Update mattermost library (#2152)

* Update mattermost library

* Fix linting
This commit is contained in:
Wim
2024-05-24 23:08:09 +02:00
committed by GitHub
parent 65d78e38af
commit d16645c952
1003 changed files with 89451 additions and 114025 deletions

View File

@@ -1,283 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package argon2 implements the key derivation function Argon2.
// Argon2 was selected as the winner of the Password Hashing Competition and can
// be used to derive cryptographic keys from passwords.
//
// For a detailed specification of Argon2 see [1].
//
// If you aren't sure which function you need, use Argon2id (IDKey) and
// the parameter recommendations for your scenario.
//
// # Argon2i
//
// Argon2i (implemented by Key) is the side-channel resistant version of Argon2.
// It uses data-independent memory access, which is preferred for password
// hashing and password-based key derivation. Argon2i requires more passes over
// memory than Argon2id to protect from trade-off attacks. The recommended
// parameters (taken from [2]) for non-interactive operations are time=3 and to
// use the maximum available memory.
//
// # Argon2id
//
// Argon2id (implemented by IDKey) is a hybrid version of Argon2 combining
// Argon2i and Argon2d. It uses data-independent memory access for the first
// half of the first iteration over the memory and data-dependent memory access
// for the rest. Argon2id is side-channel resistant and provides better brute-
// force cost savings due to time-memory tradeoffs than Argon2i. The recommended
// parameters for non-interactive operations (taken from [2]) are time=1 and to
// use the maximum available memory.
//
// [1] https://github.com/P-H-C/phc-winner-argon2/blob/master/argon2-specs.pdf
// [2] https://tools.ietf.org/html/draft-irtf-cfrg-argon2-03#section-9.3
package argon2
import (
"encoding/binary"
"sync"
"golang.org/x/crypto/blake2b"
)
// The Argon2 version implemented by this package.
const Version = 0x13
const (
argon2d = iota
argon2i
argon2id
)
// Key derives a key from the password, salt, and cost parameters using Argon2i
// returning a byte slice of length keyLen that can be used as cryptographic
// key. The CPU cost and parallelism degree must be greater than zero.
//
// For example, you can get a derived key for e.g. AES-256 (which needs a
// 32-byte key) by doing:
//
// key := argon2.Key([]byte("some password"), salt, 3, 32*1024, 4, 32)
//
// The draft RFC recommends[2] time=3, and memory=32*1024 is a sensible number.
// If using that amount of memory (32 MB) is not possible in some contexts then
// the time parameter can be increased to compensate.
//
// The time parameter specifies the number of passes over the memory and the
// memory parameter specifies the size of the memory in KiB. For example
// memory=32*1024 sets the memory cost to ~32 MB. The number of threads can be
// adjusted to the number of available CPUs. The cost parameters should be
// increased as memory latency and CPU parallelism increases. Remember to get a
// good random salt.
func Key(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
return deriveKey(argon2i, password, salt, nil, nil, time, memory, threads, keyLen)
}
// IDKey derives a key from the password, salt, and cost parameters using
// Argon2id returning a byte slice of length keyLen that can be used as
// cryptographic key. The CPU cost and parallelism degree must be greater than
// zero.
//
// For example, you can get a derived key for e.g. AES-256 (which needs a
// 32-byte key) by doing:
//
// key := argon2.IDKey([]byte("some password"), salt, 1, 64*1024, 4, 32)
//
// The draft RFC recommends[2] time=1, and memory=64*1024 is a sensible number.
// If using that amount of memory (64 MB) is not possible in some contexts then
// the time parameter can be increased to compensate.
//
// The time parameter specifies the number of passes over the memory and the
// memory parameter specifies the size of the memory in KiB. For example
// memory=64*1024 sets the memory cost to ~64 MB. The number of threads can be
// adjusted to the numbers of available CPUs. The cost parameters should be
// increased as memory latency and CPU parallelism increases. Remember to get a
// good random salt.
func IDKey(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
return deriveKey(argon2id, password, salt, nil, nil, time, memory, threads, keyLen)
}
func deriveKey(mode int, password, salt, secret, data []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
if time < 1 {
panic("argon2: number of rounds too small")
}
if threads < 1 {
panic("argon2: parallelism degree too low")
}
h0 := initHash(password, salt, secret, data, time, memory, uint32(threads), keyLen, mode)
memory = memory / (syncPoints * uint32(threads)) * (syncPoints * uint32(threads))
if memory < 2*syncPoints*uint32(threads) {
memory = 2 * syncPoints * uint32(threads)
}
B := initBlocks(&h0, memory, uint32(threads))
processBlocks(B, time, memory, uint32(threads), mode)
return extractKey(B, memory, uint32(threads), keyLen)
}
const (
blockLength = 128
syncPoints = 4
)
type block [blockLength]uint64
func initHash(password, salt, key, data []byte, time, memory, threads, keyLen uint32, mode int) [blake2b.Size + 8]byte {
var (
h0 [blake2b.Size + 8]byte
params [24]byte
tmp [4]byte
)
b2, _ := blake2b.New512(nil)
binary.LittleEndian.PutUint32(params[0:4], threads)
binary.LittleEndian.PutUint32(params[4:8], keyLen)
binary.LittleEndian.PutUint32(params[8:12], memory)
binary.LittleEndian.PutUint32(params[12:16], time)
binary.LittleEndian.PutUint32(params[16:20], uint32(Version))
binary.LittleEndian.PutUint32(params[20:24], uint32(mode))
b2.Write(params[:])
binary.LittleEndian.PutUint32(tmp[:], uint32(len(password)))
b2.Write(tmp[:])
b2.Write(password)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(salt)))
b2.Write(tmp[:])
b2.Write(salt)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(key)))
b2.Write(tmp[:])
b2.Write(key)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(data)))
b2.Write(tmp[:])
b2.Write(data)
b2.Sum(h0[:0])
return h0
}
func initBlocks(h0 *[blake2b.Size + 8]byte, memory, threads uint32) []block {
var block0 [1024]byte
B := make([]block, memory)
for lane := uint32(0); lane < threads; lane++ {
j := lane * (memory / threads)
binary.LittleEndian.PutUint32(h0[blake2b.Size+4:], lane)
binary.LittleEndian.PutUint32(h0[blake2b.Size:], 0)
blake2bHash(block0[:], h0[:])
for i := range B[j+0] {
B[j+0][i] = binary.LittleEndian.Uint64(block0[i*8:])
}
binary.LittleEndian.PutUint32(h0[blake2b.Size:], 1)
blake2bHash(block0[:], h0[:])
for i := range B[j+1] {
B[j+1][i] = binary.LittleEndian.Uint64(block0[i*8:])
}
}
return B
}
func processBlocks(B []block, time, memory, threads uint32, mode int) {
lanes := memory / threads
segments := lanes / syncPoints
processSegment := func(n, slice, lane uint32, wg *sync.WaitGroup) {
var addresses, in, zero block
if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
in[0] = uint64(n)
in[1] = uint64(lane)
in[2] = uint64(slice)
in[3] = uint64(memory)
in[4] = uint64(time)
in[5] = uint64(mode)
}
index := uint32(0)
if n == 0 && slice == 0 {
index = 2 // we have already generated the first two blocks
if mode == argon2i || mode == argon2id {
in[6]++
processBlock(&addresses, &in, &zero)
processBlock(&addresses, &addresses, &zero)
}
}
offset := lane*lanes + slice*segments + index
var random uint64
for index < segments {
prev := offset - 1
if index == 0 && slice == 0 {
prev += lanes // last block in lane
}
if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
if index%blockLength == 0 {
in[6]++
processBlock(&addresses, &in, &zero)
processBlock(&addresses, &addresses, &zero)
}
random = addresses[index%blockLength]
} else {
random = B[prev][0]
}
newOffset := indexAlpha(random, lanes, segments, threads, n, slice, lane, index)
processBlockXOR(&B[offset], &B[prev], &B[newOffset])
index, offset = index+1, offset+1
}
wg.Done()
}
for n := uint32(0); n < time; n++ {
for slice := uint32(0); slice < syncPoints; slice++ {
var wg sync.WaitGroup
for lane := uint32(0); lane < threads; lane++ {
wg.Add(1)
go processSegment(n, slice, lane, &wg)
}
wg.Wait()
}
}
}
func extractKey(B []block, memory, threads, keyLen uint32) []byte {
lanes := memory / threads
for lane := uint32(0); lane < threads-1; lane++ {
for i, v := range B[(lane*lanes)+lanes-1] {
B[memory-1][i] ^= v
}
}
var block [1024]byte
for i, v := range B[memory-1] {
binary.LittleEndian.PutUint64(block[i*8:], v)
}
key := make([]byte, keyLen)
blake2bHash(key, block[:])
return key
}
func indexAlpha(rand uint64, lanes, segments, threads, n, slice, lane, index uint32) uint32 {
refLane := uint32(rand>>32) % threads
if n == 0 && slice == 0 {
refLane = lane
}
m, s := 3*segments, ((slice+1)%syncPoints)*segments
if lane == refLane {
m += index
}
if n == 0 {
m, s = slice*segments, 0
if slice == 0 || lane == refLane {
m += index
}
}
if index == 0 || lane == refLane {
m--
}
return phi(rand, uint64(m), uint64(s), refLane, lanes)
}
func phi(rand, m, s uint64, lane, lanes uint32) uint32 {
p := rand & 0xFFFFFFFF
p = (p * p) >> 32
p = (p * m) >> 32
return lane*lanes + uint32((s+m-(p+1))%uint64(lanes))
}

View File

@@ -1,53 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package argon2
import (
"encoding/binary"
"hash"
"golang.org/x/crypto/blake2b"
)
// blake2bHash computes an arbitrary long hash value of in
// and writes the hash to out.
func blake2bHash(out []byte, in []byte) {
var b2 hash.Hash
if n := len(out); n < blake2b.Size {
b2, _ = blake2b.New(n, nil)
} else {
b2, _ = blake2b.New512(nil)
}
var buffer [blake2b.Size]byte
binary.LittleEndian.PutUint32(buffer[:4], uint32(len(out)))
b2.Write(buffer[:4])
b2.Write(in)
if len(out) <= blake2b.Size {
b2.Sum(out[:0])
return
}
outLen := len(out)
b2.Sum(buffer[:0])
b2.Reset()
copy(out, buffer[:32])
out = out[32:]
for len(out) > blake2b.Size {
b2.Write(buffer[:])
b2.Sum(buffer[:0])
copy(out, buffer[:32])
out = out[32:]
b2.Reset()
}
if outLen%blake2b.Size > 0 { // outLen > 64
r := ((outLen + 31) / 32) - 2 // ⌈τ /32⌉-2
b2, _ = blake2b.New(outLen-32*r, nil)
}
b2.Write(buffer[:])
b2.Sum(out[:0])
}

View File

@@ -1,60 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build amd64 && gc && !purego
package argon2
import "golang.org/x/sys/cpu"
func init() {
useSSE4 = cpu.X86.HasSSE41
}
//go:noescape
func mixBlocksSSE2(out, a, b, c *block)
//go:noescape
func xorBlocksSSE2(out, a, b, c *block)
//go:noescape
func blamkaSSE4(b *block)
func processBlockSSE(out, in1, in2 *block, xor bool) {
var t block
mixBlocksSSE2(&t, in1, in2, &t)
if useSSE4 {
blamkaSSE4(&t)
} else {
for i := 0; i < blockLength; i += 16 {
blamkaGeneric(
&t[i+0], &t[i+1], &t[i+2], &t[i+3],
&t[i+4], &t[i+5], &t[i+6], &t[i+7],
&t[i+8], &t[i+9], &t[i+10], &t[i+11],
&t[i+12], &t[i+13], &t[i+14], &t[i+15],
)
}
for i := 0; i < blockLength/8; i += 2 {
blamkaGeneric(
&t[i], &t[i+1], &t[16+i], &t[16+i+1],
&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
)
}
}
if xor {
xorBlocksSSE2(out, in1, in2, &t)
} else {
mixBlocksSSE2(out, in1, in2, &t)
}
}
func processBlock(out, in1, in2 *block) {
processBlockSSE(out, in1, in2, false)
}
func processBlockXOR(out, in1, in2 *block) {
processBlockSSE(out, in1, in2, true)
}

View File

@@ -1,243 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build amd64 && gc && !purego
#include "textflag.h"
DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v6, t1; \
PUNPCKLQDQ v6, t2; \
PUNPCKHQDQ v7, v6; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ v7, t2; \
MOVO t1, v7; \
MOVO v2, t1; \
PUNPCKHQDQ t2, v7; \
PUNPCKLQDQ v3, t2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v3
#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v2, t1; \
PUNPCKLQDQ v2, t2; \
PUNPCKHQDQ v3, v2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ v3, t2; \
MOVO t1, v3; \
MOVO v6, t1; \
PUNPCKHQDQ t2, v3; \
PUNPCKLQDQ v7, t2; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v7
#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, t0, c40, c48) \
MOVO v0, t0; \
PMULULQ v2, t0; \
PADDQ v2, v0; \
PADDQ t0, v0; \
PADDQ t0, v0; \
PXOR v0, v6; \
PSHUFD $0xB1, v6, v6; \
MOVO v4, t0; \
PMULULQ v6, t0; \
PADDQ v6, v4; \
PADDQ t0, v4; \
PADDQ t0, v4; \
PXOR v4, v2; \
PSHUFB c40, v2; \
MOVO v0, t0; \
PMULULQ v2, t0; \
PADDQ v2, v0; \
PADDQ t0, v0; \
PADDQ t0, v0; \
PXOR v0, v6; \
PSHUFB c48, v6; \
MOVO v4, t0; \
PMULULQ v6, t0; \
PADDQ v6, v4; \
PADDQ t0, v4; \
PADDQ t0, v4; \
PXOR v4, v2; \
MOVO v2, t0; \
PADDQ v2, t0; \
PSRLQ $63, v2; \
PXOR t0, v2; \
MOVO v1, t0; \
PMULULQ v3, t0; \
PADDQ v3, v1; \
PADDQ t0, v1; \
PADDQ t0, v1; \
PXOR v1, v7; \
PSHUFD $0xB1, v7, v7; \
MOVO v5, t0; \
PMULULQ v7, t0; \
PADDQ v7, v5; \
PADDQ t0, v5; \
PADDQ t0, v5; \
PXOR v5, v3; \
PSHUFB c40, v3; \
MOVO v1, t0; \
PMULULQ v3, t0; \
PADDQ v3, v1; \
PADDQ t0, v1; \
PADDQ t0, v1; \
PXOR v1, v7; \
PSHUFB c48, v7; \
MOVO v5, t0; \
PMULULQ v7, t0; \
PADDQ v7, v5; \
PADDQ t0, v5; \
PADDQ t0, v5; \
PXOR v5, v3; \
MOVO v3, t0; \
PADDQ v3, t0; \
PSRLQ $63, v3; \
PXOR t0, v3
#define LOAD_MSG_0(block, off) \
MOVOU 8*(off+0)(block), X0; \
MOVOU 8*(off+2)(block), X1; \
MOVOU 8*(off+4)(block), X2; \
MOVOU 8*(off+6)(block), X3; \
MOVOU 8*(off+8)(block), X4; \
MOVOU 8*(off+10)(block), X5; \
MOVOU 8*(off+12)(block), X6; \
MOVOU 8*(off+14)(block), X7
#define STORE_MSG_0(block, off) \
MOVOU X0, 8*(off+0)(block); \
MOVOU X1, 8*(off+2)(block); \
MOVOU X2, 8*(off+4)(block); \
MOVOU X3, 8*(off+6)(block); \
MOVOU X4, 8*(off+8)(block); \
MOVOU X5, 8*(off+10)(block); \
MOVOU X6, 8*(off+12)(block); \
MOVOU X7, 8*(off+14)(block)
#define LOAD_MSG_1(block, off) \
MOVOU 8*off+0*8(block), X0; \
MOVOU 8*off+16*8(block), X1; \
MOVOU 8*off+32*8(block), X2; \
MOVOU 8*off+48*8(block), X3; \
MOVOU 8*off+64*8(block), X4; \
MOVOU 8*off+80*8(block), X5; \
MOVOU 8*off+96*8(block), X6; \
MOVOU 8*off+112*8(block), X7
#define STORE_MSG_1(block, off) \
MOVOU X0, 8*off+0*8(block); \
MOVOU X1, 8*off+16*8(block); \
MOVOU X2, 8*off+32*8(block); \
MOVOU X3, 8*off+48*8(block); \
MOVOU X4, 8*off+64*8(block); \
MOVOU X5, 8*off+80*8(block); \
MOVOU X6, 8*off+96*8(block); \
MOVOU X7, 8*off+112*8(block)
#define BLAMKA_ROUND_0(block, off, t0, t1, c40, c48) \
LOAD_MSG_0(block, off); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1); \
STORE_MSG_0(block, off)
#define BLAMKA_ROUND_1(block, off, t0, t1, c40, c48) \
LOAD_MSG_1(block, off); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1); \
STORE_MSG_1(block, off)
// func blamkaSSE4(b *block)
TEXT ·blamkaSSE4(SB), 4, $0-8
MOVQ b+0(FP), AX
MOVOU ·c40<>(SB), X10
MOVOU ·c48<>(SB), X11
BLAMKA_ROUND_0(AX, 0, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 16, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 32, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 48, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 64, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 80, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 96, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 112, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 0, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 2, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 4, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 6, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 8, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 10, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 12, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 14, X8, X9, X10, X11)
RET
// func mixBlocksSSE2(out, a, b, c *block)
TEXT ·mixBlocksSSE2(SB), 4, $0-32
MOVQ out+0(FP), DX
MOVQ a+8(FP), AX
MOVQ b+16(FP), BX
MOVQ c+24(FP), CX
MOVQ $128, DI
loop:
MOVOU 0(AX), X0
MOVOU 0(BX), X1
MOVOU 0(CX), X2
PXOR X1, X0
PXOR X2, X0
MOVOU X0, 0(DX)
ADDQ $16, AX
ADDQ $16, BX
ADDQ $16, CX
ADDQ $16, DX
SUBQ $2, DI
JA loop
RET
// func xorBlocksSSE2(out, a, b, c *block)
TEXT ·xorBlocksSSE2(SB), 4, $0-32
MOVQ out+0(FP), DX
MOVQ a+8(FP), AX
MOVQ b+16(FP), BX
MOVQ c+24(FP), CX
MOVQ $128, DI
loop:
MOVOU 0(AX), X0
MOVOU 0(BX), X1
MOVOU 0(CX), X2
MOVOU 0(DX), X3
PXOR X1, X0
PXOR X2, X0
PXOR X3, X0
MOVOU X0, 0(DX)
ADDQ $16, AX
ADDQ $16, BX
ADDQ $16, CX
ADDQ $16, DX
SUBQ $2, DI
JA loop
RET

