KVM: ia64: Add optimization for some virtulization faults

optvfault.S Add optimization for some performance-critical
virtualization faults.
Signed-off-by: NAnthony Xu <anthony.xu@intel.com>
Signed-off-by: NXiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: NAvi Kivity <avi@qumranet.com>

arch/ia64/kvm/optvfault.S

+/*
+ * arch/ia64/vmx/optvfault.S
+ * optimize virtualization fault handler
+ *
+ * Copyright (C) 2006 Intel Co
+ *	Xuefei Xu (Anthony Xu) <anthony.xu@intel.com>
+ */
+
+#include <asm/asmmacro.h>
+#include <asm/processor.h>
+
+#include "vti.h"
+#include "asm-offsets.h"
+
+#define ACCE_MOV_FROM_AR
+#define ACCE_MOV_FROM_RR
+#define ACCE_MOV_TO_RR
+#define ACCE_RSM
+#define ACCE_SSM
+#define ACCE_MOV_TO_PSR
+#define ACCE_THASH
+
+//mov r1=ar3
+GLOBAL_ENTRY(kvm_asm_mov_from_ar)
+#ifndef ACCE_MOV_FROM_AR
+	br.many kvm_virtualization_fault_back
+#endif
+	add r18=VMM_VCPU_ITC_OFS_OFFSET, r21
+	add r16=VMM_VCPU_LAST_ITC_OFFSET,r21
+	extr.u r17=r25,6,7
+	;;
+	ld8 r18=[r18]
+	mov r19=ar.itc
+	mov r24=b0
+	;;
+	add r19=r19,r18
+	addl r20=@gprel(asm_mov_to_reg),gp
+	;;
+	st8 [r16] = r19
+	adds r30=kvm_resume_to_guest-asm_mov_to_reg,r20
+	shladd r17=r17,4,r20
+	;;
+	mov b0=r17
+	br.sptk.few b0
+	;;
+END(kvm_asm_mov_from_ar)
+
+
+// mov r1=rr[r3]
+GLOBAL_ENTRY(kvm_asm_mov_from_rr)
+#ifndef ACCE_MOV_FROM_RR
+	br.many kvm_virtualization_fault_back
+#endif
+	extr.u r16=r25,20,7
+	extr.u r17=r25,6,7
+	addl r20=@gprel(asm_mov_from_reg),gp
+	;;
+	adds r30=kvm_asm_mov_from_rr_back_1-asm_mov_from_reg,r20
+	shladd r16=r16,4,r20
+	mov r24=b0
+	;;
+	add r27=VMM_VCPU_VRR0_OFFSET,r21
+	mov b0=r16
+	br.many b0
+	;;
+kvm_asm_mov_from_rr_back_1:
+	adds r30=kvm_resume_to_guest-asm_mov_from_reg,r20
+	adds r22=asm_mov_to_reg-asm_mov_from_reg,r20
+	shr.u r26=r19,61
+	;;
+	shladd r17=r17,4,r22
+	shladd r27=r26,3,r27
+	;;
+	ld8 r19=[r27]
+	mov b0=r17
+	br.many b0
+END(kvm_asm_mov_from_rr)
+
+
+// mov rr[r3]=r2
+GLOBAL_ENTRY(kvm_asm_mov_to_rr)
+#ifndef ACCE_MOV_TO_RR
+	br.many kvm_virtualization_fault_back
+#endif
+	extr.u r16=r25,20,7
+	extr.u r17=r25,13,7
+	addl r20=@gprel(asm_mov_from_reg),gp
+	;;
+	adds r30=kvm_asm_mov_to_rr_back_1-asm_mov_from_reg,r20
+	shladd r16=r16,4,r20
+	mov r22=b0
+	;;
+	add r27=VMM_VCPU_VRR0_OFFSET,r21
