIFTED4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 9,10,11,16};
static const vector unsigned char VSX_NOT_SHIFTED1_4 = {16,16,16,16, 4, 5, 6, 7,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_NOT_SHIFTED2_4 = {16,16,16,16,16,16,16,16, 8, 9,10,11,16,16,16,16};
static const vector unsigned char VSX_NOT_SHIFTED3_4 = {16,16,16,16,16,16,16,16,16,16,16,16,12,13,14,15};
static const vector unsigned char VSX_NOT_SHIFTED1_3 = {16,16,16, 3, 4, 5,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_NOT_SHIFTED2_3 = {16,16,16,16,16,16, 6, 7, 8,16,16,16,16,16,16,16};
static const vector unsigned char VSX_NOT_SHIFTED3_3 = {16,16,16,16,16,16,16,16,16, 9,10,11,16,16,16,16};
static const vector unsigned char VSX_NOT_SHIFTED4_3 = {16,16,16,16,16,16,16,16,16,16,16,16,12,13,14,16};
static const vector unsigned char VSX_CHAR_ZERO = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
#ifdef __LITTLE_ENDIAN__
static const vector unsigned char VSX_CHAR_TO_SHORT1_4 = { 4,16, 5,16, 6,16, 7,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT2_4 = { 8,16, 9,16,10,16,11,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT3_4 = {12,16,13,16,14,16,15,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR1_4 = {16,16,16,16, 0, 2, 4, 6,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR2_4 = {16,16,16,16,16,16,16,16, 0, 2, 4, 6,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 0, 2, 4, 6};
static const vector unsigned char VSX_CHAR_TO_SHORT1_3 = { 3,16, 4,16, 5,16,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT2_3 = { 6,16, 7,16, 8,16,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT3_3 = { 9,16,10,16,11,16,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT4_3 = {12,16,13,16,14,16,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR1_3 = {16,16,16, 0, 2, 4,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR2_3 = {16,16,16,16,16,16, 0, 2, 4,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR3_3 = {16,16,16,16,16,16,16,16,16, 0, 2, 4,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 0, 2, 4,16};
#elif defined(__BIG_ENDIAN__)
static const vector unsigned char VSX_CHAR_TO_SHORT1_4 = {16, 4,16, 5,16, 6,16, 7,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT2_4 = {16, 8,16, 9,16,10,16,11,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT3_4 = {16,12,16,13,16,14,16,15,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR1_4 = {16,16,16,16, 1, 3, 5, 7,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR2_4 = {16,16,16,16,16,16,16,16, 1, 3, 5, 7,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 1, 3, 5, 7};
static const vector unsigned char VSX_CHAR_TO_SHORT1_3 = {16, 3,16, 4,16, 5,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT2_3 = {16, 6,16, 7,16, 8,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT3_3 = {16, 9,16,10,16,11,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_CHAR_TO_SHORT4_3 = {16,12,16,13,16,14,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR1_3 = {16,16,16, 1, 3, 5,16,16,16,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR2_3 = {16,16,16,16,16,16, 1, 3, 5,16,16,16,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR3_3 = {16,16,16,16,16,16,16,16,16, 1, 3, 5,16,16,16,16};
static const vector unsigned char VSX_SHORT_TO_CHAR4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 1, 3, 5,16};
#endif
#define vsx_char_to_short(vec,offset,bpp) (vector unsigned short)vec_perm((vec),VSX_CHAR_ZERO,VSX_CHAR_TO_SHORT##offset##_##bpp)
#define vsx_short_to_char(vec,offset,bpp) vec_perm(((vector unsigned char)(vec)),VSX_CHAR_ZERO,VSX_SHORT_TO_CHAR##offset##_##bpp)
#ifdef PNG_USE_ABS
# define vsx_abs(number) abs(number)
#else
# define vsx_abs(number) (number > 0) ? (number) : -(number)
#endif
void png_read_filter_row_sub4_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 4;
vector unsigned char rp_vec;
vector unsigned char part_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
PNG_UNUSED(pp)
/* Altivec operations require 16-byte aligned data
* but input can be unaligned. So we calculate
* unaligned part as usual.
*/
for (i = 0; i < unaligned_top; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
/* Using SIMD while we can */
while( istop >= 16 )
{
for(i=0;i < bpp ; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
rp -= bpp;
rp_vec = vec_ld(0,rp);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_4);
rp_vec = vec_add(rp_vec,part_vec);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_4);
rp_vec = vec_add(rp_vec,part_vec);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_4);
rp_vec = vec_add(rp_vec,part_vec);
vec_st(rp_vec,0,rp);
