Use faster division for matrix construction (#254)

* Use faster division for matrix construction
* Remove needless 0 check.

As note by @cristaloleg in #253 the first parameter is always 1 in uses of `galDivide`.

Implement a `galOneOver` that uses this fact and replace `galDivide`.

Make `expTable` fixed size and cast to uint8 to avoid bounds check on use.

Author: klauspost

galois.go

@@ -62,12 +62,12 @@ var logTable = [fieldSize]byte{
 
 /**
  * Inverse of the logarithm table.  Maps integer logarithms
- * to members of the field.  There is no entry for 255
- * because the highest log is 254.
+ * to members of the field.  Entry 255 is the same as entry 0 sue to mod 255.
  *
  * This table was generated by `go run gentables.go`
+ * Table has been truncated to 256 bytes, since no lookups are bigger.
  */
-var expTable = []byte{0x1, 0x2, 0x4, 0x8, 0x10, 0x20, 0x40, 0x80, 0x1d, 0x3a, 0x74, 0xe8, 0xcd, 0x87, 0x13, 0x26, 0x4c, 0x98, 0x2d, 0x5a, 0xb4, 0x75, 0xea, 0xc9, 0x8f, 0x3, 0x6, 0xc, 0x18, 0x30, 0x60, 0xc0, 0x9d, 0x27, 0x4e, 0x9c, 0x25, 0x4a, 0x94, 0x35, 0x6a, 0xd4, 0xb5, 0x77, 0xee, 0xc1, 0x9f, 0x23, 0x46, 0x8c, 0x5, 0xa, 0x14, 0x28, 0x50, 0xa0, 0x5d, 0xba, 0x69, 0xd2, 0xb9, 0x6f, 0xde, 0xa1, 0x5f, 0xbe, 0x61, 0xc2, 0x99, 0x2f, 0x5e, 0xbc, 0x65, 0xca, 0x89, 0xf, 0x1e, 0x3c, 0x78, 0xf0, 0xfd, 0xe7, 0xd3, 0xbb, 0x6b, 0xd6, 0xb1, 0x7f, 0xfe, 0xe1, 0xdf, 0xa3, 0x5b, 0xb6, 0x71, 0xe2, 0xd9, 0xaf, 0x43, 0x86, 0x11, 0x22, 0x44, 0x88, 0xd, 0x1a, 0x34, 0x68, 0xd0, 0xbd, 0x67, 0xce, 0x81, 0x1f, 0x3e, 0x7c, 0xf8, 0xed, 0xc7, 0x93, 0x3b, 0x76, 0xec, 0xc5, 0x97, 0x33, 0x66, 0xcc, 0x85, 0x17, 0x2e, 0x5c, 0xb8, 0x6d, 0xda, 0xa9, 0x4f, 0x9e, 0x21, 0x42, 0x84, 0x15, 0x2a, 0x54, 0xa8, 0x4d, 0x9a, 0x29, 0x52, 0xa4, 0x55, 0xaa, 0x49, 0x92, 0x39, 0x72, 0xe4, 0xd5, 0xb7, 0x73, 0xe6, 0xd1, 0xbf, 0x63, 0xc6, 0x91, 0x3f, 0x7e, 0xfc, 0xe5, 0xd7, 0xb3, 0x7b, 0xf6, 0xf1, 0xff, 0xe3, 0xdb, 0xab, 0x4b, 0x96, 0x31, 0x62, 0xc4, 0x95, 0x37, 0x6e, 0xdc, 0xa5, 0x57, 0xae, 0x41, 0x82, 0x19, 0x32, 0x64, 0xc8, 0x8d, 0x7, 0xe, 0x1c, 0x38, 0x70, 0xe0, 0xdd, 0xa7, 0x53, 0xa6, 0x51, 0xa2, 0x59, 0xb2, 0x79, 0xf2, 0xf9, 0xef, 0xc3, 0x9b, 0x2b, 0x56, 0xac, 0x45, 0x8a, 0x9, 0x12, 0x24, 0x48, 0x90, 0x3d, 0x7a, 0xf4, 0xf5, 0xf7, 0xf3, 0xfb, 0xeb, 0xcb, 0x8b, 0xb, 0x16, 0x2c, 0x58, 0xb0, 0x7d, 0xfa, 0xe9, 0xcf, 0x83, 0x1b, 0x36, 0x6c, 0xd8, 0xad, 0x47, 0x8e, 0x1, 0x2, 0x4, 0x8, 0x10, 0x20, 0x40, 0x80, 0x1d, 0x3a, 0x74, 0xe8, 0xcd, 0x87, 0x13, 0x26, 0x4c, 0x98, 0x2d, 0x5a, 0xb4, 0x75, 0xea, 0xc9, 0x8f, 0x3, 0x6, 0xc, 0x18, 0x30, 0x60, 0xc0, 0x9d, 0x27, 0x4e, 0x9c, 0x25, 0x4a, 0x94, 0x35, 0x6a, 0xd4, 0xb5, 0x77, 0xee, 0xc1, 0x9f, 0x23, 0x46, 0x8c, 0x5, 0xa, 0x14, 0x28, 0x50, 0xa0, 0x5d, 0xba, 0x69, 0xd2, 