cmd/internal/obj/x86: add new instructions, cleanup.

Add several instructions that were used via BYTE and use them.
Instructions added: PEXTRB, PEXTRD, PEXTRQ, PINSRB, XGETBV, POPCNT.

Change-Id: I5a80cd390dc01f3555dbbe856a475f74b5e6df65
Reviewed-on: https://go-review.googlesource.com/18593
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

src/cmd/internal/obj/x86/a.out.go

@@ -181,6 +181,7 @@ const (
 	APAUSE
 	APOPAL
 	APOPAW
+	APOPCNT
 	APOPFL
 	APOPFW
 	APOPL
@@ -500,6 +501,7 @@ const (
 	AXADDQ
 	AXCHGQ
 	AXORQ
+	AXGETBV
 
 	// media
 	AADDPD
@@ -614,6 +616,9 @@ const (
 	APCMPGTL
 	APCMPGTW
 	APEXTRW
+	APEXTRB
+	APEXTRD
+	APEXTRQ
 	APFACC
 	APFADD
 	APFCMPEQ
@@ -632,6 +637,7 @@ const (
 	APFSUB
 	APFSUBR
 	APINSRW
+	APINSRB
 	APINSRD
 	APINSRQ
 	APMADDWL

src/cmd/internal/obj/x86/anames.go

@@ -149,6 +149,7 @@ var Anames = []string{
 	"PAUSE",
 	"POPAL",
 	"POPAW",
+	"POPCNT",
 	"POPFL",
 	"POPFW",
 	"POPL",
@@ -451,6 +452,7 @@ var Anames = []string{
 	"XADDQ",
 	"XCHGQ",
 	"XORQ",
+	"XGETBV",
 	"ADDPD",
 	"ADDPS",
 	"ADDSD",
@@ -563,6 +565,9 @@ var Anames = []string{
 	"PCMPGTL",
 	"PCMPGTW",
 	"PEXTRW",
+	"PEXTRB",
+	"PEXTRD",
+	"PEXTRQ",
 	"PFACC",
 	"PFADD",
 	"PFCMPEQ",
@@ -581,6 +586,7 @@ var Anames = []string{
 	"PFSUB",
 	"PFSUBR",
 	"PINSRW",
+	"PINSRB",
 	"PINSRD",
 	"PINSRQ",
 	"PMADDWL",

src/cmd/internal/obj/x86/asm6.go

@@ -187,6 +187,7 @@ const (
 	Zm_r_xm_nr
 	Zr_m_xm_nr
 	Zibm_r /* mmx1,mmx2/mem64,imm8 */
+	Zibr_m
 	Zmb_r
 	Zaut_r
 	Zo_m
@@ -219,6 +220,7 @@ const (
 	Pf2   = 0xf2 /* xmm escape 1: f2 0f */
 	Pf3   = 0xf3 /* xmm escape 2: f3 0f */
 	Pq3   = 0x67 /* xmm escape 3: 66 48 0f */
+	Pfw   = 0xf4 /* Pf3 with Rex.w: f3 48 0f */
 	Pvex1 = 0xc5 /* 66.0f escape, vex encoding */
 	Pvex2 = 0xc6 /* f3.0f escape, vex encoding */
 	Pvex3 = 0xc7 /* 66.0f38 escape, vex encoding */
@@ -720,6 +722,10 @@ var yextrw = []ytab{
 	{Yu8, Yxr, Yrl, Zibm_r, 2},
 }
 