View File

@@ -1,163 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package argon2
var useSSE4 bool
func processBlockGeneric(out, in1, in2 *block, xor bool) {
var t block
for i := range t {
t[i] = in1[i] ^ in2[i]
}
for i := 0; i < blockLength; i += 16 {
blamkaGeneric(
&t[i+0], &t[i+1], &t[i+2], &t[i+3],
&t[i+4], &t[i+5], &t[i+6], &t[i+7],
&t[i+8], &t[i+9], &t[i+10], &t[i+11],
&t[i+12], &t[i+13], &t[i+14], &t[i+15],
)
}
for i := 0; i < blockLength/8; i += 2 {
blamkaGeneric(
&t[i], &t[i+1], &t[16+i], &t[16+i+1],
&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
)
}
if xor {
for i := range t {
out[i] ^= in1[i] ^ in2[i] ^ t[i]
}
} else {
for i := range t {
out[i] = in1[i] ^ in2[i] ^ t[i]
}
}
}
func blamkaGeneric(t00, t01, t02, t03, t04, t05, t06, t07, t08, t09, t10, t11, t12, t13, t14, t15 *uint64) {
v00, v01, v02, v03 := *t00, *t01, *t02, *t03
v04, v05, v06, v07 := *t04, *t05, *t06, *t07
v08, v09, v10, v11 := *t08, *t09, *t10, *t11
v12, v13, v14, v15 := *t12, *t13, *t14, *t15
v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
v12 ^= v00
v12 = v12>>32 | v12<<32
v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
v04 ^= v08
v04 = v04>>24 | v04<<40
v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
v12 ^= v00
v12 = v12>>16 | v12<<48
v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
v04 ^= v08
v04 = v04>>63 | v04<<1
v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
v13 ^= v01
v13 = v13>>32 | v13<<32
v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
v05 ^= v09
v05 = v05>>24 | v05<<40
v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
v13 ^= v01
v13 = v13>>16 | v13<<48
v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
v05 ^= v09
v05 = v05>>63 | v05<<1
v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
v14 ^= v02
v14 = v14>>32 | v14<<32
v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
v06 ^= v10
v06 = v06>>24 | v06<<40
v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
v14 ^= v02
v14 = v14>>16 | v14<<48
v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
v06 ^= v10
v06 = v06>>63 | v06<<1
v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
v15 ^= v03
v15 = v15>>32 | v15<<32
v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
v07 ^= v11
v07 = v07>>24 | v07<<40
v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
v15 ^= v03
v15 = v15>>16 | v15<<48
v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
v07 ^= v11
v07 = v07>>63 | v07<<1
v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
v15 ^= v00
v15 = v15>>32 | v15<<32
v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
v05 ^= v10
v05 = v05>>24 | v05<<40
v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
v15 ^= v00
v15 = v15>>16 | v15<<48
v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
v05 ^= v10
v05 = v05>>63 | v05<<1
v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
v12 ^= v01
v12 = v12>>32 | v12<<32
v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
v06 ^= v11
v06 = v06>>24 | v06<<40
v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
v12 ^= v01
v12 = v12>>16 | v12<<48
v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
v06 ^= v11
v06 = v06>>63 | v06<<1
v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
v13 ^= v02
v13 = v13>>32 | v13<<32
v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
v07 ^= v08
v07 = v07>>24 | v07<<40
v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
v13 ^= v02
v13 = v13>>16 | v13<<48
v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
v07 ^= v08
v07 = v07>>63 | v07<<1
v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
v14 ^= v03
v14 = v14>>32 | v14<<32
v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
v04 ^= v09
v04 = v04>>24 | v04<<40
v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
v14 ^= v03
v14 = v14>>16 | v14<<48
v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
v04 ^= v09
v04 = v04>>63 | v04<<1
*t00, *t01, *t02, *t03 = v00, v01, v02, v03
*t04, *t05, *t06, *t07 = v04, v05, v06, v07
*t08, *t09, *t10, *t11 = v08, v09, v10, v11
*t12, *t13, *t14, *t15 = v12, v13, v14, v15
}

View File

@@ -1,15 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build !amd64 || purego || !gc
package argon2
func processBlock(out, in1, in2 *block) {
processBlockGeneric(out, in1, in2, false)
}
func processBlockXOR(out, in1, in2 *block) {
processBlockGeneric(out, in1, in2, true)
}

View File

@@ -1,291 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package blake2b implements the BLAKE2b hash algorithm defined by RFC 7693
// and the extendable output function (XOF) BLAKE2Xb.
//
// BLAKE2b is optimized for 64-bit platforms—including NEON-enabled ARMs—and
// produces digests of any size between 1 and 64 bytes.
// For a detailed specification of BLAKE2b see https://blake2.net/blake2.pdf
// and for BLAKE2Xb see https://blake2.net/blake2x.pdf
//
// If you aren't sure which function you need, use BLAKE2b (Sum512 or New512).
// If you need a secret-key MAC (message authentication code), use the New512
// function with a non-nil key.
//
// BLAKE2X is a construction to compute hash values larger than 64 bytes. It
// can produce hash values between 0 and 4 GiB.
package blake2b
import (
"encoding/binary"
"errors"
"hash"
)
const (
// The blocksize of BLAKE2b in bytes.
BlockSize = 128
// The hash size of BLAKE2b-512 in bytes.
Size = 64
// The hash size of BLAKE2b-384 in bytes.
Size384 = 48
// The hash size of BLAKE2b-256 in bytes.
Size256 = 32
)
var (
useAVX2 bool
useAVX bool
useSSE4 bool
)
var (
errKeySize = errors.New("blake2b: invalid key size")
errHashSize = errors.New("blake2b: invalid hash size")
)
var iv = [8]uint64{
0x6a09e667f3bcc908, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
0x510e527fade682d1, 0x9b05688c2b3e6c1f, 0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
}
// Sum512 returns the BLAKE2b-512 checksum of the data.
func Sum512(data []byte) [Size]byte {
var sum [Size]byte
checkSum(&sum, Size, data)
return sum
}
// Sum384 returns the BLAKE2b-384 checksum of the data.
func Sum384(data []byte) [Size384]byte {
var sum [Size]byte
var sum384 [Size384]byte
checkSum(&sum, Size384, data)
copy(sum384[:], sum[:Size384])
return sum384
}
// Sum256 returns the BLAKE2b-256 checksum of the data.
func Sum256(data []byte) [Size256]byte {
var sum [Size]byte
var sum256 [Size256]byte
checkSum(&sum, Size256, data)
copy(sum256[:], sum[:Size256])
return sum256
}
// New512 returns a new hash.Hash computing the BLAKE2b-512 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New512(key []byte) (hash.Hash, error) { return newDigest(Size, key) }
// New384 returns a new hash.Hash computing the BLAKE2b-384 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New384(key []byte) (hash.Hash, error) { return newDigest(Size384, key) }
// New256 returns a new hash.Hash computing the BLAKE2b-256 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New256(key []byte) (hash.Hash, error) { return newDigest(Size256, key) }
// New returns a new hash.Hash computing the BLAKE2b checksum with a custom length.
// A non-nil key turns the hash into a MAC. The key must be between zero and 64 bytes long.
// The hash size can be a value between 1 and 64 but it is highly recommended to use
// values equal or greater than:
// - 32 if BLAKE2b is used as a hash function (The key is zero bytes long).
// - 16 if BLAKE2b is used as a MAC function (The key is at least 16 bytes long).
// When the key is nil, the returned hash.Hash implements BinaryMarshaler
// and BinaryUnmarshaler for state (de)serialization as documented by hash.Hash.
func New(size int, key []byte) (hash.Hash, error) { return newDigest(size, key) }
func newDigest(hashSize int, key []byte) (*digest, error) {
if hashSize < 1 || hashSize > Size {
return nil, errHashSize
}
if len(key) > Size {
return nil, errKeySize
}
d := &digest{
size: hashSize,
keyLen: len(key),
}
copy(d.key[:], key)
d.Reset()
return d, nil
}
func checkSum(sum *[Size]byte, hashSize int, data []byte) {
h := iv
h[0] ^= uint64(hashSize) | (1 << 16) | (1 << 24)
var c [2]uint64
if length := len(data); length > BlockSize {
n := length &^ (BlockSize - 1)
if length == n {
n -= BlockSize
}
hashBlocks(&h, &c, 0, data[:n])
data = data[n:]
}
var block [BlockSize]byte
offset := copy(block[:], data)
remaining := uint64(BlockSize - offset)
if c[0] < remaining {
c[1]--
}
c[0] -= remaining
hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
for i, v := range h[:(hashSize+7)/8] {
binary.LittleEndian.PutUint64(sum[8*i:], v)
}
}
type digest struct {
h [8]uint64
c [2]uint64
size int
block [BlockSize]byte
offset int
key [BlockSize]byte
keyLen int
}
const (
magic = "b2b"
marshaledSize = len(magic) + 8*8 + 2*8 + 1 + BlockSize + 1
)
func (d *digest) MarshalBinary() ([]byte, error) {
if d.keyLen != 0 {
return nil, errors.New("crypto/blake2b: cannot marshal MACs")
}
b := make([]byte, 0, marshaledSize)
b = append(b, magic...)
for i := 0; i < 8; i++ {
b = appendUint64(b, d.h[i])
}
b = appendUint64(b, d.c[0])
b = appendUint64(b, d.c[1])
// Maximum value for size is 64
b = append(b, byte(d.size))
b = append(b, d.block[:]...)
b = append(b, byte(d.offset))
return b, nil
}
func (d *digest) UnmarshalBinary(b []byte) error {
if len(b) < len(magic) || string(b[:len(magic)]) != magic {
return errors.New("crypto/blake2b: invalid hash state identifier")
}
if len(b) != marshaledSize {
return errors.New("crypto/blake2b: invalid hash state size")
}
b = b[len(magic):]
for i := 0; i < 8; i++ {
b, d.h[i] = consumeUint64(b)
}
b, d.c[0] = consumeUint64(b)
b, d.c[1] = consumeUint64(b)
d.size = int(b[0])
b = b[1:]
copy(d.block[:], b[:BlockSize])
b = b[BlockSize:]
d.offset = int(b[0])
return nil
}
func (d *digest) BlockSize() int { return BlockSize }
func (d *digest) Size() int { return d.size }
func (d *digest) Reset() {
d.h = iv
d.h[0] ^= uint64(d.size) | (uint64(d.keyLen) << 8) | (1 << 16) | (1 << 24)
d.offset, d.c[0], d.c[1] = 0, 0, 0
if d.keyLen > 0 {
d.block = d.key
d.offset = BlockSize
}
}
func (d *digest) Write(p []byte) (n int, err error) {
n = len(p)
if d.offset > 0 {
remaining := BlockSize - d.offset
if n <= remaining {
d.offset += copy(d.block[d.offset:], p)
return
}
copy(d.block[d.offset:], p[:remaining])
hashBlocks(&d.h, &d.c, 0, d.block[:])
d.offset = 0
p = p[remaining:]
}
if length := len(p); length > BlockSize {
nn := length &^ (BlockSize - 1)
if length == nn {
nn -= BlockSize
}
hashBlocks(&d.h, &d.c, 0, p[:nn])
p = p[nn:]
}
if len(p) > 0 {
d.offset += copy(d.block[:], p)
}
return
}
func (d *digest) Sum(sum []byte) []byte {
var hash [Size]byte
d.finalize(&hash)
return append(sum, hash[:d.size]...)
}
func (d *digest) finalize(hash *[Size]byte) {
var block [BlockSize]byte
copy(block[:], d.block[:d.offset])
remaining := uint64(BlockSize - d.offset)
c := d.c
if c[0] < remaining {
c[1]--
}
c[0] -= remaining
h := d.h
hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
for i, v := range h {
binary.LittleEndian.PutUint64(hash[8*i:], v)
}
}
func appendUint64(b []byte, x uint64) []byte {
var a [8]byte
binary.BigEndian.PutUint64(a[:], x)
return append(b, a[:]...)
}
func appendUint32(b []byte, x uint32) []byte {
var a [4]byte
binary.BigEndian.PutUint32(a[:], x)
return append(b, a[:]...)
}
func consumeUint64(b []byte) ([]byte, uint64) {
x := binary.BigEndian.Uint64(b)
return b[8:], x
}
func consumeUint32(b []byte) ([]byte, uint32) {
x := binary.BigEndian.Uint32(b)
return b[4:], x
}

View File

@@ -1,37 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build amd64 && gc && !purego
package blake2b
import "golang.org/x/sys/cpu"
func init() {
useAVX2 = cpu.X86.HasAVX2
useAVX = cpu.X86.HasAVX
useSSE4 = cpu.X86.HasSSE41
}
//go:noescape
func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
//go:noescape
func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
//go:noescape
func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
switch {
case useAVX2:
hashBlocksAVX2(h, c, flag, blocks)
case useAVX:
hashBlocksAVX(h, c, flag, blocks)
case useSSE4:
hashBlocksSSE4(h, c, flag, blocks)
default:
hashBlocksGeneric(h, c, flag, blocks)
}
}