+	mov b0=r16
+	br.many b0
+	;;
+kvm_asm_mov_to_rr_back_1:
+	adds r30=kvm_asm_mov_to_rr_back_2-asm_mov_from_reg,r20
+	shr.u r23=r19,61
+	shladd r17=r17,4,r20
+	;;
+	//if rr6, go back
+	cmp.eq p6,p0=6,r23
+	mov b0=r22
+	(p6) br.cond.dpnt.many kvm_virtualization_fault_back
+	;;
+	mov r28=r19
+	mov b0=r17
+	br.many b0
+kvm_asm_mov_to_rr_back_2:
+	adds r30=kvm_resume_to_guest-asm_mov_from_reg,r20
+	shladd r27=r23,3,r27
+	;; // vrr.rid<<4 |0xe
+	st8 [r27]=r19
+	mov b0=r30
+	;;
+	extr.u r16=r19,8,26
+	extr.u r18 =r19,2,6
+	mov r17 =0xe
+	;;
+	shladd r16 = r16, 4, r17
+	extr.u r19 =r19,0,8
+	;;
+	shl r16 = r16,8
+	;;
+	add r19 = r19, r16
+	;; //set ve 1
+	dep r19=-1,r19,0,1
+	cmp.lt p6,p0=14,r18
+	;;
+	(p6) mov r18=14
+	;;
+	(p6) dep r19=r18,r19,2,6
+	;;
+	cmp.eq p6,p0=0,r23
+	;;
+	cmp.eq.or p6,p0=4,r23
+	;;
+	adds r16=VMM_VCPU_MODE_FLAGS_OFFSET,r21
+	(p6) adds r17=VMM_VCPU_META_SAVED_RR0_OFFSET,r21
+	;;
+	ld4 r16=[r16]
+	cmp.eq p7,p0=r0,r0
+	(p6) shladd r17=r23,1,r17
+	;;
+	(p6) st8 [r17]=r19
+	(p6) tbit.nz p6,p7=r16,0
+	;;
+	(p7) mov rr[r28]=r19
+	mov r24=r22
+	br.many b0
+END(kvm_asm_mov_to_rr)
+
+
+//rsm
+GLOBAL_ENTRY(kvm_asm_rsm)
+#ifndef ACCE_RSM
+	br.many kvm_virtualization_fault_back
+#endif
+	add r16=VMM_VPD_BASE_OFFSET,r21
+	extr.u r26=r25,6,21
+	extr.u r27=r25,31,2
+	;;
+	ld8 r16=[r16]
+	extr.u r28=r25,36,1
+	dep r26=r27,r26,21,2
+	;;
+	add r17=VPD_VPSR_START_OFFSET,r16
+	add r22=VMM_VCPU_MODE_FLAGS_OFFSET,r21
+	//r26 is imm24
+	dep r26=r28,r26,23,1
+	;;
+	ld8 r18=[r17]
+	movl r28=IA64_PSR_IC+IA64_PSR_I+IA64_PSR_DT+IA64_PSR_SI
+	ld4 r23=[r22]
+	sub r27=-1,r26
+	mov r24=b0
+	;;
+	mov r20=cr.ipsr
+	or r28=r27,r28
+	and r19=r18,r27
+	;;
+	st8 [r17]=r19
+	and r20=r20,r28
+	/* Comment it out due to short of fp lazy alorgithm support
+	adds r27=IA64_VCPU_FP_PSR_OFFSET,r21
+	;;
+	ld8 r27=[r27]
+	;;
+	tbit.nz p8,p0= r27,IA64_PSR_DFH_BIT
+	;;
+	(p8) dep r20=-1,r20,IA64_PSR_DFH_BIT,1
+	*/
+	;;
+	mov cr.ipsr=r20
+	tbit.nz p6,p0=r23,0
+	;;
+	tbit.z.or p6,p0=r26,IA64_PSR_DT_BIT
+	(p6) br.dptk kvm_resume_to_guest
+	;;
+	add r26=VMM_VCPU_META_RR0_OFFSET,r21
+	add r27=VMM_VCPU_META_RR0_OFFSET+8,r21
+	dep r23=-1,r23,0,1
+	;;
+	ld8 r26=[r26]
+	ld8 r27=[r27]
+	st4 [r22]=r23
+	dep.z r28=4,61,3
+	;;
+	mov rr[r0]=r26
+	;;
