Implement std/crc32 ieee_hasher.up_x86_sse42

On a mid-range x86_64 laptop:

----

$ ls -lh linux-5.11.3.tar.gz | awk '{print $5 " " $9}'
178M linux-5.11.3.tar.gz

$ # Debian's /usr/bin/crc32
$ time /usr/bin/crc32 /dev/stdin < linux-5.11.3.tar.gz > /dev/null
real    0m0.411s

$ # Wuffs; before
$ time gen/bin/example-crc32     < linux-5.11.3.tar.gz > /dev/null
real    0m0.137s

$ # Wuffs; after
$ time gen/bin/example-crc32     < linux-5.11.3.tar.gz > /dev/null
real    0m0.060s

----

Wuffs benchmarks:

wuffs_crc32_ieee_10k/clang9                               1.71GB/s ± 0%   8.40GB/s ± 3%  +391.97%  (p=0.008 n=5+5)
wuffs_crc32_ieee_100k/clang9                              1.73GB/s ± 1%  10.67GB/s ± 0%  +518.47%  (p=0.008 n=5+5)

wuffs_crc32_ieee_10k/gcc10                                2.08GB/s ± 0%   8.77GB/s ± 0%  +322.05%  (p=0.008 n=5+5)
wuffs_crc32_ieee_100k/gcc10                               2.10GB/s ± 0%  11.23GB/s ± 0%  +435.56%  (p=0.008 n=5+5)

mimic_crc32_ieee_10k                                       774MB/s ± 0%    776MB/s ± 0%    +0.23%  (p=0.016 n=5+5)
mimic_crc32_ieee_100k                                      773MB/s ± 0%    775MB/s ± 0%    +0.24%  (p=0.032 n=5+5)

wuffs_gzip_decode_10k/clang9                               151MB/s ± 0%    166MB/s ± 0%    +9.81%  (p=0.008 n=5+5)
wuffs_gzip_decode_100k/clang9                              155MB/s ± 0%    178MB/s ± 0%   +14.53%  (p=0.008 n=5+5)

wuffs_gzip_decode_10k/gcc10                                150MB/s ± 0%    172MB/s ± 0%   +14.92%  (p=0.008 n=5+5)
wuffs_gzip_decode_100k/gcc10                               188MB/s ± 0%    198MB/s ± 0%    +5.37%  (p=0.008 n=5+5)

mimic_gzip_decode_10k                                      121MB/s ± 0%    121MB/s ± 0%      ~     (p=1.000 n=5+5)
mimic_gzip_decode_100k                                     143MB/s ± 0%    144MB/s ± 0%      ~     (p=0.056 n=5+5)

wuffs_png_decode_image_19k_8bpp/clang9                     111MB/s ± 0%    115MB/s ± 0%    +3.77%  (p=0.008 n=5+5)
wuffs_png_decode_image_40k_24bpp/clang9                    147MB/s ± 0%    153MB/s ± 0%    +4.66%  (p=0.008 n=5+5)
wuffs_png_decode_image_77k_8bpp/clang9                     404MB/s ± 0%    415MB/s ± 0%    +2.75%  (p=0.016 n=5+4)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/clang9   344MB/s ± 0%    345MB/s ± 1%      ~     (p=0.151 n=5+5)
wuffs_png_decode_image_552k_32bpp_verify_checksum/clang9   329MB/s ± 0%    335MB/s ± 0%    +1.78%  (p=0.016 n=5+4)
wuffs_png_decode_image_4002k_24bpp/clang9                  148MB/s ± 0%    153MB/s ± 0%    +3.38%  (p=0.008 n=5+5)

wuffs_png_decode_image_19k_8bpp/gcc10                      120MB/s ± 0%    123MB/s ± 0%    +2.62%  (p=0.008 n=5+5)
wuffs_png_decode_image_40k_24bpp/gcc10                     153MB/s ± 0%    159MB/s ± 1%    +3.54%  (p=0.008 n=5+5)
wuffs_png_decode_image_77k_8bpp/gcc10                      440MB/s ± 0%    452MB/s ± 0%    +2.82%  (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_ignore_checksum/gcc10    357MB/s ± 0%    353MB/s ± 0%    -0.95%  (p=0.008 n=5+5)
wuffs_png_decode_image_552k_32bpp_verify_checksum/gcc10    341MB/s ± 0%    341MB/s ± 0%      ~     (p=0.151 n=5+5)
wuffs_png_decode_image_4002k_24bpp/gcc10                   157MB/s ± 0%    161MB/s ± 0%    +2.85%  (p=0.016 n=5+4)

