diff --git a/plumbing/format/packfile/diff.go b/plumbing/format/packfile/diff.go
deleted file mode 100644
index ff34329..0000000
--- a/plumbing/format/packfile/diff.go
+++ /dev/null
@@ -1,397 +0,0 @@
-package packfile
-
-/*
-
-Copyright (c) 2013, Patrick Mezard
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are
-met:
-
-    Redistributions of source code must retain the above copyright
-notice, this list of conditions and the following disclaimer.
-    Redistributions in binary form must reproduce the above copyright
-notice, this list of conditions and the following disclaimer in the
-documentation and/or other materials provided with the distribution.
-    The names of its contributors may not be used to endorse or promote
-products derived from this software without specific prior written
-permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
-IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
-TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
-PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
-TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
-PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
-LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
-NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-*/
-
-// Code based on https://github.com/pmezard/go-difflib
-// Removed unnecessary code for this use case and changed string inputs to byte
-
-type match struct {
-	A    int
-	B    int
-	Size int
-}
-
-type opCode struct {
-	Tag byte
-	I1  int
-	I2  int
-	J1  int
-	J2  int
-}
-
-// SequenceMatcher compares sequence of bytes. The basic
-// algorithm predates, and is a little fancier than, an algorithm
-// published in the late 1980's by Ratcliff and Obershelp under the
-// hyperbolic name "gestalt pattern matching".  The basic idea is to find
-// the longest contiguous matching subsequence that contains no "junk"
-// elements (R-O doesn't address junk).  The same idea is then applied
-// recursively to the pieces of the sequences to the left and to the right
-// of the matching subsequence.  This does not yield minimal edit
-// sequences, but does tend to yield matches that "look right" to people.
-//
-// SequenceMatcher tries to compute a "human-friendly diff" between two
-// sequences.  Unlike e.g. UNIX(tm) diff, the fundamental notion is the
-// longest *contiguous* & junk-free matching subsequence.  That's what
-// catches peoples' eyes.  The Windows(tm) windiff has another interesting
-// notion, pairing up elements that appear uniquely in each sequence.
-// That, and the method here, appear to yield more intuitive difference
-// reports than does diff.  This method appears to be the least vulnerable
-// to synching up on blocks of "junk lines", though (like blank lines in
-// ordinary text files, or maybe "<P>" lines in HTML files).  That may be
-// because this is the only method of the 3 that has a *concept* of
-// "junk" <wink>.
-//
-// Timing:  Basic R-O is cubic time worst case and quadratic time expected
-// case.  SequenceMatcher is quadratic time for the worst case and has
-// expected-case behavior dependent in a complicated way on how many
-// elements the sequences have in common; best case time is linear.
-type sequenceMatcher struct {
-	a              []byte
-	b              []byte
-	b2j            map[byte][]int
