SHA512 Implemention with Intel assembly code

pkg/utils/crypto/sha512/sha512.go

@@ -1,29 +1,168 @@
-// +build amd64
-
+// Package sha512 implements the SHA512 hash algorithms as defined
+// in FIPS 180-2.
 package sha512
 
-// #include <stdint.h>
-// void sha512_transform_avx(const void* M, void* D, uint64_t L);
-// void sha512_transform_ssse3(const void* M, void* D, uint64_t L);
-// void sha512_transform_rorx(const void* M, void* D, uint64_t L);
-import "C"
 import (
-	gosha512 "crypto/sha512"
+	"hash"
 	"io"
+
+	"github.com/minio-io/minio/pkg/utils/cpu"
+)
+
+// The size of a SHA512 checksum in bytes.
+const Size = 64
+
+// The blocksize of SHA512 in bytes.
+const BlockSize = 128
+
+const (
+	chunk = 128
+	init0 = 0x6a09e667f3bcc908
+	init1 = 0xbb67ae8584caa73b
+	init2 = 0x3c6ef372fe94f82b
+	init3 = 0xa54ff53a5f1d36f1
+	init4 = 0x510e527fade682d1
+	init5 = 0x9b05688c2b3e6c1f
+	init6 = 0x1f83d9abfb41bd6b
+	init7 = 0x5be0cd19137e2179
 )
 
+// digest represents the partial evaluation of a checksum.
+type digest struct {
+	h   [8]uint64
+	x   [chunk]byte
+	nx  int
+	len uint64
+}
+
+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)
+	}
+}
+
+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
+}
+
+// New returns a new hash.Hash computing the SHA512 checksum.
+func New() hash.Hash {
+	d := new(digest)
+	d.Reset()
+	return d
+}
+
+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 := new(digest)
+	*d = *d0
+	hash := d.checkSum()
+	return append(in, hash[:]...)
+}
+
+func (d *digest) checkSum() [Size]byte {
+	// Padding.  Add a 1 bit and 0 bits until 112 bytes mod 128.
+	len := d.len
+	var tmp [128]byte
+	tmp[0] = 0x80
+	if len%128 < 112 {
+		d.Write(tmp[0 : 112-len%128])
+	} else {
+		d.Write(tmp[0 : 128+112-len%128])
+	}
+
+	// Length in bits.
+	len <<= 3
+	for i := uint(0); i < 16; i++ {
+		tmp[i] = byte(len >> (120 - 8*i))
+	}
+	d.Write(tmp[0:16])
+
+	if d.nx != 0 {
+		panic("d.nx != 0")
+	}
+
+	h := d.h[:]
+
+	var digest [Size]byte
+	for i, s := range h {
+		digest[i*8] = byte(s >> 56)
+		digest[i*8+1] = byte(s >> 48)
+		digest[i*8+2] = byte(s >> 40)
+		digest[i*8+3] = byte(s >> 32)
+		digest[i*8+4] = byte(s >> 24)
+		digest[i*8+5] = byte(s >> 16)
+		digest[i*8+6] = byte(s >> 8)
+		digest[i*8+7] = byte(s)
+	}
+
+	return digest
+}
+
+// Convenience functions
+
+func Sum512(data []byte) [Size]byte {
+	var d digest
+	d.Reset()
+	d.Write(data)
+	return d.checkSum()
+}
+
 func Sum(reader io.Reader) ([]byte, error) {
-	hash := gosha512.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/sha512/sha512_gen.go

@@ -0,0 +1,65 @@
+// +build ignore
+
+package main
+
+import (
+	"crypto/sha512"
+	"encoding/hex"
+	"fmt"
+	"io"
+	"os"
+	"time"
+
+	sha512intel "github.com/minio-io/minio/pkg/utils/crypto/sha512"
+)
+
+func SumIntel(reader io.Reader) ([]byte, error) {
+	h := sha512intel.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 := sha512.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/sha512/sha512_test.go

@@ -1,23 +1,112 @@
+// SHA512 hash algorithm.  See FIPS 180-2.
+
 package sha512
 
 import (
-	"bytes"
-	"encoding/hex"
+	"fmt"
+	"io"
 	"testing"
-
-	. "gopkg.in/check.v1"
 )
 
