libmp: 386/amd64 mpvec*(): replace conditional branches with ADC/SBB instructions

6 files changed, 21 insertions, 50 deletions

diff --git a/sys/src/libmp/386/mpvecadd.s b/sys/src/libmp/386/mpvecadd.s
index 50a45b140..598bf4d54 100644
--- a/sys/src/libmp/386/mpvecadd.s
+++ b/sys/src/libmp/386/mpvecadd.s
@@ -35,11 +35,8 @@ _add1:
 
 	/* sum[alen] = carry */
 _addend:
-	JC	_addcarry
-	MOVL	$0,(DI)(BP*4)
-	RET
-_addcarry:
-	MOVL	$1,(DI)(BP*4)
+	ADCL	$0, CX
+	MOVL	CX, (DI)(BP*4)
 	RET
 
 	/* sum[blen:alen-1],carry = a[blen:alen-1] + 0 */
diff --git a/sys/src/libmp/386/mpvecdigmuladd.s b/sys/src/libmp/386/mpvecdigmuladd.s
index 5a262ce8b..1a232ac91 100644
--- a/sys/src/libmp/386/mpvecdigmuladd.s
+++ b/sys/src/libmp/386/mpvecdigmuladd.s
@@ -36,13 +36,9 @@ _muladdloop:
 	MOVL	(SI)(BP*4),AX	/* lo = b[i] */
 	MULL	BX		/* hi, lo = b[i] * m */
 	ADDL	CX,AX		/* lo += oldhi */
-	JCC	_muladdnocarry1
-	INCL	DX		/* hi += carry */
-_muladdnocarry1:
+	ADCL	$0, DX		/* hi += carry */
 	ADDL	AX,(DI)(BP*4)	/* p[i] += lo */
-	JCC	_muladdnocarry2
-	INCL	DX		/* hi += carry */
-_muladdnocarry2:
+	ADCL	$0, DX		/* hi += carry */
 	MOVL	DX,CX		/* oldhi = hi */
 	INCL	BP		/* i++ */
 	JNZ	_muladdloop
diff --git a/sys/src/libmp/386/mpvecdigmulsub.s b/sys/src/libmp/386/mpvecdigmulsub.s
index 04cbfad30..42c88a72e 100644
--- a/sys/src/libmp/386/mpvecdigmulsub.s
+++ b/sys/src/libmp/386/mpvecdigmulsub.s
@@ -33,21 +33,15 @@ _mulsubloop:
 	MOVL	(SI)(BP*4),AX		/* lo = b[i] */
 	MULL	BX			/* hi, lo = b[i] * m */
 	ADDL	0(SP),AX		/* lo += oldhi */
-	JCC	_mulsubnocarry1
-	INCL	DX			/* hi += carry */
-_mulsubnocarry1:
+	ADCL	$0, DX			/* hi += carry */
 	SUBL	AX,(DI)(BP*4)
-	JCC	_mulsubnocarry2
-	INCL	DX			/* hi += carry */
-_mulsubnocarry2:
+	ADCL	$0, DX			/* hi += carry */
 	MOVL	DX,0(SP)
 	INCL	BP
 	LOOP	_mulsubloop
-	MOVL	0(SP),AX
-	SUBL	AX,(DI)(BP*4)
-	JCC	_mulsubnocarry3
-	MOVL	$-1,AX
-	RET
-_mulsubnocarry3:
-	MOVL	$1,AX
+	MOVL	CX, AX
+	MOVL	0(SP),BX
+	SUBL	BX,(DI)(BP*4)
+	SBBL	CX, AX
+	ORL	$1, AX
 	RET
diff --git a/sys/src/libmp/amd64/mpvecadd.s b/sys/src/libmp/amd64/mpvecadd.s
index 326f39dad..74a656c90 100644
--- a/sys/src/libmp/amd64/mpvecadd.s
+++ b/sys/src/libmp/amd64/mpvecadd.s
@@ -9,7 +9,6 @@ TEXT	mpvecadd(SB),$0
 
