Robot
/
lnav
mirror of https://github.com/tstack/lnav.git
2
0
Fork 0
Code Issues Releases Wiki Activity

[sysclip] support osc 52 for copying to the clipboard 

Fixes #825
This commit is contained in:
Timothy Stack 2022-07-12 13:16:13 -07:00
parent 3b9235b6e6
commit 3c54f20fce
60 changed files with 4775 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
NEWS
View File

@ -51,6 +51,10 @@ lnav v0.10.2:
* Initial indexing of large files should be faster. Decompression
and searching for line-endings are now pipelined, so they happen
in a thread that is separate from the regular expression matcher.
* Writing to the clipboard now falls back to OSC 52 escape sequence
if none of the clipboard commands could be detected. Your
terminal software will need to support the sequence and you may
need to explicitly enable it in the terminal.
Breaking Changes:
* Added a 'language' column to the lnav_view_filters table that

2
aminclude_static.am
View File

@ -1,6 +1,6 @@
# aminclude_static.am generated automatically by Autoconf
# from AX_AM_MACROS_STATIC on Fri Jul 8 15:54:09 PDT 2022
# from AX_AM_MACROS_STATIC on Tue Jul 12 09:53:25 PDT 2022
# Code coverage

1
configure.ac
View File

@ -324,6 +324,7 @@ AC_CONFIG_FILES([src/tailer/Makefile])
AC_CONFIG_FILES([src/tools/Makefile])
AC_CONFIG_FILES([src/yajl/Makefile])
AC_CONFIG_FILES([src/yajlpp/Makefile])
AC_CONFIG_FILES([src/third-party/base64/lib/Makefile])
AC_CONFIG_FILES([test/Makefile])
AC_OUTPUT

13
src/CMakeLists.txt
View File

@ -236,6 +236,15 @@ add_library(
)
target_include_directories(cppfmt PUBLIC fmtlib)
add_library(
base64 STATIC
third-party/base64/lib/lib.c
third-party/base64/lib/arch/generic/codec.c
third-party/base64/lib/tables/tables.c
)
target_include_directories(base64 PRIVATE third-party/base64/lib)
target_include_directories(base64 PUBLIC third-party/base64/include)
add_library(lnavfileio STATIC
grep_proc.hh
line_buffer.hh
@ -400,6 +409,7 @@ add_library(
lnav.management_cli.hh
lnav_config.hh
lnav_config_fwd.hh
lnav_util.hh
log.watch.hh
log_actions.hh
log_data_helper.hh
@ -513,7 +523,7 @@ add_library(
set(lnav_SRCS lnav.cc)
target_include_directories(diag PUBLIC . fmtlib ${CMAKE_CURRENT_BINARY_DIR}
third-party)
third-party third-party/base64/include)
target_link_libraries(
diag
@ -527,6 +537,7 @@ target_link_libraries(
logfmt
yajlpp
cppfmt
base64
${lnav_LIBS})
target_compile_definitions(diag PRIVATE SQLITE_OMIT_LOAD_EXTENSION)

4
src/Makefile.am
View File

@ -3,7 +3,7 @@ include $(top_srcdir)/aminclude_static.am
CXXFLAGS =
SUBDIRS = tools fmtlib pcrepp base tailer pugixml yajl yajlpp formats/logfmt .
SUBDIRS = tools fmtlib third-party/base64/lib pcrepp base tailer pugixml yajl yajlpp formats/logfmt .
bin_PROGRAMS = lnav
@ -131,6 +131,7 @@ AM_CPPFLAGS = \
-DSQLITE_OMIT_LOAD_EXTENSION \
-I$(srcdir)/fmtlib \
-I$(srcdir)/third-party \
-I$(srcdir)/third-party/base64/include \
-Wall \
$(CODE_COVERAGE_CPPFLAGS) \
$(LIBARCHIVE_CFLAGS) \
@ -151,6 +152,7 @@ LDADD = \
tailer/libtailerpp.a \
yajl/libyajl.a \
yajlpp/libyajlpp.a \
third-party/base64/lib/libbase64.a \
$(READLINE_LIBS) \
$(CURSES_LIB) \
$(SQLITE3_LIBS) \

6
src/base/auto_mem.hh
View File

@ -107,6 +107,12 @@ public:
return &this->am_ptr;
}
template<typename F>
F get_free_func() const
{
return (F) this->am_free_func;
}
void reset(T* ptr = nullptr)
{
if (this->am_ptr != ptr) {

21
src/lnav_commands.cc
View File

@ -1029,14 +1029,15 @@ com_save_to(exec_context& ec,
lnav_data.ld_stdout_used = true;
}
} else if (split_args[0] == "/dev/clipboard") {
toclose = outfile = sysclip::open(sysclip::type_t::GENERAL);
closer = pclose;
if (!outfile) {
auto open_res = sysclip::open(sysclip::type_t::GENERAL);
if (open_res.isErr()) {
alerter::singleton().chime();
return ec.make_error(
"Unable to copy to clipboard. "
"Make sure xclip or pbcopy is installed.");
return ec.make_error("Unable to copy to clipboard: {}",
open_res.unwrapErr());
}
auto holder = open_res.unwrap();
toclose = outfile = holder.release();
closer = holder.get_free_func<int(*)(FILE*)>();
} else if ((outfile = fopen(split_args[0].c_str(), mode)) == nullptr) {
return ec.make_error("unable to open file -- {}", split_args[0]);
} else {
@ -1586,14 +1587,14 @@ com_redirect_to(exec_context& ec,
ec.clear_output();
} else if (split_args[0] == "/dev/clipboard") {
auto out = sysclip::open(sysclip::type_t::GENERAL);
if (!out) {
if (out.isErr()) {
alerter::singleton().chime();
return ec.make_error(
"Unable to copy to clipboard. "
"Make sure xclip or pbcopy is installed.");
"Unable to copy to clipboard: {}", out.unwrapErr());
}
ec.set_output(split_args[0], out, pclose);
auto holder = out.unwrap();
ec.set_output(split_args[0], holder.release(), holder.get_free_func<int(*)(FILE*)>());
} else {
FILE* file = fopen(split_args[0].c_str(), "w");
if (file == nullptr) {

21
src/lnav_util.cc
View File

@ -31,8 +31,6 @@
* Dumping ground for useful functions with no other home.
*/
#include <algorithm>
#include "lnav_util.hh"
#include <stdio.h>
@ -114,16 +112,6 @@ err_to_ok(const lnav::console::user_message msg)
return Ok(msg.to_attr_line().get_string());
}
short
pollfd_revents(const std::vector<struct pollfd>& pollfds, int fd)
{
return pollfds | lnav::itertools::find_if([fd](const auto& entry) {
return entry.fd == fd;
})
| lnav::itertools::deref() | lnav::itertools::map(&pollfd::revents)
| lnav::itertools::unwrap_or((short) 0);
}
void
write_line_to(FILE* outfile, const attr_line_t& al)
{
@ -139,15 +127,6 @@ write_line_to(FILE* outfile, const attr_line_t& al)
}
}
bool
pollfd_ready(const std::vector<struct pollfd>& pollfds, int fd, short events)
{
return std::any_of(
pollfds.begin(), pollfds.end(), [fd, events](const auto& entry) {
return entry.fd == fd && entry.revents & events;
});
}
namespace lnav {
std::string

8
src/lnav_util.hh
View File

@ -137,13 +137,7 @@ is_glob(const std::string& fn)
return (fn.find('*') != std::string::npos
|| fn.find('?') != std::string::npos
|| fn.find('[') != std::string::npos);
};
short pollfd_revents(const std::vector<struct pollfd>& pollfds, int fd);
bool pollfd_ready(const std::vector<struct pollfd>& pollfds,
int fd,
short events = POLLIN | POLLHUP);
}
inline void
rusagesub(const struct rusage& left,

20
src/pollable.cc
View File

@ -31,6 +31,7 @@
#include "pollable.hh"
#include "base/itertools.hh"
#include "base/lnav_log.hh"
pollable::pollable(std::shared_ptr<pollable_supervisor> supervisor,
@ -103,3 +104,22 @@ pollable_supervisor::count(pollable::category cat)
return retval;
}
short
pollfd_revents(const std::vector<struct pollfd>& pollfds, int fd)
{
return pollfds | lnav::itertools::find_if([fd](const auto& entry) {
return entry.fd == fd;
})
| lnav::itertools::deref() | lnav::itertools::map(&pollfd::revents)
| lnav::itertools::unwrap_or((short) 0);
}
bool
pollfd_ready(const std::vector<struct pollfd>& pollfds, int fd, short events)
{
return std::any_of(
pollfds.begin(), pollfds.end(), [fd, events](const auto& entry) {
return entry.fd == fd && entry.revents & events;
});
}

6
src/pollable.hh
View File

@ -77,4 +77,10 @@ public:
size_t count(pollable::category cat);
};
short pollfd_revents(const std::vector<struct pollfd>& pollfds, int fd);
bool pollfd_ready(const std::vector<struct pollfd>& pollfds,
int fd,
short events = POLLIN | POLLHUP);
#endif

9
src/readline_callbacks.cc
View File

@ -43,7 +43,6 @@
#include "service_tags.hh"
#include "sql_help.hh"
#include "sqlite-extension-func.hh"
#include "sysclip.hh"
#include "tailer/tailer.looper.hh"
#include "view_helpers.examples.hh"
#include "vtab_module.hh"
@ -635,17 +634,11 @@ rl_callback_int(readline_curses* rc, bool is_alt)
case ln_mode_t::CAPTURE:
rl_search_internal(rc, old_mode, true);
if (!rc->get_value().empty()) {
auto_mem<FILE> pfile(pclose);
vis_bookmarks& bm = tc->get_bookmarks();
auto& bm = tc->get_bookmarks();
const auto& bv = bm[&textview_curses::BM_SEARCH];
auto vl = is_alt ? bv.prev(tc->get_top())
: bv.next(tc->get_top());
pfile = sysclip::open(sysclip::type_t::FIND);
if (pfile.in() != nullptr) {
fmt::print(
pfile, FMT_STRING("{}"), rc->get_value().get_string());
}
if (vl) {
tc->set_top(vl.value());
} else {

2
src/root-config.json
View File

@ -51,7 +51,7 @@
}
},
"tmux": {
"test": "test -n \"$TMUX\"",
"test": "test -n \"$TMUX\" -a -z \"$SSH_CLIENT\"",
"general": {
"write": "tmux load-buffer -",
"read": "tmux save-buffer -"

84
src/sysclip.cc
View File

@ -32,19 +32,23 @@
#include "sysclip.hh"
#include <stdio.h>
#include <unistd.h>
#include "base/injector.hh"
#include "base/lnav_log.hh"
#include "config.h"
#include "fmt/format.h"
#include "libbase64.h"
#include "sysclip.cfg.hh"
#define ANSI_OSC "\x1b]"
namespace sysclip {
static nonstd::optional<clipboard>
get_commands()
{
auto& cfg = injector::get<const config&>();
const auto& cfg = injector::get<const config&>();
for (const auto& pair : cfg.c_clipboard_impls) {
const auto full_cmd = fmt::format(FMT_STRING("{} > /dev/null 2>&1"),
@ -62,26 +66,76 @@ get_commands()
return nonstd::nullopt;
}
static int
osc52_close(FILE* file)
{
static const char ANSI_OSC_COPY_TO_CLIP[] = ANSI_OSC "52;c;";
log_debug("writing %d bytes of clipboard data using OSC 52", ftell(file));
write(STDOUT_FILENO, ANSI_OSC_COPY_TO_CLIP, strlen(ANSI_OSC_COPY_TO_CLIP));
base64_state b64state{};
base64_stream_encode_init(&b64state, 0);
fseek(file, 0, SEEK_SET);
auto done = false;
while (!done) {
char in_buffer[1024];
char out_buffer[2048];
size_t outlen = 0;
auto rc = fread(in_buffer, 1, sizeof(in_buffer), file);
if (rc <= 0) {
base64_stream_encode_final(&b64state, out_buffer, &outlen);
write(STDOUT_FILENO, out_buffer, outlen);
break;
}
base64_stream_encode(&b64state, in_buffer, rc, out_buffer, &outlen);
write(STDOUT_FILENO, out_buffer, outlen);
}
write(STDOUT_FILENO, "\a", 1);
fclose(file);
return 0;
}
/* XXX For one, this code is kinda crappy. For two, we should probably link
* directly with X so we don't need to have xclip installed and it'll work if
* we're ssh'd into a box.
*/
FILE*
Result<auto_mem<FILE>, std::string>
open(type_t type, op_t op)
{
const char* mode = op == op_t::WRITE ? "w" : "r";
static const auto clip_opt = sysclip::get_commands();
if (!clip_opt) {
log_error("unable to detect clipboard implementation");
return nullptr;
std::string cmd;
if (clip_opt) {
cmd = clip_opt.value().select(type).select(op);
if (cmd.empty()) {
log_info("configured clipboard does not support type/op");
}
} else {
log_info("unable to detect clipboard");
}
auto cmd = clip_opt.value().select(type).select(op);
if (cmd.empty()) {
log_error("clipboard does not support type/op");
return nullptr;
log_info(" ... falling back to OSC 52");
auto_mem<FILE> retval(osc52_close);
retval = tmpfile();
if (retval.in() == nullptr) {
return Err(
fmt::format(FMT_STRING("unable to open temporary file: {}"),
strerror(errno)));
}
return Ok(std::move(retval));
}
switch (op) {
@ -93,7 +147,17 @@ open(type_t type, op_t op)
break;
}
return popen(cmd.c_str(), mode);
auto_mem<FILE> retval(pclose);
log_debug("trying detected clipboard command: %s", cmd.c_str());
retval = popen(cmd.c_str(), mode);
if (retval.in() == nullptr) {
return Err(fmt::format(FMT_STRING("failed to open clipboard: {} -- {}"),
cmd,
strerror(errno)));
}
return Ok(std::move(retval));
}
} // namespace sysclip

6
src/sysclip.hh
View File

@ -33,6 +33,10 @@
#define sysclip_hh
#include <cstdio>
#include <string>
#include "base/auto_mem.hh"
#include "base/result.h"
namespace sysclip {
@ -46,7 +50,7 @@ enum class op_t {
READ,
};
FILE* open(type_t type, op_t op = op_t::WRITE);
Result<auto_mem<FILE>, std::string> open(type_t type, op_t op = op_t::WRITE);
} // namespace sysclip

28
src/third-party/base64/LICENSE vendored Normal file
View File

@ -0,0 +1,28 @@
Copyright (c) 2005-2007, Nick Galbreath
Copyright (c) 2013-2019, Alfred Klomp
Copyright (c) 2015-2017, Wojciech Mula
Copyright (c) 2016-2017, Matthieu Darbois
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

