TorqueGameEngines/Torque3D
1
0
Fork 0
mirror of https://github.com/TorqueGameEngines/Torque3D.git synced 2026-07-12 23:24:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

update libpng

Browse source 
update libpng, the repo now requires a vcpkg setup for integrating but skipping the install step should allow it to work for windows an linux, mac might need more
This commit is contained in:
marauder2k7 2025-12-31 20:11:14 +00:00
parent c593d860a0
commit 5d644b4ffb
300 changed files with 25573 additions and 17698 deletions
Download patch file Download diff file Expand all files Collapse all files

34
Engine/lib/lpng/arm/arm_init.c
View file

@ -1,7 +1,6 @@
/* arm_init.c - NEON optimised filter functions
*
* Copyright (c) 2018 Cosmin Truta
* Copyright (c) 2018-2022 Cosmin Truta
* Copyright (c) 2014,2016 Glenn Randers-Pehrson
* Written by Mans Rullgard, 2011.
*
@ -10,9 +9,7 @@
* and license in png.h
*/
/* Below, after checking __linux__, various non-C90 POSIX 1003.1 functions are
* called.
*/
/* This module requires POSIX 1003.1 functions. */
#define _POSIX_SOURCE 1
#include "../pngpriv.h"
@ -33,25 +30,30 @@
* has partial support is contrib/arm-neon/linux.c - a generic Linux
* implementation which reads /proc/cpufino.
*/
#include <signal.h> /* for sig_atomic_t */
#ifndef PNG_ARM_NEON_FILE
# ifdef __linux__
# define PNG_ARM_NEON_FILE "contrib/arm-neon/linux.c"
# if defined(__aarch64__) || defined(_M_ARM64)
/* ARM Neon is expected to be unconditionally available on ARM64. */
# error PNG_ARM_NEON_CHECK_SUPPORTED must not be defined on ARM64
# elif defined(__ARM_NEON__) || defined(__ARM_NEON)
/* ARM Neon is expected to be available on the target CPU architecture. */
# error PNG_ARM_NEON_CHECK_SUPPORTED must not be defined on this CPU arch
# elif defined(__linux__)
# define PNG_ARM_NEON_FILE "contrib/arm-neon/linux.c"
# else
# error No support for run-time ARM Neon checking; use compile-time options
# endif
#endif
#ifdef PNG_ARM_NEON_FILE
#include <signal.h> /* for sig_atomic_t */
static int png_have_neon(png_structp png_ptr);
#include PNG_ARM_NEON_FILE
#else /* PNG_ARM_NEON_FILE */
# error "PNG_ARM_NEON_FILE undefined: no support for run-time ARM NEON checks"
#endif /* PNG_ARM_NEON_FILE */
#ifdef PNG_ARM_NEON_FILE
# include PNG_ARM_NEON_FILE
#endif
#endif /* PNG_ARM_NEON_CHECK_SUPPORTED */
#ifndef PNG_ALIGNED_MEMORY_SUPPORTED
# error "ALIGNED_MEMORY is required; set: -DPNG_ALIGNED_MEMORY_SUPPORTED"
# error ALIGNED_MEMORY is required; please define PNG_ALIGNED_MEMORY_SUPPORTED
#endif
void