0xb9, 0x6f, 0xde, 0xa1, 0x5f, 0xbe, 0x61, 0xc2, 0x99, 0x2f, 0x5e, 0xbc, 0x65, 0xca, 0x89, 0xf, 0x1e, 0x3c, 0x78, 0xf0, 0xfd, 0xe7, 0xd3, 0xbb, 0x6b, 0xd6, 0xb1, 0x7f, 0xfe, 0xe1, 0xdf, 0xa3, 0x5b, 0xb6, 0x71, 0xe2, 0xd9, 0xaf, 0x43, 0x86, 0x11, 0x22, 0x44, 0x88, 0xd, 0x1a, 0x34, 0x68, 0xd0, 0xbd, 0x67, 0xce, 0x81, 0x1f, 0x3e, 0x7c, 0xf8, 0xed, 0xc7, 0x93, 0x3b, 0x76, 0xec, 0xc5, 0x97, 0x33, 0x66, 0xcc, 0x85, 0x17, 0x2e, 0x5c, 0xb8, 0x6d, 0xda, 0xa9, 0x4f, 0x9e, 0x21, 0x42, 0x84, 0x15, 0x2a, 0x54, 0xa8, 0x4d, 0x9a, 0x29, 0x52, 0xa4, 0x55, 0xaa, 0x49, 0x92, 0x39, 0x72, 0xe4, 0xd5, 0xb7, 0x73, 0xe6, 0xd1, 0xbf, 0x63, 0xc6, 0x91, 0x3f, 0x7e, 0xfc, 0xe5, 0xd7, 0xb3, 0x7b, 0xf6, 0xf1, 0xff, 0xe3, 0xdb, 0xab, 0x4b, 0x96, 0x31, 0x62, 0xc4, 0x95, 0x37, 0x6e, 0xdc, 0xa5, 0x57, 0xae, 0x41, 0x82, 0x19, 0x32, 0x64, 0xc8, 0x8d, 0x7, 0xe, 0x1c, 0x38, 0x70, 0xe0, 0xdd, 0xa7, 0x53, 0xa6, 0x51, 0xa2, 0x59, 0xb2, 0x79, 0xf2, 0xf9, 0xef, 0xc3, 0x9b, 0x2b, 0x56, 0xac, 0x45, 0x8a, 0x9, 0x12, 0x24, 0x48, 0x90, 0x3d, 0x7a, 0xf4, 0xf5, 0xf7, 0xf3, 0xfb, 0xeb, 0xcb, 0x8b, 0xb, 0x16, 0x2c, 0x58, 0xb0, 0x7d, 0xfa, 0xe9, 0xcf, 0x83, 0x1b, 0x36, 0x6c, 0xd8, 0xad, 0x47, 0x8e}
+var expTable = [256]byte{0x1, 0x2, 0x4, 0x8, 0x10, 0x20, 0x40, 0x80, 0x1d, 0x3a, 0x74, 0xe8, 0xcd, 0x87, 0x13, 0x26, 0x4c, 0x98, 0x2d, 0x5a, 0xb4, 0x75, 0xea, 0xc9, 0x8f, 0x3, 0x6, 0xc, 0x18, 0x30, 0x60, 0xc0, 0x9d, 0x27, 0x4e, 0x9c, 0x25, 0x4a, 0x94, 0x35, 0x6a, 0xd4, 0xb5, 0x77, 0xee, 0xc1, 0x9f, 0x23, 0x46, 0x8c, 0x5, 0xa, 0x14, 0x28, 0x50, 0xa0, 0x5d, 0xba, 0x69, 0xd2, 0xb9, 0x6f, 0xde, 0xa1, 0x5f, 0xbe, 0x61, 0xc2, 0x99, 0x2f, 0x5e, 0xbc, 0x65, 0xca, 0x89, 0xf, 0x1e, 0x3c, 0x78, 0xf0, 0xfd, 0xe7, 0xd3, 0xbb, 0x6b, 0xd6, 0xb1, 0x7f, 0xfe, 0xe1, 0xdf, 0xa3, 0x5b, 0xb6, 0x71, 0xe2, 0xd9, 0xaf, 0x43, 0x86, 0x11, 0x22, 0x44, 0x88, 0xd, 0x1a, 0x34, 0x68, 0xd0, 0xbd, 0x67, 0xce, 0x81, 0x1f, 0x3e, 0x7c, 0xf8, 0xed, 0xc7, 0x93, 0x3b, 0x76, 0xec, 0xc5, 0x97, 0x33, 0x66, 0xcc, 0x85, 0x17, 0x2e, 0x5c, 0xb8, 0x6d, 0xda, 0xa9, 0x4f, 0x9e, 0x21, 0x42, 0x84, 0x15, 0x2a, 0x54, 0xa8, 0x4d, 0x9a, 0x29, 0x52, 0xa4, 0x55, 0xaa, 0x49, 0x92, 0x39, 0x72, 0xe4, 0xd5, 0xb7, 0x73, 0xe6, 0xd1, 0xbf, 0x63, 0xc6, 0x91, 0x3f, 0x7e, 0xfc, 0xe5, 0xd7, 0xb3, 0x7b, 0xf6, 0xf1, 0xff, 0xe3, 0xdb, 0xab, 0x4b, 0x96, 0x31, 0x62, 0xc4, 0x95, 0x37, 0x6e, 0xdc, 0xa5, 0x57, 0xae, 0x41, 0x82, 0x19, 0x32, 0x64, 0xc8, 0x8d, 0x7, 0xe, 0x1c, 0x38, 0x70, 0xe0, 0xdd, 0xa7, 0x53, 0xa6, 0x51, 0xa2, 0x59, 0xb2, 0x79, 0xf2, 0xf9, 0xef, 0xc3, 0x9b, 0x2b, 0x56, 0xac, 0x45, 0x8a, 0x9, 0x12, 0x24, 0x48, 0x90, 0x3d, 0x7a, 0xf4, 0xf5, 0xf7, 0xf3, 0xfb, 0xeb, 0xcb, 0x8b, 0xb, 0x16, 0x2c, 0x58, 0xb0, 0x7d, 0xfa, 0xe9, 0xcf, 0x83, 0x1b, 0x36, 0x6c, 0xd8, 0xad, 0x47, 0x8e, 0x1}
 