145
src/third-party/base64/include/libbase64.h vendored Normal file
View File

@ -0,0 +1,145 @@
#ifndef LIBBASE64_H
#define LIBBASE64_H
#include <stddef.h> /* size_t */
#if defined(_WIN32) || defined(__CYGWIN__)
#define BASE64_SYMBOL_IMPORT __declspec(dllimport)
#define BASE64_SYMBOL_EXPORT __declspec(dllexport)
#define BASE64_SYMBOL_PRIVATE
#elif __GNUC__ >= 4
#define BASE64_SYMBOL_IMPORT __attribute__ ((visibility ("default")))
#define BASE64_SYMBOL_EXPORT __attribute__ ((visibility ("default")))
#define BASE64_SYMBOL_PRIVATE __attribute__ ((visibility ("hidden")))
#else
#define BASE64_SYMBOL_IMPORT
#define BASE64_SYMBOL_EXPORT
#define BASE64_SYMBOL_PRIVATE
#endif
#if defined(BASE64_STATIC_DEFINE)
#define BASE64_EXPORT
#define BASE64_NO_EXPORT
#else
#if defined(BASE64_EXPORTS) // defined if we are building the shared library
#define BASE64_EXPORT BASE64_SYMBOL_EXPORT
#else
#define BASE64_EXPORT BASE64_SYMBOL_IMPORT
#endif
#define BASE64_NO_EXPORT BASE64_SYMBOL_PRIVATE
#endif
#ifdef __cplusplus
extern "C" {
#endif
/* These are the flags that can be passed in the `flags` argument. The values
* below force the use of a given codec, even if that codec is a no-op in the
* current build. Used in testing. Set to 0 for the default behavior, which is
* runtime feature detection on x86, a compile-time fixed codec on ARM, and
* the plain codec on other platforms: */
#define BASE64_FORCE_AVX2 (1 << 0)
#define BASE64_FORCE_NEON32 (1 << 1)
#define BASE64_FORCE_NEON64 (1 << 2)
#define BASE64_FORCE_PLAIN (1 << 3)
#define BASE64_FORCE_SSSE3 (1 << 4)
#define BASE64_FORCE_SSE41 (1 << 5)
#define BASE64_FORCE_SSE42 (1 << 6)
#define BASE64_FORCE_AVX (1 << 7)
struct base64_state {
int eof;
int bytes;
int flags;
unsigned char carry;
};
/* Wrapper function to encode a plain string of given length. Output is written
* to *out without trailing zero. Output length in bytes is written to *outlen.
* The buffer in `out` has been allocated by the caller and is at least 4/3 the
* size of the input. See above for `flags`; set to 0 for default operation: */
void BASE64_EXPORT base64_encode
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
) ;
/* Call this before calling base64_stream_encode() to init the state. See above
* for `flags`; set to 0 for default operation: */
void BASE64_EXPORT base64_stream_encode_init
( struct base64_state *state
, int flags
) ;
/* Encodes the block of data of given length at `src`, into the buffer at
* `out`. Caller is responsible for allocating a large enough out-buffer; it
* must be at least 4/3 the size of the in-buffer, but take some margin. Places
* the number of new bytes written into `outlen` (which is set to zero when the
* function starts). Does not zero-terminate or finalize the output. */
void BASE64_EXPORT base64_stream_encode
( struct base64_state *state
, const char *src
, size_t srclen
, char *out
, size_t *outlen
) ;
/* Finalizes the output begun by previous calls to `base64_stream_encode()`.
* Adds the required end-of-stream markers if appropriate. `outlen` is modified
* and will contain the number of new bytes written at `out` (which will quite
* often be zero). */
void BASE64_EXPORT base64_stream_encode_final
( struct base64_state *state
, char *out
, size_t *outlen
) ;
/* Wrapper function to decode a plain string of given length. Output is written
* to *out without trailing zero. Output length in bytes is written to *outlen.
* The buffer in `out` has been allocated by the caller and is at least 3/4 the
* size of the input. See above for `flags`, set to 0 for default operation: */
int BASE64_EXPORT base64_decode
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
) ;
/* Call this before calling base64_stream_decode() to init the state. See above
* for `flags`; set to 0 for default operation: */
void BASE64_EXPORT base64_stream_decode_init
( struct base64_state *state
, int flags
) ;
/* Decodes the block of data of given length at `src`, into the buffer at
* `out`. Caller is responsible for allocating a large enough out-buffer; it
* must be at least 3/4 the size of the in-buffer, but take some margin. Places
* the number of new bytes written into `outlen` (which is set to zero when the
* function starts). Does not zero-terminate the output. Returns 1 if all is
* well, and 0 if a decoding error was found, such as an invalid character.
* Returns -1 if the chosen codec is not included in the current build. Used by
* the test harness to check whether a codec is available for testing. */
int BASE64_EXPORT base64_stream_decode
( struct base64_state *state
, const char *src
, size_t srclen
, char *out
, size_t *outlen
) ;
#ifdef __cplusplus
}
#endif
#endif /* LIBBASE64_H */

22
src/third-party/base64/lib/Makefile.am vendored Normal file
View File

@ -0,0 +1,22 @@
noinst_HEADERS = \
../include/libbase64.h \
arch/generic/32/enc_loop.c \
arch/generic/32/dec_loop.c \
arch/generic/enc_tail.c \
arch/generic/dec_tail.c \
arch/generic/64/enc_loop.c \
arch/generic/enc_head.c \
arch/generic/dec_head.c \
tables/tables.h \
tables/table_enc_12bit.h \
codecs.h \
config.h \
env.h
noinst_LIBRARIES = libbase64.a
libbase64_a_SOURCES = \
lib.c \
arch/generic/codec.c \
tables/tables.c

42
src/third-party/base64/lib/arch/avx/codec.c vendored Normal file
View File

@ -0,0 +1,42 @@
#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if HAVE_AVX
#include <immintrin.h>
#include "../ssse3/dec_reshuffle.c"
#include "../ssse3/dec_loop.c"
#include "../ssse3/enc_translate.c"
#include "../ssse3/enc_reshuffle.c"
#include "../ssse3/enc_loop.c"
#endif // HAVE_AVX
BASE64_ENC_FUNCTION(avx)
{
#if HAVE_AVX
#include "../generic/enc_head.c"
enc_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(avx)
{
#if HAVE_AVX
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

42
src/third-party/base64/lib/arch/avx2/codec.c vendored Normal file
View File

@ -0,0 +1,42 @@
#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if HAVE_AVX2
#include <immintrin.h>
#include "dec_reshuffle.c"
#include "dec_loop.c"
#include "enc_translate.c"
#include "enc_reshuffle.c"
#include "enc_loop.c"
#endif // HAVE_AVX2
BASE64_ENC_FUNCTION(avx2)
{
#if HAVE_AVX2
#include "../generic/enc_head.c"
enc_loop_avx2(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(avx2)
{
#if HAVE_AVX2
#include "../generic/dec_head.c"
dec_loop_avx2(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

110
src/third-party/base64/lib/arch/avx2/dec_loop.c vendored Normal file
View File

@ -0,0 +1,110 @@
static inline int
dec_loop_avx2_inner (const uint8_t **s, uint8_t **o, size_t *rounds)
{
const __m256i lut_lo = _mm256_setr_epi8(
0x15, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11,
0x11, 0x11, 0x13, 0x1A, 0x1B, 0x1B, 0x1B, 0x1A,
0x15, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11,
0x11, 0x11, 0x13, 0x1A, 0x1B, 0x1B, 0x1B, 0x1A);
const __m256i lut_hi = _mm256_setr_epi8(
0x10, 0x10, 0x01, 0x02, 0x04, 0x08, 0x04, 0x08,
0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10,
0x10, 0x10, 0x01, 0x02, 0x04, 0x08, 0x04, 0x08,
0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10);
const __m256i lut_roll = _mm256_setr_epi8(
0, 16, 19, 4, -65, -65, -71, -71,
0, 0, 0, 0, 0, 0, 0, 0,
0, 16, 19, 4, -65, -65, -71, -71,
0, 0, 0, 0, 0, 0, 0, 0);
const __m256i mask_2F = _mm256_set1_epi8(0x2F);
// Load input:
__m256i str = _mm256_loadu_si256((__m256i *) *s);
// See the SSSE3 decoder for an explanation of the algorithm.
const __m256i hi_nibbles = _mm256_and_si256(_mm256_srli_epi32(str, 4), mask_2F);
const __m256i lo_nibbles = _mm256_and_si256(str, mask_2F);
const __m256i hi = _mm256_shuffle_epi8(lut_hi, hi_nibbles);
const __m256i lo = _mm256_shuffle_epi8(lut_lo, lo_nibbles);
if (!_mm256_testz_si256(lo, hi)) {
return 0;
}
const __m256i eq_2F = _mm256_cmpeq_epi8(str, mask_2F);
const __m256i roll = _mm256_shuffle_epi8(lut_roll, _mm256_add_epi8(eq_2F, hi_nibbles));
// Now simply add the delta values to the input:
str = _mm256_add_epi8(str, roll);
// Reshuffle the input to packed 12-byte output format:
str = dec_reshuffle(str);
// Store the output:
_mm256_storeu_si256((__m256i *) *o, str);
*s += 32;
*o += 24;
*rounds -= 1;
return 1;
}
static inline void
dec_loop_avx2 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 45) {
return;
}
// Process blocks of 32 bytes per round. Because 8 extra zero bytes are
// written after the output, ensure that there will be at least 13
// bytes of input data left to cover the gap. (11 data bytes and up to
// two end-of-string markers.)
size_t rounds = (*slen - 13) / 32;
*slen -= rounds * 32; // 32 bytes consumed per round
*olen += rounds * 24; // 24 bytes produced per round
do {
if (rounds >= 8) {
if (dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 4) {
if (dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 2) {
if (dec_loop_avx2_inner(s, o, &rounds) &&
dec_loop_avx2_inner(s, o, &rounds)) {
continue;
}
break;
}
dec_loop_avx2_inner(s, o, &rounds);
break;
} while (rounds > 0);
// Adjust for any rounds that were skipped:
*slen += rounds * 32;
*olen -= rounds * 24;
}

34
src/third-party/base64/lib/arch/avx2/dec_reshuffle.c vendored Normal file
View File

@ -0,0 +1,34 @@
static inline __m256i
dec_reshuffle (const __m256i in)
{
// in, lower lane, bits, upper case are most significant bits, lower
// case are least significant bits:
// 00llllll 00kkkkLL 00jjKKKK 00JJJJJJ
// 00iiiiii 00hhhhII 00ggHHHH 00GGGGGG
// 00ffffff 00eeeeFF 00ddEEEE 00DDDDDD
// 00cccccc 00bbbbCC 00aaBBBB 00AAAAAA
const __m256i merge_ab_and_bc = _mm256_maddubs_epi16(in, _mm256_set1_epi32(0x01400140));
// 0000kkkk LLllllll 0000JJJJ JJjjKKKK
// 0000hhhh IIiiiiii 0000GGGG GGggHHHH
// 0000eeee FFffffff 0000DDDD DDddEEEE
// 0000bbbb CCcccccc 0000AAAA AAaaBBBB
__m256i out = _mm256_madd_epi16(merge_ab_and_bc, _mm256_set1_epi32(0x00011000));
// 00000000 JJJJJJjj KKKKkkkk LLllllll
// 00000000 GGGGGGgg HHHHhhhh IIiiiiii
// 00000000 DDDDDDdd EEEEeeee FFffffff
// 00000000 AAAAAAaa BBBBbbbb CCcccccc
// Pack bytes together in each lane:
out = _mm256_shuffle_epi8(out, _mm256_setr_epi8(
2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12, -1, -1, -1, -1,
2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12, -1, -1, -1, -1));
// 00000000 00000000 00000000 00000000
// LLllllll KKKKkkkk JJJJJJjj IIiiiiii
// HHHHhhhh GGGGGGgg FFffffff EEEEeeee
// DDDDDDdd CCcccccc BBBBbbbb AAAAAAaa
// Pack lanes:
return _mm256_permutevar8x32_epi32(out, _mm256_setr_epi32(0, 1, 2, 4, 5, 6, -1, -1));
}

89
src/third-party/base64/lib/arch/avx2/enc_loop.c vendored Normal file
View File

@ -0,0 +1,89 @@
static inline void
enc_loop_avx2_inner_first (const uint8_t **s, uint8_t **o)
{
// First load is done at s - 0 to not get a segfault:
__m256i src = _mm256_loadu_si256((__m256i *) *s);
// Shift by 4 bytes, as required by enc_reshuffle:
src = _mm256_permutevar8x32_epi32(src, _mm256_setr_epi32(0, 0, 1, 2, 3, 4, 5, 6));
// Reshuffle, translate, store:
src = enc_reshuffle(src);
src = enc_translate(src);
_mm256_storeu_si256((__m256i *) *o, src);
// Subsequent loads will be done at s - 4, set pointer for next round:
*s += 20;
*o += 32;
}
static inline void
enc_loop_avx2_inner (const uint8_t **s, uint8_t **o)
{
// Load input:
__m256i src = _mm256_loadu_si256((__m256i *) *s);
// Reshuffle, translate, store:
src = enc_reshuffle(src);
src = enc_translate(src);
_mm256_storeu_si256((__m256i *) *o, src);
*s += 24;
*o += 32;
}
static inline void
enc_loop_avx2 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 32) {
return;
}
// Process blocks of 24 bytes at a time. Because blocks are loaded 32
// bytes at a time an offset of -4, ensure that there will be at least
// 4 remaining bytes after the last round, so that the final read will
// not pass beyond the bounds of the input buffer:
size_t rounds = (*slen - 4) / 24;
*slen -= rounds * 24; // 24 bytes consumed per round
*olen += rounds * 32; // 32 bytes produced per round
// The first loop iteration requires special handling to ensure that
// the read, which is done at an offset, does not underflow the buffer:
enc_loop_avx2_inner_first(s, o);
rounds--;
while (rounds > 0) {
if (rounds >= 8) {
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_avx2_inner(s, o);
enc_loop_avx2_inner(s, o);
rounds -= 2;
continue;
}
enc_loop_avx2_inner(s, o);
break;
}
// Add the offset back:
*s += 4;
}

83
src/third-party/base64/lib/arch/avx2/enc_reshuffle.c vendored Normal file
View File

@ -0,0 +1,83 @@
static inline __m256i
enc_reshuffle (const __m256i input)
{
// Translation of the SSSE3 reshuffling algorithm to AVX2. This one
// works with shifted (4 bytes) input in order to be able to work
// efficiently in the two 128-bit lanes.
// Input, bytes MSB to LSB:
// 0 0 0 0 x w v u t s r q p o n m
// l k j i h g f e d c b a 0 0 0 0
const __m256i in = _mm256_shuffle_epi8(input, _mm256_set_epi8(
10, 11, 9, 10,
7, 8, 6, 7,
4, 5, 3, 4,
1, 2, 0, 1,
14, 15, 13, 14,
11, 12, 10, 11,
8, 9, 7, 8,
5, 6, 4, 5));
// in, bytes MSB to LSB:
// w x v w
// t u s t
// q r p q
// n o m n
// k l j k
// h i g h
// e f d e
// b c a b
const __m256i t0 = _mm256_and_si256(in, _mm256_set1_epi32(0x0FC0FC00));
// bits, upper case are most significant bits, lower case are least
// significant bits.
// 0000wwww XX000000 VVVVVV00 00000000
// 0000tttt UU000000 SSSSSS00 00000000
// 0000qqqq RR000000 PPPPPP00 00000000
// 0000nnnn OO000000 MMMMMM00 00000000
// 0000kkkk LL000000 JJJJJJ00 00000000
// 0000hhhh II000000 GGGGGG00 00000000
// 0000eeee FF000000 DDDDDD00 00000000
// 0000bbbb CC000000 AAAAAA00 00000000
const __m256i t1 = _mm256_mulhi_epu16(t0, _mm256_set1_epi32(0x04000040));
// 00000000 00wwwwXX 00000000 00VVVVVV
// 00000000 00ttttUU 00000000 00SSSSSS
// 00000000 00qqqqRR 00000000 00PPPPPP
// 00000000 00nnnnOO 00000000 00MMMMMM
// 00000000 00kkkkLL 00000000 00JJJJJJ
// 00000000 00hhhhII 00000000 00GGGGGG
// 00000000 00eeeeFF 00000000 00DDDDDD
// 00000000 00bbbbCC 00000000 00AAAAAA
const __m256i t2 = _mm256_and_si256(in, _mm256_set1_epi32(0x003F03F0));
// 00000000 00xxxxxx 000000vv WWWW0000
// 00000000 00uuuuuu 000000ss TTTT0000
// 00000000 00rrrrrr 000000pp QQQQ0000
// 00000000 00oooooo 000000mm NNNN0000
// 00000000 00llllll 000000jj KKKK0000
// 00000000 00iiiiii 000000gg HHHH0000
// 00000000 00ffffff 000000dd EEEE0000
// 00000000 00cccccc 000000aa BBBB0000
const __m256i t3 = _mm256_mullo_epi16(t2, _mm256_set1_epi32(0x01000010));
// 00xxxxxx 00000000 00vvWWWW 00000000
// 00uuuuuu 00000000 00ssTTTT 00000000
// 00rrrrrr 00000000 00ppQQQQ 00000000
// 00oooooo 00000000 00mmNNNN 00000000
// 00llllll 00000000 00jjKKKK 00000000
// 00iiiiii 00000000 00ggHHHH 00000000
// 00ffffff 00000000 00ddEEEE 00000000
// 00cccccc 00000000 00aaBBBB 00000000
return _mm256_or_si256(t1, t3);
// 00xxxxxx 00wwwwXX 00vvWWWW 00VVVVVV
// 00uuuuuu 00ttttUU 00ssTTTT 00SSSSSS
// 00rrrrrr 00qqqqRR 00ppQQQQ 00PPPPPP
// 00oooooo 00nnnnOO 00mmNNNN 00MMMMMM
// 00llllll 00kkkkLL 00jjKKKK 00JJJJJJ
// 00iiiiii 00hhhhII 00ggHHHH 00GGGGGG
// 00ffffff 00eeeeFF 00ddEEEE 00DDDDDD
// 00cccccc 00bbbbCC 00aaBBBB 00AAAAAA
}