267
Engine/lib/lpng/arm/filter_neon.S
View file

@ -1,253 +1,60 @@
/* filter_neon.S - NEON optimised filter functions
/* filter_neon.S - placeholder file
*
* Copyright (c) 2018 Cosmin Truta
* Copyright (c) 2014,2017 Glenn Randers-Pehrson
* Written by Mans Rullgard, 2011.
* Copyright (c) 2024 Cosmin Truta
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
/* IMPORTANT NOTE:
*
* Historically, the hand-coded assembler implementation of Neon optimizations
* in this module had not been in sync with the intrinsics-based implementation
* in filter_neon_intrinsics.c and palette_neon_intrinsics.c, at least since
* the introduction of riffled palette optimizations. Moreover, the assembler
* code used to work on 32-bit ARM only, and it caused problems, even if empty,
* on 64-bit ARM.
*
* All references to this module from our internal build scripts and projects
* have been removed.
*
* For the external projects that might still expect this module to be present,
* we leave this stub in place, for the remaining lifetime of libpng-1.6.x.
* Everything should continue to function normally, as long as there are no
* deliberate attempts to use the old hand-made assembler code. A build error
* will be raised otherwise.
*/
/* This is required to get the symbol renames, which are #defines, and the
* definitions (or not) of PNG_ARM_NEON_OPT and PNG_ARM_NEON_IMPLEMENTATION.
*/
#define PNG_VERSION_INFO_ONLY
#include "../pngpriv.h"
#if (defined(__linux__) || defined(__FreeBSD__)) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */
#endif
#ifdef PNG_READ_SUPPORTED
/* Assembler NEON support - only works for 32-bit ARM (i.e. it does not work for
* ARM64). The code in arm/filter_neon_intrinsics.c supports ARM64, however it
* only works if -mfpu=neon is specified on the GCC command line. See pngpriv.h
* for the logic which sets PNG_USE_ARM_NEON_ASM:
*/
#if PNG_ARM_NEON_IMPLEMENTATION == 2 /* hand-coded assembler */
#if PNG_ARM_NEON_OPT > 0
#ifdef __ELF__
# define ELF
#if defined(__clang__)
#define GNUC_VERSION 0 /* not gcc, although it might pretend to be */
#elif defined(__GNUC__)
#define GNUC_MAJOR (__GNUC__ + 0)
#define GNUC_MINOR (__GNUC_MINOR__ + 0)
#define GNUC_PATCHLEVEL (__GNUC_PATCHLEVEL__ + 0)
#define GNUC_VERSION (GNUC_MAJOR * 10000 + GNUC_MINOR * 100 + GNUC_PATCHLEVEL)
#else
# define ELF @
#define GNUC_VERSION 0 /* not gcc */
#endif
.arch armv7-a
.fpu neon
#if (GNUC_VERSION > 0) && (GNUC_VERSION < 40300)
#error "PNG_ARM_NEON is not supported with gcc versions earlier than 4.3.0"
#elif GNUC_VERSION == 40504
#error "PNG_ARM_NEON is not supported with gcc version 4.5.4"
#else
#error "Please use 'arm/*_neon_intrinsics.c' for PNG_ARM_NEON support"
#endif
.macro func name, export=0
.macro endfunc
ELF .size \name, . - \name
.endfunc
.purgem endfunc
.endm
.text
/* Explicitly specifying alignment here because some versions of
* GAS don't align code correctly. This is harmless in correctly
* written versions of GAS.
*/
.align 2
.if \export
.global \name
.endif
ELF .type \name, STT_FUNC
.func \name
\name:
.endm
func png_read_filter_row_sub4_neon, export=1
ldr r3, [r0, #4] @ rowbytes
vmov.i8 d3, #0
1:
vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128]
vadd.u8 d0, d3, d4
vadd.u8 d1, d0, d5
vadd.u8 d2, d1, d6
vadd.u8 d3, d2, d7
vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]!
subs r3, r3, #16
bgt 1b
bx lr
endfunc
func png_read_filter_row_sub3_neon, export=1
ldr r3, [r0, #4] @ rowbytes
vmov.i8 d3, #0
