move amd64 code for generic AES functions to separate files

This commit moves AMD64 code for generic AES functions - like
key-schedule and encrypting 128 bit blocks to separate files:
 - aes_amd64.go
 - aes_amd64.s

aes_amd64.go

@@ -0,0 +1,14 @@
+// Copyright (c) 2018 Andreas Auernhammer. All rights reserved.
+// Use of this source code is governed by a license that can be
+// found in the LICENSE file.
+
+// +build amd64,!gccgo,!appengine
+
+package siv
+
+// keySchedule performs an AES key-schedule and is implemented in aes_amd64.s
+func keySchedule(keys, key []byte)
+
+// encryptBlock encrypts one 128 bit block from src to dst using AES and is
+// implemented in aes_amd64.s
+func encryptBlock(dst, src, keys []byte, keyLen uint64)

aes_amd64.s

@@ -0,0 +1,63 @@
+// Copyright (c) 2018 Andreas Auernhammer. All rights reserved.
+// Use of this source code is governed by a license that can be
+// found in the LICENSE file.
+
+// +build amd64,!gccgo,!appengine
+
+#include "aes_macros_amd64.s"
+
+// func keySchedule(keys []uint32, key []byte)
+TEXT ·keySchedule(SB), 4, $0-48
+	MOVQ keys+0(FP), AX
+	MOVQ key+24(FP), BX
+	MOVQ keyLen+32(FP), DX
+
+	CMPQ DX, $24
+	JE   aes_192
+	JB   aes_128
+
+aes_256:
+	MOVUPS (0 * 16)(BX), X0
+	MOVUPS (1 * 16)(BX), X1
+	AES_KEY_SCHEDULE_256(AX, X0, X1, X2, X3)
+	JMP    return
+
+aes_192:
+	MOVUPS (0 * 16)(BX), X0
+	MOVQ   (1 * 16)(BX), X1
+	AES_KEY_SCHEDULE_192(AX, X0, X1, X2, X3, X4, X5, X6)
+	JMP    return
+
+aes_128:
+	MOVUPS (0 * 16)(BX), X0
+	AES_KEY_SCHEDULE_128(AX, X0, X1, X2)
+
+return:
+	RET
+
+// func encryptBlock(dst, src, keys []byte, keyLen uint64)
+TEXT ·encryptBlock(SB), 4, $0-80
+	MOVQ dst+0(FP), DI
+	MOVQ src+24(FP), SI
+	MOVQ keys+48(FP), AX
+	MOVQ keyLen+72(FP), DX
+
+	MOVUPS (0 * 16)(SI), X0
+	CMPQ DX, $24
+	JE   aes_192
+	JB   aes_128
+
+aes_256:
+	AES_256(X0, X1, AX)
+	JMP    return
+
+aes_192:
+	AES_192(X0, X1, AX)
+	JMP    return
+
+aes_128:
+	AES_128(X0, X1, AX)
+
+return:
+	MOVUPS X0, (0 * 16)(DI)
+	RET

aes_cmac_amd64.go

@@ -14,8 +14,6 @@ import (
 	"golang.org/x/sys/cpu"
 )
 
-func keySchedule(keys, key []byte)
-
 func aesCMacXORKeyStream(dst, src, iv, keys []byte, keyLen uint64)
 
 func newCMAC(key []byte) aead {

aes_cmac_amd64.s

@@ -4,7 +4,6 @@
 
 // +build amd64,!gccgo,!appengine
 
-#include "textflag.h"
 #include "aes_macros_amd64.s"
 
 #define LOAD_COUNTER(C, c0, c1, T) \
@@ -21,35 +20,6 @@
 	BSWAPQ c1;     \
 	BSWAPQ c0
 
-// func keySchedule(keys []uint32, key []byte)
-TEXT ·keySchedule(SB), 4, $0-48
-	MOVQ keys+0(FP), AX
-	MOVQ key+24(FP), BX
-	MOVQ keyLen+32(FP), DX
-
-	CMPQ DX, $24
-	JE   aes_192
-	JB   aes_128
-
-aes_256:
-	MOVUPS (0 * 16)(BX), X0
-	MOVUPS (1 * 16)(BX), X1
-	AES_KEY_SCHEDULE_256(AX, X0, X1, X2, X3)
-	JMP    return
-
-aes_192:
-	MOVUPS (0 * 16)(BX), X0
-	MOVQ   (1 * 16)(BX), X1
-	AES_KEY_SCHEDULE_192(AX, X0, X1, X2, X3, X4, X5, X6)
-	JMP    return
-
-aes_128:
-	MOVUPS (0 * 16)(BX), X0
-	AES_KEY_SCHEDULE_128(AX, X0, X1, X2)
-
-return:
-	RET
-
 // func aesCMacXORKeyStream(dst, src, iv, keys []byte, keyLen uint64)
 TEXT ·aesCMacXORKeyStream(SB), 4, $0-104
 	MOVQ dst+0(FP), DI