30
src/third-party/base64/lib/arch/avx2/enc_translate.c vendored Normal file
View File

@ -0,0 +1,30 @@
static inline __m256i
enc_translate (const __m256i in)
{
// A lookup table containing the absolute offsets for all ranges:
const __m256i lut = _mm256_setr_epi8(
65, 71, -4, -4, -4, -4, -4, -4, -4, -4, -4, -4, -19, -16, 0, 0,
65, 71, -4, -4, -4, -4, -4, -4, -4, -4, -4, -4, -19, -16, 0, 0);
// Translate values 0..63 to the Base64 alphabet. There are five sets:
// # From To Abs Index Characters
// 0 [0..25] [65..90] +65 0 ABCDEFGHIJKLMNOPQRSTUVWXYZ
// 1 [26..51] [97..122] +71 1 abcdefghijklmnopqrstuvwxyz
// 2 [52..61] [48..57] -4 [2..11] 0123456789
// 3 [62] [43] -19 12 +
// 4 [63] [47] -16 13 /
// Create LUT indices from the input. The index for range #0 is right,
// others are 1 less than expected:
__m256i indices = _mm256_subs_epu8(in, _mm256_set1_epi8(51));
// mask is 0xFF (-1) for range #[1..4] and 0x00 for range #0:
const __m256i mask = _mm256_cmpgt_epi8(in, _mm256_set1_epi8(25));
// Subtract -1, so add 1 to indices for range #[1..4]. All indices are
// now correct:
indices = _mm256_sub_epi8(indices, mask);
// Add offsets to input values:
return _mm256_add_epi8(in, _mm256_shuffle_epi8(lut, indices));
}

86
src/third-party/base64/lib/arch/generic/32/dec_loop.c vendored Normal file
View File

@ -0,0 +1,86 @@
static inline int
dec_loop_generic_32_inner (const uint8_t **s, uint8_t **o, size_t *rounds)
{
const uint32_t str
= base64_table_dec_32bit_d0[(*s)[0]]
| base64_table_dec_32bit_d1[(*s)[1]]
| base64_table_dec_32bit_d2[(*s)[2]]
| base64_table_dec_32bit_d3[(*s)[3]];
#if BASE64_LITTLE_ENDIAN
// LUTs for little-endian set MSB in case of invalid character:
if (str & UINT32_C(0x80000000)) {
return 0;
}
#else
// LUTs for big-endian set LSB in case of invalid character:
if (str & UINT32_C(1)) {
return 0;
}
#endif
// Store the output:
memcpy(*o, &str, sizeof (str));
*s += 4;
*o += 3;
*rounds -= 1;
return 1;
}
static inline void
dec_loop_generic_32 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 8) {
return;
}
// Process blocks of 4 bytes per round. Because one extra zero byte is
// written after the output, ensure that there will be at least 4 bytes
// of input data left to cover the gap. (Two data bytes and up to two
// end-of-string markers.)
size_t rounds = (*slen - 4) / 4;
*slen -= rounds * 4; // 4 bytes consumed per round
*olen += rounds * 3; // 3 bytes produced per round
do {
if (rounds >= 8) {
if (dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 4) {
if (dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 2) {
if (dec_loop_generic_32_inner(s, o, &rounds) &&
dec_loop_generic_32_inner(s, o, &rounds)) {
continue;
}
break;
}
dec_loop_generic_32_inner(s, o, &rounds);
break;
} while (rounds > 0);
// Adjust for any rounds that were skipped:
*slen += rounds * 4;
*olen -= rounds * 3;
}

73
src/third-party/base64/lib/arch/generic/32/enc_loop.c vendored Normal file
View File

@ -0,0 +1,73 @@
static inline void
enc_loop_generic_32_inner (const uint8_t **s, uint8_t **o)
{
uint32_t src;
// Load input:
memcpy(&src, *s, sizeof (src));
// Reorder to 32-bit big-endian, if not already in that format. The
// workset must be in big-endian, otherwise the shifted bits do not
// carry over properly among adjacent bytes:
src = BASE64_HTOBE32(src);
// Two indices for the 12-bit lookup table:
const size_t index0 = (src >> 20) & 0xFFFU;
const size_t index1 = (src >> 8) & 0xFFFU;
// Table lookup and store:
memcpy(*o + 0, base64_table_enc_12bit + index0, 2);
memcpy(*o + 2, base64_table_enc_12bit + index1, 2);
*s += 3;
*o += 4;
}
static inline void
enc_loop_generic_32 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 4) {
return;
}
// Process blocks of 3 bytes at a time. Because blocks are loaded 4
// bytes at a time, ensure that there will be at least one remaining
// byte after the last round, so that the final read will not pass
// beyond the bounds of the input buffer:
size_t rounds = (*slen - 1) / 3;
*slen -= rounds * 3; // 3 bytes consumed per round
*olen += rounds * 4; // 4 bytes produced per round
do {
if (rounds >= 8) {
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_generic_32_inner(s, o);
enc_loop_generic_32_inner(s, o);
rounds -= 2;
continue;
}
enc_loop_generic_32_inner(s, o);
break;
} while (rounds > 0);
}

77
src/third-party/base64/lib/arch/generic/64/enc_loop.c vendored Normal file
View File

@ -0,0 +1,77 @@
static inline void
enc_loop_generic_64_inner (const uint8_t **s, uint8_t **o)
{
uint64_t src;
// Load input:
memcpy(&src, *s, sizeof (src));
// Reorder to 64-bit big-endian, if not already in that format. The
// workset must be in big-endian, otherwise the shifted bits do not
// carry over properly among adjacent bytes:
src = BASE64_HTOBE64(src);
// Four indices for the 12-bit lookup table:
const size_t index0 = (src >> 52) & 0xFFFU;
const size_t index1 = (src >> 40) & 0xFFFU;
const size_t index2 = (src >> 28) & 0xFFFU;
const size_t index3 = (src >> 16) & 0xFFFU;
// Table lookup and store:
memcpy(*o + 0, base64_table_enc_12bit + index0, 2);
memcpy(*o + 2, base64_table_enc_12bit + index1, 2);
memcpy(*o + 4, base64_table_enc_12bit + index2, 2);
memcpy(*o + 6, base64_table_enc_12bit + index3, 2);
*s += 6;
*o += 8;
}
static inline void
enc_loop_generic_64 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 8) {
return;
}
// Process blocks of 6 bytes at a time. Because blocks are loaded 8
// bytes at a time, ensure that there will be at least 2 remaining
// bytes after the last round, so that the final read will not pass
// beyond the bounds of the input buffer:
size_t rounds = (*slen - 2) / 6;
*slen -= rounds * 6; // 6 bytes consumed per round
*olen += rounds * 8; // 8 bytes produced per round
do {
if (rounds >= 8) {
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_generic_64_inner(s, o);
enc_loop_generic_64_inner(s, o);
rounds -= 2;
continue;
}
enc_loop_generic_64_inner(s, o);
break;
} while (rounds > 0);
}

39
src/third-party/base64/lib/arch/generic/codec.c vendored Normal file
View File

@ -0,0 +1,39 @@
#include <stdint.h>
#include <stddef.h>
#include <string.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if BASE64_WORDSIZE == 32
# include "32/enc_loop.c"
#elif BASE64_WORDSIZE == 64
# include "64/enc_loop.c"
#endif
#if BASE64_WORDSIZE >= 32
# include "32/dec_loop.c"
#endif
BASE64_ENC_FUNCTION(plain)
{
#include "enc_head.c"
#if BASE64_WORDSIZE == 32
enc_loop_generic_32(&s, &slen, &o, &olen);
#elif BASE64_WORDSIZE == 64
enc_loop_generic_64(&s, &slen, &o, &olen);
#endif
#include "enc_tail.c"
}
BASE64_DEC_FUNCTION(plain)
{
#include "dec_head.c"
#if BASE64_WORDSIZE >= 32
dec_loop_generic_32(&s, &slen, &o, &olen);
#endif
#include "dec_tail.c"
}

37
src/third-party/base64/lib/arch/generic/dec_head.c vendored Normal file
View File

@ -0,0 +1,37 @@
int ret = 0;
const uint8_t *s = (const uint8_t *) src;
uint8_t *o = (uint8_t *) out;
uint8_t q;
// Use local temporaries to avoid cache thrashing:
size_t olen = 0;
size_t slen = srclen;
struct base64_state st;
st.eof = state->eof;
st.bytes = state->bytes;
st.carry = state->carry;
// If we previously saw an EOF or an invalid character, bail out:
if (st.eof) {
*outlen = 0;
ret = 0;
// If there was a trailing '=' to check, check it:
if (slen && (st.eof == BASE64_AEOF)) {
state->bytes = 0;
state->eof = BASE64_EOF;
ret = ((base64_table_dec_8bit[*s++] == 254) && (slen == 1)) ? 1 : 0;
}
return ret;
}
// Turn four 6-bit numbers into three bytes:
// out[0] = 11111122
// out[1] = 22223333
// out[2] = 33444444
// Duff's device again:
switch (st.bytes)
{
for (;;)
{
case 0:

91
src/third-party/base64/lib/arch/generic/dec_tail.c vendored Normal file
View File

@ -0,0 +1,91 @@
if (slen-- == 0) {
ret = 1;
break;
}
if ((q = base64_table_dec_8bit[*s++]) >= 254) {
st.eof = BASE64_EOF;
// Treat character '=' as invalid for byte 0:
break;
}
st.carry = q << 2;
st.bytes++;
// Deliberate fallthrough:
BASE64_FALLTHROUGH
case 1: if (slen-- == 0) {
ret = 1;
break;
}
if ((q = base64_table_dec_8bit[*s++]) >= 254) {
st.eof = BASE64_EOF;
// Treat character '=' as invalid for byte 1:
break;
}
*o++ = st.carry | (q >> 4);
st.carry = q << 4;
st.bytes++;
olen++;
// Deliberate fallthrough:
BASE64_FALLTHROUGH
case 2: if (slen-- == 0) {
ret = 1;
break;
}
if ((q = base64_table_dec_8bit[*s++]) >= 254) {
st.bytes++;
// When q == 254, the input char is '='.
// Check if next byte is also '=':
if (q == 254) {
if (slen-- != 0) {
st.bytes = 0;
// EOF:
st.eof = BASE64_EOF;
q = base64_table_dec_8bit[*s++];
ret = ((q == 254) && (slen == 0)) ? 1 : 0;
break;
}
else {
// Almost EOF
st.eof = BASE64_AEOF;
ret = 1;
break;
}
}
// If we get here, there was an error:
break;
}
*o++ = st.carry | (q >> 2);
st.carry = q << 6;
st.bytes++;
olen++;
// Deliberate fallthrough:
BASE64_FALLTHROUGH
case 3: if (slen-- == 0) {
ret = 1;
break;
}
if ((q = base64_table_dec_8bit[*s++]) >= 254) {
st.bytes = 0;
st.eof = BASE64_EOF;
// When q == 254, the input char is '='. Return 1 and EOF.
// When q == 255, the input char is invalid. Return 0 and EOF.
ret = ((q == 254) && (slen == 0)) ? 1 : 0;
break;
}
*o++ = st.carry | q;
st.carry = 0;
st.bytes = 0;
olen++;
}
}
state->eof = st.eof;
state->bytes = st.bytes;
state->carry = st.carry;
*outlen = olen;
return ret;

24
src/third-party/base64/lib/arch/generic/enc_head.c vendored Normal file
View File

@ -0,0 +1,24 @@
// Assume that *out is large enough to contain the output.
// Theoretically it should be 4/3 the length of src.
const uint8_t *s = (const uint8_t *) src;
uint8_t *o = (uint8_t *) out;
// Use local temporaries to avoid cache thrashing:
size_t olen = 0;
size_t slen = srclen;
struct base64_state st;
st.bytes = state->bytes;
st.carry = state->carry;
// Turn three bytes into four 6-bit numbers:
// in[0] = 00111111
// in[1] = 00112222
// in[2] = 00222233
// in[3] = 00333333
// Duff's device, a for() loop inside a switch() statement. Legal!
switch (st.bytes)
{
for (;;)
{
case 0:

34
src/third-party/base64/lib/arch/generic/enc_tail.c vendored Normal file
View File

@ -0,0 +1,34 @@
if (slen-- == 0) {
break;
}
*o++ = base64_table_enc_6bit[*s >> 2];
st.carry = (*s++ << 4) & 0x30;
st.bytes++;
olen += 1;
// Deliberate fallthrough:
BASE64_FALLTHROUGH
case 1: if (slen-- == 0) {
break;
}
*o++ = base64_table_enc_6bit[st.carry | (*s >> 4)];
st.carry = (*s++ << 2) & 0x3C;
st.bytes++;
olen += 1;
// Deliberate fallthrough:
BASE64_FALLTHROUGH
case 2: if (slen-- == 0) {
break;
}
*o++ = base64_table_enc_6bit[st.carry | (*s >> 6)];
*o++ = base64_table_enc_6bit[*s++ & 0x3F];
st.bytes = 0;
olen += 2;
}
}
state->bytes = st.bytes;
state->carry = st.carry;
*outlen = olen;

