bcm64: use MOVP instruction for saving and restoring registers

2 files changed, 86 insertions, 183 deletions

diff --git a/sys/src/9/bcm64/fns.h b/sys/src/9/bcm64/fns.h
index f2e357358..779fc7686 100644
--- a/sys/src/9/bcm64/fns.h
+++ b/sys/src/9/bcm64/fns.h
@@ -17,7 +17,6 @@ extern void noteret(void);
 extern void returnto(void*);
 extern void fpsaveregs(void*);
 extern void fploadregs(void*);
-extern void magic(void);
 
 extern void setttbr(uintptr pa);
 extern uintptr getfar(void);
diff --git a/sys/src/9/bcm64/l.s b/sys/src/9/bcm64/l.s
index a6a10e69f..4e64773f7 100644
--- a/sys/src/9/bcm64/l.s
+++ b/sys/src/9/bcm64/l.s
@@ -334,17 +334,6 @@ TEXT setttbr(SB), 1, $-4
 
 	B	cacheiinv(SB)
 
-TEXT magic(SB), 1, $-4
-	DSB	$SY
-	ISB	$SY
-	DSB	$SY
-	ISB	$SY
-	DSB	$SY
-	ISB	$SY
-	DSB	$SY
-	ISB	$SY
-	RETURN
-
 /*
  * TLB maintenance operations.
  * these broadcast to all cpu's in the cluser
@@ -421,10 +410,8 @@ TEXT vsys0(SB), 1, $-4
 	CMPW	$0x15, R17	// SVC trap?
 	BNE	_itsatrap	// nope.
 
-	MOV	R26, 224(RSP)	// special
-	MOV	R27, 232(RSP)	// special
-	MOV	R28, 240(RSP)	// sb
-	MOV	R29, 248(RSP)	// special
+	MOVP	R26, R27, 224(RSP)
+	MOVP	R28, R29, 240(RSP)
 
 	MRS	SP_EL0, R1
 	MRS	ELR_EL1, R2
@@ -456,10 +443,8 @@ TEXT forkret(SB), 1, $-4
 	MSR	R2, ELR_EL1
 	MSR	R3, SPSR_EL1
 
-	MOV	224(RSP), R26	// special
-	MOV	232(RSP), R27	// special
-	MOV	240(RSP), R28	// sb
-	MOV	248(RSP), R29	// special
+	MOVP	224(RSP), R26, R27
+	MOVP	240(RSP), R28, R29
 
 	MOV	256(RSP), R30	// link
 
@@ -468,38 +453,24 @@ TEXT forkret(SB), 1, $-4
 
 TEXT itsatrap<>(SB), 1, $-4
 _itsatrap:
-	MOV	R1, 24(RSP)
-	MOV	R2, 32(RSP)
-	MOV	R3, 40(RSP)
-	MOV	R4, 48(RSP)
-	MOV	R5, 56(RSP)
-	MOV	R6, 64(RSP)
-	MOV	R7, 72(RSP)
-	MOV	R8, 80(RSP)
-	MOV	R9, 88(RSP)
-	MOV	R10, 96(RSP)
-	MOV	R11, 104(RSP)
-	MOV	R12, 112(RSP)
-	MOV	R13, 120(RSP)
-	MOV	R14, 128(RSP)
-	MOV	R15, 136(RSP)
-	MOV	R16, 144(RSP)
-
-	MOV	R18, 160(RSP)
-	MOV	R19, 168(RSP)
-	MOV	R20, 176(RSP)
-	MOV	R21, 184(RSP)
-	MOV	R22, 192(RSP)
-	MOV	R23, 200(RSP)
-	MOV	R24, 208(RSP)
-	MOV	R25, 216(RSP)
+	MOVP	R1, R2, 24(RSP)
+	MOVP	R3, R4, 40(RSP)
+	MOVP	R5, R6, 56(RSP)
+	MOVP	R7, R8, 72(RSP)
+	MOVP	R9, R10, 88(RSP)
+	MOVP	R11, R12, 104(RSP)
+	MOVP	R13, R14, 120(RSP)
+	MOVP	R15, R16, 136(RSP)
+
+	MOVP	R18, R19, 160(RSP)
+	MOVP	R20, R21, 176(RSP)
+	MOVP	R22, R23, 192(RSP)
+	MOVP	R24, R25, 208(RSP)
 
 // trap/irq/fiq/serr from EL0
 TEXT vtrap0(SB), 1, $-4
-	MOV	R26, 224(RSP)	// special
-	MOV	R27, 232(RSP)	// special
-	MOV	R28, 240(RSP)	// sb
-	MOV	R29, 248(RSP)	// special
+	MOVP	R26, R27, 224(RSP)
+	MOVP	R28, R29, 240(RSP)
 
 	MRS	SP_EL0, R1
 	MRS	ELR_EL1, R2
@@ -530,38 +501,23 @@ TEXT noteret(SB), 1, $-4
 	MSR	R2, ELR_EL1
 	MSR	R3, SPSR_EL1
 
-	MOV	224(RSP), R26	// special
-	MOV	232(RSP), R27	// special
-	MOV	240(RSP), R28	// sb
-	MOV	248(RSP), R29	// special
+	MOVP	224(RSP), R26, R27
+	MOVP	240(RSP), R28, R29
 
