Edit PowerPC assembly for .and, .cms, .ior, .xor, .zer

Remove one addi instruction from some loops.  These loops had
increased 2 pointers, they now increase 1 index.  I must initialize
the index, so I add "li r6, 0" before each loop.

Change .zer to use subf instead of neg, add.

Change .xor to take the size on the real stack, as .and and .or have
done since 81c677d.

mach/powerpc/libem/and.s

@@ -5,20 +5,18 @@
 
 .define .and
 .and:
-	lwz r3, 0 (sp)      ! r3 = size
+	lwz	r3, 0(sp)		! r3 = size
-	addi sp, sp, 4
+	srwi	r7, r3, 2
-
+	mtspr	ctr, r7			! ctr = size / 4
-	mr	r4, sp			! r4 = ptr to set a
+	addi	r4, sp, 4		! r4 = ptr to set a
-	add	r5, sp, r3		! r5 = ptr to set b
+	add	r5, r4, r3		! r5 = ptr to set b
-	srwi	r6, r3, 2
+	li	r6, 0			! r6 = index
-	mtspr	ctr, r6			! ctr = r3 / 4
 1:
-	lwz	r7, 0(r4)
+	lwzx	r7, r4, r6
-	lwz	r8, 0(r5)
+	lwzx	r8, r5, r6
 	and	r8, r7, r8		! intersection of words
-	stw	r8, 0(r5)
+	stwx	r8, r5, r6
-	addi	r4, r4, 4
+	addi	r6, r6, 4
-	addi	r5, r5, 4
 	bdnz	1b			! loop ctr times
-	add	sp, sp, r3
+	mr	sp, r5
 	blr

mach/powerpc/libem/cms.s

@@ -7,24 +7,23 @@
 
 .define .cms
 .cms:
+	srwi	r7, r3, 2
+	mtspr	ctr, r7			! ctr = size / 4
 	mr	r4, sp			! r4 = ptr to set a
 	add	r5, sp, r3		! r5 = ptr to set b
-	mr	r6, r3			! r6 = size
+	li	r6, 0			! r6 = index
-	srwi	r3, r3, 2
+	add	r9, r5, r3		! r9 = future sp
-	mtspr	ctr, r3			! ctr = size / 4
 1:
-	lwz	r7, 0(r4)
+	lwzx	r7, r4, r6
-	lwz	r8, 0(r5)
+	lwzx	r8, r5, r6
 	cmpw	cr0, r7, r8		! compare words in sets
-	addi	r4, r4, 4
+	addi	r6, r6, 4
-	addi	r5, r5, 4
 	bne	cr0, 2f			! branch if not equal
 	bdnz	1b			! loop ctr times
-	addi	r3, r0, 0		! equal: return 0
+	li	r3, 0			! equal: return 0
 	b	3f
 2:
-	addi	r3, r0, 1		! not equal: return 1
+	li	r3, 1			! not equal: return 1
 3:
-	slwi	r6, r6, 1		! r6 = size * 2
+	mr	sp, r9			! remove sets from stack
-	add	sp, sp, r6		! remove sets from stack
 	blr

mach/powerpc/libem/ior.s

@@ -5,20 +5,18 @@
 
 .define .ior
 .ior:
-	lwz r3, 0 (sp)
+	lwz	r3, 0(sp)		! r3 = size
-	addi sp, sp, 4
+	srwi	r7, r3, 2
-
+	mtspr	ctr, r7			! ctr = size / 4
-	mr	r4, sp			! r4 = ptr to set a
+	addi	r4, sp, 4		! r4 = ptr to set a
-	add	r5, sp, r3		! r5 = ptr to set b
+	add	r5, r4, r3		! r5 = ptr to set b
-	srwi	r6, r3, 2
+	li	r6, 0			! r6 = index
-	mtspr	ctr, r6			! ctr = r3 / 4
 1:
-	lwz	r7, 0(r4)
+	lwzx	r7, r4, r6
-	lwz	r8, 0(r5)
+	lwzx	r8, r5, r6
 	or	r8, r7, r8		! union of words
-	stw	r8, 0(r5)
+	stwx	r8, r5, r6
-	addi	r4, r4, 4
+	addi	r6, r6, 4
-	addi	r5, r5, 4
 	bdnz	1b			! loop ctr times
-	add	sp, sp, r3
+	mr	sp, r5
 	blr

mach/powerpc/libem/xor.s

@@ -1,22 +1,22 @@
 .sect .text
 
 ! Set symmetric difference.
-!  Stack: ( b a -- a/b )
+!  Stack: ( b a size -- a/b )
-!  With r3 = size of set
 
 .define .xor
 .xor:
-	mr	r4, sp			! r4 = ptr to set a
+	lwz	r3, 0(sp)		! r3 = size
-	add	r5, sp, r3		! r5 = ptr to set b
+	srwi	r7, r3, 2
-	srwi	r6, r3, 2
+	mtspr	ctr, r7			! ctr = size / 4
-	mtspr	ctr, r6			! ctr = r3 / 4
+	addi	r4, sp, 4		! r4 = ptr to set a
+	add	r5, r4, r3		! r5 = ptr to set b
+	li	r6, 0			! r6 = index
 1:
-	lwz	r7, 0(r4)
+	lwzx	r7, r4, r6
-	lwz	r8, 0(r5)
+	lwzx	r8, r5, r6
 	xor	r8, r7, r8		! symmetric difference of words
-	stw	r8, 0(r5)
+	stwx	r8, r5, r6
-	addi	r4, r4, 4
+	addi	r6, r6, 4
-	addi	r5, r5, 4
 	bdnz	1b			! loop ctr times
-	add	sp, sp, r3
+	mr	sp, r5
 	blr

mach/powerpc/libem/zer.s

@@ -5,17 +5,15 @@
 
 .define .zer
 .zer:
-	lwz     r3, 0(sp)
+	lwz	r3, 0(sp)		! r3 = size
+	srwi	r5, r3, 2
+	mtspr	ctr, r5			! ctr = word size - 4
+	li	r4, 0			! r4 = 0
 	addi	sp, sp, 4
-
+	subf	sp, r3, sp		! sp = ptr to new set
-	srwi	r7, r3, 2
+	li	r6, 0			! r6 = index
-	li	r4, 0			! r4 = zero
-	neg	r5, r3
-	add	sp, sp, r5		! allocate set
-	mr	r6, sp			! r6 = ptr to set
-	mtspr	ctr, r7			! ctr = r3 / 4
 1:
-	stw	r4, 0(r6)		! store zero in set
+	stwx	r4, sp, r6		! store zero in set
 	addi	r6, r6, 4
 	bdnz	1b			! loop ctr times
 	blr

mach/powerpc/ncg/table

@@ -1374,11 +1374,13 @@ PATTERNS
 			yields {XOR_RC, %2, lo(%1.val)}
 
 	pat xor defined($1)                /* XOR set */
-		with STACK
+		leaving
-			kills ALL
+			loc $1
-			gen
+			cal ".xor"
-				move {CONST, $1}, R3
+
-				bl {LABEL, ".xor"}
+	pat xor !defined($1)
+		leaving
+			cal ".xor"
 
 	pat com $1==INT32                  /* NOT word */
 		with AND_RR