View File

@@ -1,744 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build amd64 && gc && !purego
#include "textflag.h"
DATA ·AVX2_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
DATA ·AVX2_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
DATA ·AVX2_iv0<>+0x10(SB)/8, $0x3c6ef372fe94f82b
DATA ·AVX2_iv0<>+0x18(SB)/8, $0xa54ff53a5f1d36f1
GLOBL ·AVX2_iv0<>(SB), (NOPTR+RODATA), $32
DATA ·AVX2_iv1<>+0x00(SB)/8, $0x510e527fade682d1
DATA ·AVX2_iv1<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
DATA ·AVX2_iv1<>+0x10(SB)/8, $0x1f83d9abfb41bd6b
DATA ·AVX2_iv1<>+0x18(SB)/8, $0x5be0cd19137e2179
GLOBL ·AVX2_iv1<>(SB), (NOPTR+RODATA), $32
DATA ·AVX2_c40<>+0x00(SB)/8, $0x0201000706050403
DATA ·AVX2_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
DATA ·AVX2_c40<>+0x10(SB)/8, $0x0201000706050403
DATA ·AVX2_c40<>+0x18(SB)/8, $0x0a09080f0e0d0c0b
GLOBL ·AVX2_c40<>(SB), (NOPTR+RODATA), $32
DATA ·AVX2_c48<>+0x00(SB)/8, $0x0100070605040302
DATA ·AVX2_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
DATA ·AVX2_c48<>+0x10(SB)/8, $0x0100070605040302
DATA ·AVX2_c48<>+0x18(SB)/8, $0x09080f0e0d0c0b0a
GLOBL ·AVX2_c48<>(SB), (NOPTR+RODATA), $32
DATA ·AVX_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
DATA ·AVX_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
GLOBL ·AVX_iv0<>(SB), (NOPTR+RODATA), $16
DATA ·AVX_iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
DATA ·AVX_iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
GLOBL ·AVX_iv1<>(SB), (NOPTR+RODATA), $16
DATA ·AVX_iv2<>+0x00(SB)/8, $0x510e527fade682d1
DATA ·AVX_iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
GLOBL ·AVX_iv2<>(SB), (NOPTR+RODATA), $16
DATA ·AVX_iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
DATA ·AVX_iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
GLOBL ·AVX_iv3<>(SB), (NOPTR+RODATA), $16
DATA ·AVX_c40<>+0x00(SB)/8, $0x0201000706050403
DATA ·AVX_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
GLOBL ·AVX_c40<>(SB), (NOPTR+RODATA), $16
DATA ·AVX_c48<>+0x00(SB)/8, $0x0100070605040302
DATA ·AVX_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
GLOBL ·AVX_c48<>(SB), (NOPTR+RODATA), $16
#define VPERMQ_0x39_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x39
#define VPERMQ_0x93_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x93
#define VPERMQ_0x4E_Y2_Y2 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2; BYTE $0x4e
#define VPERMQ_0x93_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x93
#define VPERMQ_0x39_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x39
#define ROUND_AVX2(m0, m1, m2, m3, t, c40, c48) \
VPADDQ m0, Y0, Y0; \
VPADDQ Y1, Y0, Y0; \
VPXOR Y0, Y3, Y3; \
VPSHUFD $-79, Y3, Y3; \
VPADDQ Y3, Y2, Y2; \
VPXOR Y2, Y1, Y1; \
VPSHUFB c40, Y1, Y1; \
VPADDQ m1, Y0, Y0; \
VPADDQ Y1, Y0, Y0; \
VPXOR Y0, Y3, Y3; \
VPSHUFB c48, Y3, Y3; \
VPADDQ Y3, Y2, Y2; \
VPXOR Y2, Y1, Y1; \
VPADDQ Y1, Y1, t; \
VPSRLQ $63, Y1, Y1; \
VPXOR t, Y1, Y1; \
VPERMQ_0x39_Y1_Y1; \
VPERMQ_0x4E_Y2_Y2; \
VPERMQ_0x93_Y3_Y3; \
VPADDQ m2, Y0, Y0; \
VPADDQ Y1, Y0, Y0; \
VPXOR Y0, Y3, Y3; \
VPSHUFD $-79, Y3, Y3; \
VPADDQ Y3, Y2, Y2; \
VPXOR Y2, Y1, Y1; \
VPSHUFB c40, Y1, Y1; \
VPADDQ m3, Y0, Y0; \
VPADDQ Y1, Y0, Y0; \
VPXOR Y0, Y3, Y3; \
VPSHUFB c48, Y3, Y3; \
VPADDQ Y3, Y2, Y2; \
VPXOR Y2, Y1, Y1; \
VPADDQ Y1, Y1, t; \
VPSRLQ $63, Y1, Y1; \
VPXOR t, Y1, Y1; \
VPERMQ_0x39_Y3_Y3; \
VPERMQ_0x4E_Y2_Y2; \
VPERMQ_0x93_Y1_Y1
#define VMOVQ_SI_X11_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x1E
#define VMOVQ_SI_X12_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x26
#define VMOVQ_SI_X13_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x2E
#define VMOVQ_SI_X14_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x36
#define VMOVQ_SI_X15_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x3E
#define VMOVQ_SI_X11(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x5E; BYTE $n
#define VMOVQ_SI_X12(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x66; BYTE $n
#define VMOVQ_SI_X13(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x6E; BYTE $n
#define VMOVQ_SI_X14(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x76; BYTE $n
#define VMOVQ_SI_X15(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x7E; BYTE $n
#define VPINSRQ_1_SI_X11_0 BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x1E; BYTE $0x01
#define VPINSRQ_1_SI_X12_0 BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x26; BYTE $0x01
#define VPINSRQ_1_SI_X13_0 BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x2E; BYTE $0x01
#define VPINSRQ_1_SI_X14_0 BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x36; BYTE $0x01
#define VPINSRQ_1_SI_X15_0 BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x3E; BYTE $0x01
#define VPINSRQ_1_SI_X11(n) BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x5E; BYTE $n; BYTE $0x01
#define VPINSRQ_1_SI_X12(n) BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x66; BYTE $n; BYTE $0x01
#define VPINSRQ_1_SI_X13(n) BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x6E; BYTE $n; BYTE $0x01
#define VPINSRQ_1_SI_X14(n) BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x76; BYTE $n; BYTE $0x01
#define VPINSRQ_1_SI_X15(n) BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x7E; BYTE $n; BYTE $0x01
#define VMOVQ_R8_X15 BYTE $0xC4; BYTE $0x41; BYTE $0xF9; BYTE $0x6E; BYTE $0xF8
#define VPINSRQ_1_R9_X15 BYTE $0xC4; BYTE $0x43; BYTE $0x81; BYTE $0x22; BYTE $0xF9; BYTE $0x01
// load msg: Y12 = (i0, i1, i2, i3)
// i0, i1, i2, i3 must not be 0
#define LOAD_MSG_AVX2_Y12(i0, i1, i2, i3) \
VMOVQ_SI_X12(i0*8); \
VMOVQ_SI_X11(i2*8); \
VPINSRQ_1_SI_X12(i1*8); \
VPINSRQ_1_SI_X11(i3*8); \
VINSERTI128 $1, X11, Y12, Y12
// load msg: Y13 = (i0, i1, i2, i3)
// i0, i1, i2, i3 must not be 0
#define LOAD_MSG_AVX2_Y13(i0, i1, i2, i3) \
VMOVQ_SI_X13(i0*8); \
VMOVQ_SI_X11(i2*8); \
VPINSRQ_1_SI_X13(i1*8); \
VPINSRQ_1_SI_X11(i3*8); \
VINSERTI128 $1, X11, Y13, Y13
// load msg: Y14 = (i0, i1, i2, i3)
// i0, i1, i2, i3 must not be 0
#define LOAD_MSG_AVX2_Y14(i0, i1, i2, i3) \
VMOVQ_SI_X14(i0*8); \
VMOVQ_SI_X11(i2*8); \
VPINSRQ_1_SI_X14(i1*8); \
VPINSRQ_1_SI_X11(i3*8); \
VINSERTI128 $1, X11, Y14, Y14
// load msg: Y15 = (i0, i1, i2, i3)
// i0, i1, i2, i3 must not be 0
#define LOAD_MSG_AVX2_Y15(i0, i1, i2, i3) \
VMOVQ_SI_X15(i0*8); \
VMOVQ_SI_X11(i2*8); \
VPINSRQ_1_SI_X15(i1*8); \
VPINSRQ_1_SI_X11(i3*8); \
VINSERTI128 $1, X11, Y15, Y15
#define LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15() \
VMOVQ_SI_X12_0; \
VMOVQ_SI_X11(4*8); \
VPINSRQ_1_SI_X12(2*8); \
VPINSRQ_1_SI_X11(6*8); \
VINSERTI128 $1, X11, Y12, Y12; \
LOAD_MSG_AVX2_Y13(1, 3, 5, 7); \
LOAD_MSG_AVX2_Y14(8, 10, 12, 14); \
LOAD_MSG_AVX2_Y15(9, 11, 13, 15)
#define LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3() \
LOAD_MSG_AVX2_Y12(14, 4, 9, 13); \
LOAD_MSG_AVX2_Y13(10, 8, 15, 6); \
VMOVQ_SI_X11(11*8); \
VPSHUFD $0x4E, 0*8(SI), X14; \
VPINSRQ_1_SI_X11(5*8); \
VINSERTI128 $1, X11, Y14, Y14; \
LOAD_MSG_AVX2_Y15(12, 2, 7, 3)
#define LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4() \
VMOVQ_SI_X11(5*8); \
VMOVDQU 11*8(SI), X12; \
VPINSRQ_1_SI_X11(15*8); \
VINSERTI128 $1, X11, Y12, Y12; \
VMOVQ_SI_X13(8*8); \
VMOVQ_SI_X11(2*8); \
VPINSRQ_1_SI_X13_0; \
VPINSRQ_1_SI_X11(13*8); \
VINSERTI128 $1, X11, Y13, Y13; \
LOAD_MSG_AVX2_Y14(10, 3, 7, 9); \
LOAD_MSG_AVX2_Y15(14, 6, 1, 4)
#define LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8() \
LOAD_MSG_AVX2_Y12(7, 3, 13, 11); \
LOAD_MSG_AVX2_Y13(9, 1, 12, 14); \
LOAD_MSG_AVX2_Y14(2, 5, 4, 15); \
VMOVQ_SI_X15(6*8); \
VMOVQ_SI_X11_0; \
VPINSRQ_1_SI_X15(10*8); \
VPINSRQ_1_SI_X11(8*8); \
VINSERTI128 $1, X11, Y15, Y15
#define LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13() \
LOAD_MSG_AVX2_Y12(9, 5, 2, 10); \
VMOVQ_SI_X13_0; \
VMOVQ_SI_X11(4*8); \
VPINSRQ_1_SI_X13(7*8); \
VPINSRQ_1_SI_X11(15*8); \
VINSERTI128 $1, X11, Y13, Y13; \
LOAD_MSG_AVX2_Y14(14, 11, 6, 3); \
LOAD_MSG_AVX2_Y15(1, 12, 8, 13)
#define LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9() \
VMOVQ_SI_X12(2*8); \
VMOVQ_SI_X11_0; \
VPINSRQ_1_SI_X12(6*8); \
VPINSRQ_1_SI_X11(8*8); \
VINSERTI128 $1, X11, Y12, Y12; \
LOAD_MSG_AVX2_Y13(12, 10, 11, 3); \
LOAD_MSG_AVX2_Y14(4, 7, 15, 1); \
LOAD_MSG_AVX2_Y15(13, 5, 14, 9)
#define LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11() \
LOAD_MSG_AVX2_Y12(12, 1, 14, 4); \
LOAD_MSG_AVX2_Y13(5, 15, 13, 10); \
VMOVQ_SI_X14_0; \
VPSHUFD $0x4E, 8*8(SI), X11; \
VPINSRQ_1_SI_X14(6*8); \
VINSERTI128 $1, X11, Y14, Y14; \
LOAD_MSG_AVX2_Y15(7, 3, 2, 11)
#define LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10() \
LOAD_MSG_AVX2_Y12(13, 7, 12, 3); \
LOAD_MSG_AVX2_Y13(11, 14, 1, 9); \
LOAD_MSG_AVX2_Y14(5, 15, 8, 2); \
VMOVQ_SI_X15_0; \
VMOVQ_SI_X11(6*8); \
VPINSRQ_1_SI_X15(4*8); \
VPINSRQ_1_SI_X11(10*8); \
VINSERTI128 $1, X11, Y15, Y15
#define LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5() \
VMOVQ_SI_X12(6*8); \
VMOVQ_SI_X11(11*8); \
VPINSRQ_1_SI_X12(14*8); \
VPINSRQ_1_SI_X11_0; \
VINSERTI128 $1, X11, Y12, Y12; \
LOAD_MSG_AVX2_Y13(15, 9, 3, 8); \
VMOVQ_SI_X11(1*8); \
VMOVDQU 12*8(SI), X14; \
VPINSRQ_1_SI_X11(10*8); \
VINSERTI128 $1, X11, Y14, Y14; \
VMOVQ_SI_X15(2*8); \
VMOVDQU 4*8(SI), X11; \
VPINSRQ_1_SI_X15(7*8); \
VINSERTI128 $1, X11, Y15, Y15
#define LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0() \
LOAD_MSG_AVX2_Y12(10, 8, 7, 1); \
VMOVQ_SI_X13(2*8); \
VPSHUFD $0x4E, 5*8(SI), X11; \
VPINSRQ_1_SI_X13(4*8); \
VINSERTI128 $1, X11, Y13, Y13; \
LOAD_MSG_AVX2_Y14(15, 9, 3, 13); \
VMOVQ_SI_X15(11*8); \
VMOVQ_SI_X11(12*8); \
VPINSRQ_1_SI_X15(14*8); \
VPINSRQ_1_SI_X11_0; \
VINSERTI128 $1, X11, Y15, Y15
// func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
TEXT ·hashBlocksAVX2(SB), 4, $320-48 // frame size = 288 + 32 byte alignment
MOVQ h+0(FP), AX
MOVQ c+8(FP), BX
MOVQ flag+16(FP), CX
MOVQ blocks_base+24(FP), SI
MOVQ blocks_len+32(FP), DI
MOVQ SP, DX
ADDQ $31, DX
ANDQ $~31, DX
MOVQ CX, 16(DX)
XORQ CX, CX
MOVQ CX, 24(DX)
VMOVDQU ·AVX2_c40<>(SB), Y4
VMOVDQU ·AVX2_c48<>(SB), Y5
VMOVDQU 0(AX), Y8
VMOVDQU 32(AX), Y9
VMOVDQU ·AVX2_iv0<>(SB), Y6
VMOVDQU ·AVX2_iv1<>(SB), Y7
MOVQ 0(BX), R8
MOVQ 8(BX), R9
MOVQ R9, 8(DX)
loop:
ADDQ $128, R8
MOVQ R8, 0(DX)
CMPQ R8, $128
JGE noinc
INCQ R9
MOVQ R9, 8(DX)
noinc:
VMOVDQA Y8, Y0
VMOVDQA Y9, Y1
VMOVDQA Y6, Y2
VPXOR 0(DX), Y7, Y3
LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15()
VMOVDQA Y12, 32(DX)
VMOVDQA Y13, 64(DX)
VMOVDQA Y14, 96(DX)
VMOVDQA Y15, 128(DX)
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3()
VMOVDQA Y12, 160(DX)
VMOVDQA Y13, 192(DX)
VMOVDQA Y14, 224(DX)
VMOVDQA Y15, 256(DX)
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0()
ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
ROUND_AVX2(32(DX), 64(DX), 96(DX), 128(DX), Y10, Y4, Y5)
ROUND_AVX2(160(DX), 192(DX), 224(DX), 256(DX), Y10, Y4, Y5)
VPXOR Y0, Y8, Y8
VPXOR Y1, Y9, Y9
VPXOR Y2, Y8, Y8
VPXOR Y3, Y9, Y9
LEAQ 128(SI), SI
SUBQ $128, DI
JNE loop
MOVQ R8, 0(BX)
MOVQ R9, 8(BX)
VMOVDQU Y8, 0(AX)
VMOVDQU Y9, 32(AX)
VZEROUPPER
RET
#define VPUNPCKLQDQ_X2_X2_X15 BYTE $0xC5; BYTE $0x69; BYTE $0x6C; BYTE $0xFA
#define VPUNPCKLQDQ_X3_X3_X15 BYTE $0xC5; BYTE $0x61; BYTE $0x6C; BYTE $0xFB
#define VPUNPCKLQDQ_X7_X7_X15 BYTE $0xC5; BYTE $0x41; BYTE $0x6C; BYTE $0xFF
#define VPUNPCKLQDQ_X13_X13_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x11; BYTE $0x6C; BYTE $0xFD
#define VPUNPCKLQDQ_X14_X14_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x09; BYTE $0x6C; BYTE $0xFE
#define VPUNPCKHQDQ_X15_X2_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x69; BYTE $0x6D; BYTE $0xD7
#define VPUNPCKHQDQ_X15_X3_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xDF
#define VPUNPCKHQDQ_X15_X6_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x49; BYTE $0x6D; BYTE $0xF7
#define VPUNPCKHQDQ_X15_X7_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xFF
#define VPUNPCKHQDQ_X15_X3_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xD7
#define VPUNPCKHQDQ_X15_X7_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xF7
#define VPUNPCKHQDQ_X15_X13_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xDF
#define VPUNPCKHQDQ_X15_X13_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xFF
#define SHUFFLE_AVX() \
VMOVDQA X6, X13; \
VMOVDQA X2, X14; \
VMOVDQA X4, X6; \
VPUNPCKLQDQ_X13_X13_X15; \
VMOVDQA X5, X4; \
VMOVDQA X6, X5; \
VPUNPCKHQDQ_X15_X7_X6; \
VPUNPCKLQDQ_X7_X7_X15; \
VPUNPCKHQDQ_X15_X13_X7; \
VPUNPCKLQDQ_X3_X3_X15; \
VPUNPCKHQDQ_X15_X2_X2; \
VPUNPCKLQDQ_X14_X14_X15; \
VPUNPCKHQDQ_X15_X3_X3; \
#define SHUFFLE_AVX_INV() \
VMOVDQA X2, X13; \
VMOVDQA X4, X14; \
VPUNPCKLQDQ_X2_X2_X15; \
VMOVDQA X5, X4; \
VPUNPCKHQDQ_X15_X3_X2; \
VMOVDQA X14, X5; \
VPUNPCKLQDQ_X3_X3_X15; \
VMOVDQA X6, X14; \
VPUNPCKHQDQ_X15_X13_X3; \
VPUNPCKLQDQ_X7_X7_X15; \
VPUNPCKHQDQ_X15_X6_X6; \
VPUNPCKLQDQ_X14_X14_X15; \
VPUNPCKHQDQ_X15_X7_X7; \
#define HALF_ROUND_AVX(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
VPADDQ m0, v0, v0; \
VPADDQ v2, v0, v0; \
VPADDQ m1, v1, v1; \
VPADDQ v3, v1, v1; \
VPXOR v0, v6, v6; \
VPXOR v1, v7, v7; \
VPSHUFD $-79, v6, v6; \
VPSHUFD $-79, v7, v7; \
VPADDQ v6, v4, v4; \
VPADDQ v7, v5, v5; \
VPXOR v4, v2, v2; \
VPXOR v5, v3, v3; \
VPSHUFB c40, v2, v2; \
VPSHUFB c40, v3, v3; \
VPADDQ m2, v0, v0; \
VPADDQ v2, v0, v0; \
VPADDQ m3, v1, v1; \
VPADDQ v3, v1, v1; \
VPXOR v0, v6, v6; \
VPXOR v1, v7, v7; \
VPSHUFB c48, v6, v6; \
VPSHUFB c48, v7, v7; \
VPADDQ v6, v4, v4; \
VPADDQ v7, v5, v5; \
VPXOR v4, v2, v2; \
VPXOR v5, v3, v3; \
VPADDQ v2, v2, t0; \
VPSRLQ $63, v2, v2; \
VPXOR t0, v2, v2; \
VPADDQ v3, v3, t0; \
VPSRLQ $63, v3, v3; \
VPXOR t0, v3, v3
// load msg: X12 = (i0, i1), X13 = (i2, i3), X14 = (i4, i5), X15 = (i6, i7)
// i0, i1, i2, i3, i4, i5, i6, i7 must not be 0
#define LOAD_MSG_AVX(i0, i1, i2, i3, i4, i5, i6, i7) \
VMOVQ_SI_X12(i0*8); \
VMOVQ_SI_X13(i2*8); \
VMOVQ_SI_X14(i4*8); \
VMOVQ_SI_X15(i6*8); \
VPINSRQ_1_SI_X12(i1*8); \
VPINSRQ_1_SI_X13(i3*8); \
VPINSRQ_1_SI_X14(i5*8); \
VPINSRQ_1_SI_X15(i7*8)
// load msg: X12 = (0, 2), X13 = (4, 6), X14 = (1, 3), X15 = (5, 7)
#define LOAD_MSG_AVX_0_2_4_6_1_3_5_7() \
VMOVQ_SI_X12_0; \
VMOVQ_SI_X13(4*8); \
VMOVQ_SI_X14(1*8); \
VMOVQ_SI_X15(5*8); \
VPINSRQ_1_SI_X12(2*8); \
VPINSRQ_1_SI_X13(6*8); \
VPINSRQ_1_SI_X14(3*8); \
VPINSRQ_1_SI_X15(7*8)
// load msg: X12 = (1, 0), X13 = (11, 5), X14 = (12, 2), X15 = (7, 3)
#define LOAD_MSG_AVX_1_0_11_5_12_2_7_3() \
VPSHUFD $0x4E, 0*8(SI), X12; \
VMOVQ_SI_X13(11*8); \
VMOVQ_SI_X14(12*8); \
VMOVQ_SI_X15(7*8); \
VPINSRQ_1_SI_X13(5*8); \
VPINSRQ_1_SI_X14(2*8); \
VPINSRQ_1_SI_X15(3*8)
// load msg: X12 = (11, 12), X13 = (5, 15), X14 = (8, 0), X15 = (2, 13)
#define LOAD_MSG_AVX_11_12_5_15_8_0_2_13() \
VMOVDQU 11*8(SI), X12; \
VMOVQ_SI_X13(5*8); \
VMOVQ_SI_X14(8*8); \
VMOVQ_SI_X15(2*8); \
VPINSRQ_1_SI_X13(15*8); \
VPINSRQ_1_SI_X14_0; \
VPINSRQ_1_SI_X15(13*8)
// load msg: X12 = (2, 5), X13 = (4, 15), X14 = (6, 10), X15 = (0, 8)
#define LOAD_MSG_AVX_2_5_4_15_6_10_0_8() \
VMOVQ_SI_X12(2*8); \
VMOVQ_SI_X13(4*8); \
VMOVQ_SI_X14(6*8); \
VMOVQ_SI_X15_0; \
VPINSRQ_1_SI_X12(5*8); \
VPINSRQ_1_SI_X13(15*8); \
VPINSRQ_1_SI_X14(10*8); \
VPINSRQ_1_SI_X15(8*8)
// load msg: X12 = (9, 5), X13 = (2, 10), X14 = (0, 7), X15 = (4, 15)
#define LOAD_MSG_AVX_9_5_2_10_0_7_4_15() \
VMOVQ_SI_X12(9*8); \
VMOVQ_SI_X13(2*8); \
VMOVQ_SI_X14_0; \
VMOVQ_SI_X15(4*8); \
VPINSRQ_1_SI_X12(5*8); \
VPINSRQ_1_SI_X13(10*8); \
VPINSRQ_1_SI_X14(7*8); \
VPINSRQ_1_SI_X15(15*8)
// load msg: X12 = (2, 6), X13 = (0, 8), X14 = (12, 10), X15 = (11, 3)
#define LOAD_MSG_AVX_2_6_0_8_12_10_11_3() \
VMOVQ_SI_X12(2*8); \
VMOVQ_SI_X13_0; \
VMOVQ_SI_X14(12*8); \
VMOVQ_SI_X15(11*8); \
VPINSRQ_1_SI_X12(6*8); \
VPINSRQ_1_SI_X13(8*8); \
VPINSRQ_1_SI_X14(10*8); \
VPINSRQ_1_SI_X15(3*8)
// load msg: X12 = (0, 6), X13 = (9, 8), X14 = (7, 3), X15 = (2, 11)
#define LOAD_MSG_AVX_0_6_9_8_7_3_2_11() \
MOVQ 0*8(SI), X12; \
VPSHUFD $0x4E, 8*8(SI), X13; \
MOVQ 7*8(SI), X14; \
MOVQ 2*8(SI), X15; \
VPINSRQ_1_SI_X12(6*8); \
VPINSRQ_1_SI_X14(3*8); \
VPINSRQ_1_SI_X15(11*8)
// load msg: X12 = (6, 14), X13 = (11, 0), X14 = (15, 9), X15 = (3, 8)
#define LOAD_MSG_AVX_6_14_11_0_15_9_3_8() \
MOVQ 6*8(SI), X12; \
MOVQ 11*8(SI), X13; \
MOVQ 15*8(SI), X14; \
MOVQ 3*8(SI), X15; \
VPINSRQ_1_SI_X12(14*8); \
VPINSRQ_1_SI_X13_0; \
VPINSRQ_1_SI_X14(9*8); \
VPINSRQ_1_SI_X15(8*8)
// load msg: X12 = (5, 15), X13 = (8, 2), X14 = (0, 4), X15 = (6, 10)
#define LOAD_MSG_AVX_5_15_8_2_0_4_6_10() \
MOVQ 5*8(SI), X12; \
MOVQ 8*8(SI), X13; \
MOVQ 0*8(SI), X14; \
MOVQ 6*8(SI), X15; \
VPINSRQ_1_SI_X12(15*8); \
VPINSRQ_1_SI_X13(2*8); \
VPINSRQ_1_SI_X14(4*8); \
VPINSRQ_1_SI_X15(10*8)
// load msg: X12 = (12, 13), X13 = (1, 10), X14 = (2, 7), X15 = (4, 5)
#define LOAD_MSG_AVX_12_13_1_10_2_7_4_5() \
VMOVDQU 12*8(SI), X12; \
MOVQ 1*8(SI), X13; \
MOVQ 2*8(SI), X14; \
VPINSRQ_1_SI_X13(10*8); \
VPINSRQ_1_SI_X14(7*8); \
VMOVDQU 4*8(SI), X15
// load msg: X12 = (15, 9), X13 = (3, 13), X14 = (11, 14), X15 = (12, 0)
#define LOAD_MSG_AVX_15_9_3_13_11_14_12_0() \
MOVQ 15*8(SI), X12; \
MOVQ 3*8(SI), X13; \
MOVQ 11*8(SI), X14; \
MOVQ 12*8(SI), X15; \
VPINSRQ_1_SI_X12(9*8); \
VPINSRQ_1_SI_X13(13*8); \
VPINSRQ_1_SI_X14(14*8); \
VPINSRQ_1_SI_X15_0
// func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
TEXT ·hashBlocksAVX(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
MOVQ h+0(FP), AX
MOVQ c+8(FP), BX
MOVQ flag+16(FP), CX
MOVQ blocks_base+24(FP), SI
MOVQ blocks_len+32(FP), DI
MOVQ SP, R10
ADDQ $15, R10
ANDQ $~15, R10
VMOVDQU ·AVX_c40<>(SB), X0
VMOVDQU ·AVX_c48<>(SB), X1
VMOVDQA X0, X8
VMOVDQA X1, X9
VMOVDQU ·AVX_iv3<>(SB), X0
VMOVDQA X0, 0(R10)
XORQ CX, 0(R10) // 0(R10) = ·AVX_iv3 ^ (CX || 0)
VMOVDQU 0(AX), X10
VMOVDQU 16(AX), X11
VMOVDQU 32(AX), X2
VMOVDQU 48(AX), X3
MOVQ 0(BX), R8
MOVQ 8(BX), R9
loop:
ADDQ $128, R8
CMPQ R8, $128
JGE noinc
INCQ R9
noinc:
VMOVQ_R8_X15
VPINSRQ_1_R9_X15
VMOVDQA X10, X0
VMOVDQA X11, X1
VMOVDQU ·AVX_iv0<>(SB), X4
VMOVDQU ·AVX_iv1<>(SB), X5
VMOVDQU ·AVX_iv2<>(SB), X6
VPXOR X15, X6, X6
VMOVDQA 0(R10), X7
LOAD_MSG_AVX_0_2_4_6_1_3_5_7()
VMOVDQA X12, 16(R10)
VMOVDQA X13, 32(R10)
VMOVDQA X14, 48(R10)
VMOVDQA X15, 64(R10)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX(8, 10, 12, 14, 9, 11, 13, 15)
VMOVDQA X12, 80(R10)
VMOVDQA X13, 96(R10)
VMOVDQA X14, 112(R10)
VMOVDQA X15, 128(R10)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX(14, 4, 9, 13, 10, 8, 15, 6)
VMOVDQA X12, 144(R10)
VMOVDQA X13, 160(R10)
VMOVDQA X14, 176(R10)
VMOVDQA X15, 192(R10)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_1_0_11_5_12_2_7_3()
VMOVDQA X12, 208(R10)
VMOVDQA X13, 224(R10)
VMOVDQA X14, 240(R10)
VMOVDQA X15, 256(R10)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX_11_12_5_15_8_0_2_13()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX(10, 3, 7, 9, 14, 6, 1, 4)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX(7, 3, 13, 11, 9, 1, 12, 14)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_2_5_4_15_6_10_0_8()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX_9_5_2_10_0_7_4_15()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX(14, 11, 6, 3, 1, 12, 8, 13)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX_2_6_0_8_12_10_11_3()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX(4, 7, 15, 1, 13, 5, 14, 9)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX(12, 1, 14, 4, 5, 15, 13, 10)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_0_6_9_8_7_3_2_11()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX(13, 7, 12, 3, 11, 14, 1, 9)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_5_15_8_2_0_4_6_10()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX_6_14_11_0_15_9_3_8()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_12_13_1_10_2_7_4_5()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
LOAD_MSG_AVX(10, 8, 7, 1, 2, 4, 6, 5)
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX()
LOAD_MSG_AVX_15_9_3_13_11_14_12_0()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
SHUFFLE_AVX_INV()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X15, X8, X9)
SHUFFLE_AVX()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X15, X8, X9)
SHUFFLE_AVX_INV()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X15, X8, X9)
SHUFFLE_AVX()
HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X15, X8, X9)
SHUFFLE_AVX_INV()
VMOVDQU 32(AX), X14
VMOVDQU 48(AX), X15
VPXOR X0, X10, X10
VPXOR X1, X11, X11
VPXOR X2, X14, X14
VPXOR X3, X15, X15
VPXOR X4, X10, X10
VPXOR X5, X11, X11
VPXOR X6, X14, X2
VPXOR X7, X15, X3
VMOVDQU X2, 32(AX)
VMOVDQU X3, 48(AX)
LEAQ 128(SI), SI
SUBQ $128, DI
JNE loop
VMOVDQU X10, 0(AX)
VMOVDQU X11, 16(AX)
MOVQ R8, 0(BX)
MOVQ R9, 8(BX)
VZEROUPPER
RET

