[dev.garbage] all: merge dev.power64 (7667e41f3ced) into dev.garbage

Now the only difference between dev.cc and dev.garbage
is the runtime conversion on the one side and the
garbage collection on the other. They both have the
same set of changes from default and dev.power64.

LGTM=austin
R=austin
CC=golang-codereviews
https://codereview.appspot.com/172570043

6 files changed, 41 insertions, 33 deletions

diff --git a/src/runtime/asm_power64x.s b/src/runtime/asm_power64x.s
index fd0c6be16..ba900c2b3 100644
--- a/src/runtime/asm_power64x.s
+++ b/src/runtime/asm_power64x.s
@@ -86,7 +86,7 @@ TEXT runtime·reginit(SB),NOSPLIT,$-8-0
 // void gosave(Gobuf*)
 // save state in Gobuf; setjmp
 TEXT runtime·gosave(SB), NOSPLIT, $-8-8
-	MOVD	gobuf+0(FP), R3
+	MOVD	buf+0(FP), R3
 	MOVD	R1, gobuf_sp(R3)
 	MOVD	LR, R31
 	MOVD	R31, gobuf_pc(R3)
@@ -99,7 +99,7 @@ TEXT runtime·gosave(SB), NOSPLIT, $-8-8
 // void gogo(Gobuf*)
 // restore state from Gobuf; longjmp
 TEXT runtime·gogo(SB), NOSPLIT, $-8-8
-	MOVD	gobuf+0(FP), R5
+	MOVD	buf+0(FP), R5
 	MOVD	gobuf_g(R5), g	// make sure g is not nil
 	MOVD	0(g), R4
 	MOVD	gobuf_sp(R5), R1
@@ -299,7 +299,7 @@ TEXT runtime·morestack_noctxt(SB),NOSPLIT,$-8-0
 // Note: can't just "BR NAME(SB)" - bad inlining results.
 
 TEXT ·reflectcall(SB), NOSPLIT, $-8-24
-	MOVW argsize+16(FP), R3
+	MOVWZ n+16(FP), R3
 	DISPATCH(runtime·call16, 16)
 	DISPATCH(runtime·call32, 32)
 	DISPATCH(runtime·call64, 64)
@@ -335,8 +335,8 @@ TEXT ·reflectcall(SB), NOSPLIT, $-8-24
 TEXT NAME(SB), WRAPPER, $MAXSIZE-24;		\
 	NO_LOCAL_POINTERS;			\
 	/* copy arguments to stack */		\
-	MOVD	argptr+8(FP), R3;		\
-	MOVW	argsize+16(FP), R4;		\
+	MOVD	arg+8(FP), R3;			\
+	MOVWZ	n+16(FP), R4;			\
 	MOVD	R1, R5;				\
 	ADD	$(8-1), R5;			\
 	SUB	$1, R3;				\
@@ -353,9 +353,9 @@ TEXT NAME(SB), WRAPPER, $MAXSIZE-24;		\
 	PCDATA  $PCDATA_StackMapIndex, $0;	\
 	BL	(CTR);				\
 	/* copy return values back */		\
-	MOVD	argptr+8(FP), R3;		\
-	MOVW	argsize+16(FP), R4;		\
-	MOVW	retoffset+20(FP), R6;		\
+	MOVD	arg+8(FP), R3;			\
+	MOVWZ	n+16(FP), R4;			\
+	MOVWZ	retoffset+20(FP), R6;		\
 	MOVD	R1, R5;				\
 	ADD	R6, R5; 			\
 	ADD	R6, R3;				\
@@ -398,7 +398,7 @@ CALLFN(·call268435456, 268435456)
 CALLFN(·call536870912, 536870912)
 CALLFN(·call1073741824, 1073741824)
 
-// bool cas(int32 *val, int32 old, int32 new)
+// bool cas(uint32 *ptr, uint32 old, uint32 new)
 // Atomically:
 //	if(*val == old){
 //		*val = new;
@@ -406,9 +406,9 @@ CALLFN(·call1073741824, 1073741824)
 //	} else
 //		return 0;
 TEXT runtime·cas(SB), NOSPLIT, $0-17
-	MOVD	p+0(FP), R3
-	MOVW	old+8(FP), R4
-	MOVW	new+12(FP), R5
+	MOVD	ptr+0(FP), R3
+	MOVWZ	old+8(FP), R4
+	MOVWZ	new+12(FP), R5
 cas_again:
 	SYNC
 	LWAR	(R3), R6
@@ -425,7 +425,7 @@ cas_fail:
 	MOVD	$0, R3
 	BR	-5(PC)
 