-func Test(t *testing.T) { TestingT(t) }
+type sha512Test struct {
+	out string
+	in  string
+}
+
+var golden = []sha512Test{
+	{"cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e", ""},
+	{"1f40fc92da241694750979ee6cf582f2d5d7d28e18335de05abc54d0560e0f5302860c652bf08d560252aa5e74210546f369fbbbce8c12cfc7957b2652fe9a75", "a"},
+	{"2d408a0717ec188158278a796c689044361dc6fdde28d6f04973b80896e1823975cdbf12eb63f9e0591328ee235d80e9b5bf1aa6a44f4617ff3caf6400eb172d", "ab"},
+	{"ddaf35a193617abacc417349ae20413112e6fa4e89a97ea20a9eeee64b55d39a2192992a274fc1a836ba3c23a3feebbd454d4423643ce80e2a9ac94fa54ca49f", "abc"},
+	{"d8022f2060ad6efd297ab73dcc5355c9b214054b0d1776a136a669d26a7d3b14f73aa0d0ebff19ee333368f0164b6419a96da49e3e481753e7e96b716bdccb6f", "abcd"},
+	{"878ae65a92e86cac011a570d4c30a7eaec442b85ce8eca0c2952b5e3cc0628c2e79d889ad4d5c7c626986d452dd86374b6ffaa7cd8b67665bef2289a5c70b0a1", "abcde"},
+	{"e32ef19623e8ed9d267f657a81944b3d07adbb768518068e88435745564e8d4150a0a703be2a7d88b61e3d390c2bb97e2d4c311fdc69d6b1267f05f59aa920e7", "abcdef"},
+	{"d716a4188569b68ab1b6dfac178e570114cdf0ea3a1cc0e31486c3e41241bc6a76424e8c37ab26f096fc85ef9886c8cb634187f4fddff645fb099f1ff54c6b8c", "abcdefg"},
+	{"a3a8c81bc97c2560010d7389bc88aac974a104e0e2381220c6e084c4dccd1d2d17d4f86db31c2a851dc80e6681d74733c55dcd03dd96f6062cdda12a291ae6ce", "abcdefgh"},
+	{"f22d51d25292ca1d0f68f69aedc7897019308cc9db46efb75a03dd494fc7f126c010e8ade6a00a0c1a5f1b75d81e0ed5a93ce98dc9b833db7839247b1d9c24fe", "abcdefghi"},
+	{"ef6b97321f34b1fea2169a7db9e1960b471aa13302a988087357c520be957ca119c3ba68e6b4982c019ec89de3865ccf6a3cda1fe11e59f98d99f1502c8b9745", "abcdefghij"},
+	{"2210d99af9c8bdecda1b4beff822136753d8342505ddce37f1314e2cdbb488c6016bdaa9bd2ffa513dd5de2e4b50f031393d8ab61f773b0e0130d7381e0f8a1d", "Discard medicine more than two years old."},
+	{"a687a8985b4d8d0a24f115fe272255c6afaf3909225838546159c1ed685c211a203796ae8ecc4c81a5b6315919b3a64f10713da07e341fcdbb08541bf03066ce", "He who has a shady past knows that nice guys finish last."},
+	{"8ddb0392e818b7d585ab22769a50df660d9f6d559cca3afc5691b8ca91b8451374e42bcdabd64589ed7c91d85f626596228a5c8572677eb98bc6b624befb7af8", "I wouldn't marry him with a ten foot pole."},
+	{"26ed8f6ca7f8d44b6a8a54ae39640fa8ad5c673f70ee9ce074ba4ef0d483eea00bab2f61d8695d6b34df9c6c48ae36246362200ed820448bdc03a720366a87c6", "Free! Free!/A trip/to Mars/for 900/empty jars/Burma Shave"},
+	{"e5a14bf044be69615aade89afcf1ab0389d5fc302a884d403579d1386a2400c089b0dbb387ed0f463f9ee342f8244d5a38cfbc0e819da9529fbff78368c9a982", "The days of the digital watch are numbered.  -Tom Stoppard"},
+	{"420a1faa48919e14651bed45725abe0f7a58e0f099424c4e5a49194946e38b46c1f8034b18ef169b2e31050d1648e0b982386595f7df47da4b6fd18e55333015", "Nepal premier won't resign."},
+	{"d926a863beadb20134db07683535c72007b0e695045876254f341ddcccde132a908c5af57baa6a6a9c63e6649bba0c213dc05fadcf9abccea09f23dcfb637fbe", "For every action there is an equal and opposite government program."},
+	{"9a98dd9bb67d0da7bf83da5313dff4fd60a4bac0094f1b05633690ffa7f6d61de9a1d4f8617937d560833a9aaa9ccafe3fd24db418d0e728833545cadd3ad92d", "His money is twice tainted: 'taint yours and 'taint mine."},
+	{"d7fde2d2351efade52f4211d3746a0780a26eec3df9b2ed575368a8a1c09ec452402293a8ea4eceb5a4f60064ea29b13cdd86918cd7a4faf366160b009804107", "There is no reason for any individual to have a computer in their home. -Ken Olsen, 1977"},
