Remove the last decoder allocation

zstd/README.md

@@ -34,7 +34,8 @@ For now, a high speed (fastest) and medium-fast (default) compressor has been im
 The "Fastest" compression ratio is roughly equivalent to zstd level 1. 
 The "Default" compression ration is roughly equivalent to zstd level 3 (default).
 
-In terms of speed, it is typically 2x as fast as the stdlib deflate/gzip in its fastest mode. The compression ratio compared to stdlib is around level 3, but usually 3x as fast.
+In terms of speed, it is typically 2x as fast as the stdlib deflate/gzip in its fastest mode. 
+The compression ratio compared to stdlib is around level 3, but usually 3x as fast.
 
 Compared to cgo zstd, the speed is around level 3 (default), but compression slightly worse, between level 1&2.
 
@@ -217,7 +218,8 @@ silesia.tar zstd    3   211947520   66793301    1377    146.79
 
 As part of the development process a *Snappy* -> *Zstandard* converter was also built.
 
-This can convert a *framed* [Snappy Stream](https://godoc.org/github.com/golang/snappy#Writer) to a zstd stream. Note that a single block is not framed.
+This can convert a *framed* [Snappy Stream](https://godoc.org/github.com/golang/snappy#Writer) to a zstd stream. 
+Note that a single block is not framed.
 
 Conversion is done by converting the stream directly from Snappy without intermediate full decoding.
 Therefore the compression ratio is much less than what can be done by a full decompression

zstd/decoder.go

@@ -281,17 +281,17 @@ func (d *Decoder) DecodeAll(input, dst []byte) ([]byte, error) {
 		}
 		d.decoders <- block
 		frame.rawInput = nil
+		frame.bBuf = nil
 		d.frames <- frame
 	}()
+	frame.bBuf = input
 	if cap(dst) == 0 {
 		// Allocate 1MB by default if nothing is provided.
 		dst = make([]byte, 0, 1<<20)
 	}
 
-	// Allocation here:
-	br := byteBuf(input)
 	for {
-		err := frame.reset(&br)
+		err := frame.reset(&frame.bBuf)
 		if err == io.EOF {
 			return dst, nil
 		}
@@ -313,7 +313,7 @@ func (d *Decoder) DecodeAll(input, dst []byte) ([]byte, error) {
 		if err != nil {
 			return dst, err
 		}
-		if len(br) == 0 {
+		if len(frame.bBuf) == 0 {
 			break
 		}
 	}

zstd/framedec.go

@@ -39,6 +39,9 @@ type frameDec struct {
 
 	rawInput byteBuffer
 
+	// Byte buffer that can be reused for small input blocks.
+	bBuf byteBuf
+
 	// asyncRunning indicates whether the async routine processes input on 'decoding'.
 	asyncRunning   bool
 	asyncRunningMu sync.Mutex