77
src/third-party/base64/lib/arch/neon32/codec.c vendored Normal file
View File

@ -0,0 +1,77 @@
#include <stdint.h>
#include <stddef.h>
#include <string.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#ifdef __arm__
# if (defined(__ARM_NEON__) || defined(__ARM_NEON)) && HAVE_NEON32
# define BASE64_USE_NEON32
# endif
#endif
#ifdef BASE64_USE_NEON32
#include <arm_neon.h>
// Only enable inline assembly on supported compilers.
#if defined(__GNUC__) || defined(__clang__)
#define BASE64_NEON32_USE_ASM
#endif
static inline uint8x16_t
vqtbl1q_u8 (const uint8x16_t lut, const uint8x16_t indices)
{
// NEON32 only supports 64-bit wide lookups in 128-bit tables. Emulate
// the NEON64 `vqtbl1q_u8` intrinsic to do 128-bit wide lookups.
uint8x8x2_t lut2;
uint8x8x2_t result;
lut2.val[0] = vget_low_u8(lut);
lut2.val[1] = vget_high_u8(lut);
result.val[0] = vtbl2_u8(lut2, vget_low_u8(indices));
result.val[1] = vtbl2_u8(lut2, vget_high_u8(indices));
return vcombine_u8(result.val[0], result.val[1]);
}
#include "../generic/32/dec_loop.c"
#include "../generic/32/enc_loop.c"
#include "dec_loop.c"
#include "enc_reshuffle.c"
#include "enc_translate.c"
#include "enc_loop.c"
#endif // BASE64_USE_NEON32
// Stride size is so large on these NEON 32-bit functions
// (48 bytes encode, 32 bytes decode) that we inline the
// uint32 codec to stay performant on smaller inputs.
BASE64_ENC_FUNCTION(neon32)
{
#ifdef BASE64_USE_NEON32
#include "../generic/enc_head.c"
enc_loop_neon32(&s, &slen, &o, &olen);
enc_loop_generic_32(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(neon32)
{
#ifdef BASE64_USE_NEON32
#include "../generic/dec_head.c"
dec_loop_neon32(&s, &slen, &o, &olen);
dec_loop_generic_32(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

106
src/third-party/base64/lib/arch/neon32/dec_loop.c vendored Normal file
View File

@ -0,0 +1,106 @@
static inline int
is_nonzero (const uint8x16_t v)
{
uint64_t u64;
const uint64x2_t v64 = vreinterpretq_u64_u8(v);
const uint32x2_t v32 = vqmovn_u64(v64);
vst1_u64(&u64, vreinterpret_u64_u32(v32));
return u64 != 0;
}
static inline uint8x16_t
delta_lookup (const uint8x16_t v)
{
const uint8x8_t lut = {
0, 16, 19, 4, (uint8_t) -65, (uint8_t) -65, (uint8_t) -71, (uint8_t) -71,
};
return vcombine_u8(
vtbl1_u8(lut, vget_low_u8(v)),
vtbl1_u8(lut, vget_high_u8(v)));
}
static inline uint8x16_t
dec_loop_neon32_lane (uint8x16_t *lane)
{
// See the SSSE3 decoder for an explanation of the algorithm.
const uint8x16_t lut_lo = {
0x15, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11,
0x11, 0x11, 0x13, 0x1A, 0x1B, 0x1B, 0x1B, 0x1A
};
const uint8x16_t lut_hi = {
0x10, 0x10, 0x01, 0x02, 0x04, 0x08, 0x04, 0x08,
0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10
};
const uint8x16_t mask_0F = vdupq_n_u8(0x0F);
const uint8x16_t mask_2F = vdupq_n_u8(0x2F);
const uint8x16_t hi_nibbles = vshrq_n_u8(*lane, 4);
const uint8x16_t lo_nibbles = vandq_u8(*lane, mask_0F);
const uint8x16_t eq_2F = vceqq_u8(*lane, mask_2F);
const uint8x16_t hi = vqtbl1q_u8(lut_hi, hi_nibbles);
const uint8x16_t lo = vqtbl1q_u8(lut_lo, lo_nibbles);
// Now simply add the delta values to the input:
*lane = vaddq_u8(*lane, delta_lookup(vaddq_u8(eq_2F, hi_nibbles)));
// Return the validity mask:
return vandq_u8(lo, hi);
}
static inline void
dec_loop_neon32 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 64) {
return;
}
// Process blocks of 64 bytes per round. Unlike the SSE codecs, no
// extra trailing zero bytes are written, so it is not necessary to
// reserve extra input bytes:
size_t rounds = *slen / 64;
*slen -= rounds * 64; // 64 bytes consumed per round
*olen += rounds * 48; // 48 bytes produced per round
do {
uint8x16x3_t dec;
// Load 64 bytes and deinterleave:
uint8x16x4_t str = vld4q_u8(*s);
// Decode each lane, collect a mask of invalid inputs:
const uint8x16_t classified
= dec_loop_neon32_lane(&str.val[0])
| dec_loop_neon32_lane(&str.val[1])
| dec_loop_neon32_lane(&str.val[2])
| dec_loop_neon32_lane(&str.val[3]);
// Check for invalid input: if any of the delta values are
// zero, fall back on bytewise code to do error checking and
// reporting:
if (is_nonzero(classified)) {
break;
}
// Compress four bytes into three:
dec.val[0] = vorrq_u8(vshlq_n_u8(str.val[0], 2), vshrq_n_u8(str.val[1], 4));
dec.val[1] = vorrq_u8(vshlq_n_u8(str.val[1], 4), vshrq_n_u8(str.val[2], 2));
dec.val[2] = vorrq_u8(vshlq_n_u8(str.val[2], 6), str.val[3]);
// Interleave and store decoded result:
vst3q_u8(*o, dec);
*s += 64;
*o += 48;
} while (--rounds > 0);
// Adjust for any rounds that were skipped:
*slen += rounds * 64;
*olen -= rounds * 48;
}

169
src/third-party/base64/lib/arch/neon32/enc_loop.c vendored Normal file
View File

@ -0,0 +1,169 @@
#ifdef BASE64_NEON32_USE_ASM
static inline void
enc_loop_neon32_inner_asm (const uint8_t **s, uint8_t **o)
{
// This function duplicates the functionality of enc_loop_neon32_inner,
// but entirely with inline assembly. This gives a significant speedup
// over using NEON intrinsics, which do not always generate very good
// code. The logic of the assembly is directly lifted from the
// intrinsics version, so it can be used as a guide to this code.
// Temporary registers, used as scratch space.
uint8x16_t tmp0, tmp1, tmp2, tmp3;
uint8x16_t mask0, mask1, mask2, mask3;
// A lookup table containing the absolute offsets for all ranges.
const uint8x16_t lut = {
65U, 71U, 252U, 252U,
252U, 252U, 252U, 252U,
252U, 252U, 252U, 252U,
237U, 240U, 0U, 0U
};
// Numeric constants.
const uint8x16_t n51 = vdupq_n_u8(51);
const uint8x16_t n25 = vdupq_n_u8(25);
const uint8x16_t n63 = vdupq_n_u8(63);
__asm__ (
// Load 48 bytes and deinterleave. The bytes are loaded to
// hard-coded registers q12, q13 and q14, to ensure that they
// are contiguous. Increment the source pointer.
"vld3.8 {d24, d26, d28}, [%[src]]! \n\t"
"vld3.8 {d25, d27, d29}, [%[src]]! \n\t"
// Reshuffle the bytes using temporaries.
"vshr.u8 %q[t0], q12, #2 \n\t"
"vshr.u8 %q[t1], q13, #4 \n\t"
"vshr.u8 %q[t2], q14, #6 \n\t"
"vsli.8 %q[t1], q12, #4 \n\t"
"vsli.8 %q[t2], q13, #2 \n\t"
"vand.u8 %q[t1], %q[t1], %q[n63] \n\t"
"vand.u8 %q[t2], %q[t2], %q[n63] \n\t"
"vand.u8 %q[t3], q14, %q[n63] \n\t"
// t0..t3 are the reshuffled inputs. Create LUT indices.
"vqsub.u8 q12, %q[t0], %q[n51] \n\t"
"vqsub.u8 q13, %q[t1], %q[n51] \n\t"
"vqsub.u8 q14, %q[t2], %q[n51] \n\t"
"vqsub.u8 q15, %q[t3], %q[n51] \n\t"
// Create the mask for range #0.
"vcgt.u8 %q[m0], %q[t0], %q[n25] \n\t"
"vcgt.u8 %q[m1], %q[t1], %q[n25] \n\t"
"vcgt.u8 %q[m2], %q[t2], %q[n25] \n\t"
"vcgt.u8 %q[m3], %q[t3], %q[n25] \n\t"
// Subtract -1 to correct the LUT indices.
"vsub.u8 q12, %q[m0] \n\t"
"vsub.u8 q13, %q[m1] \n\t"
"vsub.u8 q14, %q[m2] \n\t"
"vsub.u8 q15, %q[m3] \n\t"
// Lookup the delta values.
"vtbl.u8 d24, {%q[lut]}, d24 \n\t"
"vtbl.u8 d25, {%q[lut]}, d25 \n\t"
"vtbl.u8 d26, {%q[lut]}, d26 \n\t"
"vtbl.u8 d27, {%q[lut]}, d27 \n\t"
"vtbl.u8 d28, {%q[lut]}, d28 \n\t"
"vtbl.u8 d29, {%q[lut]}, d29 \n\t"
"vtbl.u8 d30, {%q[lut]}, d30 \n\t"
"vtbl.u8 d31, {%q[lut]}, d31 \n\t"
// Add the delta values.
"vadd.u8 q12, %q[t0] \n\t"
"vadd.u8 q13, %q[t1] \n\t"
"vadd.u8 q14, %q[t2] \n\t"
"vadd.u8 q15, %q[t3] \n\t"
// Store 64 bytes and interleave. Increment the dest pointer.
"vst4.8 {d24, d26, d28, d30}, [%[dst]]! \n\t"
"vst4.8 {d25, d27, d29, d31}, [%[dst]]! \n\t"
// Outputs (modified).
: [src] "+r" (*s),
[dst] "+r" (*o),
[t0] "=&w" (tmp0),
[t1] "=&w" (tmp1),
[t2] "=&w" (tmp2),
[t3] "=&w" (tmp3),
[m0] "=&w" (mask0),
[m1] "=&w" (mask1),
[m2] "=&w" (mask2),
[m3] "=&w" (mask3)
// Inputs (not modified).
: [lut] "w" (lut),
[n25] "w" (n25),
[n51] "w" (n51),
[n63] "w" (n63)
// Clobbers.
: "d24", "d25", "d26", "d27", "d28", "d29", "d30", "d31"
);
}
#endif
static inline void
enc_loop_neon32_inner (const uint8_t **s, uint8_t **o)
{
#ifdef BASE64_NEON32_USE_ASM
enc_loop_neon32_inner_asm(s, o);
#else
// Load 48 bytes and deinterleave:
uint8x16x3_t src = vld3q_u8(*s);
// Reshuffle:
uint8x16x4_t out = enc_reshuffle(src);
// Translate reshuffled bytes to the Base64 alphabet:
out = enc_translate(out);
// Interleave and store output:
vst4q_u8(*o, out);
*s += 48;
*o += 64;
#endif
}
static inline void
enc_loop_neon32 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
size_t rounds = *slen / 48;
*slen -= rounds * 48; // 48 bytes consumed per round
*olen += rounds * 64; // 64 bytes produced per round
while (rounds > 0) {
if (rounds >= 8) {
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_neon32_inner(s, o);
enc_loop_neon32_inner(s, o);
rounds -= 2;
continue;
}
enc_loop_neon32_inner(s, o);
break;
}
}

31
src/third-party/base64/lib/arch/neon32/enc_reshuffle.c vendored Normal file
View File

@ -0,0 +1,31 @@
static inline uint8x16x4_t
enc_reshuffle (uint8x16x3_t in)
{
uint8x16x4_t out;
// Input:
// in[0] = a7 a6 a5 a4 a3 a2 a1 a0
// in[1] = b7 b6 b5 b4 b3 b2 b1 b0
// in[2] = c7 c6 c5 c4 c3 c2 c1 c0
// Output:
// out[0] = 00 00 a7 a6 a5 a4 a3 a2
// out[1] = 00 00 a1 a0 b7 b6 b5 b4
// out[2] = 00 00 b3 b2 b1 b0 c7 c6
// out[3] = 00 00 c5 c4 c3 c2 c1 c0
// Move the input bits to where they need to be in the outputs. Except
// for the first output, the high two bits are not cleared.
out.val[0] = vshrq_n_u8(in.val[0], 2);
out.val[1] = vshrq_n_u8(in.val[1], 4);
out.val[2] = vshrq_n_u8(in.val[2], 6);
out.val[1] = vsliq_n_u8(out.val[1], in.val[0], 4);
out.val[2] = vsliq_n_u8(out.val[2], in.val[1], 2);
// Clear the high two bits in the second, third and fourth output.
out.val[1] = vandq_u8(out.val[1], vdupq_n_u8(0x3F));
out.val[2] = vandq_u8(out.val[2], vdupq_n_u8(0x3F));
out.val[3] = vandq_u8(in.val[2], vdupq_n_u8(0x3F));
return out;
}

57
src/third-party/base64/lib/arch/neon32/enc_translate.c vendored Normal file
View File

@ -0,0 +1,57 @@
static inline uint8x16x4_t
enc_translate (const uint8x16x4_t in)
{
// A lookup table containing the absolute offsets for all ranges:
const uint8x16_t lut = {
65U, 71U, 252U, 252U,
252U, 252U, 252U, 252U,
252U, 252U, 252U, 252U,
237U, 240U, 0U, 0U
};
const uint8x16_t offset = vdupq_n_u8(51);
uint8x16x4_t indices, mask, delta, out;
// Translate values 0..63 to the Base64 alphabet. There are five sets:
// # From To Abs Index Characters
// 0 [0..25] [65..90] +65 0 ABCDEFGHIJKLMNOPQRSTUVWXYZ
// 1 [26..51] [97..122] +71 1 abcdefghijklmnopqrstuvwxyz
// 2 [52..61] [48..57] -4 [2..11] 0123456789
// 3 [62] [43] -19 12 +
// 4 [63] [47] -16 13 /
// Create LUT indices from input:
// the index for range #0 is right, others are 1 less than expected:
indices.val[0] = vqsubq_u8(in.val[0], offset);
indices.val[1] = vqsubq_u8(in.val[1], offset);
indices.val[2] = vqsubq_u8(in.val[2], offset);
indices.val[3] = vqsubq_u8(in.val[3], offset);
// mask is 0xFF (-1) for range #[1..4] and 0x00 for range #0:
mask.val[0] = vcgtq_u8(in.val[0], vdupq_n_u8(25));
mask.val[1] = vcgtq_u8(in.val[1], vdupq_n_u8(25));
mask.val[2] = vcgtq_u8(in.val[2], vdupq_n_u8(25));
mask.val[3] = vcgtq_u8(in.val[3], vdupq_n_u8(25));
// Subtract -1, so add 1 to indices for range #[1..4], All indices are
// now correct:
indices.val[0] = vsubq_u8(indices.val[0], mask.val[0]);
indices.val[1] = vsubq_u8(indices.val[1], mask.val[1]);
indices.val[2] = vsubq_u8(indices.val[2], mask.val[2]);
indices.val[3] = vsubq_u8(indices.val[3], mask.val[3]);
// Lookup delta values:
delta.val[0] = vqtbl1q_u8(lut, indices.val[0]);
delta.val[1] = vqtbl1q_u8(lut, indices.val[1]);
delta.val[2] = vqtbl1q_u8(lut, indices.val[2]);
delta.val[3] = vqtbl1q_u8(lut, indices.val[3]);
// Add delta values:
out.val[0] = vaddq_u8(in.val[0], delta.val[0]);
out.val[1] = vaddq_u8(in.val[1], delta.val[1]);
out.val[2] = vaddq_u8(in.val[2], delta.val[2]);
out.val[3] = vaddq_u8(in.val[3], delta.val[3]);
return out;
}