View File

@@ -1,278 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build amd64 && gc && !purego
#include "textflag.h"
DATA ·iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
DATA ·iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
GLOBL ·iv0<>(SB), (NOPTR+RODATA), $16
DATA ·iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
DATA ·iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
GLOBL ·iv1<>(SB), (NOPTR+RODATA), $16
DATA ·iv2<>+0x00(SB)/8, $0x510e527fade682d1
DATA ·iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
GLOBL ·iv2<>(SB), (NOPTR+RODATA), $16
DATA ·iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
DATA ·iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
GLOBL ·iv3<>(SB), (NOPTR+RODATA), $16
DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v6, t1; \
PUNPCKLQDQ v6, t2; \
PUNPCKHQDQ v7, v6; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ v7, t2; \
MOVO t1, v7; \
MOVO v2, t1; \
PUNPCKHQDQ t2, v7; \
PUNPCKLQDQ v3, t2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v3
#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v2, t1; \
PUNPCKLQDQ v2, t2; \
PUNPCKHQDQ v3, v2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ v3, t2; \
MOVO t1, v3; \
MOVO v6, t1; \
PUNPCKHQDQ t2, v3; \
PUNPCKLQDQ v7, t2; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v7
#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
PADDQ m0, v0; \
PADDQ m1, v1; \
PADDQ v2, v0; \
PADDQ v3, v1; \
PXOR v0, v6; \
PXOR v1, v7; \
PSHUFD $0xB1, v6, v6; \
PSHUFD $0xB1, v7, v7; \
PADDQ v6, v4; \
PADDQ v7, v5; \
PXOR v4, v2; \
PXOR v5, v3; \
PSHUFB c40, v2; \
PSHUFB c40, v3; \
PADDQ m2, v0; \
PADDQ m3, v1; \
PADDQ v2, v0; \
PADDQ v3, v1; \
PXOR v0, v6; \
PXOR v1, v7; \
PSHUFB c48, v6; \
PSHUFB c48, v7; \
PADDQ v6, v4; \
PADDQ v7, v5; \
PXOR v4, v2; \
PXOR v5, v3; \
MOVOU v2, t0; \
PADDQ v2, t0; \
PSRLQ $63, v2; \
PXOR t0, v2; \
MOVOU v3, t0; \
PADDQ v3, t0; \
PSRLQ $63, v3; \
PXOR t0, v3
#define LOAD_MSG(m0, m1, m2, m3, src, i0, i1, i2, i3, i4, i5, i6, i7) \
MOVQ i0*8(src), m0; \
PINSRQ $1, i1*8(src), m0; \
MOVQ i2*8(src), m1; \
PINSRQ $1, i3*8(src), m1; \
MOVQ i4*8(src), m2; \
PINSRQ $1, i5*8(src), m2; \
MOVQ i6*8(src), m3; \
PINSRQ $1, i7*8(src), m3
// func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
TEXT ·hashBlocksSSE4(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
MOVQ h+0(FP), AX
MOVQ c+8(FP), BX
MOVQ flag+16(FP), CX
MOVQ blocks_base+24(FP), SI
MOVQ blocks_len+32(FP), DI
MOVQ SP, R10
ADDQ $15, R10
ANDQ $~15, R10
MOVOU ·iv3<>(SB), X0
MOVO X0, 0(R10)
XORQ CX, 0(R10) // 0(R10) = ·iv3 ^ (CX || 0)
MOVOU ·c40<>(SB), X13
MOVOU ·c48<>(SB), X14
MOVOU 0(AX), X12
MOVOU 16(AX), X15
MOVQ 0(BX), R8
MOVQ 8(BX), R9
loop:
ADDQ $128, R8
CMPQ R8, $128
JGE noinc
INCQ R9
noinc:
MOVQ R8, X8
PINSRQ $1, R9, X8
MOVO X12, X0
MOVO X15, X1
MOVOU 32(AX), X2
MOVOU 48(AX), X3
MOVOU ·iv0<>(SB), X4
MOVOU ·iv1<>(SB), X5
MOVOU ·iv2<>(SB), X6
PXOR X8, X6
MOVO 0(R10), X7
LOAD_MSG(X8, X9, X10, X11, SI, 0, 2, 4, 6, 1, 3, 5, 7)
MOVO X8, 16(R10)
MOVO X9, 32(R10)
MOVO X10, 48(R10)
MOVO X11, 64(R10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 8, 10, 12, 14, 9, 11, 13, 15)
MOVO X8, 80(R10)
MOVO X9, 96(R10)
MOVO X10, 112(R10)
MOVO X11, 128(R10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 14, 4, 9, 13, 10, 8, 15, 6)
MOVO X8, 144(R10)
MOVO X9, 160(R10)
MOVO X10, 176(R10)
MOVO X11, 192(R10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 1, 0, 11, 5, 12, 2, 7, 3)
MOVO X8, 208(R10)
MOVO X9, 224(R10)
MOVO X10, 240(R10)
MOVO X11, 256(R10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 11, 12, 5, 15, 8, 0, 2, 13)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 10, 3, 7, 9, 14, 6, 1, 4)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 7, 3, 13, 11, 9, 1, 12, 14)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 2, 5, 4, 15, 6, 10, 0, 8)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 9, 5, 2, 10, 0, 7, 4, 15)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 14, 11, 6, 3, 1, 12, 8, 13)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 2, 6, 0, 8, 12, 10, 11, 3)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 4, 7, 15, 1, 13, 5, 14, 9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 12, 1, 14, 4, 5, 15, 13, 10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 0, 6, 9, 8, 7, 3, 2, 11)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 13, 7, 12, 3, 11, 14, 1, 9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 5, 15, 8, 2, 0, 4, 6, 10)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 6, 14, 11, 0, 15, 9, 3, 8)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 12, 13, 1, 10, 2, 7, 4, 5)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 10, 8, 7, 1, 2, 4, 6, 5)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
LOAD_MSG(X8, X9, X10, X11, SI, 15, 9, 3, 13, 11, 14, 12, 0)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X11, X13, X14)
SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X11, X13, X14)
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
MOVOU 32(AX), X10
MOVOU 48(AX), X11
PXOR X0, X12
PXOR X1, X15
PXOR X2, X10
PXOR X3, X11
PXOR X4, X12
PXOR X5, X15
PXOR X6, X10
PXOR X7, X11
MOVOU X10, 32(AX)
MOVOU X11, 48(AX)
LEAQ 128(SI), SI
SUBQ $128, DI
JNE loop
MOVOU X12, 0(AX)
MOVOU X15, 16(AX)
MOVQ R8, 0(BX)
MOVQ R9, 8(BX)
RET

View File

@@ -1,182 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package blake2b
import (
"encoding/binary"
"math/bits"
)
// the precomputed values for BLAKE2b
// there are 12 16-byte arrays - one for each round
// the entries are calculated from the sigma constants.
var precomputed = [12][16]byte{
{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15},
{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3},
{11, 12, 5, 15, 8, 0, 2, 13, 10, 3, 7, 9, 14, 6, 1, 4},
{7, 3, 13, 11, 9, 1, 12, 14, 2, 5, 4, 15, 6, 10, 0, 8},
{9, 5, 2, 10, 0, 7, 4, 15, 14, 11, 6, 3, 1, 12, 8, 13},
{2, 6, 0, 8, 12, 10, 11, 3, 4, 7, 15, 1, 13, 5, 14, 9},
{12, 1, 14, 4, 5, 15, 13, 10, 0, 6, 9, 8, 7, 3, 2, 11},
{13, 7, 12, 3, 11, 14, 1, 9, 5, 15, 8, 2, 0, 4, 6, 10},
{6, 14, 11, 0, 15, 9, 3, 8, 12, 13, 1, 10, 2, 7, 4, 5},
{10, 8, 7, 1, 2, 4, 6, 5, 15, 9, 3, 13, 11, 14, 12, 0},
{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15}, // equal to the first
{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3}, // equal to the second
}
func hashBlocksGeneric(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
var m [16]uint64
c0, c1 := c[0], c[1]
for i := 0; i < len(blocks); {
c0 += BlockSize
if c0 < BlockSize {
c1++
}
v0, v1, v2, v3, v4, v5, v6, v7 := h[0], h[1], h[2], h[3], h[4], h[5], h[6], h[7]
v8, v9, v10, v11, v12, v13, v14, v15 := iv[0], iv[1], iv[2], iv[3], iv[4], iv[5], iv[6], iv[7]
v12 ^= c0
v13 ^= c1
v14 ^= flag
for j := range m {
m[j] = binary.LittleEndian.Uint64(blocks[i:])
i += 8
}
for j := range precomputed {
s := &(precomputed[j])
v0 += m[s[0]]
v0 += v4
v12 ^= v0
v12 = bits.RotateLeft64(v12, -32)
v8 += v12
v4 ^= v8
v4 = bits.RotateLeft64(v4, -24)
v1 += m[s[1]]
v1 += v5
v13 ^= v1
v13 = bits.RotateLeft64(v13, -32)
v9 += v13
v5 ^= v9
v5 = bits.RotateLeft64(v5, -24)
v2 += m[s[2]]
v2 += v6
v14 ^= v2
v14 = bits.RotateLeft64(v14, -32)
v10 += v14
v6 ^= v10
v6 = bits.RotateLeft64(v6, -24)
v3 += m[s[3]]
v3 += v7
v15 ^= v3
v15 = bits.RotateLeft64(v15, -32)
v11 += v15
v7 ^= v11
v7 = bits.RotateLeft64(v7, -24)
v0 += m[s[4]]
v0 += v4
v12 ^= v0
v12 = bits.RotateLeft64(v12, -16)
v8 += v12
v4 ^= v8
v4 = bits.RotateLeft64(v4, -63)
v1 += m[s[5]]
v1 += v5
v13 ^= v1
v13 = bits.RotateLeft64(v13, -16)
v9 += v13
v5 ^= v9
v5 = bits.RotateLeft64(v5, -63)
v2 += m[s[6]]
v2 += v6
v14 ^= v2
v14 = bits.RotateLeft64(v14, -16)
v10 += v14
v6 ^= v10
v6 = bits.RotateLeft64(v6, -63)
v3 += m[s[7]]
v3 += v7
v15 ^= v3
v15 = bits.RotateLeft64(v15, -16)
v11 += v15
v7 ^= v11
v7 = bits.RotateLeft64(v7, -63)
v0 += m[s[8]]
v0 += v5
v15 ^= v0
v15 = bits.RotateLeft64(v15, -32)
v10 += v15
v5 ^= v10
v5 = bits.RotateLeft64(v5, -24)
v1 += m[s[9]]
v1 += v6
v12 ^= v1
v12 = bits.RotateLeft64(v12, -32)
v11 += v12
v6 ^= v11
v6 = bits.RotateLeft64(v6, -24)
v2 += m[s[10]]
v2 += v7
v13 ^= v2
v13 = bits.RotateLeft64(v13, -32)
v8 += v13
v7 ^= v8
v7 = bits.RotateLeft64(v7, -24)
v3 += m[s[11]]
v3 += v4
v14 ^= v3
v14 = bits.RotateLeft64(v14, -32)
v9 += v14
v4 ^= v9
v4 = bits.RotateLeft64(v4, -24)
v0 += m[s[12]]
v0 += v5
v15 ^= v0
v15 = bits.RotateLeft64(v15, -16)
v10 += v15
v5 ^= v10
v5 = bits.RotateLeft64(v5, -63)
v1 += m[s[13]]
v1 += v6
v12 ^= v1
v12 = bits.RotateLeft64(v12, -16)
v11 += v12
v6 ^= v11
v6 = bits.RotateLeft64(v6, -63)
v2 += m[s[14]]
v2 += v7
v13 ^= v2
v13 = bits.RotateLeft64(v13, -16)
v8 += v13
v7 ^= v8
v7 = bits.RotateLeft64(v7, -63)
v3 += m[s[15]]
v3 += v4
v14 ^= v3
v14 = bits.RotateLeft64(v14, -16)
v9 += v14
v4 ^= v9
v4 = bits.RotateLeft64(v4, -63)
}
h[0] ^= v0 ^ v8
h[1] ^= v1 ^ v9
h[2] ^= v2 ^ v10
h[3] ^= v3 ^ v11
h[4] ^= v4 ^ v12
h[5] ^= v5 ^ v13
h[6] ^= v6 ^ v14
h[7] ^= v7 ^ v15
}
c[0], c[1] = c0, c1
}

View File

@@ -1,11 +0,0 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build !amd64 || purego || !gc
package blake2b
func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
hashBlocksGeneric(h, c, flag, blocks)
}

View File

@@ -1,177 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package blake2b
import (
"encoding/binary"
"errors"
"io"
)
// XOF defines the interface to hash functions that
// support arbitrary-length output.
type XOF interface {
// Write absorbs more data into the hash's state. It panics if called
// after Read.
io.Writer
// Read reads more output from the hash. It returns io.EOF if the limit
// has been reached.
io.Reader
// Clone returns a copy of the XOF in its current state.
Clone() XOF
// Reset resets the XOF to its initial state.
Reset()
}
// OutputLengthUnknown can be used as the size argument to NewXOF to indicate
// the length of the output is not known in advance.
const OutputLengthUnknown = 0
// magicUnknownOutputLength is a magic value for the output size that indicates
// an unknown number of output bytes.
const magicUnknownOutputLength = (1 << 32) - 1
// maxOutputLength is the absolute maximum number of bytes to produce when the
// number of output bytes is unknown.
const maxOutputLength = (1 << 32) * 64
// NewXOF creates a new variable-output-length hash. The hash either produce a
// known number of bytes (1 <= size < 2**32-1), or an unknown number of bytes
// (size == OutputLengthUnknown). In the latter case, an absolute limit of
// 256GiB applies.
//
// A non-nil key turns the hash into a MAC. The key must between
// zero and 32 bytes long.
func NewXOF(size uint32, key []byte) (XOF, error) {
if len(key) > Size {
return nil, errKeySize
}
if size == magicUnknownOutputLength {
// 2^32-1 indicates an unknown number of bytes and thus isn't a
// valid length.
return nil, errors.New("blake2b: XOF length too large")
}
if size == OutputLengthUnknown {
size = magicUnknownOutputLength
}
x := &xof{
d: digest{
size: Size,
keyLen: len(key),
},
length: size,
}
copy(x.d.key[:], key)
x.Reset()
return x, nil
}
type xof struct {
d digest
length uint32
remaining uint64
cfg, root, block [Size]byte
offset int
nodeOffset uint32
readMode bool
}
func (x *xof) Write(p []byte) (n int, err error) {
if x.readMode {
panic("blake2b: write to XOF after read")
}
return x.d.Write(p)
}
func (x *xof) Clone() XOF {
clone := *x
return &clone
}
func (x *xof) Reset() {
x.cfg[0] = byte(Size)
binary.LittleEndian.PutUint32(x.cfg[4:], uint32(Size)) // leaf length
binary.LittleEndian.PutUint32(x.cfg[12:], x.length) // XOF length
x.cfg[17] = byte(Size) // inner hash size
x.d.Reset()
x.d.h[1] ^= uint64(x.length) << 32
x.remaining = uint64(x.length)
if x.remaining == magicUnknownOutputLength {
x.remaining = maxOutputLength
}
x.offset, x.nodeOffset = 0, 0
x.readMode = false
}
func (x *xof) Read(p []byte) (n int, err error) {
if !x.readMode {
x.d.finalize(&x.root)
x.readMode = true
}
if x.remaining == 0 {
return 0, io.EOF
}
n = len(p)
if uint64(n) > x.remaining {
n = int(x.remaining)
p = p[:n]
}
if x.offset > 0 {
blockRemaining := Size - x.offset
if n < blockRemaining {
x.offset += copy(p, x.block[x.offset:])
x.remaining -= uint64(n)
return
}
copy(p, x.block[x.offset:])
p = p[blockRemaining:]
x.offset = 0
x.remaining -= uint64(blockRemaining)
}
for len(p) >= Size {
binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
x.nodeOffset++
x.d.initConfig(&x.cfg)
x.d.Write(x.root[:])
x.d.finalize(&x.block)
copy(p, x.block[:])
p = p[Size:]
x.remaining -= uint64(Size)
}
if todo := len(p); todo > 0 {
if x.remaining < uint64(Size) {
x.cfg[0] = byte(x.remaining)
}
binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
x.nodeOffset++
x.d.initConfig(&x.cfg)
x.d.Write(x.root[:])
x.d.finalize(&x.block)
x.offset = copy(p, x.block[:todo])
x.remaining -= uint64(todo)
}
return
}
func (d *digest) initConfig(cfg *[Size]byte) {
d.offset, d.c[0], d.c[1] = 0, 0, 0
for i := range d.h {
d.h[i] = iv[i] ^ binary.LittleEndian.Uint64(cfg[i*8:])
}
}