mimic_png_decode_image_19k_8bpp                           58.1MB/s ± 0%   58.1MB/s ± 0%      ~     (p=0.389 n=5+5)
mimic_png_decode_image_40k_24bpp                          73.2MB/s ± 0%   73.2MB/s ± 0%      ~     (p=0.548 n=5+5)
mimic_png_decode_image_77k_8bpp                            177MB/s ± 0%    177MB/s ± 0%    +0.14%  (p=0.008 n=5+5)
mimic_png_decode_image_552k_32bpp_ignore_checksum               skipped
mimic_png_decode_image_552k_32bpp_verify_checksum          146MB/s ± 0%    147MB/s ± 0%    +0.13%  (p=0.016 n=5+5)
mimic_png_decode_image_4002k_24bpp                         104MB/s ± 0%    104MB/s ± 0%      ~     (p=0.556 n=5+4)
This commit is contained in:
Nigel Tao committed 2021-03-07 12:27:45 +11:00
1 parent 44acd0e20a
commit 5d4fbc3ac6
6 files changed
+239 -25

No files matched your search

+7 -8
View File
@@ -78,10 +78,11 @@ polynomial is `N=4, LSB, BITS=0xC`. If `N` and `LSB`-ness is agreed beforehand,
`0xC` is all you need to specify the polynomial.
For 32 bit CRCs, the two popular polynomials (presented in `LSB` order) are
`0xEDB88320` and `0x82F63B78`, also called the IEEE and Castagnoli polynomials,
also called CRC-32 and CRC-32C. For example, the bit string representation of
the IEEE polynomial `0xEDB88320`, un-reversed and with the implicit high bit,
is `0b1_00000100_11000001_00011101_10110111`.
`0xEDB8_8320` and `0x82F6_3B78`, also called the IEEE and Castagnoli
polynomials, also called CRC-32 and CRC-32C. For example, the bit string
representation of the IEEE polynomial `0xEDB8_8320`, un-reversed and with the
implicit high bit, is `0b1_00000100_11000001_00011101_10110111 =
0x1_04C1_1DB7`.
# Worked Example
@@ -138,8 +139,8 @@ invert 01011001 00111100 01000100 10101011
hex A 9 C 3 2 2 5 D
```
The final line says that the CRC-32 checksum of "Hi\n" is 0xD5223C9A. This can
be verified by running the `/usr/bin/crc32` program:
The final line says that the CRC-32 checksum of "Hi\n" is `0xD522_3C9A`. This
can be verified by running the `/usr/bin/crc32` program:
```
$ echo Hi | xxd /dev/stdin
@@ -226,8 +227,6 @@ Instruction](https://www.intel.com/content/dam/www/public/us/en/documents/white-
by Gopal, Ozturk, Guilford, Wolrich, Feghali and Dixon of Intel Corporation and
Karakoyunlu of the Worcester Polytechnic Institute.
Wuffs does not currently implement the SIMD algorithm.
# Further Reading
+115 -3
View File
@@ -12,12 +12,30 @@
// See the License for the specific language governing permissions and
// limitations under the License.
// --------
// See "SIMD Implementations" in README.md for a link to Gopal et al. "Fast CRC
// Computation for Generic Polynomials Using PCLMULQDQ Instruction".
pri func ieee_hasher.up_x86_sse42!(x: slice base.u8),
choose cpu_arch >= x86_sse42,
{
var s : base.u32
var p : slice base.u8
var util : base.x86_sse42_utility
var k : base.x86_m128i
var x0 : base.x86_m128i
var x1 : base.x86_m128i
var x2 : base.x86_m128i
var x3 : base.x86_m128i
var y0 : base.x86_m128i
var y1 : base.x86_m128i
var y2 : base.x86_m128i
var y3 : base.x86_m128i
var tail_index : base.u64
s = 0xFFFF_FFFF ^ this.state
// Align to a 16-byte boundary.
@@ -35,10 +53,104 @@ pri func ieee_hasher.up_x86_sse42!(x: slice base.u8),
return nothing
}
// Placeholder.
iterate (p = args.x)(length: 1, advance: 1, unroll: 1) {
s = IEEE_TABLE[0][((s & 0xFF) as base.u8) ^ p[0]] ^ (s >> 8)
// Load 128×4 = 512 bits from the first 64-byte chunk.
x0 = util.make_m128i_slice128(a: args.x[0x00 .. 0x10])
x1 = util.make_m128i_slice128(a: args.x[0x10 .. 0x20])