92
src/third-party/base64/lib/arch/neon64/codec.c vendored Normal file
View File

@ -0,0 +1,92 @@
#include <stdint.h>
#include <stddef.h>
#include <string.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#ifdef __aarch64__
# if (defined(__ARM_NEON__) || defined(__ARM_NEON)) && HAVE_NEON64
# define BASE64_USE_NEON64
# endif
#endif
#ifdef BASE64_USE_NEON64
#include <arm_neon.h>
// Only enable inline assembly on supported compilers.
#if defined(__GNUC__) || defined(__clang__)
#define BASE64_NEON64_USE_ASM
#endif
static inline uint8x16x4_t
load_64byte_table (const uint8_t *p)
{
#ifdef BASE64_NEON64_USE_ASM
// Force the table to be loaded into contiguous registers. GCC will not
// normally allocate contiguous registers for a `uint8x16x4_t'. These
// registers are chosen to not conflict with the ones in the enc loop.
register uint8x16_t t0 __asm__ ("v8");
register uint8x16_t t1 __asm__ ("v9");
register uint8x16_t t2 __asm__ ("v10");
register uint8x16_t t3 __asm__ ("v11");
__asm__ (
"ld1 {%[t0].16b, %[t1].16b, %[t2].16b, %[t3].16b}, [%[src]], #64 \n\t"
: [src] "+r" (p),
[t0] "=w" (t0),
[t1] "=w" (t1),
[t2] "=w" (t2),
[t3] "=w" (t3)
);
return (uint8x16x4_t) {
.val[0] = t0,
.val[1] = t1,
.val[2] = t2,
.val[3] = t3,
};
#else
return vld1q_u8_x4(p);
#endif
}
#include "../generic/32/dec_loop.c"
#include "../generic/64/enc_loop.c"
#include "dec_loop.c"
#include "enc_reshuffle.c"
#include "enc_loop.c"
#endif // BASE64_USE_NEON64
// Stride size is so large on these NEON 64-bit functions
// (48 bytes encode, 64 bytes decode) that we inline the
// uint64 codec to stay performant on smaller inputs.
BASE64_ENC_FUNCTION(neon64)
{
#ifdef BASE64_USE_NEON64
#include "../generic/enc_head.c"
enc_loop_neon64(&s, &slen, &o, &olen);
enc_loop_generic_64(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(neon64)
{
#ifdef BASE64_USE_NEON64
#include "../generic/dec_head.c"
dec_loop_neon64(&s, &slen, &o, &olen);
dec_loop_generic_32(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

129
src/third-party/base64/lib/arch/neon64/dec_loop.c vendored Normal file
View File

@ -0,0 +1,129 @@
// The input consists of five valid character sets in the Base64 alphabet,
// which we need to map back to the 6-bit values they represent.
// There are three ranges, two singles, and then there's the rest.
//
// # From To LUT Characters
// 1 [0..42] [255] #1 invalid input
// 2 [43] [62] #1 +
// 3 [44..46] [255] #1 invalid input
// 4 [47] [63] #1 /
// 5 [48..57] [52..61] #1 0..9
// 6 [58..63] [255] #1 invalid input
// 7 [64] [255] #2 invalid input
// 8 [65..90] [0..25] #2 A..Z
// 9 [91..96] [255] #2 invalid input
// 10 [97..122] [26..51] #2 a..z
// 11 [123..126] [255] #2 invalid input
// (12) Everything else => invalid input
// The first LUT will use the VTBL instruction (out of range indices are set to
// 0 in destination).
static const uint8_t dec_lut1[] = {
255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U,
255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U,
255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 255U, 62U, 255U, 255U, 255U, 63U,
52U, 53U, 54U, 55U, 56U, 57U, 58U, 59U, 60U, 61U, 255U, 255U, 255U, 255U, 255U, 255U,
};
// The second LUT will use the VTBX instruction (out of range indices will be
// unchanged in destination). Input [64..126] will be mapped to index [1..63]
// in this LUT. Index 0 means that value comes from LUT #1.
static const uint8_t dec_lut2[] = {
0U, 255U, 0U, 1U, 2U, 3U, 4U, 5U, 6U, 7U, 8U, 9U, 10U, 11U, 12U, 13U,
14U, 15U, 16U, 17U, 18U, 19U, 20U, 21U, 22U, 23U, 24U, 25U, 255U, 255U, 255U, 255U,
255U, 255U, 26U, 27U, 28U, 29U, 30U, 31U, 32U, 33U, 34U, 35U, 36U, 37U, 38U, 39U,
40U, 41U, 42U, 43U, 44U, 45U, 46U, 47U, 48U, 49U, 50U, 51U, 255U, 255U, 255U, 255U,
};
// All input values in range for the first look-up will be 0U in the second
// look-up result. All input values out of range for the first look-up will be
// 0U in the first look-up result. Thus, the two results can be ORed without
// conflicts.
//
// Invalid characters that are in the valid range for either look-up will be
// set to 255U in the combined result. Other invalid characters will just be
// passed through with the second look-up result (using the VTBX instruction).
// Since the second LUT is 64 bytes, those passed-through values are guaranteed
// to have a value greater than 63U. Therefore, valid characters will be mapped
// to the valid [0..63] range and all invalid characters will be mapped to
// values greater than 63.
static inline void
dec_loop_neon64 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 64) {
return;
}
// Process blocks of 64 bytes per round. Unlike the SSE codecs, no
// extra trailing zero bytes are written, so it is not necessary to
// reserve extra input bytes:
size_t rounds = *slen / 64;
*slen -= rounds * 64; // 64 bytes consumed per round
*olen += rounds * 48; // 48 bytes produced per round
const uint8x16x4_t tbl_dec1 = load_64byte_table(dec_lut1);
const uint8x16x4_t tbl_dec2 = load_64byte_table(dec_lut2);
do {
const uint8x16_t offset = vdupq_n_u8(63U);
uint8x16x4_t dec1, dec2;
uint8x16x3_t dec;
// Load 64 bytes and deinterleave:
uint8x16x4_t str = vld4q_u8((uint8_t *) *s);
// Get indices for second LUT:
dec2.val[0] = vqsubq_u8(str.val[0], offset);
dec2.val[1] = vqsubq_u8(str.val[1], offset);
dec2.val[2] = vqsubq_u8(str.val[2], offset);
dec2.val[3] = vqsubq_u8(str.val[3], offset);
// Get values from first LUT:
dec1.val[0] = vqtbl4q_u8(tbl_dec1, str.val[0]);
dec1.val[1] = vqtbl4q_u8(tbl_dec1, str.val[1]);
dec1.val[2] = vqtbl4q_u8(tbl_dec1, str.val[2]);
dec1.val[3] = vqtbl4q_u8(tbl_dec1, str.val[3]);
// Get values from second LUT:
dec2.val[0] = vqtbx4q_u8(dec2.val[0], tbl_dec2, dec2.val[0]);
dec2.val[1] = vqtbx4q_u8(dec2.val[1], tbl_dec2, dec2.val[1]);
dec2.val[2] = vqtbx4q_u8(dec2.val[2], tbl_dec2, dec2.val[2]);
dec2.val[3] = vqtbx4q_u8(dec2.val[3], tbl_dec2, dec2.val[3]);
// Get final values:
str.val[0] = vorrq_u8(dec1.val[0], dec2.val[0]);
str.val[1] = vorrq_u8(dec1.val[1], dec2.val[1]);
str.val[2] = vorrq_u8(dec1.val[2], dec2.val[2]);
str.val[3] = vorrq_u8(dec1.val[3], dec2.val[3]);
// Check for invalid input, any value larger than 63:
const uint8x16_t classified
= vcgtq_u8(str.val[0], vdupq_n_u8(63))
| vcgtq_u8(str.val[1], vdupq_n_u8(63))
| vcgtq_u8(str.val[2], vdupq_n_u8(63))
| vcgtq_u8(str.val[3], vdupq_n_u8(63));
// Check that all bits are zero:
if (vmaxvq_u8(classified) != 0U) {
break;
}
// Compress four bytes into three:
dec.val[0] = vshlq_n_u8(str.val[0], 2) | vshrq_n_u8(str.val[1], 4);
dec.val[1] = vshlq_n_u8(str.val[1], 4) | vshrq_n_u8(str.val[2], 2);
dec.val[2] = vshlq_n_u8(str.val[2], 6) | str.val[3];
// Interleave and store decoded result:
vst3q_u8((uint8_t *) *o, dec);
*s += 64;
*o += 48;
} while (--rounds > 0);
// Adjust for any rounds that were skipped:
*slen += rounds * 64;
*olen -= rounds * 48;
}

133
src/third-party/base64/lib/arch/neon64/enc_loop.c vendored Normal file
View File

@ -0,0 +1,133 @@
#ifdef BASE64_NEON64_USE_ASM
static inline void
enc_loop_neon64_inner_asm (const uint8_t **s, uint8_t **o, const uint8x16x4_t tbl_enc)
{
// This function duplicates the functionality of enc_loop_neon64_inner,
// but entirely with inline assembly. This gives a significant speedup
// over using NEON intrinsics, which do not always generate very good
// code. The logic of the assembly is directly lifted from the
// intrinsics version, so it can be used as a guide to this code.
// Temporary registers, used as scratch space.
uint8x16_t tmp0, tmp1, tmp2, tmp3;
// Numeric constant.
const uint8x16_t n63 = vdupq_n_u8(63);
__asm__ (
// Load 48 bytes and deinterleave. The bytes are loaded to
// hard-coded registers v12, v13 and v14, to ensure that they
// are contiguous. Increment the source pointer.
"ld3 {v12.16b, v13.16b, v14.16b}, [%[src]], #48 \n\t"
// Reshuffle the bytes using temporaries.
"ushr %[t0].16b, v12.16b, #2 \n\t"
"ushr %[t1].16b, v13.16b, #4 \n\t"
"ushr %[t2].16b, v14.16b, #6 \n\t"
"sli %[t1].16b, v12.16b, #4 \n\t"
"sli %[t2].16b, v13.16b, #2 \n\t"
"and %[t1].16b, %[t1].16b, %[n63].16b \n\t"
"and %[t2].16b, %[t2].16b, %[n63].16b \n\t"
"and %[t3].16b, v14.16b, %[n63].16b \n\t"
// Translate the values to the Base64 alphabet.
"tbl v12.16b, {%[l0].16b, %[l1].16b, %[l2].16b, %[l3].16b}, %[t0].16b \n\t"
"tbl v13.16b, {%[l0].16b, %[l1].16b, %[l2].16b, %[l3].16b}, %[t1].16b \n\t"
"tbl v14.16b, {%[l0].16b, %[l1].16b, %[l2].16b, %[l3].16b}, %[t2].16b \n\t"
"tbl v15.16b, {%[l0].16b, %[l1].16b, %[l2].16b, %[l3].16b}, %[t3].16b \n\t"
// Store 64 bytes and interleave. Increment the dest pointer.
"st4 {v12.16b, v13.16b, v14.16b, v15.16b}, [%[dst]], #64 \n\t"
// Outputs (modified).
: [src] "+r" (*s),
[dst] "+r" (*o),
[t0] "=&w" (tmp0),
[t1] "=&w" (tmp1),
[t2] "=&w" (tmp2),
[t3] "=&w" (tmp3)
// Inputs (not modified).
: [n63] "w" (n63),
[l0] "w" (tbl_enc.val[0]),
[l1] "w" (tbl_enc.val[1]),
[l2] "w" (tbl_enc.val[2]),
[l3] "w" (tbl_enc.val[3])
// Clobbers.
: "v12", "v13", "v14", "v15"
);
}
#endif
static inline void
enc_loop_neon64_inner (const uint8_t **s, uint8_t **o, const uint8x16x4_t tbl_enc)
{
#ifdef BASE64_NEON64_USE_ASM
enc_loop_neon64_inner_asm(s, o, tbl_enc);
#else
// Load 48 bytes and deinterleave:
uint8x16x3_t src = vld3q_u8(*s);
// Divide bits of three input bytes over four output bytes:
uint8x16x4_t out = enc_reshuffle(src);
// The bits have now been shifted to the right locations;
// translate their values 0..63 to the Base64 alphabet.
// Use a 64-byte table lookup:
out.val[0] = vqtbl4q_u8(tbl_enc, out.val[0]);
out.val[1] = vqtbl4q_u8(tbl_enc, out.val[1]);
out.val[2] = vqtbl4q_u8(tbl_enc, out.val[2]);
out.val[3] = vqtbl4q_u8(tbl_enc, out.val[3]);
// Interleave and store output:
vst4q_u8(*o, out);
*s += 48;
*o += 64;
#endif
}
static inline void
enc_loop_neon64 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
size_t rounds = *slen / 48;
*slen -= rounds * 48; // 48 bytes consumed per round
*olen += rounds * 64; // 64 bytes produced per round
// Load the encoding table:
const uint8x16x4_t tbl_enc = load_64byte_table(base64_table_enc_6bit);
while (rounds > 0) {
if (rounds >= 8) {
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_neon64_inner(s, o, tbl_enc);
enc_loop_neon64_inner(s, o, tbl_enc);
rounds -= 2;
continue;
}
enc_loop_neon64_inner(s, o, tbl_enc);
break;
}
}

31
src/third-party/base64/lib/arch/neon64/enc_reshuffle.c vendored Normal file
View File

@ -0,0 +1,31 @@
static inline uint8x16x4_t
enc_reshuffle (const uint8x16x3_t in)
{
uint8x16x4_t out;
// Input:
// in[0] = a7 a6 a5 a4 a3 a2 a1 a0
// in[1] = b7 b6 b5 b4 b3 b2 b1 b0
// in[2] = c7 c6 c5 c4 c3 c2 c1 c0
// Output:
// out[0] = 00 00 a7 a6 a5 a4 a3 a2
// out[1] = 00 00 a1 a0 b7 b6 b5 b4
// out[2] = 00 00 b3 b2 b1 b0 c7 c6
// out[3] = 00 00 c5 c4 c3 c2 c1 c0
// Move the input bits to where they need to be in the outputs. Except
// for the first output, the high two bits are not cleared.
out.val[0] = vshrq_n_u8(in.val[0], 2);
out.val[1] = vshrq_n_u8(in.val[1], 4);
out.val[2] = vshrq_n_u8(in.val[2], 6);
out.val[1] = vsliq_n_u8(out.val[1], in.val[0], 4);
out.val[2] = vsliq_n_u8(out.val[2], in.val[1], 2);
// Clear the high two bits in the second, third and fourth output.
out.val[1] = vandq_u8(out.val[1], vdupq_n_u8(0x3F));
out.val[2] = vandq_u8(out.val[2], vdupq_n_u8(0x3F));
out.val[3] = vandq_u8(in.val[2], vdupq_n_u8(0x3F));
return out;
}

42
src/third-party/base64/lib/arch/sse41/codec.c vendored Normal file
View File

@ -0,0 +1,42 @@
#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if HAVE_SSE41
#include <smmintrin.h>
#include "../ssse3/dec_reshuffle.c"
#include "../ssse3/dec_loop.c"
#include "../ssse3/enc_translate.c"
#include "../ssse3/enc_reshuffle.c"
#include "../ssse3/enc_loop.c"
#endif // HAVE_SSE41
BASE64_ENC_FUNCTION(sse41)
{
#if HAVE_SSE41
#include "../generic/enc_head.c"
enc_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(sse41)
{
#if HAVE_SSE41
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

42
src/third-party/base64/lib/arch/sse42/codec.c vendored Normal file
View File

@ -0,0 +1,42 @@
#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if HAVE_SSE42
#include <nmmintrin.h>
#include "../ssse3/dec_reshuffle.c"
#include "../ssse3/dec_loop.c"
#include "../ssse3/enc_translate.c"
#include "../ssse3/enc_reshuffle.c"
#include "../ssse3/enc_loop.c"
#endif // HAVE_SSE42
BASE64_ENC_FUNCTION(sse42)
{
#if HAVE_SSE42
#include "../generic/enc_head.c"
enc_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(sse42)
{
#if HAVE_SSE42
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

