0
0
Fork 0
mirror of https://github.com/matrix-construct/construct synced 2024-12-25 23:14:13 +01:00

ircd::b64: Add parallel SIMD base64 decoder.

This commit is contained in:
Jason Volk 2020-08-10 09:00:20 -07:00
parent d51082ef17
commit f3f441ac19
3 changed files with 123 additions and 24 deletions

View file

@ -80,5 +80,5 @@ constexpr size_t
ircd::b64::decode_size(const size_t &in)
noexcept
{
return (in * 0.75) + 1; //XXX: constexpr ceil()
return in * (3.0 / 4.0);
}

View file

@ -205,7 +205,6 @@ libircd_la_SOURCES += ircd.cc
# Specific unit configurations
#
b64.lo: AM_CPPFLAGS := @BOOST_CPPFLAGS@ ${AM_CPPFLAGS}
client.lo: AM_CPPFLAGS := ${ASIO_UNIT_CPPFLAGS} ${AM_CPPFLAGS}
ctx_x86_64.lo: AM_CPPFLAGS := -I$(top_srcdir)/include
ctx.lo: AM_CPPFLAGS := ${ASIO_UNIT_CPPFLAGS} ${AM_CPPFLAGS}

View file

@ -8,9 +8,6 @@
// copyright notice and this permission notice is present in all copies. The
// full license for this software is available in the LICENSE file.
#include <boost/archive/iterators/binary_from_base64.hpp>
#include <boost/archive/iterators/transform_width.hpp>
#pragma GCC visibility push(internal)
namespace ircd::b64
{
@ -22,7 +19,15 @@ namespace ircd::b64
[[gnu::aligned(64)]]
extern const u8
encode_permute_tab[64],
encode_shift_ctrl[64];
encode_shift_ctrl[64],
decode_permute_tab[64],
decode_permute_tab_le[64];
[[gnu::aligned(64)]]
extern const i32
decode_dictionary[256];
static u8x64 decode_block(const u8x64 in) noexcept;
template<const u8 (&dict)[64]>
static u8x64 encode_block(const u8x64 in) noexcept;
@ -56,8 +61,47 @@ ircd::b64::dict_rfc4648
'w', 'x', 'y', 'z', '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '-', '_',
};
decltype(ircd::b64::decode_dictionary)
ircd::b64::decode_dictionary
{
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 7
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 15
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 23
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 31
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 39
0x00, 0x00, 0x00, 62, 63, 62, 0x00, 63, // 47
52, 53, 54, 55, 56, 57, 58, 59, // 55
60, 61, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 63
0x00, 0, 1, 2, 3, 4, 5, 6, // 71
7, 8, 9, 10, 11, 12, 13, 14, // 79
15, 16, 17, 18, 19, 20, 21, 22, // 87
23, 24, 25, 0x00, 0x00, 0x00, 0x00, 63, // 95
0x00, 26, 27, 28, 29, 30, 31, 32, // 103
33, 34, 35, 36, 37, 38, 39, 40, // 111
41, 42, 43, 44, 45, 46, 47, 48, // 119
49, 50, 51, 0x00, 0x00, 0x00, 0x00, 0x00, // 127
};
decltype(ircd::b64::decode_permute_tab)
ircd::b64::decode_permute_tab
{
6, 0, 1, 2, 9, 10, 4, 5, 12, 13, 14, 8, 22, 16, 17, 18,
25, 26, 20, 21, 28, 29, 30, 24, 38, 32, 33, 34, 41, 42, 36, 37,
44, 45, 46, 40, 54, 48, 49, 50, 57, 58, 52, 53, 60, 61, 62, 56,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
};
/// byte-order swapped for each 32-bit word from above
decltype(ircd::b64::decode_permute_tab_le)
ircd::b64::decode_permute_tab_le
{
2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12, 18, 17, 16, 22,
21, 20, 26, 25, 24, 30, 29, 28, 34, 33, 32, 38, 37, 36, 42, 41,
40, 46, 45, 44, 50, 49, 48, 54, 53, 52, 58, 57, 56, 62, 61, 60,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
};
/// For vpermb
/// From arXiv:1910.05109v1 [Mula, Lemire] 2 Oct 2019
decltype(ircd::b64::encode_permute_tab)
ircd::b64::encode_permute_tab
{
@ -72,7 +116,6 @@ ircd::b64::encode_permute_tab
};
/// For vpmultishiftqb
/// From arXiv:1910.05109v1 [Mula, Lemire] 2 Oct 2019
decltype(ircd::b64::encode_shift_ctrl)
ircd::b64::encode_shift_ctrl
{
@ -226,28 +269,85 @@ ircd::const_buffer
ircd::b64::decode(const mutable_buffer &out,
const string_view &in)
{
namespace iterators = boost::archive::iterators;
using b64bf = iterators::binary_from_base64<const char *>;
using transform = iterators::transform_width<b64bf, 8, 6>;
const auto pads
const size_t out_len
{
endswith_count(in, b64::pad)
std::min(decode_size(in), size(out))
};
const auto e
size_t i(0), j(0);
for(; i + 1 <= (size(in) / 64) && i + 1 <= (out_len / 48); ++i)
{
std::copy(transform(begin(in)), transform(begin(in) + size(in) - pads), begin(out))
};
// Destination is indexed at 48 byte stride
const auto di
{
reinterpret_cast<u512x1_u *__restrict__>(data(out) + (i * 48))
};
const auto len
{
std::distance(begin(out), e)
};
// Source is indexed at 64 byte stride
const auto si
{
reinterpret_cast<const u512x1_u *__restrict__>(data(in) + (i * 64))
};
assert(size_t(len) <= size(out));
return const_buffer
*di = decode_block(*si);
}
for(; i * 64 < size(in) && i * 48 < out_len; ++i)
{
data(out), size_t(len)
u8x64 block {0};
for(j = 0; j < 64 && i * 64 + j < size(in); ++j)
block[j] = in[i * 64 + j];
block = decode_block(block);
for(j = 0; j < 48 && i * 48 + j < out_len; ++j)
out[i * 48 + j] = block[j];
}
return string_view
{
data(out), out_len
};
}
/// Decode 64 base64 characters into a 48 byte result. The last 48 bytes of
/// the returned vector are undefined for the caller.
ircd::u8x64
ircd::b64::decode_block(const u8x64 in)
noexcept
{
size_t i, j;
i32x16 zz[4];
for(i = 0; i < 4; ++i)
for(j = 0; j < 16; ++j)
zz[i][j] = decode_dictionary[in[i * 16 + j]];
u8x64 z;
for(i = 0; i < 4; ++i)
for(j = 0; j < 16; ++j)
z[i * 16 + j] = zz[i][j];
u16x32 al, ah;
for(i = 0, j = 0; i < 32; ++i)
ah[i] = z[j++],
al[i] = z[j++];
u16x32 a;
for(i = 0; i < 32; ++i)
a[i] = ah[i] * 64 + al[i];
u32x16 bl, bh;
for(i = 0, j = 0; i < 16; ++i)
bh[i] = a[j++],
bl[i] = a[j++];
u32x16 b;
for(i = 0; i < 16; ++i)
b[i] = bh[i] * 4096 + bl[i];
u8x64 d, c(b);
for(i = 0; i < 64; ++i)
d[i] = c[decode_permute_tab_le[i]];
return d;
}