From dc05cb2dc84851ae9eedfde7da04a50e7dd41c1f Mon Sep 17 00:00:00 2001
From: George Koehler <xkernigh@netscape.net>
Date: Mon, 13 Feb 2017 16:52:32 -0500
Subject: [PATCH] Add pat cms !defined($1)

Switch .cms to pass inputs and outputs on the real stack, not in
registers; like we do with .and, .or (81c677d) and .xor (c578c49).

At this point, nearly all functions in libem use the real stack, not
registers, for passing inputs and outputs.  This simplifies the ncg
table (which needs fewer lists of specific registers) but slows calls
to libem.

For example, after ba9b021, each call to .aar4 is about 10
instructions slower.  I moved 3 inputs and 1 output from registers to
the real stack.  A program would take 4 instructions to move registers
to stack, 4 to move stack to registers, and perhaps 2 to adjust the
stack pointer.
---
 mach/powerpc/libem/cms.s | 19 ++++++++++---------
 mach/powerpc/ncg/table   | 13 +++++++------
 2 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/mach/powerpc/libem/cms.s b/mach/powerpc/libem/cms.s
index 5bcd3106d..0bcb1ab0b 100644
--- a/mach/powerpc/libem/cms.s
+++ b/mach/powerpc/libem/cms.s
@@ -1,18 +1,17 @@
 .sect .text
 
 ! Compare sets a, b.
-!  Stack: ( b a -- )
-!  With r3 = size of each set
-!  Yields r3 = 0 if equal, nonzero if not equal
+!  Stack: ( b a size -- result )
+!  Result is 0 if equal, nonzero if not equal.
 
 .define .cms
 .cms:
+	lwz	r3, 0(sp)		! r3 = size of each set
 	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
+	addi	r4, sp, 4		! r4 = ptr to set a
+	add	r5, r4, r3		! r5 = ptr to set b
 	li	r6, 0			! r6 = index
-	add	r9, r5, r3		! r9 = future sp
 1:
 	lwzx	r7, r4, r6
 	lwzx	r8, r5, r6
@@ -20,10 +19,12 @@
 	addi	r6, r6, 4
 	bne	cr0, 2f			! branch if not equal
 	bdnz	1b			! loop ctr times
-	li	r3, 0			! equal: return 0
+	li	r9, 0			! equal: return 0
 	b	3f
 2:
-	li	r3, 1			! not equal: return 1
+	li	r9, 1			! not equal: return 1
 3:
-	mr	sp, r9			! remove sets from stack
+	slwi	r7, r3, 1
+	add	sp, sp, r7		! adjust stack pointer
+	stw	r9, 0(sp)		! push result
 	blr
diff --git a/mach/powerpc/ncg/table b/mach/powerpc/ncg/table
index 7868f2a3f..84750df05 100644
--- a/mach/powerpc/ncg/table
+++ b/mach/powerpc/ncg/table
@@ -1854,12 +1854,13 @@ PATTERNS
 			cmi INT32
 
 	pat cms defined($1)
-		with STACK
-			kills ALL
-			gen
-				move {CONST, $1}, R3
-				bl {LABEL, ".cms"}
-			yields R3
+		leaving
+			loc $1
+			cal ".cms"
+
+	pat cms !defined($1)
+		leaving
+			cal ".cms"
 
 
 /* Other branching and labelling */