-// bool	runtime·cas64(uint64 *val, uint64 old, uint64 new)
+// bool	runtime·cas64(uint64 *ptr, uint64 old, uint64 new)
 // Atomically:
 //	if(*val == *old){
 //		*val = new;
@@ -434,7 +434,7 @@ cas_fail:
 //		return 0;
 //	}
 TEXT runtime·cas64(SB), NOSPLIT, $0-25
-	MOVD	p+0(FP), R3
+	MOVD	ptr+0(FP), R3
 	MOVD	old+8(FP), R4
 	MOVD	new+16(FP), R5
 cas64_again:
@@ -475,12 +475,12 @@ TEXT runtime·atomicstoreuintptr(SB), NOSPLIT, $0-16
 TEXT runtime·casp1(SB), NOSPLIT, $0-25
 	BR runtime·cas64(SB)
 
-// uint32 xadd(uint32 volatile *val, int32 delta)
+// uint32 xadd(uint32 volatile *ptr, int32 delta)
 // Atomically:
 //	*val += delta;
 //	return *val;
 TEXT runtime·xadd(SB), NOSPLIT, $0-20
-	MOVD	p+0(FP), R4
+	MOVD	ptr+0(FP), R4
 	MOVW	delta+8(FP), R5
 	SYNC
 	LWAR	(R4), R3
@@ -493,7 +493,7 @@ TEXT runtime·xadd(SB), NOSPLIT, $0-20
 	RETURN
 
 TEXT runtime·xadd64(SB), NOSPLIT, $0-24
-	MOVD	p+0(FP), R4
+	MOVD	ptr+0(FP), R4
 	MOVD	delta+8(FP), R5
 	SYNC
 	LDAR	(R4), R3
@@ -506,7 +506,7 @@ TEXT runtime·xadd64(SB), NOSPLIT, $0-24
 	RETURN
 
 TEXT runtime·xchg(SB), NOSPLIT, $0-20
-	MOVD	p+0(FP), R4
+	MOVD	ptr+0(FP), R4
 	MOVW	new+8(FP), R5
 	SYNC
 	LWAR	(R4), R3
@@ -518,7 +518,7 @@ TEXT runtime·xchg(SB), NOSPLIT, $0-20
 	RETURN
 
 TEXT runtime·xchg64(SB), NOSPLIT, $0-24
-	MOVD	p+0(FP), R4
+	MOVD	ptr+0(FP), R4
 	MOVD	new+8(FP), R5
 	SYNC
 	LDAR	(R4), R3
@@ -651,7 +651,7 @@ TEXT runtime·setcallerpc(SB),NOSPLIT,$-8-16
 	RETURN
 