-	IsJunk         func(byte) bool
-	autoJunk       bool
-	bJunk          map[byte]struct{}
-	matchingBlocks []match
-	fullBCount     map[byte]int
-	bPopular       map[byte]struct{}
-	opCodes        []opCode
-}
-
-func newMatcher(a, b []byte) *sequenceMatcher {
-	m := sequenceMatcher{autoJunk: true}
-	m.SetSeqs(a, b)
-	return &m
-}
-
-// Set two sequences to be compared.
-func (m *sequenceMatcher) SetSeqs(a, b []byte) {
-	m.SetSeq1(a)
-	m.SetSeq2(b)
-}
-
-// Set the first sequence to be compared. The second sequence to be compared is
-// not changed.
-//
-// SequenceMatcher computes and caches detailed information about the second
-// sequence, so if you want to compare one sequence S against many sequences,
-// use .SetSeq2(s) once and call .SetSeq1(x) repeatedly for each of the other
-// sequences.
-//
-// See also SetSeqs() and SetSeq2().
-func (m *sequenceMatcher) SetSeq1(a []byte) {
-	if &a == &m.a {
-		return
-	}
-	m.a = a
-	m.matchingBlocks = nil
-	m.opCodes = nil
-}
-
-// Set the second sequence to be compared. The first sequence to be compared is
-// not changed.
-func (m *sequenceMatcher) SetSeq2(b []byte) {
-	if &b == &m.b {
-		return
-	}
-	m.b = b
-	m.matchingBlocks = nil
-	m.opCodes = nil
-	m.fullBCount = nil
-	m.chainB()
-}
-
-func (m *sequenceMatcher) chainB() {
-	// Populate line -> index mapping
-	b2j := map[byte][]int{}
-	for i, s := range m.b {
-		indices := b2j[s]
-		indices = append(indices, i)
-		b2j[s] = indices
-	}
-
-	// Purge junk elements
-	m.bJunk = map[byte]struct{}{}
-	if m.IsJunk != nil {
-		junk := m.bJunk
-		for s := range b2j {
-			if m.IsJunk(s) {
-				junk[s] = struct{}{}
-			}
-		}
-		for s := range junk {
-			delete(b2j, s)
-		}
-	}
-
-	// Purge remaining popular elements
-	popular := map[byte]struct{}{}
-	n := len(m.b)
-	if m.autoJunk && n >= 200 {
-		ntest := n/100 + 1
-		for s, indices := range b2j {
-			if len(indices) > ntest {
-				popular[s] = struct{}{}
-			}
-		}
-		for s := range popular {
-			delete(b2j, s)
-		}
-	}
-	m.bPopular = popular
-	m.b2j = b2j
-}
-
-func (m *sequenceMatcher) isBJunk(s byte) bool {
-	_, ok := m.bJunk[s]
-	return ok
-}
-
-// Find longest matching block in a[alo:ahi] and b[blo:bhi].
-//
-// If IsJunk is not defined:
-//
-// Return (i,j,k) such that a[i:i+k] is equal to b[j:j+k], where
-//     alo <= i <= i+k <= ahi
-//     blo <= j <= j+k <= bhi
-// and for all (i',j',k') meeting those conditions,
-//     k >= k'
-//     i <= i'
-//     and if i == i', j <= j'
-//
-// In other words, of all maximal matching blocks, return one that
-// starts earliest in a, and of all those maximal matching blocks that
-// start earliest in a, return the one that starts earliest in b.
-//
-// If IsJunk is defined, first the longest matching block is
-// determined as above, but with the additional restriction that no
-// junk element appears in the block.  Then that block is extended as
-// far as possible by matching (only) junk elements on both sides.  So
-// the resulting block never matches on junk except as identical junk
-// happens to be adjacent to an "interesting" match.
-//
-// If no blocks match, return (alo, blo, 0).
-func (m *sequenceMatcher) findLongestMatch(alo, ahi, blo, bhi int) match {
-	// CAUTION:  stripping common prefix or suffix would be incorrect.
-	// E.g.,
-	//    ab
-	//    acab
-	// Longest matching block is "ab", but if common prefix is
