Add options (#46)

* Add options

Make constants changeable as options.

The API remains backwards compatible.

* Update documentation.

* Fix line endings

* fmt

* fmt

* Use functions for parameters.

Much neater.

.travis.yml

@@ -7,11 +7,11 @@ os:
   - osx 
 
 go:
-  - 1.3
-  - 1.4
   - 1.5
   - 1.6
-  - tip
+  - 1.7
+  - 1.8
+  - master
 
 install:
  - go get ./...
@@ -26,3 +26,8 @@ script:
  - go build examples/stream-decoder.go
  - go build examples/stream-encoder.go
  - diff <(gofmt -d .) <("")
+
+matrix:
+  allow_failures:
+    - go: 'master'
+  fast_finish: true

README.md

@@ -162,6 +162,18 @@ There is no buffering or timeouts/retry specified. If you want to add that, you
 
 For complete examples of a streaming encoder and decoder see the [examples folder](https://github.com/klauspost/reedsolomon/tree/master/examples).
 
+#Advanced Options
+
+You can modify internal options which affects how jobs are split between and processed by goroutines.
+
+To create options, use the WithXXX functions. You can supply options to `New`, `NewStream` and `NewStreamC`. If no Options are supplied, default options are used.
+
+Example of how to supply options:
+
+ ```Go
+     enc, err := reedsolomon.New(10, 3, WithMaxGoroutines(25))
+ ```
+
 
 # Performance
 Performance depends mainly on the number of parity shards. In rough terms, doubling the number of parity shards will double the encoding time.

galois_amd64.go

@@ -5,10 +5,6 @@
 
 package reedsolomon
 
-import (
-	"github.com/klauspost/cpuid"
-)
-
 //go:noescape
 func galMulSSSE3(low, high, in, out []byte)
 
@@ -40,12 +36,12 @@ func galMulSSSE3Xor(low, high, in, out []byte) {
 }
 */
 
-func galMulSlice(c byte, in, out []byte) {
+func galMulSlice(c byte, in, out []byte, ssse3, avx2 bool) {
 	var done int
-	if cpuid.CPU.AVX2() {
+	if avx2 {
 		galMulAVX2(mulTableLow[c][:], mulTableHigh[c][:], in, out)
 		done = (len(in) >> 5) << 5
-	} else if cpuid.CPU.SSSE3() {
+	} else if ssse3 {
 		galMulSSSE3(mulTableLow[c][:], mulTableHigh[c][:], in, out)
 		done = (len(in) >> 4) << 4
 	}
@@ -58,12 +54,12 @@ func galMulSlice(c byte, in, out []byte) {
 	}
 }
 
-func galMulSliceXor(c byte, in, out []byte) {
+func galMulSliceXor(c byte, in, out []byte, ssse3, avx2 bool) {
 	var done int
-	if cpuid.CPU.AVX2() {
+	if avx2 {
 		galMulAVX2Xor(mulTableLow[c][:], mulTableHigh[c][:], in, out)
 		done = (len(in) >> 5) << 5
-	} else if cpuid.CPU.SSSE3() {
+	} else if ssse3 {
 		galMulSSSE3Xor(mulTableLow[c][:], mulTableHigh[c][:], in, out)
 		done = (len(in) >> 4) << 4
 	}

galois_noasm.go

@@ -4,14 +4,14 @@
 
 package reedsolomon
 
-func galMulSlice(c byte, in, out []byte) {
+func galMulSlice(c byte, in, out []byte, ssse3, avx2 bool) {
 	mt := mulTable[c]
 	for n, input := range in {
 		out[n] = mt[input]
 	}
 }
 
-func galMulSliceXor(c byte, in, out []byte) {
+func galMulSliceXor(c byte, in, out []byte, ssse3, avx2 bool) {
 	mt := mulTable[c]
 	for n, input := range in {
 		out[n] ^= mt[input]

galois_test.go

@@ -131,13 +131,13 @@ func TestGalois(t *testing.T) {
 	// Test slices (>16 entries to test assembler)
 	in := []byte{0, 1, 2, 3, 4, 5, 6, 10, 50, 100, 150, 174, 201, 255, 99, 32, 67, 85}
 	out := make([]byte, len(in))
-	galMulSlice(25, in, out)
+	galMulSlice(25, in, out, false, false)
 	expect := []byte{0x0, 0x19, 0x32, 0x2b, 0x64, 0x7d, 0x56, 0xfa, 0xb8, 0x6d, 0xc7, 0x85, 0xc3, 0x1f, 0x22, 0x7, 0x25, 0xfe}
 	if 0 != bytes.Compare(out, expect) {
 		t.Errorf("got %#v, expected %#v", out, expect)
 	}
 