x2 = util.make_m128i_slice128(a: args.x[0x20 .. 0x30])
x3 = util.make_m128i_slice128(a: args.x[0x30 .. 0x40])
// Combine with the initial state.
x0 = x0._mm_xor_si128(b: util.make_m128i_single_u32(a: s))
// Process the remaining 64-byte chunks.
k = util.make_m128i_slice128(a: IEEE_X86_SSE42_K1K2[.. 16])
iterate (p = args.x[64 ..])(length: 64, advance: 64, unroll: 1) {
y0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x00)
y1 = x1._mm_clmulepi64_si128(b: k, imm8: 0x00)
y2 = x2._mm_clmulepi64_si128(b: k, imm8: 0x00)
y3 = x3._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x11)
x1 = x1._mm_clmulepi64_si128(b: k, imm8: 0x11)
x2 = x2._mm_clmulepi64_si128(b: k, imm8: 0x11)
x3 = x3._mm_clmulepi64_si128(b: k, imm8: 0x11)
x0 = x0._mm_xor_si128(b: y0)._mm_xor_si128(b: util.make_m128i_slice128(a: p[0x00 .. 0x10]))
x1 = x1._mm_xor_si128(b: y1)._mm_xor_si128(b: util.make_m128i_slice128(a: p[0x10 .. 0x20]))
x2 = x2._mm_xor_si128(b: y2)._mm_xor_si128(b: util.make_m128i_slice128(a: p[0x20 .. 0x30]))
x3 = x3._mm_xor_si128(b: y3)._mm_xor_si128(b: util.make_m128i_slice128(a: p[0x30 .. 0x40]))
}
// Reduce 128×4 = 512 bits to 128 bits.
k = util.make_m128i_slice128(a: IEEE_X86_SSE42_K3K4[.. 16])
y0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x11)
x0 = x0._mm_xor_si128(b: x1)
x0 = x0._mm_xor_si128(b: y0)
y0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x11)
x0 = x0._mm_xor_si128(b: x2)
x0 = x0._mm_xor_si128(b: y0)
y0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x11)
x0 = x0._mm_xor_si128(b: x3)
x0 = x0._mm_xor_si128(b: y0)
// Reduce 128 bits to 64 bits.
x1 = x0._mm_clmulepi64_si128(b: k, imm8: 0x10)
x2 = util.make_m128i_multiple_u32(
a00: 0xFFFF_FFFF,
a01: 0x0000_0000,
a02: 0xFFFF_FFFF,
a03: 0x0000_0000)
x0 = x0._mm_srli_si128(imm8: 8)
x0 = x0._mm_xor_si128(b: x1)
k = util.make_m128i_slice128(a: IEEE_X86_SSE42_K5ZZ[.. 16])
x1 = x0._mm_srli_si128(imm8: 4)
x0 = x0._mm_and_si128(b: x2)
x0 = x0._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_xor_si128(b: x1)
// Reduce 64 bits to 32 bits and extract.
k = util.make_m128i_slice128(a: IEEE_X86_SSE42_K6MU[.. 16])
x1 = x0._mm_and_si128(b: x2)
x1 = x1._mm_clmulepi64_si128(b: k, imm8: 0x10)
x1 = x1._mm_and_si128(b: x2)
x1 = x1._mm_clmulepi64_si128(b: k, imm8: 0x00)
x0 = x0._mm_xor_si128(b: x1)
s = x0._mm_extract_epi32(imm8: 1)
// Handle the tail of args.x that wasn't a complete 64-byte chunk.
tail_index = args.x.length() & 0xFFFF_FFFF_FFFF_FFC0 // And-not 64.
if tail_index < args.x.length() {
iterate (p = args.x[tail_index ..])(length: 1, advance: 1, unroll: 1) {
s = IEEE_TABLE[0][((s & 0xFF) as base.u8) ^ p[0]] ^ (s >> 8)
}
}
this.state = 0xFFFF_FFFF ^ s
}
// These constants come from page 22 of Gopal et al.
//
// (†) TODO: explain the discrepancy.
pri const IEEE_X86_SSE42_K1K2 : array[16] base.u8 = [
0xD4, 0x2B, 0x44, 0x54, 0x01, 0x00, 0x00, 0x00, // k1' = 0x1_5444_2BD4
0x96, 0x15, 0xE4, 0xC6, 0x01, 0x00, 0x00, 0x00, // k2' = 0x1_C6E4_1596
]
pri const IEEE_X86_SSE42_K3K4 : array[16] base.u8 = [
0xD0, 0x97, 0x19, 0x75, 0x01, 0x00, 0x00, 0x00, // k3' = 0x1_7519_97D0
0x9E, 0x00, 0xAA, 0xCC, 0x00, 0x00, 0x00, 0x00, // k4' = 0x0_CCAA_009E
]
pri const IEEE_X86_SSE42_K5ZZ : array[16] base.u8 = [
0x24, 0x61, 0xCD, 0x63, 0x01, 0x00, 0x00, 0x00, // k5' = 0x1_63CD_6124
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // Unused
]
pri const IEEE_X86_SSE42_K6MU : array[16] base.u8 = [
0x41, 0x06, 0x71, 0xDB, 0x01, 0x00, 0x00, 0x00, // k6' = 0x1_DB71_0640 (†)
0x41, 0x16, 0x01, 0xF7, 0x01, 0x00, 0x00, 0x00, // μ' = 0x1_F701_1641
]