From c5dac2387d8c3e3de36bc9d69aaacbf748e8d395 Mon Sep 17 00:00:00 2001
From: frankw <fwessels@xs4all.nl>
Date: Sat, 25 Jun 2016 08:42:30 +0200
Subject: [PATCH] G2 macro completed

---
 block.go         | 28 ++++++++++++++--------------
 compress_amd64.s | 12 ++++++++++++
 2 files changed, 26 insertions(+), 14 deletions(-)

diff --git a/block.go b/block.go
index b23cccf..a2d0f86 100644
--- a/block.go
+++ b/block.go
@@ -66,20 +66,20 @@ func blocks(d *digest, p []uint8) {
 		v11 += v15
 		v7 ^= v11
 		v7 = v7<<(64-24) | v7>>24
-		//v2 += m[5]
-		//v2 += v6
-		//v14 ^= v2
-		//v14 = v14<<(64-16) | v14>>16
-		//v10 += v14
-		//v6 ^= v10
-		//v6 = v6<<(64-63) | v6>>63
-		//v3 += m[7]
-		//v3 += v7
-		//v15 ^= v3
-		//v15 = v15<<(64-16) | v15>>16
-		//v11 += v15
-		//v7 ^= v11
-		//v7 = v7<<(64-63) | v7>>63
+		v2 += m[5]
+		v2 += v6
+		v14 ^= v2
+		v14 = v14<<(64-16) | v14>>16
+		v10 += v14
+		v6 ^= v10
+		v6 = v6<<(64-63) | v6>>63
+		v3 += m[7]
+		v3 += v7
+		v15 ^= v3
+		v15 = v15<<(64-16) | v15>>16
+		v11 += v15
+		v7 ^= v11
+		v7 = v7<<(64-63) | v7>>63
 		v1 += m[3]
 		v1 += v5
 		v13 ^= v1
diff --git a/compress_amd64.s b/compress_amd64.s
index 6815511..8aae6ef 100644
--- a/compress_amd64.s
+++ b/compress_amd64.s
@@ -77,6 +77,7 @@ TEXT ·compressSSE(SB), 7, $0
     ///////////////////////////////////////////////////////////////////////////
 
     // LOAD_MSG_ ##r ##_1(b0, b1);
+    // LOAD_MSG_ ##r ##_2(b0, b1);
     //   (X12 used as additional temp register)
     MOVQ   message+0(FP), DX  // DX: &p (message)
     MOVOU   0(DX), X12        // X12 = m[0]+m[1]
@@ -86,6 +87,7 @@ TEXT ·compressSSE(SB), 7, $0
     BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6c; BYTE $0xc5   // VPUNPCKLQDQ  XMM8, XMM12, XMM13  /* m[0], m[2] */
     BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6c; BYTE $0xcf   // VPUNPCKLQDQ  XMM9, XMM14, XMM15  /* m[4], m[6] */
     BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xd5   // VPUNPCKHQDQ XMM10, XMM12, XMM13  /* m[1], m[3] */
+    BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xdf   // VPUNPCKHQDQ XMM11, XMM14, XMM15  /* m[5], m[7] */
 
     // Load shuffle value
     MOVQ   shffle+120(FP), SI // SI: &shuffle
@@ -109,15 +111,25 @@ TEXT ·compressSSE(SB), 7, $0
 