-	// stripped, it's "a" (tied with "b").  UNIX(tm) diff does so
-	// strip, so ends up claiming that ab is changed to acab by
-	// inserting "ca" in the middle.  That's minimal but unintuitive:
-	// "it's obvious" that someone inserted "ac" at the front.
-	// Windiff ends up at the same place as diff, but by pairing up
-	// the unique 'b's and then matching the first two 'a's.
-	besti, bestj, bestsize := alo, blo, 0
-
-	// find longest junk-free match
-	// during an iteration of the loop, j2len[j] = length of longest
-	// junk-free match ending with a[i-1] and b[j]
-	j2len := map[int]int{}
-	for i := alo; i != ahi; i++ {
-		// look at all instances of a[i] in b; note that because
-		// b2j has no junk keys, the loop is skipped if a[i] is junk
-		newj2len := map[int]int{}
-		for _, j := range m.b2j[m.a[i]] {
-			// a[i] matches b[j]
-			if j < blo {
-				continue
-			}
-			if j >= bhi {
-				break
-			}
-			k := j2len[j-1] + 1
-			newj2len[j] = k
-			if k > bestsize {
-				besti, bestj, bestsize = i-k+1, j-k+1, k
-			}
-		}
-		j2len = newj2len
-	}
-
-	// Extend the best by non-junk elements on each end.  In particular,
-	// "popular" non-junk elements aren't in b2j, which greatly speeds
-	// the inner loop above, but also means "the best" match so far
-	// doesn't contain any junk *or* popular non-junk elements.
-	for besti > alo && bestj > blo && !m.isBJunk(m.b[bestj-1]) &&
-		m.a[besti-1] == m.b[bestj-1] {
-		besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
-	}
-	for besti+bestsize < ahi && bestj+bestsize < bhi &&
-		!m.isBJunk(m.b[bestj+bestsize]) &&
-		m.a[besti+bestsize] == m.b[bestj+bestsize] {
-		bestsize += 1
-	}
-
-	// Now that we have a wholly interesting match (albeit possibly
-	// empty!), we may as well suck up the matching junk on each
-	// side of it too.  Can't think of a good reason not to, and it
-	// saves post-processing the (possibly considerable) expense of
-	// figuring out what to do with it.  In the case of an empty
-	// interesting match, this is clearly the right thing to do,
-	// because no other kind of match is possible in the regions.
-	for besti > alo && bestj > blo && m.isBJunk(m.b[bestj-1]) &&
-		m.a[besti-1] == m.b[bestj-1] {
-		besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
-	}
-	for besti+bestsize < ahi && bestj+bestsize < bhi &&
-		m.isBJunk(m.b[bestj+bestsize]) &&
-		m.a[besti+bestsize] == m.b[bestj+bestsize] {
-		bestsize += 1
-	}
-
-	return match{A: besti, B: bestj, Size: bestsize}
-}
-
-// Return list of triples describing matching subsequences.
-//
-// Each triple is of the form (i, j, n), and means that
-// a[i:i+n] == b[j:j+n].  The triples are monotonically increasing in
-// i and in j. It's also guaranteed that if (i, j, n) and (i', j', n') are
-// adjacent triples in the list, and the second is not the last triple in the
-// list, then i+n != i' or j+n != j'. IOW, adjacent triples never describe
-// adjacent equal blocks.
-//
-// The last triple is a dummy, (len(a), len(b), 0), and is the only
-// triple with n==0.
-func (m *sequenceMatcher) GetMatchingBlocks() []match {
-	if m.matchingBlocks != nil {
-		return m.matchingBlocks
-	}
-
-	var matchBlocks func(alo, ahi, blo, bhi int, matched []match) []match
-	matchBlocks = func(alo, ahi, blo, bhi int, matched []match) []match {
-		match := m.findLongestMatch(alo, ahi, blo, bhi)