mov r0, r1
mov r2, #3
mov r12, #12
vld1.8 {q11}, [r0], r12
1:
vext.8 d5, d22, d23, #3
vadd.u8 d0, d3, d22
vext.8 d6, d22, d23, #6
vadd.u8 d1, d0, d5
vext.8 d7, d23, d23, #1
vld1.8 {q11}, [r0], r12
vst1.32 {d0[0]}, [r1,:32], r2
vadd.u8 d2, d1, d6
vst1.32 {d1[0]}, [r1], r2
vadd.u8 d3, d2, d7
vst1.32 {d2[0]}, [r1], r2
vst1.32 {d3[0]}, [r1], r2
subs r3, r3, #12
bgt 1b
bx lr
endfunc
func png_read_filter_row_up_neon, export=1
ldr r3, [r0, #4] @ rowbytes
1:
vld1.8 {q0}, [r1,:128]
vld1.8 {q1}, [r2,:128]!
vadd.u8 q0, q0, q1
vst1.8 {q0}, [r1,:128]!
subs r3, r3, #16
bgt 1b
bx lr
endfunc
func png_read_filter_row_avg4_neon, export=1
ldr r12, [r0, #4] @ rowbytes
vmov.i8 d3, #0
1:
vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128]
vld4.32 {d16[],d17[],d18[],d19[]},[r2,:128]!
vhadd.u8 d0, d3, d16
vadd.u8 d0, d0, d4
vhadd.u8 d1, d0, d17
vadd.u8 d1, d1, d5
vhadd.u8 d2, d1, d18
vadd.u8 d2, d2, d6
vhadd.u8 d3, d2, d19
vadd.u8 d3, d3, d7
vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]!
subs r12, r12, #16
bgt 1b
bx lr
endfunc
func png_read_filter_row_avg3_neon, export=1
push {r4,lr}
ldr r12, [r0, #4] @ rowbytes
vmov.i8 d3, #0
mov r0, r1
mov r4, #3
mov lr, #12
vld1.8 {q11}, [r0], lr
1:
vld1.8 {q10}, [r2], lr
vext.8 d5, d22, d23, #3
vhadd.u8 d0, d3, d20
vext.8 d17, d20, d21, #3
vadd.u8 d0, d0, d22
vext.8 d6, d22, d23, #6
vhadd.u8 d1, d0, d17
vext.8 d18, d20, d21, #6
vadd.u8 d1, d1, d5
vext.8 d7, d23, d23, #1
vld1.8 {q11}, [r0], lr
vst1.32 {d0[0]}, [r1,:32], r4
vhadd.u8 d2, d1, d18
vst1.32 {d1[0]}, [r1], r4
vext.8 d19, d21, d21, #1
vadd.u8 d2, d2, d6
vhadd.u8 d3, d2, d19
vst1.32 {d2[0]}, [r1], r4
vadd.u8 d3, d3, d7
vst1.32 {d3[0]}, [r1], r4
subs r12, r12, #12
bgt 1b
pop {r4,pc}
endfunc
.macro paeth rx, ra, rb, rc
vaddl.u8 q12, \ra, \rb @ a + b
vaddl.u8 q15, \rc, \rc @ 2*c
vabdl.u8 q13, \rb, \rc @ pa
vabdl.u8 q14, \ra, \rc @ pb
vabd.u16 q15, q12, q15 @ pc
vcle.u16 q12, q13, q14 @ pa <= pb
vcle.u16 q13, q13, q15 @ pa <= pc
vcle.u16 q14, q14, q15 @ pb <= pc
vand q12, q12, q13 @ pa <= pb && pa <= pc
vmovn.u16 d28, q14
vmovn.u16 \rx, q12
vbsl d28, \rb, \rc
vbsl \rx, \ra, d28
.endm
func png_read_filter_row_paeth4_neon, export=1
ldr r12, [r0, #4] @ rowbytes
vmov.i8 d3, #0
vmov.i8 d20, #0
1:
vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128]
vld4.32 {d16[],d17[],d18[],d19[]},[r2,:128]!
paeth d0, d3, d16, d20
vadd.u8 d0, d0, d4
paeth d1, d0, d17, d16
vadd.u8 d1, d1, d5
paeth d2, d1, d18, d17
vadd.u8 d2, d2, d6
paeth d3, d2, d19, d18
vmov d20, d19
vadd.u8 d3, d3, d7
vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]!
subs r12, r12, #16
bgt 1b
bx lr
endfunc
func png_read_filter_row_paeth3_neon, export=1
push {r4,lr}
ldr r12, [r0, #4] @ rowbytes
vmov.i8 d3, #0
vmov.i8 d4, #0
mov r0, r1
mov r4, #3
mov lr, #12
vld1.8 {q11}, [r0], lr
1:
vld1.8 {q10}, [r2], lr
paeth d0, d3, d20, d4
vext.8 d5, d22, d23, #3
vadd.u8 d0, d0, d22
vext.8 d17, d20, d21, #3
paeth d1, d0, d17, d20
vst1.32 {d0[0]}, [r1,:32], r4
vext.8 d6, d22, d23, #6
vadd.u8 d1, d1, d5
vext.8 d18, d20, d21, #6
paeth d2, d1, d18, d17
vext.8 d7, d23, d23, #1
vld1.8 {q11}, [r0], lr
vst1.32 {d1[0]}, [r1], r4
vadd.u8 d2, d2, d6
vext.8 d19, d21, d21, #1
paeth d3, d2, d19, d18
vst1.32 {d2[0]}, [r1], r4
vmov d4, d19
vadd.u8 d3, d3, d7
vst1.32 {d3[0]}, [r1], r4
subs r12, r12, #12
bgt 1b
pop {r4,pc}
endfunc
#endif /* PNG_ARM_NEON_OPT > 0 */
#endif /* PNG_ARM_NEON_IMPLEMENTATION == 2 (assembler) */
#endif /* PNG_ARM_NEON_IMPLEMENTATION == 2 */
#endif /* READ */