     // G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1);
     BYTE $0xc4; BYTE $0xc1; BYTE $0x79; BYTE $0xd4; BYTE $0xc2   // VPADDQ  XMM0,XMM0,XMM10  /* v0 += m[1], v1 += m[3] */
+    BYTE $0xc4; BYTE $0xc1; BYTE $0x71; BYTE $0xd4; BYTE $0xcb   // VPADDQ  XMM1,XMM1,XMM11  /* v2 += m[5], v3 += m[7] */
     BYTE $0xc5; BYTE $0xf9; BYTE $0xd4; BYTE $0xc2               // VPADDQ  XMM0,XMM0,XMM2   /* v0 += v4, v1 += v5 */
+    BYTE $0xc5; BYTE $0xf1; BYTE $0xd4; BYTE $0xcb               // VPADDQ  XMM1,XMM1,XMM3   /* v2 += v6, v3 += v7 */
     BYTE $0xc5; BYTE $0xc9; BYTE $0xef; BYTE $0xf0               // VPXOR   XMM6,XMM6,XMM0   /* v12 ^= v0, v13 ^= v1 */
+    BYTE $0xc5; BYTE $0xc1; BYTE $0xef; BYTE $0xf9               // VPXOR   XMM7,XMM7,XMM1   /* v14 ^= v2, v15 ^= v3 */
     BYTE $0xc5; BYTE $0xfb; BYTE $0x70; BYTE $0xf6; BYTE $0x39   // VPSHUFLW XMM6,XMM6,0x39  /* combined with next ... */
     BYTE $0xc5; BYTE $0xfa; BYTE $0x70; BYTE $0xf6; BYTE $0x39   // VPSHUFHW XMM6,XMM6,0x39  /* v12 = v12<<(64-16) | v12>>16, v13 = v13<<(64-16) | v13>>16 */
+    BYTE $0xc5; BYTE $0xfb; BYTE $0x70; BYTE $0xff; BYTE $0x39   // VPSHUFLW XMM7,XMM7,0x39  /* combined with next ... */
+    BYTE $0xc5; BYTE $0xfa; BYTE $0x70; BYTE $0xff; BYTE $0x39   // VPSHUFHW XMM7,XMM7,0x39  /* v14 = v14<<(64-16) | v14>>16, v15 = v15<<(64-16) | v15>>16 */
     BYTE $0xc5; BYTE $0xd9; BYTE $0xd4; BYTE $0xe6               // VPADDQ  XMM4,XMM4,XMM6   /* v8 += v12, v9 += v13 */
+    BYTE $0xc5; BYTE $0xd1; BYTE $0xd4; BYTE $0xef               // VPADDQ  XMM5,XMM5,XMM7   /* v10 += v14, v11 += v15 */
     BYTE $0xc5; BYTE $0xe9; BYTE $0xef; BYTE $0xd4               // VPXOR   XMM2,XMM2,XMM4   /* v4 ^= v8, v5 ^= v9 */
+    BYTE $0xc5; BYTE $0xe1; BYTE $0xef; BYTE $0xdd               // VPXOR   XMM3,XMM3,XMM5   /* v6 ^= v10, v7 ^= v11 */
     BYTE $0xc5; BYTE $0x69; BYTE $0xd4; BYTE $0xfa               // VPADDQ  XMM15,XMM2,XMM2  /* temp reg = reg*2   */
     BYTE $0xc5; BYTE $0xe9; BYTE $0x73; BYTE $0xd2; BYTE $0x3f   // VPSRLQ  XMM2,XMM2,0x3f   /*      reg = reg>>63 */
     BYTE $0xc4; BYTE $0xc1; BYTE $0x69; BYTE $0xef; BYTE $0xd7   // VPXOR   XMM2,XMM2,XMM15  /* ORed together: v4 = v4<<(64-63) | v4>>63, v5 = v5<<(64-63) | v5>>63 */
+    BYTE $0xc5; BYTE $0x61; BYTE $0xd4; BYTE $0xfb               // VPADDQ XMM15,XMM3,XMM3   /* temp reg = reg*2   */
+    BYTE $0xc5; BYTE $0xe1; BYTE $0x73; BYTE $0xd3; BYTE $0x3f   // VPSRLQ XMM3,XMM3,0x3f    /*      reg = reg>>63 */
+    BYTE $0xc4; BYTE $0xc1; BYTE $0x61; BYTE $0xef; BYTE $0xdf   // VPXOR  XMM3,XMM3,XMM15   /* ORed together: v6 = v6<<(64-63) | v6>>63, v7 = v7<<(64-63) | v7>>63 */
 
     // Reload digest
     MOVQ   in+24(FP),  SI     // SI: &in