 _intrreturn:
-	MOV	16(RSP), R0
-	MOV	24(RSP), R1
-	MOV	32(RSP), R2
-	MOV	40(RSP), R3
-	MOV	48(RSP), R4
-	MOV	56(RSP), R5
-	MOV	64(RSP), R6
-	MOV	72(RSP), R7
-	MOV	80(RSP), R8
-	MOV	88(RSP), R9
-	MOV	96(RSP), R10
-	MOV	104(RSP), R11
-	MOV	112(RSP), R12
-	MOV	120(RSP), R13
-	MOV	128(RSP), R14
-	MOV	136(RSP), R15
-	MOV	144(RSP), R16
-	MOV	152(RSP), R17
-	MOV	160(RSP), R18
-	MOV	168(RSP), R19
-	MOV	176(RSP), R20
-	MOV	184(RSP), R21
-	MOV	192(RSP), R22
-	MOV	200(RSP), R23
-	MOV	208(RSP), R24
-	MOV	216(RSP), R25
+	MOVP	16(RSP), R0, R1
+	MOVP	32(RSP), R2, R3
+	MOVP	48(RSP), R4, R5
+	MOVP	64(RSP), R6, R7
+	MOVP	80(RSP), R8, R9
+	MOVP	96(RSP), R10, R11
+	MOVP	112(RSP), R12, R13
+	MOVP	128(RSP), R14, R15
+	MOVP	144(RSP), R16, R17
+	MOVP	160(RSP), R18, R19
+	MOVP	176(RSP), R20, R21
+	MOVP	192(RSP), R22, R23
+	MOVP	208(RSP), R24, R25
 
 	MOV	256(RSP), R30	// link
 
@@ -612,32 +568,19 @@ _vsyspatch:
 TEXT vtrap(SB), 1, $-4
 	SUB	$TRAPFRAMESIZE, RSP
 
-	MOV	R0, 16(RSP)
-	MOV	R1, 24(RSP)
-	MOV	R2, 32(RSP)
-	MOV	R3, 40(RSP)
-	MOV	R4, 48(RSP)
-	MOV	R5, 56(RSP)
-	MOV	R6, 64(RSP)
-	MOV	R7, 72(RSP)
-	MOV	R8, 80(RSP)
-	MOV	R9, 88(RSP)
-	MOV	R10, 96(RSP)
-	MOV	R11, 104(RSP)
-	MOV	R12, 112(RSP)
-	MOV	R13, 120(RSP)
-	MOV	R14, 128(RSP)
-	MOV	R15, 136(RSP)
-	MOV	R16, 144(RSP)
-	MOV	R17, 152(RSP)
-	MOV	R18, 160(RSP)
-	MOV	R19, 168(RSP)
-	MOV	R20, 176(RSP)
-	MOV	R21, 184(RSP)
-	MOV	R22, 192(RSP)
-	MOV	R23, 200(RSP)
-	MOV	R24, 208(RSP)
-	MOV	R25, 216(RSP)
+	MOVP	R0, R1, 16(RSP)
+	MOVP	R2, R3, 32(RSP)
+	MOVP	R4, R5, 48(RSP)
+	MOVP	R6, R7, 64(RSP)
+	MOVP	R8, R9, 80(RSP)
+	MOVP	R10, R11, 96(RSP)
+	MOVP	R12, R13, 112(RSP)
+	MOVP	R14, R15, 128(RSP)
+	MOVP	R16, R17, 144(RSP)
+	MOVP	R18, R19, 160(RSP)
+	MOVP	R20, R21, 176(RSP)
+	MOVP	R22, R23, 192(RSP)
+	MOVP	R24, R25, 208(RSP)
 
 	MOV	R30, 256(RSP)	// link
 
@@ -649,32 +592,19 @@ _vtrappatch:
 TEXT virq(SB), 1, $-4
 	SUB	$TRAPFRAMESIZE, RSP
 
-	MOV	R0, 16(RSP)
-	MOV	R1, 24(RSP)
-	MOV	R2, 32(RSP)
-	MOV	R3, 40(RSP)
-	MOV	R4, 48(RSP)
-	MOV	R5, 56(RSP)
-	MOV	R6, 64(RSP)
-	MOV	R7, 72(RSP)
-	MOV	R8, 80(RSP)
-	MOV	R9, 88(RSP)
-	MOV	R10, 96(RSP)
-	MOV	R11, 104(RSP)
-	MOV	R12, 112(RSP)
-	MOV	R13, 120(RSP)
-	MOV	R14, 128(RSP)
-	MOV	R15, 136(RSP)
-	MOV	R16, 144(RSP)
-	MOV	R17, 152(RSP)
-	MOV	R18, 160(RSP)
-	MOV	R19, 168(RSP)
-	MOV	R20, 176(RSP)
-	MOV	R21, 184(RSP)
-	MOV	R22, 192(RSP)
-	MOV	R23, 200(RSP)
-	MOV	R24, 208(RSP)
-	MOV	R25, 216(RSP)
+	MOVP	R0, R1, 16(RSP)
+	MOVP	R2, R3, 32(RSP)
+	MOVP	R4, R5, 48(RSP)
+	MOVP	R6, R7, 64(RSP)
+	MOVP	R8, R9, 80(RSP)
+	MOVP	R10, R11, 96(RSP)
+	MOVP	R12, R13, 112(RSP)
+	MOVP	R14, R15, 128(RSP)
+	MOVP	R16, R17, 144(RSP)
+	MOVP	R18, R19, 160(RSP)
+	MOVP	R20, R21, 176(RSP)
+	MOVP	R22, R23, 192(RSP)
+	MOVP	R24, R25, 208(RSP)
 