3
Engine/lib/lpng/arm/filter_neon_intrinsics.c
View file

@ -1,4 +1,3 @@
/* filter_neon_intrinsics.c - NEON optimised filter functions
*
* Copyright (c) 2018 Cosmin Truta
@ -18,7 +17,7 @@
/* This code requires -mfpu=neon on the command line: */
#if PNG_ARM_NEON_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */
#if defined(_MSC_VER) && defined(_M_ARM64)
#if defined(_MSC_VER) && !defined(__clang__) && defined(_M_ARM64)
# include <arm64_neon.h>
#else
# include <arm_neon.h>

17
Engine/lib/lpng/arm/palette_neon_intrinsics.c
View file

@ -1,4 +1,3 @@
/* palette_neon_intrinsics.c - NEON optimised palette expansion functions
*
* Copyright (c) 2018-2019 Cosmin Truta
@ -14,7 +13,7 @@
#if PNG_ARM_NEON_IMPLEMENTATION == 1
#if defined(_MSC_VER) && defined(_M_ARM64)
#if defined(_MSC_VER) && !defined(__clang__) && defined(_M_ARM64)
# include <arm64_neon.h>
#else
# include <arm_neon.h>
@ -30,8 +29,6 @@ png_riffle_palette_neon(png_structrp png_ptr)
int num_trans = png_ptr->num_trans;
int i;
png_debug(1, "in png_riffle_palette_neon");
/* Initially black, opaque. */
uint8x16x4_t w = {{
vdupq_n_u8(0x00),
@ -40,6 +37,8 @@ png_riffle_palette_neon(png_structrp png_ptr)
vdupq_n_u8(0xff),
}};
png_debug(1, "in png_riffle_palette_neon");
/* First, riffle the RGB colours into an RGBA8 palette.
* The alpha component is set to opaque for now.
*/
@ -64,12 +63,13 @@ png_do_expand_palette_rgba8_neon(png_structrp png_ptr, png_row_infop row_info,
{
png_uint_32 row_width = row_info->width;
const png_uint_32 *riffled_palette =
(const png_uint_32 *)png_ptr->riffled_palette;
const png_int_32 pixels_per_chunk = 4;
int i;
png_aligncastconst(png_const_uint_32p, png_ptr->riffled_palette);
const png_uint_32 pixels_per_chunk = 4;
png_uint_32 i;
png_debug(1, "in png_do_expand_palette_rgba8_neon");
PNG_UNUSED(row)
if (row_width < pixels_per_chunk)
return 0;
@ -109,10 +109,11 @@ png_do_expand_palette_rgb8_neon(png_structrp png_ptr, png_row_infop row_info,
png_uint_32 row_width = row_info->width;
png_const_bytep palette = (png_const_bytep)png_ptr->palette;
const png_uint_32 pixels_per_chunk = 8;
int i;
png_uint_32 i;
png_debug(1, "in png_do_expand_palette_rgb8_neon");
PNG_UNUSED(row)
if (row_width <= pixels_per_chunk)
return 0;