 	MOVL	alen+8(FP),DX
 	MOVL	blen+24(FP),CX
-/*	MOVL	a+0(FP),SI */
 	MOVQ	RARG, SI
 	MOVQ	b+16(FP),BX
 	SUBL	CX,DX
@@ -35,12 +34,10 @@ _add1:
 	LOOP	_addloop2
 
 	/* sum[alen] = carry */
+
 _addend:
-	JC	_addcarry
-	MOVL	$0,(DI)(BP*4)
-	RET
-_addcarry:
-	MOVL	$1,(DI)(BP*4)
+	ADCL	$0, CX
+	MOVL	CX,(DI)(BP*4)
 	RET
 
 	/* sum[blen:alen-1],carry = a[blen:alen-1] + 0 */
diff --git a/sys/src/libmp/amd64/mpvecdigmuladd.s b/sys/src/libmp/amd64/mpvecdigmuladd.s
index 6599a42c8..1ddd6487e 100644
--- a/sys/src/libmp/amd64/mpvecdigmuladd.s
+++ b/sys/src/libmp/amd64/mpvecdigmuladd.s
@@ -22,7 +22,6 @@
  */
 TEXT	mpvecdigmuladd(SB),$0
 
-/*	MOVQ	b+0(FP),SI	*/
 	MOVQ	RARG,SI
 	MOVL	n+8(FP),CX
 	MOVL	m+16(FP),BX
@@ -37,13 +36,9 @@ _muladdloop:
 	MOVL	(SI)(BP*4),AX	/* lo = b[i] */
 	MULL	BX		/* hi, lo = b[i] * m */
 	ADDL	CX,AX		/* lo += oldhi */
-	JCC	_muladdnocarry1
-	INCL	DX		/* hi += carry */
-_muladdnocarry1:
+	ADCL	$0, DX		/* hi += carry */
 	ADDL	AX,(DI)(BP*4)	/* p[i] += lo */
-	JCC	_muladdnocarry2
-	INCL	DX		/* hi += carry */
-_muladdnocarry2:
+	ADCL	$0, DX		/* hi += carry */
 	MOVL	DX,CX		/* oldhi = hi */
 	INCQ	BP		/* i++ */
 	JNZ	_muladdloop
diff --git a/sys/src/libmp/amd64/mpvecdigmulsub.s b/sys/src/libmp/amd64/mpvecdigmulsub.s
index 0b5a35761..635bc493e 100644
--- a/sys/src/libmp/amd64/mpvecdigmulsub.s
+++ b/sys/src/libmp/amd64/mpvecdigmulsub.s
@@ -22,8 +22,6 @@
  *		
  */
 TEXT	mpvecdigmulsub(SB),$0
-
-/*	MOVL	b+0(FP),SI	*/
 	MOVQ	RARG,SI
 	MOVL	n+8(FP),CX
 	MOVL	m+16(FP),BX
@@ -34,20 +32,14 @@ _mulsubloop:
 	MOVL	(SI)(BP*4),AX		/* lo = b[i] */
 	MULL	BX			/* hi, lo = b[i] * m */
 	ADDL	R8,AX		/* lo += oldhi */
-	JCC	_mulsubnocarry1
-	INCL	DX			/* hi += carry */
-_mulsubnocarry1:
+	ADCL	$0, DX		/* hi += carry */
 	SUBL	AX,(DI)(BP*4)
-	JCC	_mulsubnocarry2
-	INCL	DX			/* hi += carry */
-_mulsubnocarry2:
+	ADCL	$0, DX		/* hi += carry */
 	MOVL	DX,R8
 	INCL	BP
 	LOOP	_mulsubloop
+	MOVL	CX, AX
 	SUBL	R8,(DI)(BP*4)
-	JCC	_mulsubnocarry3
-	MOVQ	$-1,AX
-	RET
-_mulsubnocarry3:
-	MOVQ	$1,AX
+	SBBQ	CX, AX
+	ORQ	$1, AX
 	RET