42
src/third-party/base64/lib/arch/ssse3/codec.c vendored Normal file
View File

@ -0,0 +1,42 @@
#include <stdint.h>
#include <stddef.h>
#include <stdlib.h>
#include "../../../include/libbase64.h"
#include "../../tables/tables.h"
#include "../../codecs.h"
#include "config.h"
#include "../../env.h"
#if HAVE_SSSE3
#include <tmmintrin.h>
#include "dec_reshuffle.c"
#include "dec_loop.c"
#include "enc_reshuffle.c"
#include "enc_translate.c"
#include "enc_loop.c"
#endif // HAVE_SSSE3
BASE64_ENC_FUNCTION(ssse3)
{
#if HAVE_SSSE3
#include "../generic/enc_head.c"
enc_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/enc_tail.c"
#else
BASE64_ENC_STUB
#endif
}
BASE64_DEC_FUNCTION(ssse3)
{
#if HAVE_SSSE3
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#else
BASE64_DEC_STUB
#endif
}

173
src/third-party/base64/lib/arch/ssse3/dec_loop.c vendored Normal file
View File

@ -0,0 +1,173 @@
// The input consists of six character sets in the Base64 alphabet, which we
// need to map back to the 6-bit values they represent. There are three ranges,
// two singles, and then there's the rest.
//
// # From To Add Characters
// 1 [43] [62] +19 +
// 2 [47] [63] +16 /
// 3 [48..57] [52..61] +4 0..9
// 4 [65..90] [0..25] -65 A..Z
// 5 [97..122] [26..51] -71 a..z
// (6) Everything else => invalid input
//
// We will use lookup tables for character validation and offset computation.
// Remember that 0x2X and 0x0X are the same index for _mm_shuffle_epi8, this
// allows to mask with 0x2F instead of 0x0F and thus save one constant
// declaration (register and/or memory access).
//
// For offsets:
// Perfect hash for lut = ((src >> 4) & 0x2F) + ((src == 0x2F) ? 0xFF : 0x00)
// 0000 = garbage
// 0001 = /
// 0010 = +
// 0011 = 0-9
// 0100 = A-Z
// 0101 = A-Z
// 0110 = a-z
// 0111 = a-z
// 1000 >= garbage
//
// For validation, here's the table.
// A character is valid if and only if the AND of the 2 lookups equals 0:
//
// hi \ lo 0000 0001 0010 0011 0100 0101 0110 0111 1000 1001 1010 1011 1100 1101 1110 1111
// LUT 0x15 0x11 0x11 0x11 0x11 0x11 0x11 0x11 0x11 0x11 0x13 0x1A 0x1B 0x1B 0x1B 0x1A
//
// 0000 0x10 char NUL SOH STX ETX EOT ENQ ACK BEL BS HT LF VT FF CR SO SI
// andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
//
// 0001 0x10 char DLE DC1 DC2 DC3 DC4 NAK SYN ETB CAN EM SUB ESC FS GS RS US
// andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
//
// 0010 0x01 char ! " # $ % & ' ( ) * + , - . /
// andlut 0x01 0x01 0x01 0x01 0x01 0x01 0x01 0x01 0x01 0x01 0x01 0x00 0x01 0x01 0x01 0x00
//
// 0011 0x02 char 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
// andlut 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x02 0x02 0x02 0x02 0x02 0x02
//
// 0100 0x04 char @ A B C D E F G H I J K L M N O
// andlut 0x04 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00
//
// 0101 0x08 char P Q R S T U V W X Y Z [ \ ] ^ _
// andlut 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x08 0x08 0x08 0x08 0x08
//
// 0110 0x04 char ` a b c d e f g h i j k l m n o
// andlut 0x04 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00
// 0111 0x08 char p q r s t u v w x y z { | } ~
// andlut 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x08 0x08 0x08 0x08 0x08
//
// 1000 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1001 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1010 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1011 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1100 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1101 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1110 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
// 1111 0x10 andlut 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10 0x10
static inline int
dec_loop_ssse3_inner (const uint8_t **s, uint8_t **o, size_t *rounds)
{
const __m128i lut_lo = _mm_setr_epi8(
0x15, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11,
0x11, 0x11, 0x13, 0x1A, 0x1B, 0x1B, 0x1B, 0x1A);
const __m128i lut_hi = _mm_setr_epi8(
0x10, 0x10, 0x01, 0x02, 0x04, 0x08, 0x04, 0x08,
0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10);
const __m128i lut_roll = _mm_setr_epi8(
0, 16, 19, 4, -65, -65, -71, -71,
0, 0, 0, 0, 0, 0, 0, 0);
const __m128i mask_2F = _mm_set1_epi8(0x2F);
// Load input:
__m128i str = _mm_loadu_si128((__m128i *) *s);
// Table lookups:
const __m128i hi_nibbles = _mm_and_si128(_mm_srli_epi32(str, 4), mask_2F);
const __m128i lo_nibbles = _mm_and_si128(str, mask_2F);
const __m128i hi = _mm_shuffle_epi8(lut_hi, hi_nibbles);
const __m128i lo = _mm_shuffle_epi8(lut_lo, lo_nibbles);
// Check for invalid input: if any "and" values from lo and hi are not
// zero, fall back on bytewise code to do error checking and reporting:
if (_mm_movemask_epi8(_mm_cmpgt_epi8(_mm_and_si128(lo, hi), _mm_setzero_si128())) != 0) {
return 0;
}
const __m128i eq_2F = _mm_cmpeq_epi8(str, mask_2F);
const __m128i roll = _mm_shuffle_epi8(lut_roll, _mm_add_epi8(eq_2F, hi_nibbles));
// Now simply add the delta values to the input:
str = _mm_add_epi8(str, roll);
// Reshuffle the input to packed 12-byte output format:
str = dec_reshuffle(str);
// Store the output:
_mm_storeu_si128((__m128i *) *o, str);
*s += 16;
*o += 12;
*rounds -= 1;
return 1;
}
static inline void
dec_loop_ssse3 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 24) {
return;
}
// Process blocks of 16 bytes per round. Because 4 extra zero bytes are
// written after the output, ensure that there will be at least 8 bytes
// of input data left to cover the gap. (6 data bytes and up to two
// end-of-string markers.)
size_t rounds = (*slen - 8) / 16;
*slen -= rounds * 16; // 16 bytes consumed per round
*olen += rounds * 12; // 12 bytes produced per round
do {
if (rounds >= 8) {
if (dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 4) {
if (dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds)) {
continue;
}
break;
}
if (rounds >= 2) {
if (dec_loop_ssse3_inner(s, o, &rounds) &&
dec_loop_ssse3_inner(s, o, &rounds)) {
continue;
}
break;
}
dec_loop_ssse3_inner(s, o, &rounds);
break;
} while (rounds > 0);
// Adjust for any rounds that were skipped:
*slen += rounds * 16;
*olen -= rounds * 12;
}

33
src/third-party/base64/lib/arch/ssse3/dec_reshuffle.c vendored Normal file
View File

@ -0,0 +1,33 @@
static inline __m128i
dec_reshuffle (const __m128i in)
{
// in, bits, upper case are most significant bits, lower case are least significant bits
// 00llllll 00kkkkLL 00jjKKKK 00JJJJJJ
// 00iiiiii 00hhhhII 00ggHHHH 00GGGGGG
// 00ffffff 00eeeeFF 00ddEEEE 00DDDDDD
// 00cccccc 00bbbbCC 00aaBBBB 00AAAAAA
const __m128i merge_ab_and_bc = _mm_maddubs_epi16(in, _mm_set1_epi32(0x01400140));
// 0000kkkk LLllllll 0000JJJJ JJjjKKKK
// 0000hhhh IIiiiiii 0000GGGG GGggHHHH
// 0000eeee FFffffff 0000DDDD DDddEEEE
// 0000bbbb CCcccccc 0000AAAA AAaaBBBB
const __m128i out = _mm_madd_epi16(merge_ab_and_bc, _mm_set1_epi32(0x00011000));
// 00000000 JJJJJJjj KKKKkkkk LLllllll
// 00000000 GGGGGGgg HHHHhhhh IIiiiiii
// 00000000 DDDDDDdd EEEEeeee FFffffff
// 00000000 AAAAAAaa BBBBbbbb CCcccccc
// Pack bytes together:
return _mm_shuffle_epi8(out, _mm_setr_epi8(
2, 1, 0,
6, 5, 4,
10, 9, 8,
14, 13, 12,
-1, -1, -1, -1));
// 00000000 00000000 00000000 00000000
// LLllllll KKKKkkkk JJJJJJjj IIiiiiii
// HHHHhhhh GGGGGGgg FFffffff EEEEeeee
// DDDDDDdd CCcccccc BBBBbbbb AAAAAAaa
}

67
src/third-party/base64/lib/arch/ssse3/enc_loop.c vendored Normal file
View File

@ -0,0 +1,67 @@
static inline void
enc_loop_ssse3_inner (const uint8_t **s, uint8_t **o)
{
// Load input:
__m128i str = _mm_loadu_si128((__m128i *) *s);
// Reshuffle:
str = enc_reshuffle(str);
// Translate reshuffled bytes to the Base64 alphabet:
str = enc_translate(str);
// Store:
_mm_storeu_si128((__m128i *) *o, str);
*s += 12;
*o += 16;
}
static inline void
enc_loop_ssse3 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
if (*slen < 16) {
return;
}
// Process blocks of 12 bytes at a time. Because blocks are loaded 16
// bytes at a time, ensure that there will be at least 4 remaining
// bytes after the last round, so that the final read will not pass
// beyond the bounds of the input buffer:
size_t rounds = (*slen - 4) / 12;
*slen -= rounds * 12; // 12 bytes consumed per round
*olen += rounds * 16; // 16 bytes produced per round
do {
if (rounds >= 8) {
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
rounds -= 8;
continue;
}
if (rounds >= 4) {
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
rounds -= 4;
continue;
}
if (rounds >= 2) {
enc_loop_ssse3_inner(s, o);
enc_loop_ssse3_inner(s, o);
rounds -= 2;
continue;
}
enc_loop_ssse3_inner(s, o);
break;
} while (rounds > 0);
}

48
src/third-party/base64/lib/arch/ssse3/enc_reshuffle.c vendored Normal file
View File

@ -0,0 +1,48 @@
static inline __m128i
enc_reshuffle (__m128i in)
{
// Input, bytes MSB to LSB:
// 0 0 0 0 l k j i h g f e d c b a
in = _mm_shuffle_epi8(in, _mm_set_epi8(
10, 11, 9, 10,
7, 8, 6, 7,
4, 5, 3, 4,
1, 2, 0, 1));
// in, bytes MSB to LSB:
// k l j k
// h i g h
// e f d e
// b c a b
const __m128i t0 = _mm_and_si128(in, _mm_set1_epi32(0x0FC0FC00));
// bits, upper case are most significant bits, lower case are least significant bits
// 0000kkkk LL000000 JJJJJJ00 00000000
// 0000hhhh II000000 GGGGGG00 00000000
// 0000eeee FF000000 DDDDDD00 00000000
// 0000bbbb CC000000 AAAAAA00 00000000
const __m128i t1 = _mm_mulhi_epu16(t0, _mm_set1_epi32(0x04000040));
// 00000000 00kkkkLL 00000000 00JJJJJJ
// 00000000 00hhhhII 00000000 00GGGGGG
// 00000000 00eeeeFF 00000000 00DDDDDD
// 00000000 00bbbbCC 00000000 00AAAAAA
const __m128i t2 = _mm_and_si128(in, _mm_set1_epi32(0x003F03F0));
// 00000000 00llllll 000000jj KKKK0000
// 00000000 00iiiiii 000000gg HHHH0000
// 00000000 00ffffff 000000dd EEEE0000
// 00000000 00cccccc 000000aa BBBB0000
const __m128i t3 = _mm_mullo_epi16(t2, _mm_set1_epi32(0x01000010));
// 00llllll 00000000 00jjKKKK 00000000
// 00iiiiii 00000000 00ggHHHH 00000000
// 00ffffff 00000000 00ddEEEE 00000000
// 00cccccc 00000000 00aaBBBB 00000000
return _mm_or_si128(t1, t3);
// 00llllll 00kkkkLL 00jjKKKK 00JJJJJJ
// 00iiiiii 00hhhhII 00ggHHHH 00GGGGGG
// 00ffffff 00eeeeFF 00ddEEEE 00DDDDDD
// 00cccccc 00bbbbCC 00aaBBBB 00AAAAAA
}

33
src/third-party/base64/lib/arch/ssse3/enc_translate.c vendored Normal file
View File

@ -0,0 +1,33 @@
static inline __m128i
enc_translate (const __m128i in)
{
// A lookup table containing the absolute offsets for all ranges:
const __m128i lut = _mm_setr_epi8(
65, 71, -4, -4,
-4, -4, -4, -4,
-4, -4, -4, -4,
-19, -16, 0, 0
);
// Translate values 0..63 to the Base64 alphabet. There are five sets:
// # From To Abs Index Characters
// 0 [0..25] [65..90] +65 0 ABCDEFGHIJKLMNOPQRSTUVWXYZ
// 1 [26..51] [97..122] +71 1 abcdefghijklmnopqrstuvwxyz
// 2 [52..61] [48..57] -4 [2..11] 0123456789
// 3 [62] [43] -19 12 +
// 4 [63] [47] -16 13 /
// Create LUT indices from the input. The index for range #0 is right,
// others are 1 less than expected:
__m128i indices = _mm_subs_epu8(in, _mm_set1_epi8(51));
// mask is 0xFF (-1) for range #[1..4] and 0x00 for range #0:
__m128i mask = _mm_cmpgt_epi8(in, _mm_set1_epi8(25));
// Subtract -1, so add 1 to indices for range #[1..4]. All indices are
// now correct:
indices = _mm_sub_epi8(indices, mask);
// Add offsets to input values:
return _mm_add_epi8(in, _mm_shuffle_epi8(lut, indices));
}

