Use Intel optimized code (ssse3, avx, avx2) to calculate sha256

pkg/utils/crypto/sha256/sha256.go

@@ -1,59 +1,162 @@
-// +build amd64
-
+// Package sha256 SHA256SSE3, SHA256AVX, SHA256AVX2
 package sha256
 
-// #include <stdint.h>
-// void sha256_transform_avx (uint8_t *input_data, uint32_t digest[8], uint64_t num_blks);
-// void sha256_transform_ssse3 (uint8_t *input_data, uint32_t digest[8], uint64_t num_blks);
-// void sha256_transform_rorx (uint8_t *input_data, uint32_t digest[8], uint64_t num_blks);
-// #define SHA256_DIGEST_SIZE      32
-// #define SHA256_BLOCK_SIZE       64
-// #define SHA256_H0       0x6a09e667UL
-// #define SHA256_H1       0xbb67ae85UL
-// #define SHA256_H2       0x3c6ef372UL
-// #define SHA256_H3       0xa54ff53aUL
-// #define SHA256_H4       0x510e527fUL
-// #define SHA256_H5       0x9b05688cUL
-// #define SHA256_H6       0x1f83d9abUL
-// #define SHA256_H7       0x5be0cd19UL
-import "C"
 import (
-	gosha256 "crypto/sha256"
+	"hash"
 	"io"
+
+	"github.com/minio-io/minio/pkg/utils/cpu"
+)
+
+// The size of a SHA256 checksum in bytes.
+const Size = 32
+
+// The blocksize of SHA256 in bytes.
+const BlockSize = 64
+
+const (
+	chunk = 64
+	init0 = 0x6A09E667
+	init1 = 0xBB67AE85
+	init2 = 0x3C6EF372
+	init3 = 0xA54FF53A
+	init4 = 0x510E527F
+	init5 = 0x9B05688C
+	init6 = 0x1F83D9AB
+	init7 = 0x5BE0CD19
 )
 
-/*
-func Sha256(buffer []byte) ([]uint32, error) {
+// digest represents the partial evaluation of a checksum.
+type digest struct {
+	h   [8]uint32
+	x   [chunk]byte
+	nx  int
+	len uint64
+}
+
+func (d *digest) Reset() {
+	d.h[0] = init0
+	d.h[1] = init1
+	d.h[2] = init2
+	d.h[3] = init3
+	d.h[4] = init4
+	d.h[5] = init5
+	d.h[6] = init6
+	d.h[7] = init7
+	d.nx = 0
+	d.len = 0
+}
 
-	if cpu.HasSSE41() {
-		C.sha256_transform_ssse3()
-		return 0, nil
+func block(dig *digest, p []byte) {
+	switch true {
+	case cpu.HasAVX2() == true:
+		blockAVX2(dig, p)
+	case cpu.HasAVX() == true:
+		blockAVX(dig, p)
+	case cpu.HasSSE41() == true:
+		blockSSE(dig, p)
 	}
+}
+
+// New returns a new hash.Hash computing the SHA256 checksum.
+func New() hash.Hash {
+	d := new(digest)
+	d.Reset()
+	return d
+}
 
-	if cpu.HasAVX() {
-		C.sha256_transform_avx()
-		return 0, nil
+func (d *digest) Size() int {
+	return Size
+}
+
+func (d *digest) BlockSize() int { return BlockSize }
+
+func (d *digest) Write(p []byte) (nn int, err error) {
+	nn = len(p)
+	d.len += uint64(nn)
+	if d.nx > 0 {
+		n := copy(d.x[d.nx:], p)
+		d.nx += n
+		if d.nx == chunk {
+			block(d, d.x[:])
+			d.nx = 0
+		}
+		p = p[n:]
+	}
+	if len(p) >= chunk {
+		n := len(p) &^ (chunk - 1)
+		block(d, p[:n])
+		p = p[n:]
 	}
+	if len(p) > 0 {
+		d.nx = copy(d.x[:], p)
+	}
+	return
+}
+
+func (d0 *digest) Sum(in []byte) []byte {
+	// Make a copy of d0 so that caller can keep writing and summing.
+	d := *d0
+	hash := d.checkSum()
+	return append(in, hash[:]...)
+}
+
+func (d *digest) checkSum() [Size]byte {
+	len := d.len
+	// Padding.  Add a 1 bit and 0 bits until 56 bytes mod 64.
+	var tmp [64]byte
+	tmp[0] = 0x80
+	if len%64 < 56 {
+		d.Write(tmp[0 : 56-len%64])
+	} else {
+		d.Write(tmp[0 : 64+56-len%64])
+	}
+
+	// Length in bits.
+	len <<= 3
+	for i := uint(0); i < 8; i++ {
+		tmp[i] = byte(len >> (56 - 8*i))
+	}
+	d.Write(tmp[0:8])
 