 func galAdd(a, b byte) byte {
 	return a ^ b
@@ -883,6 +883,15 @@ func galDivide(a, b byte) byte {
 	if logResult < 0 {
 		logResult += 255
 	}
+	return expTable[uint8(logResult)]
+}
+
+// galOneOver is the same as galDivide(1, a).
+func galOneOver(a byte) byte {
+	if a == 0 {
+		panic("Argument 'divisor' is 0")
+	}
+	logResult := logTable[a] ^ 255
 	return expTable[logResult]
 }
 
@@ -902,7 +911,7 @@ func galExp(a byte, n int) byte {
 	for logResult >= 255 {
 		logResult -= 255
 	}
-	return expTable[logResult]
+	return expTable[uint8(logResult)]
 }
 
 func genAvx2Matrix(matrixRows [][]byte, inputs, inIdx, outputs int, dst []byte) []byte {

galois_test.go

@@ -62,6 +62,10 @@ func TestInverse(t *testing.T) {
 			if c != 1 {
 				t.Fatal("inverse div/mul", c, "!=", 1)
 			}
+			b2 := galOneOver(a)
+			if b != b2 {
+				t.Fatal("inverse div/mul", b, "!=", b2)
+			}
 		}
 	}
 }

matrix.go

@@ -231,7 +231,7 @@ func (m matrix) gaussianElimination() error {
 		}
 		// Scale to 1.
 		if m[r][r] != 1 {
-			scale := galDivide(1, m[r][r])
+			scale := galOneOver(m[r][r])
 			for c := 0; c < columns; c++ {
 				m[r][c] = galMultiply(m[r][c], scale)
 			}

reedsolomon.go

@@ -283,7 +283,7 @@ func buildMatrixJerasure(dataShards, totalShards int) (matrix, error) {
 		// Multiply by the inverted matrix (same as vm.Multiply(vm[0:dataShards].Invert()))
 		if vm[i][i] != 1 {
 			// Make vm[i][i] = 1 by dividing the column by vm[i][i]
-			tmp := galDivide(1, vm[i][i])
+			tmp := galOneOver(vm[i][i])
 			for j := 0; j < totalShards; j++ {
 				vm[j][i] = galMultiply(vm[j][i], tmp)
 			}
@@ -303,7 +303,7 @@ func buildMatrixJerasure(dataShards, totalShards int) (matrix, error) {
 	for j := 0; j < dataShards; j++ {
 		tmp := vm[dataShards][j]
 		if tmp != 1 {
-			tmp = galDivide(1, tmp)
+			tmp = galOneOver(tmp)
 			for i := dataShards; i < totalShards; i++ {
 				vm[i][j] = galMultiply(vm[i][j], tmp)
 			}
@@ -314,7 +314,7 @@ func buildMatrixJerasure(dataShards, totalShards int) (matrix, error) {
 	for i := dataShards + 1; i < totalShards; i++ {
 		tmp := vm[i][0]
 		if tmp != 1 {
-			tmp = galDivide(1, tmp)
+			tmp = galOneOver(tmp)
 			for j := 0; j < dataShards; j++ {
 				vm[i][j] = galMultiply(vm[i][j], tmp)
 			}