 	MOV	R30, 256(RSP)	// link
 
@@ -686,32 +616,19 @@ _virqpatch:
 TEXT vfiq(SB), 1, $-4
 	SUB	$TRAPFRAMESIZE, RSP
 
-	MOV	R0, 16(RSP)
-	MOV	R1, 24(RSP)
-	MOV	R2, 32(RSP)
-	MOV	R3, 40(RSP)
-	MOV	R4, 48(RSP)
-	MOV	R5, 56(RSP)
-	MOV	R6, 64(RSP)
-	MOV	R7, 72(RSP)
-	MOV	R8, 80(RSP)
-	MOV	R9, 88(RSP)
-	MOV	R10, 96(RSP)
-	MOV	R11, 104(RSP)
-	MOV	R12, 112(RSP)
-	MOV	R13, 120(RSP)
-	MOV	R14, 128(RSP)
-	MOV	R15, 136(RSP)
-	MOV	R16, 144(RSP)
-	MOV	R17, 152(RSP)
-	MOV	R18, 160(RSP)
-	MOV	R19, 168(RSP)
-	MOV	R20, 176(RSP)
-	MOV	R21, 184(RSP)
-	MOV	R22, 192(RSP)
-	MOV	R23, 200(RSP)
-	MOV	R24, 208(RSP)
-	MOV	R25, 216(RSP)
+	MOVP	R0, R1, 16(RSP)
+	MOVP	R2, R3, 32(RSP)
+	MOVP	R4, R5, 48(RSP)
+	MOVP	R6, R7, 64(RSP)
+	MOVP	R8, R9, 80(RSP)
+	MOVP	R10, R11, 96(RSP)
+	MOVP	R12, R13, 112(RSP)
+	MOVP	R14, R15, 128(RSP)
+	MOVP	R16, R17, 144(RSP)
+	MOVP	R18, R19, 160(RSP)
+	MOVP	R20, R21, 176(RSP)
+	MOVP	R22, R23, 192(RSP)
+	MOVP	R24, R25, 208(RSP)
 
 	MOV	R30, 256(RSP)	// link
 	MOV	$(2<<32), R0	// type fiq
@@ -722,32 +639,19 @@ _vfiqpatch:
 TEXT vserr(SB), 1, $-4
 	SUB	$TRAPFRAMESIZE, RSP
 
-	MOV	R0, 16(RSP)
-	MOV	R1, 24(RSP)
-	MOV	R2, 32(RSP)
-	MOV	R3, 40(RSP)
-	MOV	R4, 48(RSP)
-	MOV	R5, 56(RSP)
-	MOV	R6, 64(RSP)
-	MOV	R7, 72(RSP)
-	MOV	R8, 80(RSP)
-	MOV	R9, 88(RSP)
-	MOV	R10, 96(RSP)
-	MOV	R11, 104(RSP)
-	MOV	R12, 112(RSP)
-	MOV	R13, 120(RSP)
-	MOV	R14, 128(RSP)
-	MOV	R15, 136(RSP)
-	MOV	R16, 144(RSP)
-	MOV	R17, 152(RSP)
-	MOV	R18, 160(RSP)
-	MOV	R19, 168(RSP)
-	MOV	R20, 176(RSP)
-	MOV	R21, 184(RSP)
-	MOV	R22, 192(RSP)
-	MOV	R23, 200(RSP)
-	MOV	R24, 208(RSP)
-	MOV	R25, 216(RSP)
+	MOVP	R0, R1, 16(RSP)
+	MOVP	R2, R3, 32(RSP)
+	MOVP	R4, R5, 48(RSP)
+	MOVP	R6, R7, 64(RSP)
+	MOVP	R8, R9, 80(RSP)
+	MOVP	R10, R11, 96(RSP)
+	MOVP	R12, R13, 112(RSP)
+	MOVP	R14, R15, 128(RSP)
+	MOVP	R16, R17, 144(RSP)
+	MOVP	R18, R19, 160(RSP)
+	MOVP	R20, R21, 176(RSP)
+	MOVP	R22, R23, 192(RSP)
+	MOVP	R24, R25, 208(RSP)
 
 	MOV	R30, 256(RSP)	// link