rp += 16;
istop -= 16;
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp - bpp))) & 0xff);
rp++;
}
}
void png_read_filter_row_sub3_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 3;
vector unsigned char rp_vec;
vector unsigned char part_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
PNG_UNUSED(pp)
/* Altivec operations require 16-byte aligned data
* but input can be unaligned. So we calculate
* unaligned part as usual.
*/
for (i = 0; i < unaligned_top; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
/* Using SIMD while we can */
while( istop >= 16 )
{
for(i=0;i < bpp ; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
rp -= bpp;
rp_vec = vec_ld(0,rp);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_3);
rp_vec = vec_add(rp_vec,part_vec);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_3);
rp_vec = vec_add(rp_vec,part_vec);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_3);
rp_vec = vec_add(rp_vec,part_vec);
part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED4_3);
rp_vec = vec_add(rp_vec,part_vec);
vec_st(rp_vec,0,rp);
rp += 15;
istop -= 16;
/* Since 16 % bpp = 16 % 3 = 1, last element of array must
* be proceeded manually
*/
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
*rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff);
rp++;
}
}
void png_read_filter_row_avg4_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 4;
vector unsigned char rp_vec;
vector unsigned char pp_vec;
vector unsigned char pp_part_vec;
vector unsigned char rp_part_vec;
vector unsigned char avg_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
rp -= bpp;
if(istop >= bpp)
istop -= bpp;
for (i = 0; i < bpp; i++)
{
*rp = (png_byte)(((int)(*rp) +
((int)(*pp++) / 2 )) & 0xff);
rp++;
}
/* Altivec operations require 16-byte aligned data
* but input can be unaligned. So we calculate
* unaligned part as usual.
*/
for (i = 0; i < unaligned_top; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
/* Using SIMD while we can */
while( istop >= 16 )
{
for(i=0;i < bpp ; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
rp -= bpp;
pp -= bpp;
vec_ld_unaligned(pp_vec,pp);
rp_vec = vec_ld(0,rp);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_4);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED1_4);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_4);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED2_4);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_4);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED3_4);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
vec_st(rp_vec,0,rp);
rp += 16;
pp += 16;
istop -= 16;
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
}
void png_read_filter_row_avg3_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 3;
vector unsigned char rp_vec;
vector unsigned char pp_vec;
vector unsigned char pp_part_vec;
vector unsigned char rp_part_vec;
vector unsigned char avg_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
rp -= bpp;
if(istop >= bpp)
istop -= bpp;
for (i = 0; i < bpp; i++)
{
*rp = (png_byte)(((int)(*rp) +
((int)(*pp++) / 2 )) & 0xff);
rp++;
}
/* Altivec operations require 16-byte aligned data
* but input can be unaligned. So we calculate
* unaligned part as usual.
*/
for (i = 0; i < unaligned_top; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
/* Using SIMD while we can */
while( istop >= 16 )
{
for(i=0;i < bpp ; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
rp -= bpp;
pp -= bpp;
vec_ld_unaligned(pp_vec,pp);
rp_vec = vec_ld(0,rp);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_3);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED1_3);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_3);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED2_3);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_3);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED3_3);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED4_3);
pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED4_3);
avg_vec = vec_avg(rp_part_vec,pp_part_vec);
avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1)));
rp_vec = vec_add(rp_vec,avg_vec);
vec_st(rp_vec,0,rp);
rp += 15;
pp += 15;
istop -= 16;
/* Since 16 % bpp = 16 % 3 = 1, last element of array must
* be proceeded manually
*/
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
*rp = (png_byte)(((int)(*rp) +
(int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff);
rp++;
}
}
/* Bytewise c ? t : e. */
#define if_then_else(c,t,e) vec_sel(e,t,c)
#define vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) {\
c = *(pp - bpp);\
a = *(rp - bpp);\
b = *pp++;\