View File

@@ -1,30 +0,0 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package blake2b
import (
"crypto"
"hash"
)
func init() {
newHash256 := func() hash.Hash {
h, _ := New256(nil)
return h
}
newHash384 := func() hash.Hash {
h, _ := New384(nil)
return h
}
newHash512 := func() hash.Hash {
h, _ := New512(nil)
return h
}
crypto.RegisterHash(crypto.BLAKE2b_256, newHash256)
crypto.RegisterHash(crypto.BLAKE2b_384, newHash384)
crypto.RegisterHash(crypto.BLAKE2b_512, newHash512)
}

View File

@@ -0,0 +1,525 @@
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package timeseries implements a time series structure for stats collection.
package timeseries // import "golang.org/x/net/internal/timeseries"
import (
"fmt"
"log"
"time"
)
const (
timeSeriesNumBuckets = 64
minuteHourSeriesNumBuckets = 60
)
var timeSeriesResolutions = []time.Duration{
1 * time.Second,
10 * time.Second,
1 * time.Minute,
10 * time.Minute,
1 * time.Hour,
6 * time.Hour,
24 * time.Hour, // 1 day
7 * 24 * time.Hour, // 1 week
4 * 7 * 24 * time.Hour, // 4 weeks
16 * 7 * 24 * time.Hour, // 16 weeks
}
var minuteHourSeriesResolutions = []time.Duration{
1 * time.Second,
1 * time.Minute,
}
// An Observable is a kind of data that can be aggregated in a time series.
type Observable interface {
Multiply(ratio float64) // Multiplies the data in self by a given ratio
Add(other Observable) // Adds the data from a different observation to self
Clear() // Clears the observation so it can be reused.
CopyFrom(other Observable) // Copies the contents of a given observation to self
}
// Float attaches the methods of Observable to a float64.
type Float float64
// NewFloat returns a Float.
func NewFloat() Observable {
f := Float(0)
return &f
}
// String returns the float as a string.
func (f *Float) String() string { return fmt.Sprintf("%g", f.Value()) }
// Value returns the float's value.
func (f *Float) Value() float64 { return float64(*f) }
func (f *Float) Multiply(ratio float64) { *f *= Float(ratio) }
func (f *Float) Add(other Observable) {
o := other.(*Float)
*f += *o
}
func (f *Float) Clear() { *f = 0 }
func (f *Float) CopyFrom(other Observable) {
o := other.(*Float)
*f = *o
}
// A Clock tells the current time.
type Clock interface {
Time() time.Time
}
type defaultClock int
var defaultClockInstance defaultClock
func (defaultClock) Time() time.Time { return time.Now() }
// Information kept per level. Each level consists of a circular list of
// observations. The start of the level may be derived from end and the
// len(buckets) * sizeInMillis.
type tsLevel struct {
oldest int // index to oldest bucketed Observable
newest int // index to newest bucketed Observable
end time.Time // end timestamp for this level
size time.Duration // duration of the bucketed Observable
buckets []Observable // collections of observations
provider func() Observable // used for creating new Observable
}
func (l *tsLevel) Clear() {
l.oldest = 0
l.newest = len(l.buckets) - 1
l.end = time.Time{}
for i := range l.buckets {
if l.buckets[i] != nil {
l.buckets[i].Clear()
l.buckets[i] = nil
}
}
}
func (l *tsLevel) InitLevel(size time.Duration, numBuckets int, f func() Observable) {
l.size = size
l.provider = f
l.buckets = make([]Observable, numBuckets)
}
// Keeps a sequence of levels. Each level is responsible for storing data at
// a given resolution. For example, the first level stores data at a one
// minute resolution while the second level stores data at a one hour
// resolution.
// Each level is represented by a sequence of buckets. Each bucket spans an
// interval equal to the resolution of the level. New observations are added
// to the last bucket.
type timeSeries struct {
provider func() Observable // make more Observable
numBuckets int // number of buckets in each level
levels []*tsLevel // levels of bucketed Observable
lastAdd time.Time // time of last Observable tracked
total Observable // convenient aggregation of all Observable
clock Clock // Clock for getting current time
pending Observable // observations not yet bucketed
pendingTime time.Time // what time are we keeping in pending
dirty bool // if there are pending observations
}
// init initializes a level according to the supplied criteria.
func (ts *timeSeries) init(resolutions []time.Duration, f func() Observable, numBuckets int, clock Clock) {
ts.provider = f
ts.numBuckets = numBuckets
ts.clock = clock
ts.levels = make([]*tsLevel, len(resolutions))
for i := range resolutions {
if i > 0 && resolutions[i-1] >= resolutions[i] {
log.Print("timeseries: resolutions must be monotonically increasing")
break
}
newLevel := new(tsLevel)
newLevel.InitLevel(resolutions[i], ts.numBuckets, ts.provider)
ts.levels[i] = newLevel
}
ts.Clear()
}
// Clear removes all observations from the time series.
func (ts *timeSeries) Clear() {
ts.lastAdd = time.Time{}
ts.total = ts.resetObservation(ts.total)
ts.pending = ts.resetObservation(ts.pending)
ts.pendingTime = time.Time{}
ts.dirty = false
for i := range ts.levels {
ts.levels[i].Clear()
}
}
// Add records an observation at the current time.
func (ts *timeSeries) Add(observation Observable) {
ts.AddWithTime(observation, ts.clock.Time())
}
// AddWithTime records an observation at the specified time.
func (ts *timeSeries) AddWithTime(observation Observable, t time.Time) {
smallBucketDuration := ts.levels[0].size
if t.After(ts.lastAdd) {
ts.lastAdd = t
}
if t.After(ts.pendingTime) {
ts.advance(t)
ts.mergePendingUpdates()
ts.pendingTime = ts.levels[0].end
ts.pending.CopyFrom(observation)
ts.dirty = true
} else if t.After(ts.pendingTime.Add(-1 * smallBucketDuration)) {
// The observation is close enough to go into the pending bucket.
// This compensates for clock skewing and small scheduling delays
// by letting the update stay in the fast path.
ts.pending.Add(observation)
ts.dirty = true
} else {
ts.mergeValue(observation, t)
}
}
// mergeValue inserts the observation at the specified time in the past into all levels.
func (ts *timeSeries) mergeValue(observation Observable, t time.Time) {
for _, level := range ts.levels {
index := (ts.numBuckets - 1) - int(level.end.Sub(t)/level.size)
if 0 <= index && index < ts.numBuckets {
bucketNumber := (level.oldest + index) % ts.numBuckets
if level.buckets[bucketNumber] == nil {
level.buckets[bucketNumber] = level.provider()
}
level.buckets[bucketNumber].Add(observation)
}
}
ts.total.Add(observation)
}
// mergePendingUpdates applies the pending updates into all levels.
func (ts *timeSeries) mergePendingUpdates() {
if ts.dirty {
ts.mergeValue(ts.pending, ts.pendingTime)
ts.pending = ts.resetObservation(ts.pending)
ts.dirty = false
}
}
// advance cycles the buckets at each level until the latest bucket in
// each level can hold the time specified.
func (ts *timeSeries) advance(t time.Time) {
if !t.After(ts.levels[0].end) {
return
}
for i := 0; i < len(ts.levels); i++ {
level := ts.levels[i]
if !level.end.Before(t) {
break
}
// If the time is sufficiently far, just clear the level and advance
// directly.
if !t.Before(level.end.Add(level.size * time.Duration(ts.numBuckets))) {
for _, b := range level.buckets {
ts.resetObservation(b)
}
level.end = time.Unix(0, (t.UnixNano()/level.size.Nanoseconds())*level.size.Nanoseconds())
}
for t.After(level.end) {
level.end = level.end.Add(level.size)
level.newest = level.oldest
level.oldest = (level.oldest + 1) % ts.numBuckets
ts.resetObservation(level.buckets[level.newest])
}
t = level.end
}
}
// Latest returns the sum of the num latest buckets from the level.
func (ts *timeSeries) Latest(level, num int) Observable {
now := ts.clock.Time()
if ts.levels[0].end.Before(now) {
ts.advance(now)
}
ts.mergePendingUpdates()
result := ts.provider()
l := ts.levels[level]
index := l.newest
for i := 0; i < num; i++ {
if l.buckets[index] != nil {
result.Add(l.buckets[index])
}
if index == 0 {
index = ts.numBuckets
}
index--
}
return result
}
// LatestBuckets returns a copy of the num latest buckets from level.
func (ts *timeSeries) LatestBuckets(level, num int) []Observable {
if level < 0 || level > len(ts.levels) {
log.Print("timeseries: bad level argument: ", level)
return nil
}
if num < 0 || num >= ts.numBuckets {
log.Print("timeseries: bad num argument: ", num)
return nil
}
results := make([]Observable, num)
now := ts.clock.Time()
if ts.levels[0].end.Before(now) {
ts.advance(now)
}
ts.mergePendingUpdates()
l := ts.levels[level]
index := l.newest
for i := 0; i < num; i++ {
result := ts.provider()
results[i] = result
if l.buckets[index] != nil {
result.CopyFrom(l.buckets[index])
}
if index == 0 {
index = ts.numBuckets
}
index -= 1
}
return results
}
// ScaleBy updates observations by scaling by factor.
func (ts *timeSeries) ScaleBy(factor float64) {
for _, l := range ts.levels {
for i := 0; i < ts.numBuckets; i++ {
l.buckets[i].Multiply(factor)
}
}
ts.total.Multiply(factor)
ts.pending.Multiply(factor)
}
// Range returns the sum of observations added over the specified time range.
// If start or finish times don't fall on bucket boundaries of the same
// level, then return values are approximate answers.
func (ts *timeSeries) Range(start, finish time.Time) Observable {
return ts.ComputeRange(start, finish, 1)[0]
}
// Recent returns the sum of observations from the last delta.
func (ts *timeSeries) Recent(delta time.Duration) Observable {
now := ts.clock.Time()
return ts.Range(now.Add(-delta), now)
}
// Total returns the total of all observations.
func (ts *timeSeries) Total() Observable {
ts.mergePendingUpdates()
return ts.total
}
// ComputeRange computes a specified number of values into a slice using
// the observations recorded over the specified time period. The return
// values are approximate if the start or finish times don't fall on the
// bucket boundaries at the same level or if the number of buckets spanning
// the range is not an integral multiple of num.
func (ts *timeSeries) ComputeRange(start, finish time.Time, num int) []Observable {
if start.After(finish) {
log.Printf("timeseries: start > finish, %v>%v", start, finish)
return nil
}
if num < 0 {
log.Printf("timeseries: num < 0, %v", num)
return nil
}
results := make([]Observable, num)
for _, l := range ts.levels {
if !start.Before(l.end.Add(-l.size * time.Duration(ts.numBuckets))) {
ts.extract(l, start, finish, num, results)
return results
}
}
// Failed to find a level that covers the desired range. So just
// extract from the last level, even if it doesn't cover the entire
// desired range.
ts.extract(ts.levels[len(ts.levels)-1], start, finish, num, results)
return results
}
// RecentList returns the specified number of values in slice over the most
// recent time period of the specified range.
func (ts *timeSeries) RecentList(delta time.Duration, num int) []Observable {
if delta < 0 {
return nil
}
now := ts.clock.Time()
return ts.ComputeRange(now.Add(-delta), now, num)
}
// extract returns a slice of specified number of observations from a given
// level over a given range.
func (ts *timeSeries) extract(l *tsLevel, start, finish time.Time, num int, results []Observable) {
ts.mergePendingUpdates()
srcInterval := l.size
dstInterval := finish.Sub(start) / time.Duration(num)
dstStart := start
srcStart := l.end.Add(-srcInterval * time.Duration(ts.numBuckets))
srcIndex := 0
// Where should scanning start?
if dstStart.After(srcStart) {
advance := int(dstStart.Sub(srcStart) / srcInterval)
srcIndex += advance
srcStart = srcStart.Add(time.Duration(advance) * srcInterval)
}
// The i'th value is computed as show below.
// interval = (finish/start)/num
// i'th value = sum of observation in range
// [ start + i * interval,
// start + (i + 1) * interval )
for i := 0; i < num; i++ {
results[i] = ts.resetObservation(results[i])
dstEnd := dstStart.Add(dstInterval)
for srcIndex < ts.numBuckets && srcStart.Before(dstEnd) {
srcEnd := srcStart.Add(srcInterval)
if srcEnd.After(ts.lastAdd) {
srcEnd = ts.lastAdd
}
if !srcEnd.Before(dstStart) {
srcValue := l.buckets[(srcIndex+l.oldest)%ts.numBuckets]
if !srcStart.Before(dstStart) && !srcEnd.After(dstEnd) {
// dst completely contains src.
if srcValue != nil {
results[i].Add(srcValue)
}
} else {
// dst partially overlaps src.
overlapStart := maxTime(srcStart, dstStart)
overlapEnd := minTime(srcEnd, dstEnd)
base := srcEnd.Sub(srcStart)
fraction := overlapEnd.Sub(overlapStart).Seconds() / base.Seconds()
used := ts.provider()
if srcValue != nil {
used.CopyFrom(srcValue)
}
used.Multiply(fraction)
results[i].Add(used)
}
if srcEnd.After(dstEnd) {
break
}
}
srcIndex++
srcStart = srcStart.Add(srcInterval)
}
dstStart = dstStart.Add(dstInterval)
}
}
// resetObservation clears the content so the struct may be reused.
func (ts *timeSeries) resetObservation(observation Observable) Observable {
if observation == nil {
observation = ts.provider()
} else {
observation.Clear()
}
return observation
}
// TimeSeries tracks data at granularities from 1 second to 16 weeks.
type TimeSeries struct {
timeSeries
}
// NewTimeSeries creates a new TimeSeries using the function provided for creating new Observable.
func NewTimeSeries(f func() Observable) *TimeSeries {
return NewTimeSeriesWithClock(f, defaultClockInstance)
}
// NewTimeSeriesWithClock creates a new TimeSeries using the function provided for creating new Observable and the clock for
// assigning timestamps.
func NewTimeSeriesWithClock(f func() Observable, clock Clock) *TimeSeries {
ts := new(TimeSeries)
ts.timeSeries.init(timeSeriesResolutions, f, timeSeriesNumBuckets, clock)
return ts
}
// MinuteHourSeries tracks data at granularities of 1 minute and 1 hour.
type MinuteHourSeries struct {
timeSeries
}
// NewMinuteHourSeries creates a new MinuteHourSeries using the function provided for creating new Observable.
func NewMinuteHourSeries(f func() Observable) *MinuteHourSeries {
return NewMinuteHourSeriesWithClock(f, defaultClockInstance)
}
// NewMinuteHourSeriesWithClock creates a new MinuteHourSeries using the function provided for creating new Observable and the clock for
// assigning timestamps.
func NewMinuteHourSeriesWithClock(f func() Observable, clock Clock) *MinuteHourSeries {
ts := new(MinuteHourSeries)
ts.timeSeries.init(minuteHourSeriesResolutions, f,
minuteHourSeriesNumBuckets, clock)
return ts
}
func (ts *MinuteHourSeries) Minute() Observable {
return ts.timeSeries.Latest(0, 60)
}
func (ts *MinuteHourSeries) Hour() Observable {
return ts.timeSeries.Latest(1, 60)
}
func minTime(a, b time.Time) time.Time {
if a.Before(b) {
return a
}
return b
}
func maxTime(a, b time.Time) time.Time {
if a.After(b) {
return a
}
return b
}

Binary file not shown.

Binary file not shown.