-	if cpu.HasAVX2() {
-		C.sha256_transform_rorx()
-		return 0, nil
+	if d.nx != 0 {
+		panic("d.nx != 0")
 	}
+
+	h := d.h[:]
+
+	var digest [Size]byte
+	for i, s := range h {
+		digest[i*4] = byte(s >> 24)
+		digest[i*4+1] = byte(s >> 16)
+		digest[i*4+2] = byte(s >> 8)
+		digest[i*4+3] = byte(s)
+	}
+
+	return digest
+}
+
+// Convenience functions
+
+func Sum256(data []byte) [Size]byte {
+	var d digest
+	d.Reset()
+	d.Write(data)
+	return d.checkSum()
 }
-*/
 
 func Sum(reader io.Reader) ([]byte, error) {
-	hash := gosha256.New()
+	h := New()
 	var err error
 	for err == nil {
 		length := 0
 		byteBuffer := make([]byte, 1024*1024)
 		length, err = reader.Read(byteBuffer)
 		byteBuffer = byteBuffer[0:length]
-		hash.Write(byteBuffer)
+		h.Write(byteBuffer)
 	}
 	if err != io.EOF {
 		return nil, err
 	}
-	return hash.Sum(nil), nil
+	return h.Sum(nil), nil
 }

pkg/utils/crypto/sha256/sha256_gen.go

@@ -0,0 +1,65 @@
+// +build ignore
+
+package main
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"fmt"
+	"io"
+	"os"
+	"time"
+
+	sha256intel "github.com/minio-io/minio/pkg/utils/crypto/sha256"
+)
+
+func SumIntel(reader io.Reader) ([]byte, error) {
+	h := sha256intel.New()
+	var err error
+	for err == nil {
+		length := 0
+		byteBuffer := make([]byte, 1024*1024)
+		length, err = reader.Read(byteBuffer)
+		byteBuffer = byteBuffer[0:length]
+		h.Write(byteBuffer)
+	}
+	if err != io.EOF {
+		return nil, err
+	}
+	return h.Sum(nil), nil
+}
+
+func Sum(reader io.Reader) ([]byte, error) {
+	k := sha256.New()
+	var err error
+	for err == nil {
+		length := 0
+		byteBuffer := make([]byte, 1024*1024)
+		length, err = reader.Read(byteBuffer)
+		byteBuffer = byteBuffer[0:length]
+		k.Write(byteBuffer)
+	}
+	if err != io.EOF {
+		return nil, err
+	}
+	return k.Sum(nil), nil
+}
+
+func main() {
+	fmt.Println("-- start")
+
+	file1, _ := os.Open("filename1")
+	defer file1.Close()
+	stark := time.Now()
+	sum, _ := Sum(file1)
+	endk := time.Since(stark)
+
+	file2, _ := os.Open("filename2")
+	defer file2.Close()
+	starth := time.Now()
+	sumSSE, _ := SumIntel(file2)
+	endh := time.Since(starth)
+
+	fmt.Println("std(", endk, ")", "ssse3(", endh, ")")
+	fmt.Println(hex.EncodeToString(sum), hex.EncodeToString(sumSSE))
+}

pkg/utils/crypto/sha256/sha256_test.go

@@ -1,23 +1,112 @@
+// SHA256 hash algorithm.  See FIPS 180-2.
+
 package sha256
 
 import (
-	"bytes"
-	"encoding/hex"
+	"fmt"
+	"io"
 	"testing"
-
-	. "gopkg.in/check.v1"
 )
 
