big: assembly routines for 386 long shifts

R=rsc
CC=golang-dev
https://golang.org/cl/974043

src/pkg/big/arith_386.s

@@ -11,8 +11,8 @@ TEXT ·addVV(SB),7,$0
 	MOVL x+4(FP), SI
 	MOVL y+8(FP), CX
 	MOVL n+12(FP), BP
-	MOVL $0, BX         // i = 0
-	MOVL $0, DX         // c = 0
+	MOVL $0, BX		// i = 0
+	MOVL $0, DX		// c = 0
 	JMP E1
 
 L1:	MOVL (SI)(BX*4), AX
@@ -20,7 +20,7 @@ L1:	MOVL (SI)(BX*4), AX
 	ADCL (CX)(BX*4), AX
 	RCLL $1, DX
 	MOVL AX, (DI)(BX*4)
-	ADDL $1, BX			// i++
+	ADDL $1, BX		// i++
 
 E1:	CMPL BX, BP		// i < n
 	JL L1
@@ -36,8 +36,8 @@ TEXT ·subVV(SB),7,$0
 	MOVL x+4(FP), SI
 	MOVL y+8(FP), CX
 	MOVL n+12(FP), BP
-	MOVL $0, BX         // i = 0
-	MOVL $0, DX         // c = 0
+	MOVL $0, BX		// i = 0
+	MOVL $0, DX		// c = 0
 	JMP E2
 
 L2:	MOVL (SI)(BX*4), AX
@@ -45,9 +45,9 @@ L2:	MOVL (SI)(BX*4), AX
 	SBBL (CX)(BX*4), AX
 	RCLL $1, DX
 	MOVL AX, (DI)(BX*4)
-	ADDL $1, BX         // i++
+	ADDL $1, BX		// i++
 
-E2:	CMPL BX, BP        // i < n
+E2:	CMPL BX, BP		// i < n
 	JL L2
 
 	MOVL DX, c+16(FP)
@@ -58,18 +58,18 @@ E2:	CMPL BX, BP        // i < n
 TEXT ·addVW(SB),7,$0
 	MOVL z+0(FP), DI
 	MOVL x+4(FP), SI
-	MOVL y+8(FP), AX   // c = y
+	MOVL y+8(FP), AX	// c = y
 	MOVL n+12(FP), BP
-	MOVL $0, BX         // i = 0
+	MOVL $0, BX		// i = 0
 	JMP E3
 
 L3:	ADDL (SI)(BX*4), AX
 	MOVL AX, (DI)(BX*4)
 	RCLL $1, AX
 	ANDL $1, AX
-	ADDL $1, BX         // i++
+	ADDL $1, BX		// i++
 
-E3:	CMPL BX, BP        // i < n
+E3:	CMPL BX, BP		// i < n
 	JL L3
 
 	MOVL AX, c+16(FP)
@@ -80,9 +80,9 @@ E3:	CMPL BX, BP        // i < n
 TEXT ·subVW(SB),7,$0
 	MOVL z+0(FP), DI
 	MOVL x+4(FP), SI
-	MOVL y+8(FP), AX   // c = y
+	MOVL y+8(FP), AX	// c = y
 	MOVL n+12(FP), BP
-	MOVL $0, BX         // i = 0
+	MOVL $0, BX		// i = 0
 	JMP E4
 
 L4:	MOVL (SI)(BX*4), DX	// TODO(gri) is there a reverse SUBL?
@@ -90,9 +90,9 @@ L4:	MOVL (SI)(BX*4), DX	// TODO(gri) is there a reverse SUBL?
 	MOVL DX, (DI)(BX*4)
 	RCLL $1, AX
 	ANDL $1, AX
-	ADDL $1, BX          // i++
+	ADDL $1, BX		// i++
 
-E4:	CMPL BX, BP         // i < n
+E4:	CMPL BX, BP		// i < n
 	JL L4
 
 	MOVL AX, c+16(FP)
@@ -100,17 +100,52 @@ E4:	CMPL BX, BP         // i < n
 
 
 // func shlVW(z, x *Word, s Word, n int) (c Word)
-// TODO(gri) implement this routine
 TEXT ·shlVW(SB),7,$0
-	NOP			// work around bug in linker
-	JMP ·shlVW_g(SB)
+	MOVL z+0(FP), DI
+	MOVL x+4(FP), SI
+	MOVL s+8(FP), CX
+	MOVL n+12(FP), BP
+	MOVL $0, AX		// c = 0
+	MOVL $0, BX		// i = 0
+	JMP E8
+
+L8:	MOVL (SI)(BX*8), DX
+	SHLL CX, DX:AX
+	MOVL DX, (DI)(BX*8)
+	MOVL (SI)(BX*8), AX	// reload (not enough regs to save original DX)
+	ADDL $1, BX		// i++
+
+E8:	CMPL BX, BP		// i < n
+	JL L8
+
+	MOVL $0, DX
+	SHLL CX, DX:AX
+	MOVL DX, c+16(FP)
+	RET
 