-		i, j, k := match.A, match.B, match.Size
-		if match.Size > 0 {
-			if alo < i && blo < j {
-				matched = matchBlocks(alo, i, blo, j, matched)
-			}
-			matched = append(matched, match)
-			if i+k < ahi && j+k < bhi {
-				matched = matchBlocks(i+k, ahi, j+k, bhi, matched)
-			}
-		}
-		return matched
-	}
-	matched := matchBlocks(0, len(m.a), 0, len(m.b), nil)
-
-	// It's possible that we have adjacent equal blocks in the
-	// matching_blocks list now.
-	nonAdjacent := []match{}
-	i1, j1, k1 := 0, 0, 0
-	for _, b := range matched {
-		// Is this block adjacent to i1, j1, k1?
-		i2, j2, k2 := b.A, b.B, b.Size
-		if i1+k1 == i2 && j1+k1 == j2 {
-			// Yes, so collapse them -- this just increases the length of
-			// the first block by the length of the second, and the first
-			// block so lengthened remains the block to compare against.
-			k1 += k2
-		} else {
-			// Not adjacent.  Remember the first block (k1==0 means it's
-			// the dummy we started with), and make the second block the
-			// new block to compare against.
-			if k1 > 0 {
-				nonAdjacent = append(nonAdjacent, match{i1, j1, k1})
-			}
-			i1, j1, k1 = i2, j2, k2
-		}
-	}
-	if k1 > 0 {
-		nonAdjacent = append(nonAdjacent, match{i1, j1, k1})
-	}
-
-	nonAdjacent = append(nonAdjacent, match{len(m.a), len(m.b), 0})
-	m.matchingBlocks = nonAdjacent
-	return m.matchingBlocks
-}
-
-const (
-	tagReplace = 'r'
-	tagDelete  = 'd'
-	tagInsert  = 'i'
-	tagEqual   = 'e'
-)
-
-// Return list of 5-tuples describing how to turn a into b.
-//
-// Each tuple is of the form (tag, i1, i2, j1, j2).  The first tuple
-// has i1 == j1 == 0, and remaining tuples have i1 == the i2 from the
-// tuple preceding it, and likewise for j1 == the previous j2.
-//
-// The tags are characters, with these meanings:
-//
-// 'r' (replace):  a[i1:i2] should be replaced by b[j1:j2]
-//
-// 'd' (delete):   a[i1:i2] should be deleted, j1==j2 in this case.
-//
-// 'i' (insert):   b[j1:j2] should be inserted at a[i1:i1], i1==i2 in this case.
-//
-// 'e' (equal):    a[i1:i2] == b[j1:j2]
-func (m *sequenceMatcher) GetOpCodes() []opCode {
-	if m.opCodes != nil {
-		return m.opCodes
-	}
-	i, j := 0, 0
-	matching := m.GetMatchingBlocks()
-	opCodes := make([]opCode, 0, len(matching))
-	for _, m := range matching {
-		//  invariant:  we've pumped out correct diffs to change
-		//  a[:i] into b[:j], and the next matching block is
-		//  a[ai:ai+size] == b[bj:bj+size]. So we need to pump
-		//  out a diff to change a[i:ai] into b[j:bj], pump out
-		//  the matching block, and move (i,j) beyond the match
-		ai, bj, size := m.A, m.B, m.Size
-		tag := byte(0)
-		if i < ai && j < bj {
-			tag = tagReplace
-		} else if i < ai {
-			tag = tagDelete
-		} else if j < bj {
-			tag = tagInsert
-		}
-		if tag > 0 {
-			opCodes = append(opCodes, opCode{tag, i, ai, j, bj})
-		}
-		i, j = ai+size, bj+size
-		// the list of matching blocks is terminated by a
-		// sentinel with size 0
-		if size > 0 {
-			opCodes = append(opCodes, opCode{tagEqual, ai, i, bj, j})
-		}
-	}
-	m.opCodes = opCodes
-	return m.opCodes
-}
diff --git a/plumbing/format/packfile/diff_delta.go b/plumbing/format/packfile/diff_delta.go
index 40d450f..e3438aa 100644
--- a/plumbing/format/packfile/diff_delta.go
+++ b/plumbing/format/packfile/diff_delta.go
@@ -1,6 +1,8 @@
 package packfile
 