+	{"b0f35ffa2697359c33a56f5c0cf715c7aeed96da9905ca2698acadb08fbc9e669bf566b6bd5d61a3e86dc22999bcc9f2224e33d1d4f32a228cf9d0349e2db518", "It's a tiny change to the code and not completely disgusting. - Bob Manchek"},
+	{"3d2e5f91778c9e66f7e061293aaa8a8fc742dd3b2e4f483772464b1144189b49273e610e5cccd7a81a19ca1fa70f16b10f1a100a4d8c1372336be8484c64b311", "size:  a.out:  bad magic"},
+	{"b2f68ff58ac015efb1c94c908b0d8c2bf06f491e4de8e6302c49016f7f8a33eac3e959856c7fddbc464de618701338a4b46f76dbfaf9a1e5262b5f40639771c7", "The major problem is with sendmail.  -Mark Horton"},
+	{"d8c92db5fdf52cf8215e4df3b4909d29203ff4d00e9ad0b64a6a4e04dec5e74f62e7c35c7fb881bd5de95442123df8f57a489b0ae616bd326f84d10021121c57", "Give me a rock, paper and scissors and I will move the world.  CCFestoon"},
+	{"19a9f8dc0a233e464e8566ad3ca9b91e459a7b8c4780985b015776e1bf239a19bc233d0556343e2b0a9bc220900b4ebf4f8bdf89ff8efeaf79602d6849e6f72e", "If the enemy is within range, then so are you."},
+	{"00b4c41f307bde87301cdc5b5ab1ae9a592e8ecbb2021dd7bc4b34e2ace60741cc362560bec566ba35178595a91932b8d5357e2c9cec92d393b0fa7831852476", "It's well we cannot hear the screams/That we create in others' dreams."},
+	{"91eccc3d5375fd026e4d6787874b1dce201cecd8a27dbded5065728cb2d09c58a3d467bb1faf353bf7ba567e005245d5321b55bc344f7c07b91cb6f26c959be7", "You remind me of a TV show, but that's all right: I watch it anyway."},
+	{"fabbbe22180f1f137cfdc9556d2570e775d1ae02a597ded43a72a40f9b485d500043b7be128fb9fcd982b83159a0d99aa855a9e7cc4240c00dc01a9bdf8218d7", "C is as portable as Stonehedge!!"},
+	{"2ecdec235c1fa4fc2a154d8fba1dddb8a72a1ad73838b51d792331d143f8b96a9f6fcb0f34d7caa351fe6d88771c4f105040e0392f06e0621689d33b2f3ba92e", "Even if I could be Shakespeare, I think I should still choose to be Faraday. - A. Huxley"},
+	{"7ad681f6f96f82f7abfa7ecc0334e8fa16d3dc1cdc45b60b7af43fe4075d2357c0c1d60e98350f1afb1f2fe7a4d7cd2ad55b88e458e06b73c40b437331f5dab4", "The fugacity of a constituent in a mixture of gases at a given temperature is proportional to its mole fraction.  Lewis-Randall Rule"},
+	{"833f9248ab4a3b9e5131f745fda1ffd2dd435b30e965957e78291c7ab73605fd1912b0794e5c233ab0a12d205a39778d19b83515d6a47003f19cdee51d98c7e0", "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", Sum512([]byte(g.in)))
+		if s != g.out {
+			t.Fatalf("Sum512 function: sha512(%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("sha512[%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) TestSha512Stream(c *C) {
-	testString := []byte("Test string")
-	expectedHash, _ := hex.DecodeString("811aa0c53c0039b6ead0ca878b096eed1d39ed873fd2d2d270abfb9ca620d3ed561c565d6dbd1114c323d38e3f59c00df475451fc9b30074f2abda3529df2fa7")
-	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/sha512/sha512block.go

@@ -0,0 +1,23 @@
+// +build amd64
+
+package sha512
+
+// #cgo CFLAGS: -DHAS_SSE41 -DHAS_AVX -DHAS_AVX2
+// #include <stdint.h>
+// void sha512_transform_ssse3 (const void* M, void* D, uint64_t L);
+// void sha512_transform_avx (const void* M, void* D, uint64_t L);
+// void sha512_transform_rorx (const void* M, void* D, uint64_t L);
+import "C"
+import "unsafe"
+
+func blockSSE(dig *digest, p []byte) {
+	C.sha512_transform_ssse3(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
+}
+
+func blockAVX(dig *digest, p []byte) {
+	C.sha512_transform_avx(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
+}
+
+func blockAVX2(dig *digest, p []byte) {
+	C.sha512_transform_rorx(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
+}