-	galMulSlice(177, in, out)
+	galMulSlice(177, in, out, false, false)
 	expect = []byte{0x0, 0xb1, 0x7f, 0xce, 0xfe, 0x4f, 0x81, 0x9e, 0x3, 0x6, 0xe8, 0x75, 0xbd, 0x40, 0x36, 0xa3, 0x95, 0xcb}
 	if 0 != bytes.Compare(out, expect) {
 		t.Errorf("got %#v, expected %#v", out, expect)

options.go

@@ -0,0 +1,67 @@
+package reedsolomon
+
+import (
+	"runtime"
+
+	"github.com/klauspost/cpuid"
+)
+
+// Option allows to override processing parameters.
+type Option func(*options)
+
+type options struct {
+	maxGoroutines     int
+	minSplitSize      int
+	useAVX2, useSSSE3 bool
+}
+
+var defaultOptions = options{
+	maxGoroutines: 50,
+	minSplitSize:  512,
+}
+
+func init() {
+	if runtime.GOMAXPROCS(0) <= 1 {
+		defaultOptions.maxGoroutines = 1
+	}
+	// Detect CPU capabilities.
+	defaultOptions.useSSSE3 = cpuid.CPU.SSSE3()
+	defaultOptions.useAVX2 = cpuid.CPU.AVX2()
+}
+
+// WithMaxGoroutines is the maximum number of goroutines number for encoding & decoding.
+// Jobs will be split into this many parts, unless each goroutine would have to process
+// less than minSplitSize bytes (set with WithMinSplitSize).
+// For the best speed, keep this well above the GOMAXPROCS number for more fine grained
+// scheduling.
+// If n <= 0, it is ignored.
+func WithMaxGoroutines(n int) Option {
+	return func(o *options) {
+		if n > 0 {
+			o.maxGoroutines = n
+		}
+	}
+}
+
+// MinSplitSize Is the minimum encoding size in bytes per goroutine.
+// See WithMaxGoroutines on how jobs are split.
+// If n <= 0, it is ignored.
+func WithMinSplitSize(n int) Option {
+	return func(o *options) {
+		if n > 0 {
+			o.maxGoroutines = n
+		}
+	}
+}
+
+func withSSE3(enabled bool) Option {
+	return func(o *options) {
+		o.useSSSE3 = enabled
+	}
+}
+
+func withAVX2(enabled bool) Option {
+	return func(o *options) {
+		o.useAVX2 = enabled
+	}
+}

reedsolomon.go

@@ -15,7 +15,6 @@ import (
 	"bytes"
 	"errors"
 	"io"
-	"runtime"
 	"sync"
 )
 
@@ -83,6 +82,7 @@ type reedSolomon struct {
 	m            matrix
 	tree         inversionTree
 	parity       [][]byte
+	o            options
 }
 
 // ErrInvShardNum will be returned by New, if you attempt to create
@@ -98,13 +98,18 @@ var ErrMaxShardNum = errors.New("cannot create Encoder with 255 or more data+par
 // the number of data shards and parity shards that
 // you want to use. You can reuse this encoder.
 // Note that the maximum number of data shards is 256.
-func New(dataShards, parityShards int) (Encoder, error) {
+// If no options are supplied, default options are used.
+func New(dataShards, parityShards int, opts ...Option) (Encoder, error) {
 	r := reedSolomon{
 		DataShards:   dataShards,
 		ParityShards: parityShards,
 		Shards:       dataShards + parityShards,
+		o:            defaultOptions,
 	}
 
+	for _, opt := range opts {
+		opt(&r.o)
+	}
 	if dataShards <= 0 || parityShards <= 0 {
 		return nil, ErrInvShardNum
 	}
@@ -201,34 +206,29 @@ func (r reedSolomon) Verify(shards [][]byte) (bool, error) {
 // number of matrix rows used, is determined by
 // outputCount, which is the number of outputs to compute.
 func (r reedSolomon) codeSomeShards(matrixRows, inputs, outputs [][]byte, outputCount, byteCount int) {
-	if runtime.GOMAXPROCS(0) > 1 && len(inputs[0]) > minSplitSize {
+	if r.o.maxGoroutines > 1 && byteCount > r.o.minSplitSize {
 		r.codeSomeShardsP(matrixRows, inputs, outputs, outputCount, byteCount)
 		return
 	}
 	for c := 0; c < r.DataShards; c++ {
 		in := inputs[c]
 		for iRow := 0; iRow < outputCount; iRow++ {
 			if c == 0 {
-				galMulSlice(matrixRows[iRow][c], in, outputs[iRow])
+				galMulSlice(matrixRows[iRow][c], in, outputs[iRow], r.o.useSSSE3, r.o.useAVX2)
 			} else {
-				galMulSliceXor(matrixRows[iRow][c], in, outputs[iRow])
+				galMulSliceXor(matrixRows[iRow][c], in, outputs[iRow], r.o.useSSSE3, r.o.useAVX2)
 			}
 		}
 	}
 }
 
