cmd/asm: add amd64 EXTRACTPS instruction

Adds last missing SSE4 instruction.
Also introduces additional ytab set 'yextractps'.

See https://golang.org/cl/57470 that adds other SSE4 instructions
but skips this one due to 'yextractps'.

To make EXTRACTPS less "sloppy", Yu2 oclass added to forbid
usage of invalid offset values in immediate operand.

Part of the mission to add missing amd64 SSE4 instructions to Go asm.

Change-Id: I0e67e3497054f53257dd8eb4c6268da5118b4853
Reviewed-on: https://go-review.googlesource.com/57490
TryBot-Result: Gobot Gobot <gobot@golang.org>
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
Reviewed-by: Russ Cox <rsc@golang.org>

src/cmd/asm/internal/asm/testdata/amd64enc.s

@@ -1640,14 +1640,14 @@ TEXT asmtest(SB),DUPOK|NOSPLIT,$0
 	DPPS $7, X11, X11                       // 66450f3a40db07
 	EMMS                                    // 0f77
 	//TODO: ENTERQ $0x12, $0xf123           // c823f112
-	//TODO: EXTRACTPS $7, X2, (BX)          // 660f3a171307
-	//TODO: EXTRACTPS $7, X11, (BX)         // 66440f3a171b07
-	//TODO: EXTRACTPS $7, X2, (R11)         // 66410f3a171307
-	//TODO: EXTRACTPS $7, X11, (R11)        // 66450f3a171b07
-	//TODO: EXTRACTPS $7, X2, DX            // 660f3a17d207
-	//TODO: EXTRACTPS $7, X11, DX           // 66440f3a17da07
-	//TODO: EXTRACTPS $7, X2, R11           // 66410f3a17d307
-	//TODO: EXTRACTPS $7, X11, R11          // 66450f3a17db07
+	EXTRACTPS $0, X2, (BX)                  // 660f3a171300
+	EXTRACTPS $1, X11, (BX)                 // 66440f3a171b01
+	EXTRACTPS $2, X2, (R11)                 // 66410f3a171302
+	EXTRACTPS $3, X11, (R11)                // 66450f3a171b03
+	EXTRACTPS $3, X2, DX                    // 660f3a17d203
+	EXTRACTPS $2, X11, DX                   // 66440f3a17da02
+	EXTRACTPS $1, X2, R11                   // 66410f3a17d301
+	EXTRACTPS $0, X11, R11                  // 66450f3a17db00
 	F2XM1                                   // d9f0
 	FABS                                    // d9e1
 	FADDD F2, F0                            // d8c2

src/cmd/asm/internal/asm/testdata/amd64error.s

@@ -5,4 +5,6 @@
 TEXT errors(SB),$0
 	MOVL	foo<>(SB)(AX), AX	// ERROR "invalid instruction"
 	MOVL	(AX)(SP*1), AX		// ERROR "invalid instruction"
+	EXTRACTPS $4, X2, (BX)          // ERROR "invalid instruction"
+	EXTRACTPS $-1, X2, (BX)         // ERROR "invalid instruction"
 	RET

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

@@ -579,6 +579,7 @@ const (
 	ADPPD
 	ADPPS
 	AEMMS
+	AEXTRACTPS
 	AFXRSTOR
 	AFXRSTOR64
 	AFXSAVE

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

@@ -522,6 +522,7 @@ var Anames = []string{
 	"DPPD",
 	"DPPS",
 	"EMMS",
+	"EXTRACTPS",
 	"FXRSTOR",
 	"FXRSTOR64",
 	"FXSAVE",

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

@@ -86,6 +86,7 @@ const (
 	Ynone
 	Yi0 // $0
 	Yi1 // $1
+	Yu2 // $x, x fits in uint2
 	Yi8 // $x, x fits in int8
 	Yu8 // $x, x fits in uint8
 	Yu7 // $x, x in 0..127 (fits in both int8 and uint8)
@@ -894,6 +895,10 @@ var ymmxmm0f38 = []ytab{
 	{Zlitm_r, 5, argList{Yxm, Yxr}},
 }
 
