Rewrite feed-in code, reducing code size and making it work for v8plus.

2 files changed, 42 insertions, 83 deletions

diff --git a/mpn/sparc32/v8/addmul_1.asm b/mpn/sparc32/v8/addmul_1.asm
index 005209278..0bf1b247f 100644
--- a/mpn/sparc32/v8/addmul_1.asm
+++ b/mpn/sparc32/v8/addmul_1.asm
@@ -40,50 +40,32 @@ C s2_limb	o3
 
 ASM_START()
 PROLOGUE(mpn_addmul_1)
+	ld	[%o1+0],%o4
+	andcc	%o2,1,%g0
+	be	L(bx0)
+	andcc	%o2,2,%g0
+L(bx1):	be	L(01)
 	orcc	%g0,%g0,%g2
-	ld	[%o1+0],%o4	C 1
-
-	sll	%o2,4,%g1
-	and	%g1,(4-1)<<4,%g1
-ifdef(`PIC',
-`	mov	%o7,%g4		C Save return address register
-0:	call	1f
-	add	%o7,L(1)-0b,%g3
-1:	mov	%g4,%o7		C Restore return address register
-',
-`	sethi	%hi(L(1)),%g3
-	or	%g3,%lo(L(1)),%g3
-')
-	jmp	%g3+%g1
-	nop
-L(1):
-L(L00):	add	%o0,-4,%o0
-	b	L(loop00)	C 4, 8, 12, ...
-	add	%o1,-4,%o1
-	nop
-L(L01):	b	L(loop01)	C 1, 5, 9, ...
-	nop
-	nop
-	nop
-L(L10):	add	%o0,-12,%o0	C 2, 6, 10, ...
-	b	L(loop10)
-	add	%o1,4,%o1
-	nop
-L(L11):	add	%o0,-8,%o0	C 3, 7, 11, ...
-	b	L(loop11)
+L(b11):	add	%o0,-8,%o0
+	b	L(11)
 	add	%o1,-8,%o1
-	nop
+L(bx0):	be	L(b00)
+	orcc	%g0,%g0,%g2
+L(b10):	add	%o0,-12,%o0
+	b	L(10)
+	add	%o1,4,%o1
+L(b00):	add	%o0,-4,%o0
+	b	L(00)
+	add	%o1,-4,%o1
 
-L(loop):
-	addcc	%g3,%g2,%g3	C 1
+L(top):	addcc	%g3,%g2,%g3	C 1
 	ld	[%o1+4],%o4	C 2
 	rd	%y,%g2		C 1
 	addx	%g0,%g2,%g2
 	ld	[%o0+0],%g1	C 2
 	addcc	%g1,%g3,%g3
 	st	%g3,[%o0+0]	C 1
-L(loop00):
-	umul	%o4,%o3,%g3	C 2
+L(00):	umul	%o4,%o3,%g3	C 2
 	ld	[%o0+4],%g1	C 2
 	addxcc	%g3,%g2,%g3	C 2
 	ld	[%o1+8],%o4	C 3
@@ -92,8 +74,7 @@ L(loop00):
 	nop
 	addcc	%g1,%g3,%g3
 	st	%g3,[%o0+4]	C 2
-L(loop11):
-	umul	%o4,%o3,%g3	C 3
+L(11):	umul	%o4,%o3,%g3	C 3
 	addxcc	%g3,%g2,%g3	C 3
 	ld	[%o1+12],%o4	C 4
 	rd	%y,%g2		C 3
@@ -102,8 +83,7 @@ L(loop11):
 	ld	[%o0+8],%g1	C 2
 	addcc	%g1,%g3,%g3
 	st	%g3,[%o0+8]	C 3
-L(loop10):
-	umul	%o4,%o3,%g3	C 4
+L(10):	umul	%o4,%o3,%g3	C 4
 	addxcc	%g3,%g2,%g3	C 4
 	ld	[%o1+0],%o4	C 1
 	rd	%y,%g2		C 4
@@ -113,9 +93,8 @@ L(loop10):
 	st	%g3,[%o0+12]	C 4
 	add	%o0,16,%o0
 	addx	%g0,%g2,%g2
-L(loop01):
-	addcc	%o2,-4,%o2
-	bg	L(loop)
+L(01):	addcc	%o2,-4,%o2
+	bg	L(top)
 	umul	%o4,%o3,%g3	C 1
 