+var yextr = []ytab{
+	{Yu8, Yxr, Ymm, Zibr_m, 3},
+}
+
 var yinsrw = []ytab{
 	{Yu8, Yml, Yxr, Zibm_r, 2},
 }
@@ -1162,6 +1168,9 @@ var optab =
 	{APCMPGTL, ymm, Py1, [23]uint8{0x66, Pe, 0x66}},
 	{APCMPGTW, ymm, Py1, [23]uint8{0x65, Pe, 0x65}},
 	{APEXTRW, yextrw, Pq, [23]uint8{0xc5, 00}},
+	{APEXTRB, yextr, Pq, [23]uint8{0x3a, 0x14, 00}},
+	{APEXTRD, yextr, Pq, [23]uint8{0x3a, 0x16, 00}},
+	{APEXTRQ, yextr, Pq3, [23]uint8{0x3a, 0x16, 00}},
 	{APF2IL, ymfp, Px, [23]uint8{0x1d}},
 	{APF2IW, ymfp, Px, [23]uint8{0x1c}},
 	{API2FL, ymfp, Px, [23]uint8{0x0d}},
@@ -1183,6 +1192,7 @@ var optab =
 	{APFSUB, ymfp, Px, [23]uint8{0x9a}},
 	{APFSUBR, ymfp, Px, [23]uint8{0xaa}},
 	{APINSRW, yinsrw, Pq, [23]uint8{0xc4, 00}},
+	{APINSRB, yinsr, Pq, [23]uint8{0x3a, 0x20, 00}},
 	{APINSRD, yinsr, Pq, [23]uint8{0x3a, 0x22, 00}},
 	{APINSRQ, yinsr, Pq3, [23]uint8{0x3a, 0x22, 00}},
 	{APMADDWL, ymm, Py1, [23]uint8{0xf5, Pe, 0xf5}},
@@ -1198,6 +1208,7 @@ var optab =
 	{APMULULQ, ymm, Py1, [23]uint8{0xf4, Pe, 0xf4}},
 	{APOPAL, ynone, P32, [23]uint8{0x61}},
 	{APOPAW, ynone, Pe, [23]uint8{0x61}},
+	{APOPCNT, yml_rl, Pfw, [23]uint8{0xb8}},
 	{APOPFL, ynone, P32, [23]uint8{0x9d}},
 	{APOPFQ, ynone, Py, [23]uint8{0x9d}},
 	{APOPFW, ynone, Pe, [23]uint8{0x9d}},
@@ -1533,6 +1544,7 @@ var optab =
 	{AXABORT, yxabort, Px, [23]uint8{0xc6, 0xf8}},
 	{AXEND, ynone, Px, [23]uint8{0x0f, 01, 0xd5}},
 	{AXTEST, ynone, Px, [23]uint8{0x0f, 01, 0xd6}},
+	{AXGETBV, ynone, Pm, [23]uint8{01, 0xd0}},
 	{obj.AUSEFIELD, ynop, Px, [23]uint8{0, 0}},
 	{obj.ATYPE, nil, 0, [23]uint8{}},
 	{obj.AFUNCDATA, yfuncdata, Px, [23]uint8{0, 0}},
@@ -3194,6 +3206,15 @@ func doasm(ctxt *obj.Link, p *obj.Prog) {
 				ctxt.Andptr[0] = Pm
 				ctxt.Andptr = ctxt.Andptr[1:]
 
+			case Pfw: /* first escape, Rex.w, and second escape */
+				ctxt.Andptr[0] = Pf3
+				ctxt.Andptr = ctxt.Andptr[1:]
+
+				ctxt.Andptr[0] = Pw
+				ctxt.Andptr = ctxt.Andptr[1:]
+				ctxt.Andptr[0] = Pm
+				ctxt.Andptr = ctxt.Andptr[1:]
+
 			case Pm: /* opcode escape */
 				ctxt.Andptr[0] = Pm
 				ctxt.Andptr = ctxt.Andptr[1:]
@@ -3343,7 +3364,7 @@ func doasm(ctxt *obj.Link, p *obj.Prog) {
 				ctxt.Andptr[0] = byte(op)
 				ctxt.Andptr = ctxt.Andptr[1:]
 
-			case Zibm_r:
+			case Zibm_r, Zibr_m:
 				for {
 					tmp1 := z
 					z++
@@ -3354,7 +3375,11 @@ func doasm(ctxt *obj.Link, p *obj.Prog) {
 					ctxt.Andptr[0] = byte(op)
 					ctxt.Andptr = ctxt.Andptr[1:]
 				}
+				if yt.zcase == Zibr_m {
+					asmand(ctxt, p, &p.To, p.From3)
+				} else {
 					asmand(ctxt, p, p.From3, &p.To)
+				}
 				ctxt.Andptr[0] = byte(p.From.Offset)
 				ctxt.Andptr = ctxt.Andptr[1:]
 

src/crypto/aes/asm_amd64.s

@@ -217,8 +217,6 @@ Lexp_dec_loop:
 	MOVUPS X0, 16(DX)
 	RET
 
-#define PSHUFD_X0_X0_ BYTE $0x66; BYTE $0x0f; BYTE $0x70; BYTE $0xc0
-#define PSHUFD_X1_X1_ BYTE $0x66; BYTE $0x0f; BYTE $0x70; BYTE $0xc9
 TEXT _expand_key_128<>(SB),NOSPLIT,$0
 	PSHUFD $0xff, X1, X1
 	SHUFPS $0x10, X0, X4
@@ -230,8 +228,6 @@ TEXT _expand_key_128<>(SB),NOSPLIT,$0
 	ADDQ $16, BX
 	RET
 