File diff suppressed because one or more lines are too long

View File

@@ -1,203 +0,0 @@
// Copyright 2012 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:generate go run gen.go
// Package publicsuffix provides a public suffix list based on data from
// https://publicsuffix.org/
//
// A public suffix is one under which Internet users can directly register
// names. It is related to, but different from, a TLD (top level domain).
//
// "com" is a TLD (top level domain). Top level means it has no dots.
//
// "com" is also a public suffix. Amazon and Google have registered different
// siblings under that domain: "amazon.com" and "google.com".
//
// "au" is another TLD, again because it has no dots. But it's not "amazon.au".
// Instead, it's "amazon.com.au".
//
// "com.au" isn't an actual TLD, because it's not at the top level (it has
// dots). But it is an eTLD (effective TLD), because that's the branching point
// for domain name registrars.
//
// Another name for "an eTLD" is "a public suffix". Often, what's more of
// interest is the eTLD+1, or one more label than the public suffix. For
// example, browsers partition read/write access to HTTP cookies according to
// the eTLD+1. Web pages served from "amazon.com.au" can't read cookies from
// "google.com.au", but web pages served from "maps.google.com" can share
// cookies from "www.google.com", so you don't have to sign into Google Maps
// separately from signing into Google Web Search. Note that all four of those
// domains have 3 labels and 2 dots. The first two domains are each an eTLD+1,
// the last two are not (but share the same eTLD+1: "google.com").
//
// All of these domains have the same eTLD+1:
// - "www.books.amazon.co.uk"
// - "books.amazon.co.uk"
// - "amazon.co.uk"
//
// Specifically, the eTLD+1 is "amazon.co.uk", because the eTLD is "co.uk".
//
// There is no closed form algorithm to calculate the eTLD of a domain.
// Instead, the calculation is data driven. This package provides a
// pre-compiled snapshot of Mozilla's PSL (Public Suffix List) data at
// https://publicsuffix.org/
package publicsuffix // import "golang.org/x/net/publicsuffix"
// TODO: specify case sensitivity and leading/trailing dot behavior for
// func PublicSuffix and func EffectiveTLDPlusOne.
import (
"fmt"
"net/http/cookiejar"
"strings"
)
// List implements the cookiejar.PublicSuffixList interface by calling the
// PublicSuffix function.
var List cookiejar.PublicSuffixList = list{}
type list struct{}
func (list) PublicSuffix(domain string) string {
ps, _ := PublicSuffix(domain)
return ps
}
func (list) String() string {
return version
}
// PublicSuffix returns the public suffix of the domain using a copy of the
// publicsuffix.org database compiled into the library.
//
// icann is whether the public suffix is managed by the Internet Corporation
// for Assigned Names and Numbers. If not, the public suffix is either a
// privately managed domain (and in practice, not a top level domain) or an
// unmanaged top level domain (and not explicitly mentioned in the
// publicsuffix.org list). For example, "foo.org" and "foo.co.uk" are ICANN
// domains, "foo.dyndns.org" and "foo.blogspot.co.uk" are private domains and
// "cromulent" is an unmanaged top level domain.
//
// Use cases for distinguishing ICANN domains like "foo.com" from private
// domains like "foo.appspot.com" can be found at
// https://wiki.mozilla.org/Public_Suffix_List/Use_Cases
func PublicSuffix(domain string) (publicSuffix string, icann bool) {
lo, hi := uint32(0), uint32(numTLD)
s, suffix, icannNode, wildcard := domain, len(domain), false, false
loop:
for {
dot := strings.LastIndex(s, ".")
if wildcard {
icann = icannNode
suffix = 1 + dot
}
if lo == hi {
break
}
f := find(s[1+dot:], lo, hi)
if f == notFound {
break
}
u := uint32(nodes.get(f) >> (nodesBitsTextOffset + nodesBitsTextLength))
icannNode = u&(1<<nodesBitsICANN-1) != 0
u >>= nodesBitsICANN
u = children.get(u & (1<<nodesBitsChildren - 1))
lo = u & (1<<childrenBitsLo - 1)
u >>= childrenBitsLo
hi = u & (1<<childrenBitsHi - 1)
u >>= childrenBitsHi
switch u & (1<<childrenBitsNodeType - 1) {
case nodeTypeNormal:
suffix = 1 + dot
case nodeTypeException:
suffix = 1 + len(s)
break loop
}
u >>= childrenBitsNodeType
wildcard = u&(1<<childrenBitsWildcard-1) != 0
if !wildcard {
icann = icannNode
}
if dot == -1 {
break
}
s = s[:dot]
}
if suffix == len(domain) {
// If no rules match, the prevailing rule is "*".
return domain[1+strings.LastIndex(domain, "."):], icann
}
return domain[suffix:], icann
}
const notFound uint32 = 1<<32 - 1
// find returns the index of the node in the range [lo, hi) whose label equals
// label, or notFound if there is no such node. The range is assumed to be in
// strictly increasing node label order.
func find(label string, lo, hi uint32) uint32 {
for lo < hi {
mid := lo + (hi-lo)/2
s := nodeLabel(mid)
if s < label {
lo = mid + 1
} else if s == label {
return mid
} else {
hi = mid
}
}
return notFound
}
// nodeLabel returns the label for the i'th node.
func nodeLabel(i uint32) string {
x := nodes.get(i)
length := x & (1<<nodesBitsTextLength - 1)
x >>= nodesBitsTextLength
offset := x & (1<<nodesBitsTextOffset - 1)
return text[offset : offset+length]
}
// EffectiveTLDPlusOne returns the effective top level domain plus one more
// label. For example, the eTLD+1 for "foo.bar.golang.org" is "golang.org".
func EffectiveTLDPlusOne(domain string) (string, error) {
if strings.HasPrefix(domain, ".") || strings.HasSuffix(domain, ".") || strings.Contains(domain, "..") {
return "", fmt.Errorf("publicsuffix: empty label in domain %q", domain)
}
suffix, _ := PublicSuffix(domain)
if len(domain) <= len(suffix) {
return "", fmt.Errorf("publicsuffix: cannot derive eTLD+1 for domain %q", domain)
}
i := len(domain) - len(suffix) - 1
if domain[i] != '.' {
return "", fmt.Errorf("publicsuffix: invalid public suffix %q for domain %q", suffix, domain)
}
return domain[1+strings.LastIndex(domain[:i], "."):], nil
}
type uint32String string
func (u uint32String) get(i uint32) uint32 {
off := i * 4
return (uint32(u[off])<<24 |
uint32(u[off+1])<<16 |
uint32(u[off+2])<<8 |
uint32(u[off+3]))
}
type uint40String string
func (u uint40String) get(i uint32) uint64 {
off := uint64(i * (nodesBits / 8))
return uint64(u[off])<<32 |
uint64(u[off+1])<<24 |
uint64(u[off+2])<<16 |
uint64(u[off+3])<<8 |
uint64(u[off+4])
}

View File

@@ -1,70 +0,0 @@
// generated by go run gen.go; DO NOT EDIT
package publicsuffix
import _ "embed"
const version = "publicsuffix.org's public_suffix_list.dat, git revision 63cbc63d470d7b52c35266aa96c4c98c96ec499c (2023-08-03T10:01:25Z)"
const (
nodesBits = 40
nodesBitsChildren = 10
nodesBitsICANN = 1
nodesBitsTextOffset = 16
nodesBitsTextLength = 6
childrenBitsWildcard = 1
childrenBitsNodeType = 2
childrenBitsHi = 14
childrenBitsLo = 14
)
const (
nodeTypeNormal = 0
nodeTypeException = 1
nodeTypeParentOnly = 2
)
// numTLD is the number of top level domains.
const numTLD = 1474
// text is the combined text of all labels.
//
//go:embed data/text
var text string
// nodes is the list of nodes. Each node is represented as a 40-bit integer,
// which encodes the node's children, wildcard bit and node type (as an index
// into the children array), ICANN bit and text.
//
// The layout within the node, from MSB to LSB, is:
//
// [ 7 bits] unused
// [10 bits] children index
// [ 1 bits] ICANN bit
// [16 bits] text index
// [ 6 bits] text length
//
//go:embed data/nodes
var nodes uint40String
// children is the list of nodes' children, the parent's wildcard bit and the
// parent's node type. If a node has no children then their children index
// will be in the range [0, 6), depending on the wildcard bit and node type.
//
// The layout within the uint32, from MSB to LSB, is:
//
// [ 1 bits] unused
// [ 1 bits] wildcard bit
// [ 2 bits] node type
// [14 bits] high nodes index (exclusive) of children
// [14 bits] low nodes index (inclusive) of children
//
//go:embed data/children
var children uint32String
// max children 743 (capacity 1023)
// max text offset 30876 (capacity 65535)
// max text length 31 (capacity 63)
// max hi 9322 (capacity 16383)
// max lo 9317 (capacity 16383)

532
vendor/golang.org/x/net/trace/events.go generated vendored Normal file
View File

@@ -0,0 +1,532 @@
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package trace
import (
"bytes"
"fmt"
"html/template"
"io"
"log"
"net/http"
"runtime"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"text/tabwriter"
"time"
)
const maxEventsPerLog = 100
type bucket struct {
MaxErrAge time.Duration
String string
}
var buckets = []bucket{
{0, "total"},
{10 * time.Second, "errs<10s"},
{1 * time.Minute, "errs<1m"},
{10 * time.Minute, "errs<10m"},
{1 * time.Hour, "errs<1h"},
{10 * time.Hour, "errs<10h"},
{24000 * time.Hour, "errors"},
}
// RenderEvents renders the HTML page typically served at /debug/events.
// It does not do any auth checking. The request may be nil.
//
// Most users will use the Events handler.
func RenderEvents(w http.ResponseWriter, req *http.Request, sensitive bool) {
now := time.Now()
data := &struct {
Families []string // family names
Buckets []bucket
Counts [][]int // eventLog count per family/bucket
// Set when a bucket has been selected.
Family string
Bucket int
EventLogs eventLogs
Expanded bool
}{
Buckets: buckets,
}
data.Families = make([]string, 0, len(families))
famMu.RLock()
for name := range families {
data.Families = append(data.Families, name)
}
famMu.RUnlock()
sort.Strings(data.Families)
// Count the number of eventLogs in each family for each error age.
data.Counts = make([][]int, len(data.Families))
for i, name := range data.Families {
// TODO(sameer): move this loop under the family lock.
f := getEventFamily(name)
data.Counts[i] = make([]int, len(data.Buckets))
for j, b := range data.Buckets {
data.Counts[i][j] = f.Count(now, b.MaxErrAge)
}
}
if req != nil {
var ok bool
data.Family, data.Bucket, ok = parseEventsArgs(req)
if !ok {
// No-op
} else {
data.EventLogs = getEventFamily(data.Family).Copy(now, buckets[data.Bucket].MaxErrAge)
}
if data.EventLogs != nil {
defer data.EventLogs.Free()
sort.Sort(data.EventLogs)
}
if exp, err := strconv.ParseBool(req.FormValue("exp")); err == nil {
data.Expanded = exp
}
}
famMu.RLock()
defer famMu.RUnlock()
if err := eventsTmpl().Execute(w, data); err != nil {
log.Printf("net/trace: Failed executing template: %v", err)
}
}
func parseEventsArgs(req *http.Request) (fam string, b int, ok bool) {
fam, bStr := req.FormValue("fam"), req.FormValue("b")
if fam == "" || bStr == "" {
return "", 0, false
}
b, err := strconv.Atoi(bStr)
if err != nil || b < 0 || b >= len(buckets) {
return "", 0, false
}
return fam, b, true
}
// An EventLog provides a log of events associated with a specific object.
type EventLog interface {
// Printf formats its arguments with fmt.Sprintf and adds the
// result to the event log.
Printf(format string, a ...interface{})
// Errorf is like Printf, but it marks this event as an error.
Errorf(format string, a ...interface{})
// Finish declares that this event log is complete.
// The event log should not be used after calling this method.
Finish()
}
// NewEventLog returns a new EventLog with the specified family name
// and title.
func NewEventLog(family, title string) EventLog {
el := newEventLog()
el.ref()
el.Family, el.Title = family, title
el.Start = time.Now()
el.events = make([]logEntry, 0, maxEventsPerLog)
el.stack = make([]uintptr, 32)
n := runtime.Callers(2, el.stack)
el.stack = el.stack[:n]
getEventFamily(family).add(el)
return el
}
func (el *eventLog) Finish() {
getEventFamily(el.Family).remove(el)
el.unref() // matches ref in New
}
var (
famMu sync.RWMutex
families = make(map[string]*eventFamily) // family name => family
)
func getEventFamily(fam string) *eventFamily {
famMu.Lock()
defer famMu.Unlock()
f := families[fam]
if f == nil {
f = &eventFamily{}
families[fam] = f
}
return f
}
type eventFamily struct {
mu sync.RWMutex
eventLogs eventLogs
}
func (f *eventFamily) add(el *eventLog) {
f.mu.Lock()
f.eventLogs = append(f.eventLogs, el)
f.mu.Unlock()
}
func (f *eventFamily) remove(el *eventLog) {
f.mu.Lock()
defer f.mu.Unlock()
for i, el0 := range f.eventLogs {
if el == el0 {
copy(f.eventLogs[i:], f.eventLogs[i+1:])
f.eventLogs = f.eventLogs[:len(f.eventLogs)-1]
return
}
}
}
func (f *eventFamily) Count(now time.Time, maxErrAge time.Duration) (n int) {
f.mu.RLock()
defer f.mu.RUnlock()
for _, el := range f.eventLogs {
if el.hasRecentError(now, maxErrAge) {
n++
}
}
return
}
func (f *eventFamily) Copy(now time.Time, maxErrAge time.Duration) (els eventLogs) {
f.mu.RLock()
defer f.mu.RUnlock()
els = make(eventLogs, 0, len(f.eventLogs))
for _, el := range f.eventLogs {
if el.hasRecentError(now, maxErrAge) {
el.ref()
els = append(els, el)
}
}
return
}
type eventLogs []*eventLog
// Free calls unref on each element of the list.
func (els eventLogs) Free() {
for _, el := range els {
el.unref()
}
}
// eventLogs may be sorted in reverse chronological order.
func (els eventLogs) Len() int { return len(els) }
func (els eventLogs) Less(i, j int) bool { return els[i].Start.After(els[j].Start) }
func (els eventLogs) Swap(i, j int) { els[i], els[j] = els[j], els[i] }
// A logEntry is a timestamped log entry in an event log.
type logEntry struct {
When time.Time
Elapsed time.Duration // since previous event in log
NewDay bool // whether this event is on a different day to the previous event
What string
IsErr bool
}
// WhenString returns a string representation of the elapsed time of the event.
// It will include the date if midnight was crossed.
func (e logEntry) WhenString() string {
if e.NewDay {
return e.When.Format("2006/01/02 15:04:05.000000")
}
return e.When.Format("15:04:05.000000")
}
// An eventLog represents an active event log.
type eventLog struct {
// Family is the top-level grouping of event logs to which this belongs.
Family string
// Title is the title of this event log.
Title string
// Timing information.
Start time.Time
// Call stack where this event log was created.
stack []uintptr
// Append-only sequence of events.
//
// TODO(sameer): change this to a ring buffer to avoid the array copy
// when we hit maxEventsPerLog.
mu sync.RWMutex
events []logEntry
LastErrorTime time.Time
discarded int
refs int32 // how many buckets this is in
}
func (el *eventLog) reset() {
// Clear all but the mutex. Mutexes may not be copied, even when unlocked.
el.Family = ""
el.Title = ""
el.Start = time.Time{}
el.stack = nil
el.events = nil
el.LastErrorTime = time.Time{}
el.discarded = 0
el.refs = 0
}
func (el *eventLog) hasRecentError(now time.Time, maxErrAge time.Duration) bool {
if maxErrAge == 0 {
return true
}
el.mu.RLock()
defer el.mu.RUnlock()
return now.Sub(el.LastErrorTime) < maxErrAge
}
// delta returns the elapsed time since the last event or the log start,
// and whether it spans midnight.
// L >= el.mu
func (el *eventLog) delta(t time.Time) (time.Duration, bool) {
if len(el.events) == 0 {
return t.Sub(el.Start), false
}
prev := el.events[len(el.events)-1].When
return t.Sub(prev), prev.Day() != t.Day()
}
func (el *eventLog) Printf(format string, a ...interface{}) {
el.printf(false, format, a...)
}
func (el *eventLog) Errorf(format string, a ...interface{}) {
el.printf(true, format, a...)
}
func (el *eventLog) printf(isErr bool, format string, a ...interface{}) {
e := logEntry{When: time.Now(), IsErr: isErr, What: fmt.Sprintf(format, a...)}
el.mu.Lock()
e.Elapsed, e.NewDay = el.delta(e.When)
if len(el.events) < maxEventsPerLog {
el.events = append(el.events, e)
} else {
// Discard the oldest event.
if el.discarded == 0 {
// el.discarded starts at two to count for the event it
// is replacing, plus the next one that we are about to
// drop.
el.discarded = 2
} else {
el.discarded++
}
// TODO(sameer): if this causes allocations on a critical path,
// change eventLog.What to be a fmt.Stringer, as in trace.go.
el.events[0].What = fmt.Sprintf("(%d events discarded)", el.discarded)
// The timestamp of the discarded meta-event should be
// the time of the last event it is representing.
el.events[0].When = el.events[1].When
copy(el.events[1:], el.events[2:])
el.events[maxEventsPerLog-1] = e
}
if e.IsErr {
el.LastErrorTime = e.When
}
el.mu.Unlock()
}
func (el *eventLog) ref() {
atomic.AddInt32(&el.refs, 1)
}
func (el *eventLog) unref() {
if atomic.AddInt32(&el.refs, -1) == 0 {
freeEventLog(el)
}
}
func (el *eventLog) When() string {
return el.Start.Format("2006/01/02 15:04:05.000000")
}
func (el *eventLog) ElapsedTime() string {
elapsed := time.Since(el.Start)
return fmt.Sprintf("%.6f", elapsed.Seconds())
}
func (el *eventLog) Stack() string {
buf := new(bytes.Buffer)
tw := tabwriter.NewWriter(buf, 1, 8, 1, '\t', 0)
printStackRecord(tw, el.stack)
tw.Flush()
return buf.String()
}
// printStackRecord prints the function + source line information
// for a single stack trace.
// Adapted from runtime/pprof/pprof.go.
func printStackRecord(w io.Writer, stk []uintptr) {
for _, pc := range stk {
f := runtime.FuncForPC(pc)
if f == nil {
continue
}
file, line := f.FileLine(pc)
name := f.Name()
// Hide runtime.goexit and any runtime functions at the beginning.
if strings.HasPrefix(name, "runtime.") {
continue
}
fmt.Fprintf(w, "# %s\t%s:%d\n", name, file, line)
}
}
func (el *eventLog) Events() []logEntry {
el.mu.RLock()
defer el.mu.RUnlock()
return el.events
}
// freeEventLogs is a freelist of *eventLog
var freeEventLogs = make(chan *eventLog, 1000)
// newEventLog returns a event log ready to use.
func newEventLog() *eventLog {
select {
case el := <-freeEventLogs:
return el
default:
return new(eventLog)
}
}
// freeEventLog adds el to freeEventLogs if there's room.
// This is non-blocking.
func freeEventLog(el *eventLog) {
el.reset()
select {
case freeEventLogs <- el:
default:
}
}
var eventsTmplCache *template.Template
var eventsTmplOnce sync.Once
func eventsTmpl() *template.Template {
eventsTmplOnce.Do(func() {
eventsTmplCache = template.Must(template.New("events").Funcs(template.FuncMap{
"elapsed": elapsed,
"trimSpace": strings.TrimSpace,
}).Parse(eventsHTML))
})
return eventsTmplCache
}
const eventsHTML = `
<html>
<head>
<title>events</title>
</head>
<style type="text/css">
body {
font-family: sans-serif;
}
table#req-status td.family {
padding-right: 2em;
}
table#req-status td.active {
padding-right: 1em;
}
table#req-status td.empty {
color: #aaa;
}
table#reqs {
margin-top: 1em;
}
table#reqs tr.first {
{{if $.Expanded}}font-weight: bold;{{end}}
}
table#reqs td {
font-family: monospace;
}
table#reqs td.when {
text-align: right;
white-space: nowrap;
}
table#reqs td.elapsed {
padding: 0 0.5em;
text-align: right;
white-space: pre;
width: 10em;
}
address {
font-size: smaller;
margin-top: 5em;
}
</style>
<body>
<h1>/debug/events</h1>
<table id="req-status">
{{range $i, $fam := .Families}}
<tr>
<td class="family">{{$fam}}</td>
{{range $j, $bucket := $.Buckets}}
{{$n := index $.Counts $i $j}}
<td class="{{if not $bucket.MaxErrAge}}active{{end}}{{if not $n}}empty{{end}}">
{{if $n}}<a href="?fam={{$fam}}&b={{$j}}{{if $.Expanded}}&exp=1{{end}}">{{end}}
[{{$n}} {{$bucket.String}}]
{{if $n}}</a>{{end}}
</td>
{{end}}
</tr>{{end}}
</table>
{{if $.EventLogs}}
<hr />
<h3>Family: {{$.Family}}</h3>
{{if $.Expanded}}<a href="?fam={{$.Family}}&b={{$.Bucket}}">{{end}}
[Summary]{{if $.Expanded}}</a>{{end}}
{{if not $.Expanded}}<a href="?fam={{$.Family}}&b={{$.Bucket}}&exp=1">{{end}}
[Expanded]{{if not $.Expanded}}</a>{{end}}
<table id="reqs">
<tr><th>When</th><th>Elapsed</th></tr>
{{range $el := $.EventLogs}}
<tr class="first">
<td class="when">{{$el.When}}</td>
<td class="elapsed">{{$el.ElapsedTime}}</td>
<td>{{$el.Title}}
</tr>
{{if $.Expanded}}
<tr>
<td class="when"></td>
<td class="elapsed"></td>
<td><pre>{{$el.Stack|trimSpace}}</pre></td>
</tr>
{{range $el.Events}}
<tr>
<td class="when">{{.WhenString}}</td>
<td class="elapsed">{{elapsed .Elapsed}}</td>
<td>.{{if .IsErr}}E{{else}}.{{end}}. {{.What}}</td>
</tr>
{{end}}
{{end}}
{{end}}
</table>
{{end}}
</body>
</html>
`