p = b - c;\
pc = a - c;\
pa = vsx_abs(p);\
pb = vsx_abs(pc);\
pc = vsx_abs(p + pc);\
if (pb < pa) pa = pb, a = b;\
if (pc < pa) a = c;\
a += *rp;\
*rp++ = (png_byte)a;\
}
void png_read_filter_row_paeth4_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 4;
int a, b, c, pa, pb, pc, p;
vector unsigned char rp_vec;
vector unsigned char pp_vec;
vector unsigned short a_vec,b_vec,c_vec,nearest_vec;
vector signed short pa_vec,pb_vec,pc_vec,smallest_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
rp -= bpp;
if(istop >= bpp)
istop -= bpp;
/* Process the first pixel in the row completely (this is the same as 'up'
* because there is only one candidate predictor for the first row).
*/
for(i = 0; i < bpp ; i++)
{
*rp = (png_byte)( *rp + *pp);
rp++;
pp++;
}
for(i = 0; i < unaligned_top ; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
while( istop >= 16)
{
for(i = 0; i < bpp ; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
rp -= bpp;
pp -= bpp;
rp_vec = vec_ld(0,rp);
vec_ld_unaligned(pp_vec,pp);
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_4),1,4);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED1_4),1,4);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_4),1,4);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,1,4)));
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_4),2,4);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED2_4),2,4);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_4),2,4);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,2,4)));
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_4),3,4);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED3_4),3,4);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_4),3,4);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,3,4)));
vec_st(rp_vec,0,rp);
rp += 16;
pp += 16;
istop -= 16;
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
}
void png_read_filter_row_paeth3_vsx(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
const png_byte bpp = 3;
int a, b, c, pa, pb, pc, p;
vector unsigned char rp_vec;
vector unsigned char pp_vec;
vector unsigned short a_vec,b_vec,c_vec,nearest_vec;
vector signed short pa_vec,pb_vec,pc_vec,smallest_vec;
vsx_declare_common_vars(row_info,row,prev_row,bpp)
rp -= bpp;
if(istop >= bpp)
istop -= bpp;
/* Process the first pixel in the row completely (this is the same as 'up'
* because there is only one candidate predictor for the first row).
*/
for(i = 0; i < bpp ; i++)
{
*rp = (png_byte)( *rp + *pp);
rp++;
pp++;
}
for(i = 0; i < unaligned_top ; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
while( istop >= 16)
{
for(i = 0; i < bpp ; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
rp -= bpp;
pp -= bpp;
rp_vec = vec_ld(0,rp);
vec_ld_unaligned(pp_vec,pp);
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_3),1,3);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED1_3),1,3);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_3),1,3);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,1,3)));
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_3),2,3);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED2_3),2,3);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_3),2,3);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,2,3)));
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_3),3,3);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED3_3),3,3);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_3),3,3);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,3,3)));
a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED4_3),4,3);
b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED4_3),4,3);
c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED4_3),4,3);
pa_vec = (vector signed short) vec_sub(b_vec,c_vec);
pb_vec = (vector signed short) vec_sub(a_vec , c_vec);
pc_vec = vec_add(pa_vec,pb_vec);
pa_vec = vec_abs(pa_vec);
pb_vec = vec_abs(pb_vec);
pc_vec = vec_abs(pc_vec);
smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec));
nearest_vec = if_then_else(
vec_cmpeq(pa_vec,smallest_vec),
a_vec,
if_then_else(
vec_cmpeq(pb_vec,smallest_vec),
b_vec,
c_vec
)
);
rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,4,3)));
vec_st(rp_vec,0,rp);
rp += 15;
pp += 15;
istop -= 16;
/* Since 16 % bpp = 16 % 3 = 1, last element of array must
* be proceeded manually
*/
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
if(istop > 0)
for (i = 0; i < istop % 16; i++)
{
vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp)
}
}
#endif /* PNG_POWERPC_VSX_OPT > 0 */
#endif /* PNG_POWERPC_VSX_IMPLEMENTATION == 1 (intrinsics) */
#endif /* READ */