 	addcc	%g3,%g2,%g3	C 4
@@ -124,8 +103,7 @@ L(loop01):
 	ld	[%o0+0],%g1	C 2
 	addcc	%g1,%g3,%g3
 	st	%g3,[%o0+0]	C 4
-	addx	%g0,%g2,%o0
 
 	retl
-	 nop
+	addx	%g0,%g2,%o0
 EPILOGUE(mpn_addmul_1)
diff --git a/mpn/sparc32/v8/mul_1.asm b/mpn/sparc32/v8/mul_1.asm
index e26c853ae..d03a0e6c0 100644
--- a/mpn/sparc32/v8/mul_1.asm
+++ b/mpn/sparc32/v8/mul_1.asm
@@ -40,67 +40,48 @@ C s2_limb	o3
 
 ASM_START()
 PROLOGUE(mpn_mul_1)
-	sll	%o2,4,%g1
-	and	%g1,(4-1)<<4,%g1
-ifdef(`PIC',
-`	mov	%o7,%g4		C Save return address register
-0:	call	1f
-	add	%o7,L(1)-0b,%g3
-1:	mov	%g4,%o7		C Restore return address register
-',
-`	sethi	%hi(L(1)),%g3
-	or	%g3,%lo(L(1)),%g3
-')
-	jmp	%g3+%g1
-	ld	[%o1+0],%o4	C 1
-L(1):
-L(L00):	add	%o0,-4,%o0
-	add	%o1,-4,%o1
-	b	L(loop00)	C 4, 8, 12, ...
+	ld	[%o1+0],%o4
+	andcc	%o2,1,%g0
+	be	L(bx0)
+	andcc	%o2,2,%g0
+L(bx1):	be	L(01)
 	orcc	%g0,%g0,%g2
-L(L01):	b	L(loop01)	C 1, 5, 9, ...
-	orcc	%g0,%g0,%g2
-	nop
-	nop
-L(L10):	add	%o0,-12,%o0	C 2, 6, 10, ...
-	add	%o1,4,%o1
-	b	L(loop10)
-	orcc	%g0,%g0,%g2
-	nop
-L(L11):	add	%o0,-8,%o0	C 3, 7, 11, ...
+L(b11):	add	%o0,-8,%o0
+	b	L(11)
 	add	%o1,-8,%o1
-	b	L(loop11)
+L(bx0):	be	L(b00)
 	orcc	%g0,%g0,%g2
+L(b10):	add	%o0,-12,%o0
+	b	L(10)
+	add	%o1,4,%o1
+L(b00):	add	%o0,-4,%o0
+	b	L(00)
+	add	%o1,-4,%o1
 
-L(loop):
-	addcc	%g3,%g2,%g3	C 1
+L(top):	addcc	%g3,%g2,%g3	C 1
 	ld	[%o1+4],%o4	C 2
 	st	%g3,[%o0+0]	C 1
 	rd	%y,%g2		C 1
-L(loop00):
-	umul	%o4,%o3,%g3	C 2
+L(00):	umul	%o4,%o3,%g3	C 2
 	addxcc	%g3,%g2,%g3	C 2
 	ld	[%o1+8],%o4	C 3
 	st	%g3,[%o0+4]	C 2
 	rd	%y,%g2		C 2
-L(loop11):
-	umul	%o4,%o3,%g3	C 3
+L(11):	umul	%o4,%o3,%g3	C 3
 	addxcc	%g3,%g2,%g3	C 3
 	ld	[%o1+12],%o4	C 4
 	add	%o1,16,%o1
 	st	%g3,[%o0+8]	C 3
 	rd	%y,%g2		C 3
-L(loop10):
-	umul	%o4,%o3,%g3	C 4
+L(10):	umul	%o4,%o3,%g3	C 4
 	addxcc	%g3,%g2,%g3	C 4
 	ld	[%o1+0],%o4	C 1
 	st	%g3,[%o0+12]	C 4
 	add	%o0,16,%o0
 	rd	%y,%g2		C 4
 	addx	%g0,%g2,%g2
-L(loop01):
-	addcc	%o2,-4,%o2
-	bg	L(loop)
+L(01):	addcc	%o2,-4,%o2
+	bg	L(top)
 	umul	%o4,%o3,%g3	C 1
 
 	addcc	%g3,%g2,%g3	C 4