-func Test(t *testing.T) { TestingT(t) }
+type sha256Test struct {
+	out string
+	in  string
+}
+
+var golden = []sha256Test{
+	{"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", ""},
+	{"ca978112ca1bbdcafac231b39a23dc4da786eff8147c4e72b9807785afee48bb", "a"},
+	{"fb8e20fc2e4c3f248c60c39bd652f3c1347298bb977b8b4d5903b85055620603", "ab"},
+	{"ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad", "abc"},
+	{"88d4266fd4e6338d13b845fcf289579d209c897823b9217da3e161936f031589", "abcd"},
+	{"36bbe50ed96841d10443bcb670d6554f0a34b761be67ec9c4a8ad2c0c44ca42c", "abcde"},
+	{"bef57ec7f53a6d40beb640a780a639c83bc29ac8a9816f1fc6c5c6dcd93c4721", "abcdef"},
+	{"7d1a54127b222502f5b79b5fb0803061152a44f92b37e23c6527baf665d4da9a", "abcdefg"},
+	{"9c56cc51b374c3ba189210d5b6d4bf57790d351c96c47c02190ecf1e430635ab", "abcdefgh"},
+	{"19cc02f26df43cc571bc9ed7b0c4d29224a3ec229529221725ef76d021c8326f", "abcdefghi"},
+	{"72399361da6a7754fec986dca5b7cbaf1c810a28ded4abaf56b2106d06cb78b0", "abcdefghij"},
+	{"a144061c271f152da4d151034508fed1c138b8c976339de229c3bb6d4bbb4fce", "Discard medicine more than two years old."},
+	{"6dae5caa713a10ad04b46028bf6dad68837c581616a1589a265a11288d4bb5c4", "He who has a shady past knows that nice guys finish last."},
+	{"ae7a702a9509039ddbf29f0765e70d0001177914b86459284dab8b348c2dce3f", "I wouldn't marry him with a ten foot pole."},
+	{"6748450b01c568586715291dfa3ee018da07d36bb7ea6f180c1af6270215c64f", "Free! Free!/A trip/to Mars/for 900/empty jars/Burma Shave"},
+	{"14b82014ad2b11f661b5ae6a99b75105c2ffac278cd071cd6c05832793635774", "The days of the digital watch are numbered.  -Tom Stoppard"},
+	{"7102cfd76e2e324889eece5d6c41921b1e142a4ac5a2692be78803097f6a48d8", "Nepal premier won't resign."},
+	{"23b1018cd81db1d67983c5f7417c44da9deb582459e378d7a068552ea649dc9f", "For every action there is an equal and opposite government program."},
+	{"8001f190dfb527261c4cfcab70c98e8097a7a1922129bc4096950e57c7999a5a", "His money is twice tainted: 'taint yours and 'taint mine."},
+	{"8c87deb65505c3993eb24b7a150c4155e82eee6960cf0c3a8114ff736d69cad5", "There is no reason for any individual to have a computer in their home. -Ken Olsen, 1977"},
+	{"bfb0a67a19cdec3646498b2e0f751bddc41bba4b7f30081b0b932aad214d16d7", "It's a tiny change to the code and not completely disgusting. - Bob Manchek"},
+	{"7f9a0b9bf56332e19f5a0ec1ad9c1425a153da1c624868fda44561d6b74daf36", "size:  a.out:  bad magic"},
+	{"b13f81b8aad9e3666879af19886140904f7f429ef083286195982a7588858cfc", "The major problem is with sendmail.  -Mark Horton"},
+	{"b26c38d61519e894480c70c8374ea35aa0ad05b2ae3d6674eec5f52a69305ed4", "Give me a rock, paper and scissors and I will move the world.  CCFestoon"},
+	{"049d5e26d4f10222cd841a119e38bd8d2e0d1129728688449575d4ff42b842c1", "If the enemy is within range, then so are you."},
+	{"0e116838e3cc1c1a14cd045397e29b4d087aa11b0853fc69ec82e90330d60949", "It's well we cannot hear the screams/That we create in others' dreams."},
+	{"4f7d8eb5bcf11de2a56b971021a444aa4eafd6ecd0f307b5109e4e776cd0fe46", "You remind me of a TV show, but that's all right: I watch it anyway."},
+	{"61c0cc4c4bd8406d5120b3fb4ebc31ce87667c162f29468b3c779675a85aebce", "C is as portable as Stonehedge!!"},
+	{"1fb2eb3688093c4a3f80cd87a5547e2ce940a4f923243a79a2a1e242220693ac", "Even if I could be Shakespeare, I think I should still choose to be Faraday. - A. Huxley"},
+	{"395585ce30617b62c80b93e8208ce866d4edc811a177fdb4b82d3911d8696423", "The fugacity of a constituent in a mixture of gases at a given temperature is proportional to its mole fraction.  Lewis-Randall Rule"},
+	{"4f9b189a13d030838269dce846b16a1ce9ce81fe63e65de2f636863336a98fe6", "How can you write a big system without C++?  -Paul Glick"},
+}
+
+func TestGolden(t *testing.T) {
+	for i := 0; i < len(golden); i++ {
+		g := golden[i]
+		s := fmt.Sprintf("%x", Sum256([]byte(g.in)))
+		if s != g.out {
+			t.Fatalf("Sum256 function: sha256(%s) = %s want %s", g.in, s, g.out)
+		}
+		c := New()
+		for j := 0; j < 3; j++ {
+			if j < 2 {
+				io.WriteString(c, g.in)
+			} else {
+				io.WriteString(c, g.in[0:len(g.in)/2])
+				c.Sum(nil)
+				io.WriteString(c, g.in[len(g.in)/2:])
+			}
+			s := fmt.Sprintf("%x", c.Sum(nil))
+			if s != g.out {
+				t.Fatalf("sha256[%d](%s) = %s want %s", j, g.in, s, g.out)
+			}
+			c.Reset()
+		}
+	}
+}
+
+func TestSize(t *testing.T) {
+	c := New()
+	if got := c.Size(); got != Size {
+		t.Errorf("Size = %d; want %d", got, Size)
+	}
+}
+
+func TestBlockSize(t *testing.T) {
+	c := New()
+	if got := c.BlockSize(); got != BlockSize {
+		t.Errorf("BlockSize = %d want %d", got, BlockSize)
+	}
+}
+
+var bench = New()
+var buf = make([]byte, 8192)
 