281
src/third-party/base64/lib/codec_choose.c vendored Normal file
View File

@ -0,0 +1,281 @@
#include <stdbool.h>
#include <stdint.h>
#include <stddef.h>
#include <stdint.h>
#include "../include/libbase64.h"
#include "codecs.h"
#include "config.h"
#include "env.h"
#if (__x86_64__ || __i386__ || _M_X86 || _M_X64)
#define BASE64_X86
#if (HAVE_SSSE3 || HAVE_SSE41 || HAVE_SSE42 || HAVE_AVX || HAVE_AVX2)
#define BASE64_X86_SIMD
#endif
#endif
#ifdef BASE64_X86
#ifdef _MSC_VER
#include <intrin.h>
#define __cpuid_count(__level, __count, __eax, __ebx, __ecx, __edx) \
{ \
int info[4]; \
__cpuidex(info, __level, __count); \
__eax = info[0]; \
__ebx = info[1]; \
__ecx = info[2]; \
__edx = info[3]; \
}
#define __cpuid(__level, __eax, __ebx, __ecx, __edx) \
__cpuid_count(__level, 0, __eax, __ebx, __ecx, __edx)
#else
#include <cpuid.h>
#if HAVE_AVX2 || HAVE_AVX
#if ((__GNUC__ > 4 || __GNUC__ == 4 && __GNUC_MINOR__ >= 2) || (__clang_major__ >= 3))
static inline uint64_t _xgetbv (uint32_t index)
{
uint32_t eax, edx;
__asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
return ((uint64_t)edx << 32) | eax;
}
#else
#error "Platform not supported"
#endif
#endif
#endif
#ifndef bit_AVX2
#define bit_AVX2 (1 << 5)
#endif
#ifndef bit_SSSE3
#define bit_SSSE3 (1 << 9)
#endif
#ifndef bit_SSE41
#define bit_SSE41 (1 << 19)
#endif
#ifndef bit_SSE42
#define bit_SSE42 (1 << 20)
#endif
#ifndef bit_AVX
#define bit_AVX (1 << 28)
#endif
#define bit_XSAVE_XRSTORE (1 << 27)
#ifndef _XCR_XFEATURE_ENABLED_MASK
#define _XCR_XFEATURE_ENABLED_MASK 0
#endif
#define _XCR_XMM_AND_YMM_STATE_ENABLED_BY_OS 0x6
#endif
// Function declarations:
#define BASE64_CODEC_FUNCS(arch) \
BASE64_ENC_FUNCTION(arch); \
BASE64_DEC_FUNCTION(arch); \
BASE64_CODEC_FUNCS(avx2)
BASE64_CODEC_FUNCS(neon32)
BASE64_CODEC_FUNCS(neon64)
BASE64_CODEC_FUNCS(plain)
BASE64_CODEC_FUNCS(ssse3)
BASE64_CODEC_FUNCS(sse41)
BASE64_CODEC_FUNCS(sse42)
BASE64_CODEC_FUNCS(avx)
static bool
codec_choose_forced (struct codec *codec, int flags)
{
// If the user wants to use a certain codec,
// always allow it, even if the codec is a no-op.
// For testing purposes.
if (!(flags & 0xFF)) {
return false;
}
if (flags & BASE64_FORCE_AVX2) {
codec->enc = base64_stream_encode_avx2;
codec->dec = base64_stream_decode_avx2;
return true;
}
if (flags & BASE64_FORCE_NEON32) {
codec->enc = base64_stream_encode_neon32;
codec->dec = base64_stream_decode_neon32;
return true;
}
if (flags & BASE64_FORCE_NEON64) {
codec->enc = base64_stream_encode_neon64;
codec->dec = base64_stream_decode_neon64;
return true;
}
if (flags & BASE64_FORCE_PLAIN) {
codec->enc = base64_stream_encode_plain;
codec->dec = base64_stream_decode_plain;
return true;
}
if (flags & BASE64_FORCE_SSSE3) {
codec->enc = base64_stream_encode_ssse3;
codec->dec = base64_stream_decode_ssse3;
return true;
}
if (flags & BASE64_FORCE_SSE41) {
codec->enc = base64_stream_encode_sse41;
codec->dec = base64_stream_decode_sse41;
return true;
}
if (flags & BASE64_FORCE_SSE42) {
codec->enc = base64_stream_encode_sse42;
codec->dec = base64_stream_decode_sse42;
return true;
}
if (flags & BASE64_FORCE_AVX) {
codec->enc = base64_stream_encode_avx;
codec->dec = base64_stream_decode_avx;
return true;
}
return false;
}
static bool
codec_choose_arm (struct codec *codec)
{
#if (defined(__ARM_NEON__) || defined(__ARM_NEON)) && ((defined(__aarch64__) && HAVE_NEON64) || HAVE_NEON32)
// Unfortunately there is no portable way to check for NEON
// support at runtime from userland in the same way that x86
// has cpuid, so just stick to the compile-time configuration:
#if defined(__aarch64__) && HAVE_NEON64
codec->enc = base64_stream_encode_neon64;
codec->dec = base64_stream_decode_neon64;
#else
codec->enc = base64_stream_encode_neon32;
codec->dec = base64_stream_decode_neon32;
#endif
return true;
#else
(void)codec;
return false;
#endif
}
static bool
codec_choose_x86 (struct codec *codec)
{
#ifdef BASE64_X86_SIMD
unsigned int eax, ebx = 0, ecx = 0, edx;
unsigned int max_level;
#ifdef _MSC_VER
int info[4];
__cpuidex(info, 0, 0);
max_level = info[0];
#else
max_level = __get_cpuid_max(0, NULL);
#endif
#if HAVE_AVX2 || HAVE_AVX
// Check for AVX/AVX2 support:
// Checking for AVX requires 3 things:
// 1) CPUID indicates that the OS uses XSAVE and XRSTORE instructions
// (allowing saving YMM registers on context switch)
// 2) CPUID indicates support for AVX
// 3) XGETBV indicates the AVX registers will be saved and restored on
// context switch
//
// Note that XGETBV is only available on 686 or later CPUs, so the
// instruction needs to be conditionally run.
if (max_level >= 1) {
__cpuid_count(1, 0, eax, ebx, ecx, edx);
if (ecx & bit_XSAVE_XRSTORE) {
uint64_t xcr_mask;
xcr_mask = _xgetbv(_XCR_XFEATURE_ENABLED_MASK);
if (xcr_mask & _XCR_XMM_AND_YMM_STATE_ENABLED_BY_OS) {
#if HAVE_AVX2
if (max_level >= 7) {
__cpuid_count(7, 0, eax, ebx, ecx, edx);
if (ebx & bit_AVX2) {
codec->enc = base64_stream_encode_avx2;
codec->dec = base64_stream_decode_avx2;
return true;
}
}
#endif
#if HAVE_AVX
__cpuid_count(1, 0, eax, ebx, ecx, edx);
if (ecx & bit_AVX) {
codec->enc = base64_stream_encode_avx;
codec->dec = base64_stream_decode_avx;
return true;
}
#endif
}
}
}
#endif
#if HAVE_SSE42
// Check for SSE42 support:
if (max_level >= 1) {
__cpuid(1, eax, ebx, ecx, edx);
if (ecx & bit_SSE42) {
codec->enc = base64_stream_encode_sse42;
codec->dec = base64_stream_decode_sse42;
return true;
}
}
#endif
#if HAVE_SSE41
// Check for SSE41 support:
if (max_level >= 1) {
__cpuid(1, eax, ebx, ecx, edx);
if (ecx & bit_SSE41) {
codec->enc = base64_stream_encode_sse41;
codec->dec = base64_stream_decode_sse41;
return true;
}
}
#endif
#if HAVE_SSSE3
// Check for SSSE3 support:
if (max_level >= 1) {
__cpuid(1, eax, ebx, ecx, edx);
if (ecx & bit_SSSE3) {
codec->enc = base64_stream_encode_ssse3;
codec->dec = base64_stream_decode_ssse3;
return true;
}
}
#endif
#else
(void)codec;
#endif
return false;
}
void
codec_choose (struct codec *codec, int flags)
{
// User forced a codec:
if (codec_choose_forced(codec, flags)) {
return;
}
// Runtime feature detection:
if (codec_choose_arm(codec)) {
return;
}
if (codec_choose_x86(codec)) {
return;
}
codec->enc = base64_stream_encode_plain;
codec->dec = base64_stream_decode_plain;
}

65
src/third-party/base64/lib/codecs.h vendored Normal file
View File

@ -0,0 +1,65 @@
#include <stdint.h>
#include <stddef.h>
#include "../include/libbase64.h"
#include "config.h"
// Function parameters for encoding functions:
#define BASE64_ENC_PARAMS \
( struct base64_state *state \
, const char *src \
, size_t srclen \
, char *out \
, size_t *outlen \
)
// Function parameters for decoding functions:
#define BASE64_DEC_PARAMS \
( struct base64_state *state \
, const char *src \
, size_t srclen \
, char *out \
, size_t *outlen \
)
// Function signature for encoding functions:
#define BASE64_ENC_FUNCTION(arch) \
void \
base64_stream_encode_ ## arch \
BASE64_ENC_PARAMS
// Function signature for decoding functions:
#define BASE64_DEC_FUNCTION(arch) \
int \
base64_stream_decode_ ## arch \
BASE64_DEC_PARAMS
// Cast away unused variable, silence compiler:
#define UNUSED(x) ((void)(x))
// Stub function when encoder arch unsupported:
#define BASE64_ENC_STUB \
UNUSED(state); \
UNUSED(src); \
UNUSED(srclen); \
UNUSED(out); \
\
*outlen = 0;
// Stub function when decoder arch unsupported:
#define BASE64_DEC_STUB \
UNUSED(state); \
UNUSED(src); \
UNUSED(srclen); \
UNUSED(out); \
UNUSED(outlen); \
\
return -1;
struct codec
{
void (* enc) BASE64_ENC_PARAMS;
int (* dec) BASE64_DEC_PARAMS;
};
extern void codec_choose (struct codec *, int flags);

7
src/third-party/base64/lib/config.h vendored Normal file
View File

@ -0,0 +1,7 @@
#define HAVE_AVX2 0
#define HAVE_NEON32 0
#define HAVE_NEON64 0
#define HAVE_SSSE3 0
#define HAVE_SSE41 0
#define HAVE_SSE42 0
#define HAVE_AVX 0

74
src/third-party/base64/lib/env.h vendored Normal file
View File

@ -0,0 +1,74 @@
#ifndef BASE64_ENV_H
#define BASE64_ENV_H
// This header file contains macro definitions that describe certain aspects of
// the compile-time environment. Compatibility and portability macros go here.
// Define machine endianness. This is for GCC:
#if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
# define BASE64_LITTLE_ENDIAN 1
#else
# define BASE64_LITTLE_ENDIAN 0
#endif
// This is for Clang:
#ifdef __LITTLE_ENDIAN__
# define BASE64_LITTLE_ENDIAN 1
#endif
#ifdef __BIG_ENDIAN__
# define BASE64_LITTLE_ENDIAN 0
#endif
// MSVC++ needs intrin.h for _byteswap_uint64 (issue #68):
#if BASE64_LITTLE_ENDIAN && defined(_MSC_VER)
# include <intrin.h>
#endif
// Endian conversion functions:
#if BASE64_LITTLE_ENDIAN
# ifdef _MSC_VER
// Microsoft Visual C++:
# define BASE64_HTOBE32(x) _byteswap_ulong(x)
# define BASE64_HTOBE64(x) _byteswap_uint64(x)
# else
// GCC and Clang:
# define BASE64_HTOBE32(x) __builtin_bswap32(x)
# define BASE64_HTOBE64(x) __builtin_bswap64(x)
# endif
#else
// No conversion needed:
# define BASE64_HTOBE32(x) (x)
# define BASE64_HTOBE64(x) (x)
#endif
// Detect word size:
#if defined (__x86_64__)
// This also works for the x32 ABI, which has a 64-bit word size.
# define BASE64_WORDSIZE 64
#elif defined (_INTEGRAL_MAX_BITS)
# define BASE64_WORDSIZE _INTEGRAL_MAX_BITS
#elif defined (__WORDSIZE)
# define BASE64_WORDSIZE __WORDSIZE
#elif defined (__SIZE_WIDTH__)
# define BASE64_WORDSIZE __SIZE_WIDTH__
#else
# error BASE64_WORDSIZE_NOT_DEFINED
#endif
// End-of-file definitions.
// Almost end-of-file when waiting for the last '=' character:
#define BASE64_AEOF 1
// End-of-file when stream end has been reached or invalid input provided:
#define BASE64_EOF 2
// GCC 7 defaults to issuing a warning for fallthrough in switch statements,
// unless the fallthrough cases are marked with an attribute. As we use
// fallthrough deliberately, define an alias for the attribute:
#if __GNUC__ >= 7
# define BASE64_FALLTHROUGH __attribute__((fallthrough));
#else
# define BASE64_FALLTHROUGH
#endif
#endif // BASE64_ENV_H

175
src/third-party/base64/lib/lib.c vendored Normal file
View File

@ -0,0 +1,175 @@
#include <stdint.h>
#include <stddef.h>
#ifdef _OPENMP
#include <omp.h>
#endif
#include "../include/libbase64.h"
#include "tables/tables.h"
#include "codecs.h"
#include "env.h"
// These static function pointers are initialized once when the library is
// first used, and remain in use for the remaining lifetime of the program.
// The idea being that CPU features don't change at runtime.
static struct codec codec = { NULL, NULL };
// Function declarations:
#define BASE64_CODEC_FUNCS(arch) \
BASE64_ENC_FUNCTION(arch); \
BASE64_DEC_FUNCTION(arch); \
BASE64_CODEC_FUNCS(plain)
void
base64_stream_encode_init (struct base64_state *state, int flags)
{
// If any of the codec flags are set, redo choice:
if (codec.enc == NULL || flags & 0xFF) {
// codec_choose(&codec, flags);
codec.enc = base64_stream_encode_plain;
codec.dec = base64_stream_decode_plain;
}
state->eof = 0;
state->bytes = 0;
state->carry = 0;
state->flags = flags;
}
void
base64_stream_encode
( struct base64_state *state
, const char *src
, size_t srclen
, char *out
, size_t *outlen
)
{
codec.enc(state, src, srclen, out, outlen);
}
void
base64_stream_encode_final
( struct base64_state *state
, char *out
, size_t *outlen
)
{
uint8_t *o = (uint8_t *)out;
if (state->bytes == 1) {
*o++ = base64_table_enc_6bit[state->carry];
*o++ = '=';
*o++ = '=';
*outlen = 3;
return;
}
if (state->bytes == 2) {
*o++ = base64_table_enc_6bit[state->carry];
*o++ = '=';
*outlen = 2;
return;
}
*outlen = 0;
}
void
base64_stream_decode_init (struct base64_state *state, int flags)
{
// If any of the codec flags are set, redo choice:
if (codec.dec == NULL || flags & 0xFF) {
// codec_choose(&codec, flags);
codec.enc = base64_stream_encode_plain;
codec.dec = base64_stream_decode_plain;
}
state->eof = 0;
state->bytes = 0;
state->carry = 0;
state->flags = flags;
}
int
base64_stream_decode
( struct base64_state *state
, const char *src
, size_t srclen
, char *out
, size_t *outlen
)
{
return codec.dec(state, src, srclen, out, outlen);
}
#ifdef _OPENMP
// Due to the overhead of initializing OpenMP and creating a team of
// threads, we require the data length to be larger than a threshold:
#define OMP_THRESHOLD 20000
// Conditionally include OpenMP-accelerated codec implementations:
#include "lib_openmp.c"
#endif
void
base64_encode
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
)
{
size_t s;
size_t t;
struct base64_state state;
#ifdef _OPENMP
if (srclen >= OMP_THRESHOLD) {
base64_encode_openmp(src, srclen, out, outlen, flags);
return;
}
#endif
// Init the stream reader:
base64_stream_encode_init(&state, flags);
// Feed the whole string to the stream reader:
base64_stream_encode(&state, src, srclen, out, &s);
// Finalize the stream by writing trailer if any:
base64_stream_encode_final(&state, out + s, &t);
// Final output length is stream length plus tail:
*outlen = s + t;
}
int
base64_decode
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
)
{
int ret;
struct base64_state state;
#ifdef _OPENMP
if (srclen >= OMP_THRESHOLD) {
return base64_decode_openmp(src, srclen, out, outlen, flags);
}
#endif
// Init the stream reader:
base64_stream_decode_init(&state, flags);
// Feed the whole string to the stream reader:
ret = base64_stream_decode(&state, src, srclen, out, outlen);
// If when decoding a whole block, we're still waiting for input then fail:
if (ret && (state.bytes == 0)) {
return ret;
}
return 0;
}