125
powerpc/powerpc_init.c Normal file
View File

@ -0,0 +1,125 @@
/* powerpc_init.c - POWERPC optimised filter functions
*
* Copyright (c) 2017 Glenn Randers-Pehrson
* Written by Vadim Barkov, 2017.
* Last changed in libpng 1.6.29 [March 16, 2017]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
/* Below, after checking __linux__, various non-C90 POSIX 1003.1 functions are
* called.
*/
#define _POSIX_SOURCE 1
#include <stdio.h>
#include "../pngpriv.h"
#ifdef PNG_READ_SUPPORTED
#if PNG_POWERPC_VSX_OPT > 0
#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED /* Do run-time checks */
/* WARNING: it is strongly recommended that you do not build libpng with
* run-time checks for CPU features if at all possible. In the case of the PowerPC
* VSX instructions there is no processor-specific way of detecting the
* presence of the required support, therefore run-time detection is extremely
* OS specific.
*
* You may set the macro PNG_POWERPC_VSX_FILE to the file name of file containing
* a fragment of C source code which defines the png_have_vsx function. There
* are a number of implementations in contrib/powerpc-vsx, but the only one that
* has partial support is contrib/powerpc-vsx/linux.c - a generic Linux
* implementation which reads /proc/cpufino.
*/
#ifndef PNG_POWERPC_VSX_FILE
# ifdef __linux__
# define PNG_POWERPC_VSX_FILE "contrib/powerpc-vsx/linux_aux.c"
# endif
#endif
#ifdef PNG_POWERPC_VSX_FILE
#include <signal.h> /* for sig_atomic_t */
static int png_have_vsx(png_structp png_ptr);
#include PNG_POWERPC_VSX_FILE
#else /* PNG_POWERPC_VSX_FILE */
# error "PNG_POWERPC_VSX_FILE undefined: no support for run-time POWERPC VSX checks"
#endif /* PNG_POWERPC_VSX_FILE */
#endif /* PNG_POWERPC_VSX_CHECK_SUPPORTED */
void
png_init_filter_functions_vsx(png_structp pp, unsigned int bpp)
{
/* The switch statement is compiled in for POWERPC_VSX_API, the call to
* png_have_vsx is compiled in for POWERPC_VSX_CHECK. If both are defined
* the check is only performed if the API has not set the PowerPC option on
* or off explicitly. In this case the check controls what happens.
*/
#ifdef PNG_POWERPC_VSX_API_SUPPORTED
switch ((pp->options >> PNG_POWERPC_VSX) & 3)
{
case PNG_OPTION_UNSET:
/* Allow the run-time check to execute if it has been enabled -
* thus both API and CHECK can be turned on. If it isn't supported
* this case will fall through to the 'default' below, which just
* returns.
*/
#endif /* PNG_POWERPC_VSX_API_SUPPORTED */
#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED
{
static volatile sig_atomic_t no_vsx = -1; /* not checked */
if (no_vsx < 0)
no_vsx = !png_have_vsx(pp);
if (no_vsx)
return;
}
#ifdef PNG_POWERPC_VSX_API_SUPPORTED
break;
#endif
#endif /* PNG_POWERPC_VSX_CHECK_SUPPORTED */
#ifdef PNG_POWERPC_VSX_API_SUPPORTED
default: /* OFF or INVALID */
return;
case PNG_OPTION_ON:
/* Option turned on */
break;
}
#endif
/* IMPORTANT: any new internal functions used here must be declared using
* PNG_INTERNAL_FUNCTION in ../pngpriv.h. This is required so that the
* 'prefix' option to configure works:
*
* ./configure --with-libpng-prefix=foobar_
*
* Verify you have got this right by running the above command, doing a build
* and examining pngprefix.h; it must contain a #define for every external
* function you add. (Notice that this happens automatically for the
* initialization function.)
*/
pp->read_filter[PNG_FILTER_VALUE_UP-1] = png_read_filter_row_up_vsx;
if (bpp == 3)
{
pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub3_vsx;
pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg3_vsx;
pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth3_vsx;
}
else if (bpp == 4)
{
pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub4_vsx;
pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg4_vsx;
pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth4_vsx;
}
}
#endif /* PNG_POWERPC_VSX_OPT > 0 */
#endif /* READ */

2
projects/vstudio/README.txt
View File

@ -1,7 +1,7 @@
VisualStudio instructions
libpng version 1.6.28 - January 5, 2017
libpng version 1.6.29 - March 16, 2017
Copyright (c) 2010,2013,2015 Glenn Randers-Pehrson

