cmd/asm: add amd64 HADDPD/HADDPS

Was part of #13822 but not in the first message, so I missed it.

Fixes #13822 again.

Change-Id: I775004fa8d47b6af293124605521ec396573e267
Reviewed-on: https://go-review.googlesource.com/18900
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

4 files changed, 44 insertions, 32 deletions

diff --git a/src/cmd/asm/internal/asm/testdata/amd64enc.s b/src/cmd/asm/internal/asm/testdata/amd64enc.s
index 63ba7cafea..63fdcac27d 100644
--- a/src/cmd/asm/internal/asm/testdata/amd64enc.s
+++ b/src/cmd/asm/internal/asm/testdata/amd64enc.s
@@ -1878,39 +1878,39 @@ TEXT asmtest(SB),7,$0
 	FXTRACT                                 // d9f4
 	FYL2X                                   // d9f1
 	FYL2XP1                                 // d9f9
-	//TODO: HADDPD (BX), X2                 // 660f7c13
-	//TODO: HADDPD (R11), X2                // 66410f7c13
-	//TODO: HADDPD X2, X2                   // 660f7cd2
-	//TODO: HADDPD X11, X2                  // 66410f7cd3
-	//TODO: HADDPD (BX), X11                // 66440f7c1b
-	//TODO: HADDPD (R11), X11               // 66450f7c1b
-	//TODO: HADDPD X2, X11                  // 66440f7cda
-	//TODO: HADDPD X11, X11                 // 66450f7cdb
-	//TODO: HADDPS (BX), X2                 // f20f7c13
-	//TODO: HADDPS (R11), X2                // f2410f7c13
-	//TODO: HADDPS X2, X2                   // f20f7cd2
-	//TODO: HADDPS X11, X2                  // f2410f7cd3
-	//TODO: HADDPS (BX), X11                // f2440f7c1b
-	//TODO: HADDPS (R11), X11               // f2450f7c1b
-	//TODO: HADDPS X2, X11                  // f2440f7cda
-	//TODO: HADDPS X11, X11                 // f2450f7cdb
+	HADDPD (BX), X2                         // 660f7c13
+	HADDPD (R11), X2                        // 66410f7c13
+	HADDPD X2, X2                           // 660f7cd2
+	HADDPD X11, X2                          // 66410f7cd3
+	HADDPD (BX), X11                        // 66440f7c1b
+	HADDPD (R11), X11                       // 66450f7c1b
+	HADDPD X2, X11                          // 66440f7cda
+	HADDPD X11, X11                         // 66450f7cdb
+	HADDPS (BX), X2                         // f20f7c13
+	HADDPS (R11), X2                        // f2410f7c13
+	HADDPS X2, X2                           // f20f7cd2
+	HADDPS X11, X2                          // f2410f7cd3
+	HADDPS (BX), X11                        // f2440f7c1b
+	HADDPS (R11), X11                       // f2450f7c1b
+	HADDPS X2, X11                          // f2440f7cda
+	HADDPS X11, X11                         // f2450f7cdb
 	HLT                                     // f4
-	//TODO: HSUBPD (BX), X2                 // 660f7d13
-	//TODO: HSUBPD (R11), X2                // 66410f7d13
-	//TODO: HSUBPD X2, X2                   // 660f7dd2
-	//TODO: HSUBPD X11, X2                  // 66410f7dd3
-	//TODO: HSUBPD (BX), X11                // 66440f7d1b
-	//TODO: HSUBPD (R11), X11               // 66450f7d1b