+	mov rr[r28]=r27
+	;;
+	srlz.d
+	br.many kvm_resume_to_guest
+END(kvm_asm_rsm)
+
+
+//ssm
+GLOBAL_ENTRY(kvm_asm_ssm)
+#ifndef ACCE_SSM
+	br.many kvm_virtualization_fault_back
+#endif
+	add r16=VMM_VPD_BASE_OFFSET,r21
+	extr.u r26=r25,6,21
+	extr.u r27=r25,31,2
+	;;
+	ld8 r16=[r16]
+	extr.u r28=r25,36,1
+	dep r26=r27,r26,21,2
+	;;  //r26 is imm24
+	add r27=VPD_VPSR_START_OFFSET,r16
+	dep r26=r28,r26,23,1
+	;;  //r19 vpsr
+	ld8 r29=[r27]
+	mov r24=b0
+	;;
+	add r22=VMM_VCPU_MODE_FLAGS_OFFSET,r21
+	mov r20=cr.ipsr
+	or r19=r29,r26
+	;;
+	ld4 r23=[r22]
+	st8 [r27]=r19
+	or r20=r20,r26
+	;;
+	mov cr.ipsr=r20
+	movl r28=IA64_PSR_DT+IA64_PSR_RT+IA64_PSR_IT
+	;;
+	and r19=r28,r19
+	tbit.z p6,p0=r23,0
+	;;
+	cmp.ne.or p6,p0=r28,r19
+	(p6) br.dptk kvm_asm_ssm_1
+	;;
+	add r26=VMM_VCPU_META_SAVED_RR0_OFFSET,r21
+	add r27=VMM_VCPU_META_SAVED_RR0_OFFSET+8,r21
+	dep r23=0,r23,0,1
+	;;
+	ld8 r26=[r26]
+	ld8 r27=[r27]
+	st4 [r22]=r23
+	dep.z r28=4,61,3
+	;;
+	mov rr[r0]=r26
+	;;
+	mov rr[r28]=r27
+	;;
+	srlz.d
+	;;
+kvm_asm_ssm_1:
+	tbit.nz p6,p0=r29,IA64_PSR_I_BIT
+	;;
+	tbit.z.or p6,p0=r19,IA64_PSR_I_BIT
+	(p6) br.dptk kvm_resume_to_guest
+	;;
+	add r29=VPD_VTPR_START_OFFSET,r16
+	add r30=VPD_VHPI_START_OFFSET,r16
+	;;
+	ld8 r29=[r29]
+	ld8 r30=[r30]
+	;;
+	extr.u r17=r29,4,4
+	extr.u r18=r29,16,1
+	;;
+	dep r17=r18,r17,4,1
+	;;
+	cmp.gt p6,p0=r30,r17
+	(p6) br.dpnt.few kvm_asm_dispatch_vexirq
+	br.many kvm_resume_to_guest
+END(kvm_asm_ssm)
+
+
+//mov psr.l=r2
+GLOBAL_ENTRY(kvm_asm_mov_to_psr)
+#ifndef ACCE_MOV_TO_PSR
+	br.many kvm_virtualization_fault_back
+#endif
+	add r16=VMM_VPD_BASE_OFFSET,r21
+	extr.u r26=r25,13,7 //r2
+	;;
+	ld8 r16=[r16]
+	addl r20=@gprel(asm_mov_from_reg),gp
+	;;
+	adds r30=kvm_asm_mov_to_psr_back-asm_mov_from_reg,r20
+	shladd r26=r26,4,r20
+	mov r24=b0
+	;;
+	add r27=VPD_VPSR_START_OFFSET,r16
+	mov b0=r26
+	br.many b0
+	;;
+kvm_asm_mov_to_psr_back:
+	ld8 r17=[r27]
+	add r22=VMM_VCPU_MODE_FLAGS_OFFSET,r21
+	dep r19=0,r19,32,32
+	;;
+	ld4 r23=[r22]
+	dep r18=0,r17,0,32
+	;;
+	add r30=r18,r19
+	movl r28=IA64_PSR_DT+IA64_PSR_RT+IA64_PSR_IT
+	;;
+	st8 [r27]=r30
+	and r27=r28,r30
+	and r29=r28,r17
+	;;
+	cmp.eq p5,p0=r29,r27
+	cmp.eq p6,p7=r28,r27
+	(p5) br.many kvm_asm_mov_to_psr_1
+	;;