 TEXT runtime·getcallersp(SB),NOSPLIT,$0-16
-	MOVD	sp+0(FP), R3
+	MOVD	argp+0(FP), R3
 	SUB	$8, R3
 	MOVD	R3, ret+8(FP)
 	RETURN
@@ -695,16 +695,17 @@ TEXT runtime·aeshashstr(SB),NOSPLIT,$-8-0
 TEXT runtime·memeq(SB),NOSPLIT,$-8-25
 	MOVD	a+0(FP), R3
 	MOVD	b+8(FP), R4
-	MOVD	count+16(FP), R5
+	MOVD	size+16(FP), R5
 	SUB	$1, R3
 	SUB	$1, R4
 	ADD	R3, R5, R8
 loop:
 	CMP	R3, R8
-	BNE	4(PC)
+	BNE	test
 	MOVD	$1, R3
 	MOVB	R3, ret+24(FP)
 	RETURN
+test:
 	MOVBZU	1(R3), R6
 	MOVBZU	1(R4), R7
 	CMP	R6, R7
@@ -828,7 +829,7 @@ notfound:
 // in ../../cmd/9g/ggen.c:/^clearfat.
 // R0: always zero
 // R3 (aka REGRT1): ptr to memory to be zeroed - 8
-// R3 is updated as a side effect.
+// On return, R3 points to the last zeroed dword.
 TEXT runtime·duffzero(SB), NOSPLIT, $-8-0
 	MOVDU	R0, 8(R3)
 	MOVDU	R0, 8(R3)
@@ -964,7 +965,7 @@ TEXT runtime·fastrand1(SB), NOSPLIT, $0-4
 	MOVD	g_m(g), R4
 	MOVWZ	m_fastrand(R4), R3
 	ADD	R3, R3
-	CMP	R3, $0
+	CMPW	R3, $0
 	BGE	2(PC)
 	XOR	$0x88888eef, R3
 	MOVW	R3, m_fastrand(R4)
@@ -979,3 +980,9 @@ TEXT runtime·return0(SB), NOSPLIT, $0
 // Must obey the gcc calling convention.
 TEXT _cgo_topofstack(SB),NOSPLIT,$0
 	MOVD	R0, 26(R0)
+
+// The top-most function running on a goroutine
+// returns to goexit+PCQuantum.
+TEXT runtime·goexit(SB),NOSPLIT,$-8-0
+	MOVD	R0, R0	// NOP
+	BL	runtime·goexit1(SB)	// does not return
diff --git a/src/runtime/gcinfo_test.go b/src/runtime/gcinfo_test.go
index 1a33f3b3b..662b7546d 100644
--- a/src/runtime/gcinfo_test.go
+++ b/src/runtime/gcinfo_test.go
@@ -137,7 +137,7 @@ func infoBigStruct() []byte {
 			BitsScalar, BitsScalar, BitsScalar, BitsScalar, // t int; y uint16; u uint64
 			BitsPointer, BitsDead, // i string
 		}
-	case "amd64":
+	case "amd64", "power64", "power64le":
 		return []byte{
 			BitsPointer,                        // q *int
 			BitsScalar, BitsScalar, BitsScalar, // w byte; e [17]byte
@@ -153,12 +153,6 @@ func infoBigStruct() []byte {
 			BitsScalar, BitsScalar, BitsDead, BitsScalar, BitsScalar, // t int; y uint16; u uint64
 			BitsPointer, BitsDead, // i string
 		}
-	case "power64", "power64le":
-		return []byte{
-			BitsPointer, BitsScalar, BitsScalar, BitsScalar,
-			BitsMultiWord, BitsSlice, BitsScalar, BitsScalar,
-			BitsScalar, BitsScalar, BitsMultiWord, BitsString,
-		}
 	default:
 		panic("unknown arch")
 	}
diff --git a/src/runtime/mgc0.c b/src/runtime/mgc0.c
index 3248b0f49..f37c01af0 100644
--- a/src/runtime/mgc0.c
+++ b/src/runtime/mgc0.c
@@ -122,6 +122,7 @@
 
 enum {
 	Debug		= 0,
+	DebugPtrs	= 0, // if 1, print trace of every pointer load during GC
 	ConcurrentSweep	= 1,
 
 	FinBlockSize	= 4*1024,
diff --git a/src/runtime/panic.c b/src/runtime/panic.c
index 46683b2b0..b19fdd0e1 100644
--- a/src/runtime/panic.c
+++ b/src/runtime/panic.c
@@ -69,7 +69,7 @@ runtime·recovery_m(G *gp)
 	// each call to deferproc.
 	// (The pc we're returning to does pop pop
 	// before it tests the return value.)
-	// On the arm there are 2 saved LRs mixed in too.
+	// On the arm and power there are 2 saved LRs mixed in too.
 	if(thechar == '5' || thechar == '9')
 		gp->sched.sp = (uintptr)argp - 4*sizeof(uintptr);
 	else
diff --git a/src/runtime/runtime.c b/src/runtime/runtime.c
index f19f8e4be..a68414284 100644
--- a/src/runtime/runtime.c
+++ b/src/runtime/runtime.c
@@ -226,6 +226,12 @@ runtime·check(void)
 	if(z != 4)
 		runtime·throw("cas4");
 
+	z = 0xffffffff;
+	if(!runtime·cas(&z, 0xffffffff, 0xfffffffe))
+		runtime·throw("cas5");
+	if(z != 0xfffffffe)
+		runtime·throw("cas6");
+
 	k = (byte*)0xfedcb123;
 	if(sizeof(void*) == 8)
 		k = (byte*)((uintptr)k<<10);
diff --git a/src/runtime/signal_power64x.c b/src/runtime/signal_power64x.c
index 89c5c7848..c0bf1c4a5 100644
--- a/src/runtime/signal_power64x.c
+++ b/src/runtime/signal_power64x.c
@@ -124,7 +124,7 @@ runtime·sighandler(int32 sig, Siginfo *info, void *ctxt, G *gp)
 
 	if(runtime·gotraceback(&crash)){
 		runtime·goroutineheader(gp);
-		runtime·traceback(SIG_PC(info, ctxt), SIG_SP(info, ctxt), SIG_LINK(info, ctxt), gp);
+		runtime·tracebacktrap(SIG_PC(info, ctxt), SIG_SP(info, ctxt), SIG_LINK(info, ctxt), gp);
 		runtime·tracebackothers(gp);
 		runtime·printf("\n");
 		runtime·dumpregs(info, ctxt);