-	//TODO: HSUBPD X2, X11                  // 66440f7dda
-	//TODO: HSUBPD X11, X11                 // 66450f7ddb
-	//TODO: HSUBPS (BX), X2                 // f20f7d13
-	//TODO: HSUBPS (R11), X2                // f2410f7d13
-	//TODO: HSUBPS X2, X2                   // f20f7dd2
-	//TODO: HSUBPS X11, X2                  // f2410f7dd3
-	//TODO: HSUBPS (BX), X11                // f2440f7d1b
-	//TODO: HSUBPS (R11), X11               // f2450f7d1b
-	//TODO: HSUBPS X2, X11                  // f2440f7dda
-	//TODO: HSUBPS X11, X11                 // f2450f7ddb
+	HSUBPD (BX), X2                         // 660f7d13
+	HSUBPD (R11), X2                        // 66410f7d13
+	HSUBPD X2, X2                           // 660f7dd2
+	HSUBPD X11, X2                          // 66410f7dd3
+	HSUBPD (BX), X11                        // 66440f7d1b
+	HSUBPD (R11), X11                       // 66450f7d1b
+	HSUBPD X2, X11                          // 66440f7dda
+	HSUBPD X11, X11                         // 66450f7ddb
+	HSUBPS (BX), X2                         // f20f7d13
+	HSUBPS (R11), X2                        // f2410f7d13
+	HSUBPS X2, X2                           // f20f7dd2
+	HSUBPS X11, X2                          // f2410f7dd3
+	HSUBPS (BX), X11                        // f2440f7d1b
+	HSUBPS (R11), X11                       // f2450f7d1b
+	HSUBPS X2, X11                          // f2440f7dda
+	HSUBPS X11, X11                         // f2450f7ddb
 	//TODO: ICEBP                           // f1
 	IDIVW (BX)                              // 66f73b
 	IDIVW (R11)                             // 6641f73b
diff --git a/src/cmd/internal/obj/x86/a.out.go b/src/cmd/internal/obj/x86/a.out.go
index efce9867d5..12eaa90bf6 100644
--- a/src/cmd/internal/obj/x86/a.out.go
+++ b/src/cmd/internal/obj/x86/a.out.go
@@ -89,7 +89,11 @@ const (
 	ADIVL
 	ADIVW
 	AENTER
+	AHADDPD
+	AHADDPS
 	AHLT
+	AHSUBPD
+	AHSUBPS
 	AIDIVB
 	AIDIVL
 	AIDIVW
diff --git a/src/cmd/internal/obj/x86/anames.go b/src/cmd/internal/obj/x86/anames.go
index b1c3f00f7c..1875eae418 100644
--- a/src/cmd/internal/obj/x86/anames.go
+++ b/src/cmd/internal/obj/x86/anames.go
@@ -57,7 +57,11 @@ var Anames = []string{
 	"DIVL",
 	"DIVW",
 	"ENTER",
+	"HADDPD",
+	"HADDPS",
 	"HLT",
+	"HSUBPD",
+	"HSUBPS",
 	"IDIVB",
 	"IDIVL",
 	"IDIVW",
diff --git a/src/cmd/internal/obj/x86/asm6.go b/src/cmd/internal/obj/x86/asm6.go
index fdc25faf98..4ed1d8790b 100644
--- a/src/cmd/internal/obj/x86/asm6.go
+++ b/src/cmd/internal/obj/x86/asm6.go
@@ -1171,6 +1171,10 @@ var optab =
 	{AJPC, yjcond, Px, [23]uint8{0x7b, 0x8b}},
 	{AJPL, yjcond, Px, [23]uint8{0x79, 0x89}},
 	{AJPS, yjcond, Px, [23]uint8{0x7a, 0x8a}},
+	{AHADDPD, yxm, Pq, [23]uint8{0x7c}},
+	{AHADDPS, yxm, Pf2, [23]uint8{0x7c}},
+	{AHSUBPD, yxm, Pq, [23]uint8{0x7d}},
+	{AHSUBPS, yxm, Pf2, [23]uint8{0x7d}},
 	{ALAHF, ynone, Px, [23]uint8{0x9f}},
 	{ALARL, yml_rl, Pm, [23]uint8{0x02}},
 	{ALARW, yml_rl, Pq, [23]uint8{0x02}},