 import (
+	"bytes"
+	"hash/adler32"
 	"io/ioutil"
 
 	"gopkg.in/src-d/go-git.v4/plumbing"
@@ -11,7 +13,8 @@
 // for more info
 
 const (
-	maxCopyLen = 0xffff
+	// Standard chunk size used to generate fingerprints
+	s = 16
 )
 
 // GetDelta returns an EncodedObject of type OFSDeltaObject. Base and Target object,
@@ -51,52 +54,99 @@
 	return delta, nil
 }
 
-// DiffDelta returns the delta that transforms baseBuf into targetBuf.
-func DiffDelta(baseBuf []byte, targetBuf []byte) []byte {
-	var outBuff []byte
+// DiffDelta returns the delta that transforms src into tgt.
+func DiffDelta(src []byte, tgt []byte) []byte {
+	buf := bytes.NewBuffer(nil)
+	buf.Write(deltaEncodeSize(len(src)))
+	buf.Write(deltaEncodeSize(len(tgt)))
 
-	outBuff = append(outBuff, deltaEncodeSize(len(baseBuf))...)
-	outBuff = append(outBuff, deltaEncodeSize(len(targetBuf))...)
+	sindex := initMatch(src)
 
-	sm := newMatcher(baseBuf, targetBuf)
-	for _, op := range sm.GetOpCodes() {
-		switch {
-		case op.Tag == tagEqual:
-			copyStart := op.I1
-			copyLen := op.I2 - op.I1
-			for {
-				if copyLen <= 0 {
-					break
-				}
-				var toCopy int
-				if copyLen < maxCopyLen {
-					toCopy = copyLen
-				} else {
-					toCopy = maxCopyLen
-				}
+	ibuf := bytes.NewBuffer(nil)
+	for i := 0; i < len(tgt); i++ {
+		offset, l := findMatch(src, tgt, sindex, i)
 
-				outBuff = append(outBuff, encodeCopyOperation(copyStart, toCopy)...)
-				copyStart += toCopy
-				copyLen -= toCopy
-			}
-		case op.Tag == tagReplace || op.Tag == tagInsert:
-			s := op.J2 - op.J1
-			o := op.J1
-			for {
-				if s <= 127 {
-					break
-				}
-				outBuff = append(outBuff, byte(127))
-				outBuff = append(outBuff, targetBuf[o:o+127]...)
-				s -= 127
-				o += 127
-			}
-			outBuff = append(outBuff, byte(s))
-			outBuff = append(outBuff, targetBuf[o:o+s]...)
+		if l < s {
+			ibuf.WriteByte(tgt[i])
+		} else {
+			encodeInsertOperation(ibuf, buf)
+			buf.Write(encodeCopyOperation(offset, l))
+			i += l - 1
 		}
 	}
 
-	return outBuff
+	encodeInsertOperation(ibuf, buf)
+
+	return buf.Bytes()
+}
+
+func encodeInsertOperation(ibuf, buf *bytes.Buffer) {
+	if ibuf.Len() == 0 {
+		return
+	}
+
+	b := ibuf.Bytes()
+	s := ibuf.Len()
+	o := 0
+	for {
+		if s <= 127 {
+			break
+		}
+		buf.WriteByte(byte(127))
+		buf.Write(b[o : o+127])
+		s -= 127
+		o += 127
+	}
+	buf.WriteByte(byte(s))
+	buf.Write(b[o : o+s])
+
+	ibuf.Reset()
+}
+
+func initMatch(src []byte) map[uint32]int {
+	i := 0
+	index := make(map[uint32]int)
+	for {
+		if i+s > len(src) {
+			break
+		}
+
+		ch := adler32.Checksum(src[i : i+s])
+		index[ch] = i
+		i += s
+	}
+
+	return index
+}
+
+func findMatch(src, tgt []byte, sindex map[uint32]int, tgtOffset int) (srcOffset, l int) {
+	if len(tgt) >= tgtOffset+s {
+		ch := adler32.Checksum(tgt[tgtOffset : tgtOffset+s])
+		var ok bool
+		srcOffset, ok = sindex[ch]
+		if !ok {
+			return
+		}
+
+		l = matchLength(src, tgt, tgtOffset, srcOffset)
+	}
+
+	return
+}
+
+func matchLength(src, tgt []byte, otgt, osrc int) int {
+	l := 0
+	for {
+		if (osrc >= len(src) || otgt >= len(tgt)) || src[osrc] != tgt[otgt] {
+			break
+		}
+
+		l++
+		osrc++
+		otgt++
+	}
+
+	return l
 }
 
 func deltaEncodeSize(size int) []byte {