365
vendor/golang.org/x/net/trace/histogram.go generated vendored Normal file
View File

@@ -0,0 +1,365 @@
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package trace
// This file implements histogramming for RPC statistics collection.
import (
"bytes"
"fmt"
"html/template"
"log"
"math"
"sync"
"golang.org/x/net/internal/timeseries"
)
const (
bucketCount = 38
)
// histogram keeps counts of values in buckets that are spaced
// out in powers of 2: 0-1, 2-3, 4-7...
// histogram implements timeseries.Observable
type histogram struct {
sum int64 // running total of measurements
sumOfSquares float64 // square of running total
buckets []int64 // bucketed values for histogram
value int // holds a single value as an optimization
valueCount int64 // number of values recorded for single value
}
// addMeasurement records a value measurement observation to the histogram.
func (h *histogram) addMeasurement(value int64) {
// TODO: assert invariant
h.sum += value
h.sumOfSquares += float64(value) * float64(value)
bucketIndex := getBucket(value)
if h.valueCount == 0 || (h.valueCount > 0 && h.value == bucketIndex) {
h.value = bucketIndex
h.valueCount++
} else {
h.allocateBuckets()
h.buckets[bucketIndex]++
}
}
func (h *histogram) allocateBuckets() {
if h.buckets == nil {
h.buckets = make([]int64, bucketCount)
h.buckets[h.value] = h.valueCount
h.value = 0
h.valueCount = -1
}
}
func log2(i int64) int {
n := 0
for ; i >= 0x100; i >>= 8 {
n += 8
}
for ; i > 0; i >>= 1 {
n += 1
}
return n
}
func getBucket(i int64) (index int) {
index = log2(i) - 1
if index < 0 {
index = 0
}
if index >= bucketCount {
index = bucketCount - 1
}
return
}
// Total returns the number of recorded observations.
func (h *histogram) total() (total int64) {
if h.valueCount >= 0 {
total = h.valueCount
}
for _, val := range h.buckets {
total += int64(val)
}
return
}
// Average returns the average value of recorded observations.
func (h *histogram) average() float64 {
t := h.total()
if t == 0 {
return 0
}
return float64(h.sum) / float64(t)
}
// Variance returns the variance of recorded observations.
func (h *histogram) variance() float64 {
t := float64(h.total())
if t == 0 {
return 0
}
s := float64(h.sum) / t
return h.sumOfSquares/t - s*s
}
// StandardDeviation returns the standard deviation of recorded observations.
func (h *histogram) standardDeviation() float64 {
return math.Sqrt(h.variance())
}
// PercentileBoundary estimates the value that the given fraction of recorded
// observations are less than.
func (h *histogram) percentileBoundary(percentile float64) int64 {
total := h.total()
// Corner cases (make sure result is strictly less than Total())
if total == 0 {
return 0
} else if total == 1 {
return int64(h.average())
}
percentOfTotal := round(float64(total) * percentile)
var runningTotal int64
for i := range h.buckets {
value := h.buckets[i]
runningTotal += value
if runningTotal == percentOfTotal {
// We hit an exact bucket boundary. If the next bucket has data, it is a
// good estimate of the value. If the bucket is empty, we interpolate the
// midpoint between the next bucket's boundary and the next non-zero
// bucket. If the remaining buckets are all empty, then we use the
// boundary for the next bucket as the estimate.
j := uint8(i + 1)
min := bucketBoundary(j)
if runningTotal < total {
for h.buckets[j] == 0 {
j++
}
}
max := bucketBoundary(j)
return min + round(float64(max-min)/2)
} else if runningTotal > percentOfTotal {
// The value is in this bucket. Interpolate the value.
delta := runningTotal - percentOfTotal
percentBucket := float64(value-delta) / float64(value)
bucketMin := bucketBoundary(uint8(i))
nextBucketMin := bucketBoundary(uint8(i + 1))
bucketSize := nextBucketMin - bucketMin
return bucketMin + round(percentBucket*float64(bucketSize))
}
}
return bucketBoundary(bucketCount - 1)
}
// Median returns the estimated median of the observed values.
func (h *histogram) median() int64 {
return h.percentileBoundary(0.5)
}
// Add adds other to h.
func (h *histogram) Add(other timeseries.Observable) {
o := other.(*histogram)
if o.valueCount == 0 {
// Other histogram is empty
} else if h.valueCount >= 0 && o.valueCount > 0 && h.value == o.value {
// Both have a single bucketed value, aggregate them
h.valueCount += o.valueCount
} else {
// Two different values necessitate buckets in this histogram
h.allocateBuckets()
if o.valueCount >= 0 {
h.buckets[o.value] += o.valueCount
} else {
for i := range h.buckets {
h.buckets[i] += o.buckets[i]
}
}
}
h.sumOfSquares += o.sumOfSquares
h.sum += o.sum
}
// Clear resets the histogram to an empty state, removing all observed values.
func (h *histogram) Clear() {
h.buckets = nil
h.value = 0
h.valueCount = 0
h.sum = 0
h.sumOfSquares = 0
}
// CopyFrom copies from other, which must be a *histogram, into h.
func (h *histogram) CopyFrom(other timeseries.Observable) {
o := other.(*histogram)
if o.valueCount == -1 {
h.allocateBuckets()
copy(h.buckets, o.buckets)
}
h.sum = o.sum
h.sumOfSquares = o.sumOfSquares
h.value = o.value
h.valueCount = o.valueCount
}
// Multiply scales the histogram by the specified ratio.
func (h *histogram) Multiply(ratio float64) {
if h.valueCount == -1 {
for i := range h.buckets {
h.buckets[i] = int64(float64(h.buckets[i]) * ratio)
}
} else {
h.valueCount = int64(float64(h.valueCount) * ratio)
}
h.sum = int64(float64(h.sum) * ratio)
h.sumOfSquares = h.sumOfSquares * ratio
}
// New creates a new histogram.
func (h *histogram) New() timeseries.Observable {
r := new(histogram)
r.Clear()
return r
}
func (h *histogram) String() string {
return fmt.Sprintf("%d, %f, %d, %d, %v",
h.sum, h.sumOfSquares, h.value, h.valueCount, h.buckets)
}
// round returns the closest int64 to the argument
func round(in float64) int64 {
return int64(math.Floor(in + 0.5))
}
// bucketBoundary returns the first value in the bucket.
func bucketBoundary(bucket uint8) int64 {
if bucket == 0 {
return 0
}
return 1 << bucket
}
// bucketData holds data about a specific bucket for use in distTmpl.
type bucketData struct {
Lower, Upper int64
N int64
Pct, CumulativePct float64
GraphWidth int
}
// data holds data about a Distribution for use in distTmpl.
type data struct {
Buckets []*bucketData
Count, Median int64
Mean, StandardDeviation float64
}
// maxHTMLBarWidth is the maximum width of the HTML bar for visualizing buckets.
const maxHTMLBarWidth = 350.0
// newData returns data representing h for use in distTmpl.
func (h *histogram) newData() *data {
// Force the allocation of buckets to simplify the rendering implementation
h.allocateBuckets()
// We scale the bars on the right so that the largest bar is
// maxHTMLBarWidth pixels in width.
maxBucket := int64(0)
for _, n := range h.buckets {
if n > maxBucket {
maxBucket = n
}
}
total := h.total()
barsizeMult := maxHTMLBarWidth / float64(maxBucket)
var pctMult float64
if total == 0 {
pctMult = 1.0
} else {
pctMult = 100.0 / float64(total)
}
buckets := make([]*bucketData, len(h.buckets))
runningTotal := int64(0)
for i, n := range h.buckets {
if n == 0 {
continue
}
runningTotal += n
var upperBound int64
if i < bucketCount-1 {
upperBound = bucketBoundary(uint8(i + 1))
} else {
upperBound = math.MaxInt64
}
buckets[i] = &bucketData{
Lower: bucketBoundary(uint8(i)),
Upper: upperBound,
N: n,
Pct: float64(n) * pctMult,
CumulativePct: float64(runningTotal) * pctMult,
GraphWidth: int(float64(n) * barsizeMult),
}
}
return &data{
Buckets: buckets,
Count: total,
Median: h.median(),
Mean: h.average(),
StandardDeviation: h.standardDeviation(),
}
}
func (h *histogram) html() template.HTML {
buf := new(bytes.Buffer)
if err := distTmpl().Execute(buf, h.newData()); err != nil {
buf.Reset()
log.Printf("net/trace: couldn't execute template: %v", err)
}
return template.HTML(buf.String())
}
var distTmplCache *template.Template
var distTmplOnce sync.Once
func distTmpl() *template.Template {
distTmplOnce.Do(func() {
// Input: data
distTmplCache = template.Must(template.New("distTmpl").Parse(`
<table>
<tr>
<td style="padding:0.25em">Count: {{.Count}}</td>
<td style="padding:0.25em">Mean: {{printf "%.0f" .Mean}}</td>
<td style="padding:0.25em">StdDev: {{printf "%.0f" .StandardDeviation}}</td>
<td style="padding:0.25em">Median: {{.Median}}</td>
</tr>
</table>
<hr>
<table>
{{range $b := .Buckets}}
{{if $b}}
<tr>
<td style="padding:0 0 0 0.25em">[</td>
<td style="text-align:right;padding:0 0.25em">{{.Lower}},</td>
<td style="text-align:right;padding:0 0.25em">{{.Upper}})</td>
<td style="text-align:right;padding:0 0.25em">{{.N}}</td>
<td style="text-align:right;padding:0 0.25em">{{printf "%#.3f" .Pct}}%</td>
<td style="text-align:right;padding:0 0.25em">{{printf "%#.3f" .CumulativePct}}%</td>
<td><div style="background-color: blue; height: 1em; width: {{.GraphWidth}};"></div></td>
</tr>
{{end}}
{{end}}
</table>
`))
})
return distTmplCache
}

1130
vendor/golang.org/x/net/trace/trace.go generated vendored Normal file

File diff suppressed because it is too large Load Diff