149
src/third-party/base64/lib/lib_openmp.c vendored Normal file
View File

@ -0,0 +1,149 @@
// This code makes some assumptions on the implementation of
// base64_stream_encode_init(), base64_stream_encode() and base64_stream_decode().
// Basically these assumptions boil down to that when breaking the src into
// parts, out parts can be written without side effects.
// This is met when:
// 1) base64_stream_encode() and base64_stream_decode() don't use globals;
// 2) the shared variables src and out are not read or written outside of the
// bounds of their parts, i.e. when base64_stream_encode() reads a multiple
// of 3 bytes, it must write no more then a multiple of 4 bytes, not even
// temporarily;
// 3) the state flag can be discarded after base64_stream_encode() and
// base64_stream_decode() on the parts.
static inline void
base64_encode_openmp
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
)
{
size_t s;
size_t t;
size_t sum = 0, len, last_len;
struct base64_state state, initial_state;
int num_threads, i;
// Request a number of threads but not necessarily get them:
#pragma omp parallel
{
// Get the number of threads used from one thread only,
// as num_threads is a shared var:
#pragma omp single
{
num_threads = omp_get_num_threads();
// Split the input string into num_threads parts, each
// part a multiple of 3 bytes. The remaining bytes will
// be done later:
len = srclen / (num_threads * 3);
len *= 3;
last_len = srclen - num_threads * len;
// Init the stream reader:
base64_stream_encode_init(&state, flags);
initial_state = state;
}
// Single has an implicit barrier for all threads to wait here
// for the above to complete:
#pragma omp for firstprivate(state) private(s) reduction(+:sum) schedule(static,1)
for (i = 0; i < num_threads; i++)
{
// Feed each part of the string to the stream reader:
base64_stream_encode(&state, src + i * len, len, out + i * len * 4 / 3, &s);
sum += s;
}
}
// As encoding should never fail and we encode an exact multiple
// of 3 bytes, we can discard state:
state = initial_state;
// Encode the remaining bytes:
base64_stream_encode(&state, src + num_threads * len, last_len, out + num_threads * len * 4 / 3, &s);
// Finalize the stream by writing trailer if any:
base64_stream_encode_final(&state, out + num_threads * len * 4 / 3 + s, &t);
// Final output length is stream length plus tail:
sum += s + t;
*outlen = sum;
}
static inline int
base64_decode_openmp
( const char *src
, size_t srclen
, char *out
, size_t *outlen
, int flags
)
{
int num_threads, result = 0, i;
size_t sum = 0, len, last_len, s;
struct base64_state state, initial_state;
// Request a number of threads but not necessarily get them:
#pragma omp parallel
{
// Get the number of threads used from one thread only,
// as num_threads is a shared var:
#pragma omp single
{
num_threads = omp_get_num_threads();
// Split the input string into num_threads parts, each
// part a multiple of 4 bytes. The remaining bytes will
// be done later:
len = srclen / (num_threads * 4);
len *= 4;
last_len = srclen - num_threads * len;
// Init the stream reader:
base64_stream_decode_init(&state, flags);
initial_state = state;
}
// Single has an implicit barrier to wait here for the above to
// complete:
#pragma omp for firstprivate(state) private(s) reduction(+:sum, result) schedule(static,1)
for (i = 0; i < num_threads; i++)
{
int this_result;
// Feed each part of the string to the stream reader:
this_result = base64_stream_decode(&state, src + i * len, len, out + i * len * 3 / 4, &s);
sum += s;
result += this_result;
}
}
// If `result' equals `-num_threads', then all threads returned -1,
// indicating that the requested codec is not available:
if (result == -num_threads) {
return -1;
}
// If `result' does not equal `num_threads', then at least one of the
// threads hit a decode error:
if (result != num_threads) {
return 0;
}
// So far so good, now decode whatever remains in the buffer. Reuse the
// initial state, since we are at a 4-byte boundary:
state = initial_state;
result = base64_stream_decode(&state, src + num_threads * len, last_len, out + num_threads * len * 3 / 4, &s);
sum += s;
*outlen = sum;
// If when decoding a whole block, we're still waiting for input then fail:
if (result && (state.bytes == 0)) {
return result;
}
return 0;
}

393
src/third-party/base64/lib/tables/table_dec_32bit.h vendored Normal file
View File

@ -0,0 +1,393 @@
#include <stdint.h>
#define CHAR62 '+'
#define CHAR63 '/'
#define CHARPAD '='
#if BASE64_LITTLE_ENDIAN
/* SPECIAL DECODE TABLES FOR LITTLE ENDIAN (INTEL) CPUS */
const uint32_t base64_table_dec_32bit_d0[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x000000f8, 0xffffffff, 0xffffffff, 0xffffffff, 0x000000fc,
0x000000d0, 0x000000d4, 0x000000d8, 0x000000dc, 0x000000e0, 0x000000e4,
0x000000e8, 0x000000ec, 0x000000f0, 0x000000f4, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00000004, 0x00000008, 0x0000000c, 0x00000010, 0x00000014, 0x00000018,
0x0000001c, 0x00000020, 0x00000024, 0x00000028, 0x0000002c, 0x00000030,
0x00000034, 0x00000038, 0x0000003c, 0x00000040, 0x00000044, 0x00000048,
0x0000004c, 0x00000050, 0x00000054, 0x00000058, 0x0000005c, 0x00000060,
0x00000064, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00000068, 0x0000006c, 0x00000070, 0x00000074, 0x00000078,
0x0000007c, 0x00000080, 0x00000084, 0x00000088, 0x0000008c, 0x00000090,
0x00000094, 0x00000098, 0x0000009c, 0x000000a0, 0x000000a4, 0x000000a8,
0x000000ac, 0x000000b0, 0x000000b4, 0x000000b8, 0x000000bc, 0x000000c0,
0x000000c4, 0x000000c8, 0x000000cc, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d1[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x0000e003, 0xffffffff, 0xffffffff, 0xffffffff, 0x0000f003,
0x00004003, 0x00005003, 0x00006003, 0x00007003, 0x00008003, 0x00009003,
0x0000a003, 0x0000b003, 0x0000c003, 0x0000d003, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00001000, 0x00002000, 0x00003000, 0x00004000, 0x00005000, 0x00006000,
0x00007000, 0x00008000, 0x00009000, 0x0000a000, 0x0000b000, 0x0000c000,
0x0000d000, 0x0000e000, 0x0000f000, 0x00000001, 0x00001001, 0x00002001,
0x00003001, 0x00004001, 0x00005001, 0x00006001, 0x00007001, 0x00008001,
0x00009001, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x0000a001, 0x0000b001, 0x0000c001, 0x0000d001, 0x0000e001,
0x0000f001, 0x00000002, 0x00001002, 0x00002002, 0x00003002, 0x00004002,
0x00005002, 0x00006002, 0x00007002, 0x00008002, 0x00009002, 0x0000a002,
0x0000b002, 0x0000c002, 0x0000d002, 0x0000e002, 0x0000f002, 0x00000003,
0x00001003, 0x00002003, 0x00003003, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d2[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00800f00, 0xffffffff, 0xffffffff, 0xffffffff, 0x00c00f00,
0x00000d00, 0x00400d00, 0x00800d00, 0x00c00d00, 0x00000e00, 0x00400e00,
0x00800e00, 0x00c00e00, 0x00000f00, 0x00400f00, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00400000, 0x00800000, 0x00c00000, 0x00000100, 0x00400100, 0x00800100,
0x00c00100, 0x00000200, 0x00400200, 0x00800200, 0x00c00200, 0x00000300,
0x00400300, 0x00800300, 0x00c00300, 0x00000400, 0x00400400, 0x00800400,
0x00c00400, 0x00000500, 0x00400500, 0x00800500, 0x00c00500, 0x00000600,
0x00400600, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00800600, 0x00c00600, 0x00000700, 0x00400700, 0x00800700,
0x00c00700, 0x00000800, 0x00400800, 0x00800800, 0x00c00800, 0x00000900,
0x00400900, 0x00800900, 0x00c00900, 0x00000a00, 0x00400a00, 0x00800a00,
0x00c00a00, 0x00000b00, 0x00400b00, 0x00800b00, 0x00c00b00, 0x00000c00,
0x00400c00, 0x00800c00, 0x00c00c00, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d3[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x003e0000, 0xffffffff, 0xffffffff, 0xffffffff, 0x003f0000,
0x00340000, 0x00350000, 0x00360000, 0x00370000, 0x00380000, 0x00390000,
0x003a0000, 0x003b0000, 0x003c0000, 0x003d0000, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00010000, 0x00020000, 0x00030000, 0x00040000, 0x00050000, 0x00060000,
0x00070000, 0x00080000, 0x00090000, 0x000a0000, 0x000b0000, 0x000c0000,
0x000d0000, 0x000e0000, 0x000f0000, 0x00100000, 0x00110000, 0x00120000,
0x00130000, 0x00140000, 0x00150000, 0x00160000, 0x00170000, 0x00180000,
0x00190000, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x001a0000, 0x001b0000, 0x001c0000, 0x001d0000, 0x001e0000,
0x001f0000, 0x00200000, 0x00210000, 0x00220000, 0x00230000, 0x00240000,
0x00250000, 0x00260000, 0x00270000, 0x00280000, 0x00290000, 0x002a0000,
0x002b0000, 0x002c0000, 0x002d0000, 0x002e0000, 0x002f0000, 0x00300000,
0x00310000, 0x00320000, 0x00330000, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
#else
/* SPECIAL DECODE TABLES FOR BIG ENDIAN (IBM/MOTOROLA/SUN) CPUS */
const uint32_t base64_table_dec_32bit_d0[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xf8000000, 0xffffffff, 0xffffffff, 0xffffffff, 0xfc000000,
0xd0000000, 0xd4000000, 0xd8000000, 0xdc000000, 0xe0000000, 0xe4000000,
0xe8000000, 0xec000000, 0xf0000000, 0xf4000000, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x04000000, 0x08000000, 0x0c000000, 0x10000000, 0x14000000, 0x18000000,
0x1c000000, 0x20000000, 0x24000000, 0x28000000, 0x2c000000, 0x30000000,
0x34000000, 0x38000000, 0x3c000000, 0x40000000, 0x44000000, 0x48000000,
0x4c000000, 0x50000000, 0x54000000, 0x58000000, 0x5c000000, 0x60000000,
0x64000000, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x68000000, 0x6c000000, 0x70000000, 0x74000000, 0x78000000,
0x7c000000, 0x80000000, 0x84000000, 0x88000000, 0x8c000000, 0x90000000,
0x94000000, 0x98000000, 0x9c000000, 0xa0000000, 0xa4000000, 0xa8000000,
0xac000000, 0xb0000000, 0xb4000000, 0xb8000000, 0xbc000000, 0xc0000000,
0xc4000000, 0xc8000000, 0xcc000000, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d1[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x03e00000, 0xffffffff, 0xffffffff, 0xffffffff, 0x03f00000,
0x03400000, 0x03500000, 0x03600000, 0x03700000, 0x03800000, 0x03900000,
0x03a00000, 0x03b00000, 0x03c00000, 0x03d00000, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00100000, 0x00200000, 0x00300000, 0x00400000, 0x00500000, 0x00600000,
0x00700000, 0x00800000, 0x00900000, 0x00a00000, 0x00b00000, 0x00c00000,
0x00d00000, 0x00e00000, 0x00f00000, 0x01000000, 0x01100000, 0x01200000,
0x01300000, 0x01400000, 0x01500000, 0x01600000, 0x01700000, 0x01800000,
0x01900000, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x01a00000, 0x01b00000, 0x01c00000, 0x01d00000, 0x01e00000,
0x01f00000, 0x02000000, 0x02100000, 0x02200000, 0x02300000, 0x02400000,
0x02500000, 0x02600000, 0x02700000, 0x02800000, 0x02900000, 0x02a00000,
0x02b00000, 0x02c00000, 0x02d00000, 0x02e00000, 0x02f00000, 0x03000000,
0x03100000, 0x03200000, 0x03300000, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d2[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x000f8000, 0xffffffff, 0xffffffff, 0xffffffff, 0x000fc000,
0x000d0000, 0x000d4000, 0x000d8000, 0x000dc000, 0x000e0000, 0x000e4000,
0x000e8000, 0x000ec000, 0x000f0000, 0x000f4000, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00004000, 0x00008000, 0x0000c000, 0x00010000, 0x00014000, 0x00018000,
0x0001c000, 0x00020000, 0x00024000, 0x00028000, 0x0002c000, 0x00030000,
0x00034000, 0x00038000, 0x0003c000, 0x00040000, 0x00044000, 0x00048000,
0x0004c000, 0x00050000, 0x00054000, 0x00058000, 0x0005c000, 0x00060000,
0x00064000, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00068000, 0x0006c000, 0x00070000, 0x00074000, 0x00078000,
0x0007c000, 0x00080000, 0x00084000, 0x00088000, 0x0008c000, 0x00090000,
0x00094000, 0x00098000, 0x0009c000, 0x000a0000, 0x000a4000, 0x000a8000,
0x000ac000, 0x000b0000, 0x000b4000, 0x000b8000, 0x000bc000, 0x000c0000,
0x000c4000, 0x000c8000, 0x000cc000, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
const uint32_t base64_table_dec_32bit_d3[256] = {
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00003e00, 0xffffffff, 0xffffffff, 0xffffffff, 0x00003f00,
0x00003400, 0x00003500, 0x00003600, 0x00003700, 0x00003800, 0x00003900,
0x00003a00, 0x00003b00, 0x00003c00, 0x00003d00, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0x00000000,
0x00000100, 0x00000200, 0x00000300, 0x00000400, 0x00000500, 0x00000600,
0x00000700, 0x00000800, 0x00000900, 0x00000a00, 0x00000b00, 0x00000c00,
0x00000d00, 0x00000e00, 0x00000f00, 0x00001000, 0x00001100, 0x00001200,
0x00001300, 0x00001400, 0x00001500, 0x00001600, 0x00001700, 0x00001800,
0x00001900, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0x00001a00, 0x00001b00, 0x00001c00, 0x00001d00, 0x00001e00,
0x00001f00, 0x00002000, 0x00002100, 0x00002200, 0x00002300, 0x00002400,
0x00002500, 0x00002600, 0x00002700, 0x00002800, 0x00002900, 0x00002a00,
0x00002b00, 0x00002c00, 0x00002d00, 0x00002e00, 0x00002f00, 0x00003000,
0x00003100, 0x00003200, 0x00003300, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff,
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
};
#endif

1031
src/third-party/base64/lib/tables/table_enc_12bit.h vendored Normal file
View File

File diff suppressed because it is too large Load Diff

40
src/third-party/base64/lib/tables/tables.c vendored Normal file
View File

@ -0,0 +1,40 @@
#include "tables.h"
const uint8_t
base64_table_enc_6bit[] =
"ABCDEFGHIJKLMNOPQRSTUVWXYZ"
"abcdefghijklmnopqrstuvwxyz"
"0123456789"
"+/";
// In the lookup table below, note that the value for '=' (character 61) is
// 254, not 255. This character is used for in-band signaling of the end of
// the datastream, and we will use that later. The characters A-Z, a-z, 0-9
// and + / are mapped to their "decoded" values. The other bytes all map to
// the value 255, which flags them as "invalid input".
const uint8_t
base64_table_dec_8bit[] =
{
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, // 0..15
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, // 16..31
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 62, 255, 255, 255, 63, // 32..47
52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 255, 255, 255, 254, 255, 255, // 48..63
255, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, // 64..79
15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 255, 255, 255, 255, 255, // 80..95
255, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, // 96..111
41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 255, 255, 255, 255, 255, // 112..127
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, // 128..143
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
};
#if BASE64_WORDSIZE >= 32
# include "table_dec_32bit.h"
# include "table_enc_12bit.h"
#endif

23
src/third-party/base64/lib/tables/tables.h vendored Normal file
View File

@ -0,0 +1,23 @@
#ifndef BASE64_TABLES_H
#define BASE64_TABLES_H
#include <stdint.h>
#include "../env.h"
// These tables are used by all codecs for fallback plain encoding/decoding:
extern const uint8_t base64_table_enc_6bit[];
extern const uint8_t base64_table_dec_8bit[];
// These tables are used for the 32-bit and 64-bit generic decoders:
#if BASE64_WORDSIZE >= 32
extern const uint32_t base64_table_dec_32bit_d0[];
extern const uint32_t base64_table_dec_32bit_d1[];
extern const uint32_t base64_table_dec_32bit_d2[];
extern const uint32_t base64_table_dec_32bit_d3[];
// This table is used by the 32 and 64-bit generic encoders:
extern const uint16_t base64_table_enc_12bit[];
#endif
#endif // BASE64_TABLES_H