+var yextractps = []ytab{
+	{Yu2, Yxr, Yml, Zibr_m, 2},
+}
+
 /*
  * You are doasm, holding in your hand a *obj.Prog with p.As set to, say,
  * ACRC32, and p.From and p.To as operands (obj.Addr).  The linker scans optab
@@ -1117,6 +1122,7 @@ var optab =
 	{ADPPD, yxshuf, Pq, [23]uint8{0x3a, 0x41, 0}},
 	{ADPPS, yxshuf, Pq, [23]uint8{0x3a, 0x40, 0}},
 	{AEMMS, ynone, Pm, [23]uint8{0x77}},
+	{AEXTRACTPS, yextractps, Pq, [23]uint8{0x3a, 0x17, 0}},
 	{AENTER, nil, 0, [23]uint8{}}, /* botch */
 	{AFXRSTOR, ysvrs_mo, Pm, [23]uint8{0xae, 01, 0xae, 01}},
 	{AFXSAVE, ysvrs_om, Pm, [23]uint8{0xae, 00, 0xae, 00}},
@@ -2049,25 +2055,33 @@ func instinit(ctxt *obj.Link) {
 		ycover[i*Ymax+i] = 1
 	}
 
+	ycover[Yi0*Ymax+Yu2] = 1
+	ycover[Yi1*Ymax+Yu2] = 1
+
 	ycover[Yi0*Ymax+Yi8] = 1
 	ycover[Yi1*Ymax+Yi8] = 1
+	ycover[Yu2*Ymax+Yi8] = 1
 	ycover[Yu7*Ymax+Yi8] = 1
 
 	ycover[Yi0*Ymax+Yu7] = 1
 	ycover[Yi1*Ymax+Yu7] = 1
+	ycover[Yu2*Ymax+Yu7] = 1
 
 	ycover[Yi0*Ymax+Yu8] = 1
 	ycover[Yi1*Ymax+Yu8] = 1
+	ycover[Yu2*Ymax+Yu8] = 1
 	ycover[Yu7*Ymax+Yu8] = 1
 
 	ycover[Yi0*Ymax+Ys32] = 1
 	ycover[Yi1*Ymax+Ys32] = 1
+	ycover[Yu2*Ymax+Ys32] = 1
 	ycover[Yu7*Ymax+Ys32] = 1
 	ycover[Yu8*Ymax+Ys32] = 1
 	ycover[Yi8*Ymax+Ys32] = 1
 
 	ycover[Yi0*Ymax+Yi32] = 1
 	ycover[Yi1*Ymax+Yi32] = 1
+	ycover[Yu2*Ymax+Yi32] = 1
 	ycover[Yu7*Ymax+Yi32] = 1
 	ycover[Yu8*Ymax+Yi32] = 1
 	ycover[Yi8*Ymax+Yi32] = 1
@@ -2076,6 +2090,7 @@ func instinit(ctxt *obj.Link) {
 	ycover[Yi0*Ymax+Yi64] = 1
 	ycover[Yi1*Ymax+Yi64] = 1
 	ycover[Yu7*Ymax+Yi64] = 1
+	ycover[Yu2*Ymax+Yi64] = 1
 	ycover[Yu8*Ymax+Yi64] = 1
 	ycover[Yi8*Ymax+Yi64] = 1
 	ycover[Ys32*Ymax+Yi64] = 1
@@ -2406,6 +2421,9 @@ func oclass(ctxt *obj.Link, p *obj.Prog, a *obj.Addr) int {
 		if v == 1 {
 			return Yi1
 		}
+		if v >= 0 && v <= 3 {
+			return Yu2
+		}
 		if v >= 0 && v <= 127 {
 			return Yu7
 		}