-const (
-	minSplitSize  = 512 // min split size per goroutine
-	maxGoroutines = 50  // max goroutines number for encoding & decoding
-)
-
 // Perform the same as codeSomeShards, but split the workload into
 // several goroutines.
 func (r reedSolomon) codeSomeShardsP(matrixRows, inputs, outputs [][]byte, outputCount, byteCount int) {
 	var wg sync.WaitGroup
-	do := byteCount / maxGoroutines
-	if do < minSplitSize {
-		do = minSplitSize
+	do := byteCount / r.o.maxGoroutines
+	if do < r.o.minSplitSize {
+		do = r.o.minSplitSize
 	}
 	start := 0
 	for start < byteCount {
@@ -241,9 +241,9 @@ func (r reedSolomon) codeSomeShardsP(matrixRows, inputs, outputs [][]byte, outpu
 				in := inputs[c]
 				for iRow := 0; iRow < outputCount; iRow++ {
 					if c == 0 {
-						galMulSlice(matrixRows[iRow][c], in[start:stop], outputs[iRow][start:stop])
+						galMulSlice(matrixRows[iRow][c], in[start:stop], outputs[iRow][start:stop], r.o.useSSSE3, r.o.useAVX2)
 					} else {
-						galMulSliceXor(matrixRows[iRow][c], in[start:stop], outputs[iRow][start:stop])
+						galMulSliceXor(matrixRows[iRow][c], in[start:stop], outputs[iRow][start:stop], r.o.useSSSE3, r.o.useAVX2)
 					}
 				}
 			}
@@ -258,13 +258,36 @@ func (r reedSolomon) codeSomeShardsP(matrixRows, inputs, outputs [][]byte, outpu
 // except this will check values and return
 // as soon as a difference is found.
 func (r reedSolomon) checkSomeShards(matrixRows, inputs, toCheck [][]byte, outputCount, byteCount int) bool {
+	if r.o.maxGoroutines > 1 && byteCount > r.o.minSplitSize {
+		return r.checkSomeShardsP(matrixRows, inputs, toCheck, outputCount, byteCount)
+	}
+	outputs := make([][]byte, len(toCheck))
+	for i := range outputs {
+		outputs[i] = make([]byte, byteCount)
+	}
+	for c := 0; c < r.DataShards; c++ {
+		in := inputs[c]
+		for iRow := 0; iRow < outputCount; iRow++ {
+			galMulSliceXor(matrixRows[iRow][c], in, outputs[iRow], r.o.useSSSE3, r.o.useAVX2)
+		}
+	}
+
+	for i, calc := range outputs {
+		if !bytes.Equal(calc, toCheck[i]) {
+			return false
+		}
+	}
+	return true
+}
+
+func (r reedSolomon) checkSomeShardsP(matrixRows, inputs, toCheck [][]byte, outputCount, byteCount int) bool {
 	same := true
 	var mu sync.RWMutex // For above
 
 	var wg sync.WaitGroup
-	do := byteCount / maxGoroutines
-	if do < minSplitSize {
-		do = minSplitSize
+	do := byteCount / r.o.maxGoroutines
+	if do < r.o.minSplitSize {
+		do = r.o.minSplitSize
 	}
 	start := 0
 	for start < byteCount {
@@ -287,7 +310,7 @@ func (r reedSolomon) checkSomeShards(matrixRows, inputs, toCheck [][]byte, outpu
 				mu.RUnlock()
 				in := inputs[c][start : start+do]
 				for iRow := 0; iRow < outputCount; iRow++ {
-					galMulSliceXor(matrixRows[iRow][c], in, outputs[iRow])
+					galMulSliceXor(matrixRows[iRow][c], in, outputs[iRow], r.o.useSSSE3, r.o.useAVX2)
 				}
 			}