-type MySuite struct{}
+func benchmarkSize(b *testing.B, size int) {
+	b.SetBytes(int64(size))
+	sum := make([]byte, bench.Size())
+	for i := 0; i < b.N; i++ {
+		bench.Reset()
+		bench.Write(buf[:size])
+		bench.Sum(sum[:0])
+	}
+}
+
+func BenchmarkHash8Bytes(b *testing.B) {
+	benchmarkSize(b, 8)
+}
 
-var _ = Suite(&MySuite{})
+func BenchmarkHash1K(b *testing.B) {
+	benchmarkSize(b, 1024)
+}
 
-func (s *MySuite) TestSha256Stream(c *C) {
-	testString := []byte("Test string")
-	expectedHash, _ := hex.DecodeString("a3e49d843df13c2e2a7786f6ecd7e0d184f45d718d1ac1a8a63e570466e489dd")
-	hash, err := Sum(bytes.NewBuffer(testString))
-	c.Assert(err, IsNil)
-	c.Assert(bytes.Equal(expectedHash, hash), Equals, true)
+func BenchmarkHash8K(b *testing.B) {
+	benchmarkSize(b, 8192)
 }

pkg/utils/crypto/sha256/sha256block.go

@@ -0,0 +1,23 @@
+// +build amd64
+
+package sha256
+
+// #cgo CFLAGS: -DHAS_SSE41 -DHAS_AVX -DHAS_AVX2
+// #include <stdint.h>
+// void sha256_transform_ssse3 (const char *input_data, uint32_t *digest, unsigned long num_blks);
+// void sha256_transform_avx (const char *input_data, uint32_t *digest, unsigned long num_blks);
+// void sha256_transform_rorx (const char *input_data, uint32_t *digest, unsigned long num_blks);
+import "C"
+import "unsafe"
+
+func blockSSE(dig *digest, p []byte) {
+	C.sha256_transform_ssse3((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
+}
+
+func blockAVX(dig *digest, p []byte) {
+	C.sha256_transform_avx((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
+}
+
+func blockAVX2(dig *digest, p []byte) {
+	C.sha256_transform_rorx((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
+}