2
projects/vstudio/zlib.props
View File

@ -2,7 +2,7 @@
<!--
* zlib.props - location of zlib source
*
* libpng version 1.6.28 - January 5, 2017
* libpng version 1.6.29 - March 16, 2017
*
* Copyright (c) 1998-2011 Glenn Randers-Pehrson
*

10
scripts/README.txt
View File

@ -1,9 +1,9 @@
Makefiles for libpng version 1.6.28 - January 5, 2017
Makefiles for libpng version 1.6.29 - March 16, 2017
pnglibconf.h.prebuilt => Stores configuration settings
makefile.linux => Linux/ELF makefile
(gcc, creates libpng16.so.16.1.6.28)
(gcc, creates libpng16.so.16.1.6.29)
makefile.gcc => Generic makefile (gcc, creates static libpng.a)
makefile.knr => Archaic UNIX Makefile that converts files with
ansi2knr (Requires ansi2knr.c from
@ -33,12 +33,12 @@ pnglibconf.h.prebuilt => Stores configuration settings
makefile.os2 => OS/2 Makefile (gcc and emx, requires libpng.def)
makefile.sco => For SCO OSr5 ELF and Unixware 7 with Native cc
makefile.sggcc => Silicon Graphics (gcc,
creates libpng16.so.16.1.6.28)
creates libpng16.so.16.1.6.29)
makefile.sgi => Silicon Graphics IRIX makefile (cc, creates static lib)
makefile.solaris => Solaris 2.X makefile (gcc,
creates libpng16.so.16.1.6.28)
creates libpng16.so.16.1.6.29)
makefile.so9 => Solaris 9 makefile (gcc,
creates libpng16.so.16.1.6.28)
creates libpng16.so.16.1.6.29)
makefile.std => Generic UNIX makefile (cc, creates static libpng.a)
makefile.sunos => Sun makefile
makefile.32sunu => Sun Ultra 32-bit makefile

2
scripts/def.c
View File

@ -21,7 +21,7 @@ PNG_DFN "OS2 DESCRIPTION "PNG image compression library""
PNG_DFN "OS2 CODE PRELOAD MOVEABLE DISCARDABLE"
PNG_DFN ""
PNG_DFN "EXPORTS"
PNG_DFN ";Version 1.6.28"
PNG_DFN ";Version 1.6.29"
#define PNG_EXPORTA(ordinal, type, name, args, attributes)\
PNG_DFN "@" SYMBOL_PREFIX "@@" name "@"

2
scripts/libpng-config-head.in
View File

@ -11,7 +11,7 @@
# Modeled after libxml-config.
version=1.6.28
version=1.6.29
prefix=""
libdir=""
libs=""

2
scripts/libpng.pc.in
View File

@ -5,6 +5,6 @@ includedir=@includedir@/libpng16
Name: libpng
Description: Loads and saves PNG files
Version: 1.6.28
Version: 1.6.29
Libs: -L${libdir} -lpng16
Cflags: -I${includedir}

2
scripts/makefile.cegcc
View File

@ -23,7 +23,7 @@
VERMAJ = 1
VERMIN = 6
VERMIC = 28
VERMIC = 29
VER = $(VERMAJ).$(VERMIN).$(VERMIC)
NAME = libpng
PACKAGE = $(NAME)-$(VER)

2
scripts/makefile.linux
View File

@ -10,7 +10,7 @@
# Library name:
LIBNAME = libpng16
PNGMAJ = 16
RELEASE = 28
RELEASE = 29
# Shared library names:
LIBSO=$(LIBNAME).so

2
scripts/makefile.msys
View File

@ -18,7 +18,7 @@ exec_prefix=$(prefix)
# Library name:
LIBNAME = libpng16
PNGMAJ = 16
RELEASE = 28
RELEASE = 29
# Shared library names:
LIBSO=$(LIBNAME).dll

2
scripts/makefile.ne12bsd
View File

@ -17,7 +17,7 @@ INCSDIR=${LOCALBASE}/include/libpng16
LIB= png16
SHLIB_MAJOR= 0
SHLIB_MINOR= 1.6.28
SHLIB_MINOR= 1.6.29
SRCS= png.c pngset.c pngget.c pngrutil.c pngtrans.c pngwutil.c \
pngread.c pngrio.c pngwio.c pngwrite.c pngrtran.c \
pngwtran.c pngmem.c pngerror.c pngpread.c