-#define PSLLDQ_X5_ BYTE $0x66; BYTE $0x0f; BYTE $0x73; BYTE $0xfd
-#define PSHUFD_X0_X3_ BYTE $0x66; BYTE $0x0f; BYTE $0x70; BYTE $0xd8
 TEXT _expand_key_192a<>(SB),NOSPLIT,$0
 	PSHUFD $0x55, X1, X1
 	SHUFPS $0x10, X0, X4
@@ -242,7 +238,7 @@ TEXT _expand_key_192a<>(SB),NOSPLIT,$0
 
 	MOVAPS X2, X5
 	MOVAPS X2, X6
-	PSLLDQ_X5_; BYTE $0x4
+	PSLLDQ $0x4, X5
 	PSHUFD $0xff, X0, X3
 	PXOR X3, X2
 	PXOR X5, X2
@@ -264,7 +260,7 @@ TEXT _expand_key_192b<>(SB),NOSPLIT,$0
 	PXOR X1, X0
 
 	MOVAPS X2, X5
-	PSLLDQ_X5_; BYTE $0x4
+	PSLLDQ $0x4, X5
 	PSHUFD $0xff, X0, X3
 	PXOR X3, X2
 	PXOR X5, X2

src/crypto/aes/gcm_amd64.s

@@ -345,7 +345,7 @@ TEXT ·gcmAesData(SB),NOSPLIT,$0
 	PXOR B0, B0
 	MOVQ (aut), B0
 	PINSRD $2, 8(aut), B0
-	BYTE $0x66; BYTE $0x0f; BYTE $0x3a; BYTE $0x20; BYTE $0x46; BYTE $0x0c; BYTE $0x0c  //PINSRB $12, 12(aut), B0
+	PINSRB $12, 12(aut), B0
 	XORQ autLen, autLen
 	JMP dataMul
 
@@ -404,7 +404,7 @@ dataEnd:
 dataLoadLoop:
 
 		PSLLDQ $1, B0
-		BYTE $0x66; BYTE $0x0f; BYTE $0x3a; BYTE $0x20; BYTE $0x06; BYTE $0x00   //PINSRB $0, (aut), B0
+		PINSRB $0, (aut), B0
 
 		LEAQ -1(aut), aut
 		DECQ autLen
@@ -892,7 +892,7 @@ encLast4:
 	PXOR B0, B0
 ptxLoadLoop:
 		PSLLDQ $1, B0
-		BYTE $0x66; BYTE $0x0f; BYTE $0x3a; BYTE $0x20; BYTE $0x06; BYTE $0x00  //PINSRB $0, (ptx), B0
+		PINSRB $0, (ptx), B0
 		LEAQ -1(ptx), ptx
 		DECQ ptxLen
 	JNE ptxLoadLoop
@@ -1264,7 +1264,7 @@ decLast3:
 	PXOR T1, B0
 
 ptxStoreLoop:
-		BYTE $0x66; BYTE $0x0f; BYTE $0x3a; BYTE $0x14; BYTE $0x06; BYTE $0x00  // PEXTRB $0, B0, (ptx)
+		PEXTRB $0, B0, (ptx)
 		PSRLDQ $1, B0
 		LEAQ 1(ptx), ptx
 		DECQ ptxLen

src/hash/crc32/crc32_amd64.s

@@ -225,9 +225,7 @@ finish:
 	PCLMULQDQ   $0, X0, X1
 	PXOR        X2, X1
 
-	/* PEXTRD   $1, X1, AX  (SSE 4.1) */
-	BYTE $0x66; BYTE $0x0f; BYTE $0x3a;
-	BYTE $0x16; BYTE $0xc8; BYTE $0x01;
+	PEXTRD	$1, X1, AX
 	MOVL        AX, ret+32(FP)
 
 	RET

src/runtime/asm_amd64.s

@@ -56,7 +56,7 @@ notintel:
 	JNE     noavx
 	MOVL    $0, CX
 	// For XGETBV, OSXSAVE bit is required and sufficient
-	BYTE $0x0F; BYTE $0x01; BYTE $0xD0
+	XGETBV
 	ANDL    $6, AX
 	CMPL    AX, $6 // Check for OS support of YMM registers
 	JNE     noavx
@@ -822,10 +822,10 @@ TEXT runtime·getcallersp(SB),NOSPLIT,$0-16
 TEXT runtime·cputicks(SB),NOSPLIT,$0-0
 	CMPB	runtime·lfenceBeforeRdtsc(SB), $1
 	JNE	mfence
-	BYTE	$0x0f; BYTE $0xae; BYTE $0xe8 // LFENCE
+	LFENCE
 	JMP	done
 mfence:
-	BYTE	$0x0f; BYTE $0xae; BYTE $0xf0 // MFENCE
+	MFENCE
 done:
 	RDTSC
 	SHLQ	$32, DX