 
 // func shrVW(z, x *Word, s Word, n int) (c Word)
-// TODO(gri) implement this routine
 TEXT ·shrVW(SB),7,$0
-	NOP			// work around bug in linker
-	JMP ·shrVW_g(SB)
+	MOVL z+0(FP), DI
+	MOVL x+4(FP), SI
+	MOVL s+8(FP), CX
+	MOVL n+12(FP), BX	// i = n
+	MOVL $0, AX		// c = 0
+	JMP E9
+
+L9:	MOVL (SI)(BX*8), DX
+	MOVL DX, BP
+	SHRL CX, DX:AX
+	MOVL DX, (DI)(BX*8)
+	MOVL BP, AX
+
+E9:	SUBL $1, BX		// i--
+	JGE L9
+
+	MOVL $0, DX
+	SHRL CX, DX:AX
+	MOVL DX, c+16(FP)
+	RET
 
 
 // func mulAddVWW(z, x *Word, y, r Word, n int) (c Word)
@@ -118,7 +153,7 @@ TEXT ·mulAddVWW(SB),7,$0
 	MOVL z+0(FP), DI
 	MOVL x+4(FP), SI
 	MOVL y+8(FP), BP
-	MOVL r+12(FP), CX   // c = r
+	MOVL r+12(FP), CX	// c = r
 	MOVL n+16(FP), BX
 	LEAL	(SI)(BX*4), SI
 	LEAL	(DI)(BX*4), DI
@@ -131,9 +166,9 @@ L5:	MOVL (SI)(BX*4), AX
 	ADCL $0, DX
 	MOVL AX, (DI)(BX*4)
 	MOVL DX, CX
-	ADDL $1, BX         // i++
+	ADDL $1, BX		// i++
 
-E5:	CMPL BX, $0        // i < 0
+E5:	CMPL BX, $0		// i < 0
 	JL L5
 
 	MOVL CX, c+20(FP)
@@ -148,8 +183,8 @@ TEXT ·addMulVVW(SB),7,$0
 	MOVL n+12(FP), BX
 	LEAL	(SI)(BX*4), SI
 	LEAL	(DI)(BX*4), DI
-	NEGL BX  // i = -n
-	MOVL $0, CX         // c = 0
+	NEGL BX			// i = -n
+	MOVL $0, CX		// c = 0
 	JMP E6
 
 L6:	MOVL (SI)(BX*4), AX
@@ -160,9 +195,9 @@ L6:	MOVL (SI)(BX*4), AX
 	ADCL $0, DX
 	MOVL AX, (DI)(BX*4)
 	MOVL DX, CX
-	ADDL $1, BX         // i++
+	ADDL $1, BX		// i++
 
-E6:	CMPL BX, $0        // i < 0
+E6:	CMPL BX, $0		// i < 0
 	JL L6
 
 	MOVL CX, c+16(FP)
@@ -172,18 +207,18 @@ E6:	CMPL BX, $0        // i < 0
 // divWVW(z* Word, xn Word, x *Word, y Word, n int) (r Word)
 TEXT ·divWVW(SB),7,$0
 	MOVL z+0(FP), DI
-	MOVL xn+4(FP), DX   // r = xn
+	MOVL xn+4(FP), DX	// r = xn
 	MOVL x+8(FP), SI
 	MOVL y+12(FP), CX
-	MOVL n+16(FP), BX   // i = n
+	MOVL n+16(FP), BX	// i = n
 	JMP E7
 
 L7:	MOVL (SI)(BX*4), AX
 	DIVL CX
 	MOVL AX, (DI)(BX*4)
 
-E7:	SUBL $1, BX         // i--
-	JGE L7              // i >= 0
+E7:	SUBL $1, BX		// i--
+	JGE L7			// i >= 0
 
 	MOVL DX, r+20(FP)
 	RET