2
scripts/makefile.netbsd
View File

@ -17,7 +17,7 @@ INCSDIR=${LOCALBASE}/include
LIB= png
SHLIB_MAJOR= 16
SHLIB_MINOR= 1.6.28
SHLIB_MINOR= 1.6.29
SRCS= png.c pngset.c pngget.c pngrutil.c pngtrans.c pngwutil.c \
pngread.c pngrio.c pngwio.c pngwrite.c pngrtran.c \
pngwtran.c pngmem.c pngerror.c pngpread.c

2
scripts/makefile.openbsd
View File

@ -11,7 +11,7 @@ LIBDIR= ${PREFIX}/lib
MANDIR= ${PREFIX}/man/cat
SHLIB_MAJOR= 16
SHLIB_MINOR= 1.6.28
SHLIB_MINOR= 1.6.29
LIB= png
SRCS= png.c pngerror.c pngget.c pngmem.c pngpread.c \

27
scripts/pnglibconf.dfa
View File

@ -229,6 +229,33 @@ option ARM_NEON_API disabled requires ALIGNED_MEMORY enables SET_OPTION,
option ARM_NEON_CHECK disabled requires ALIGNED_MEMORY,
sets ARM_NEON_OPT 1
# These options are specific to the PowerPC VSX hardware optimizations.
#
# POWERPC_VSX_OPT: unset: check at compile time (__PPC64__,__ALTIVEC__,__VSX__
# must be defined by the compiler, typically as a result
# of specifying
# "-mvsx -maltivec" compiler flags)
# 0: disable (even if the CPU supports VSX.)
# 1: check at run time (via POWERPC_VSX_{API,CHECK})
# 2: switch on unconditionally (inadvisable - instead pass
# -mvsx -maltivec to compiler options)
# When building libpng avoid using any setting other than '0'; '1' is
# set automatically when either 'API' or 'CHECK' are configured in,
# '2' should not be necessary as "-mvsx -maltivec" will achieve the same
# effect as well as applying VSX optimizations to the rest of the
# libpng code.
# POWERPC_VSX_API: (PNG_POWERPC_VSX == 1) allow the optimization to be switched on
# with png_set_option
# POWERPC_VSX_CHECK: (PNG_POWERPC_VSX == 1) compile a run-time check to see if VSX
# extensions are supported. This is supported not for all OSes
# (see contrib/powerpc/README)
setting POWERPC_VSX_OPT
option POWERPC_VSX_API disabled enables SET_OPTION,
sets POWERPC_VSX_OPT 1
option POWERPC_VSX_CHECK disabled,
sets POWERPC_VSX_OPT 1
# These settings configure the default compression level (0-9) and 'strategy';
# strategy is as defined by the implementors of zlib. It describes the input
# data and modifies the zlib parameters in an attempt to optimize the balance

6
scripts/pnglibconf.h.prebuilt
View File

@ -1,8 +1,8 @@
/* libpng 1.6.28 STANDARD API DEFINITION */
/* libpng 1.6.29 STANDARD API DEFINITION */
/* pnglibconf.h - library build configuration */
/* Libpng version 1.6.28 - January 5, 2017 */
/* Libpng version 1.6.29 - March 16, 2017 */
/* Copyright (c) 1998-2015 Glenn Randers-Pehrson */
@ -20,6 +20,8 @@
#define PNG_ALIGNED_MEMORY_SUPPORTED
/*#undef PNG_ARM_NEON_API_SUPPORTED*/
/*#undef PNG_ARM_NEON_CHECK_SUPPORTED*/
/*#undef PNG_POWERPC_VSX_API_SUPPORTED*/
/*#undef PNG_POWERPC_VSX_CHECK_SUPPORTED*/
#define PNG_BENIGN_ERRORS_SUPPORTED
#define PNG_BENIGN_READ_ERRORS_SUPPORTED
/*#undef PNG_BENIGN_WRITE_ERRORS_SUPPORTED*/

2
scripts/symbols.def
View File

@ -1,4 +1,4 @@
;Version 1.6.28
;Version 1.6.29
;--------------------------------------------------------------
; LIBPNG symbol list as a Win32 DEF file
; Contains all the symbols that can be exported from libpng