+	//virtual to physical
+	(p7) add r26=VMM_VCPU_META_RR0_OFFSET,r21
+	(p7) add r27=VMM_VCPU_META_RR0_OFFSET+8,r21
+	(p7) dep r23=-1,r23,0,1
+	;;
+	//physical to virtual
+	(p6) add r26=VMM_VCPU_META_SAVED_RR0_OFFSET,r21
+	(p6) add r27=VMM_VCPU_META_SAVED_RR0_OFFSET+8,r21
+	(p6) dep r23=0,r23,0,1
+	;;
+	ld8 r26=[r26]
+	ld8 r27=[r27]
+	st4 [r22]=r23
+	dep.z r28=4,61,3
+	;;
+	mov rr[r0]=r26
+	;;
+	mov rr[r28]=r27
+	;;
+	srlz.d
+	;;
+kvm_asm_mov_to_psr_1:
+	mov r20=cr.ipsr
+	movl r28=IA64_PSR_IC+IA64_PSR_I+IA64_PSR_DT+IA64_PSR_SI+IA64_PSR_RT
+	;;
+	or r19=r19,r28
+	dep r20=0,r20,0,32
+	;;
+	add r20=r19,r20
+	mov b0=r24
+	;;
+	/* Comment it out due to short of fp lazy algorithm support
+	adds r27=IA64_VCPU_FP_PSR_OFFSET,r21
+	;;
+	ld8 r27=[r27]
+	;;
+	tbit.nz p8,p0=r27,IA64_PSR_DFH_BIT
+	;;
+	(p8) dep r20=-1,r20,IA64_PSR_DFH_BIT,1
+	;;
+	*/
+	mov cr.ipsr=r20
+	cmp.ne p6,p0=r0,r0
+	;;
+	tbit.nz.or p6,p0=r17,IA64_PSR_I_BIT
+	tbit.z.or p6,p0=r30,IA64_PSR_I_BIT
+	(p6) br.dpnt.few kvm_resume_to_guest
+	;;
+	add r29=VPD_VTPR_START_OFFSET,r16
+	add r30=VPD_VHPI_START_OFFSET,r16
+	;;
+	ld8 r29=[r29]
+	ld8 r30=[r30]
+	;;
+	extr.u r17=r29,4,4
+	extr.u r18=r29,16,1
+	;;
+	dep r17=r18,r17,4,1
+	;;
+	cmp.gt p6,p0=r30,r17
+	(p6) br.dpnt.few kvm_asm_dispatch_vexirq
+	br.many kvm_resume_to_guest
+END(kvm_asm_mov_to_psr)
+
+
+ENTRY(kvm_asm_dispatch_vexirq)
+//increment iip
+	mov r16=cr.ipsr
+	;;
+	extr.u r17=r16,IA64_PSR_RI_BIT,2
+	tbit.nz p6,p7=r16,IA64_PSR_RI_BIT+1
+	;;
+	(p6) mov r18=cr.iip
+	(p6) mov r17=r0
+	(p7) add r17=1,r17
+	;;
+	(p6) add r18=0x10,r18
+	dep r16=r17,r16,IA64_PSR_RI_BIT,2
+	;;
+	(p6) mov cr.iip=r18
+	mov cr.ipsr=r16
+	mov r30 =1
+	br.many kvm_dispatch_vexirq
+END(kvm_asm_dispatch_vexirq)
+
+// thash
+// TODO: add support when pta.vf = 1
+GLOBAL_ENTRY(kvm_asm_thash)
+#ifndef ACCE_THASH
+	br.many kvm_virtualization_fault_back
+#endif
+	extr.u r17=r25,20,7		// get r3 from opcode in r25
+	extr.u r18=r25,6,7		// get r1 from opcode in r25
+	addl r20=@gprel(asm_mov_from_reg),gp
+	;;
+	adds r30=kvm_asm_thash_back1-asm_mov_from_reg,r20
+	shladd r17=r17,4,r20	// get addr of MOVE_FROM_REG(r17)
+	adds r16=VMM_VPD_BASE_OFFSET,r21	// get vcpu.arch.priveregs
+	;;
+	mov r24=b0
+	;;
+	ld8 r16=[r16]		// get VPD addr
+	mov b0=r17
+	br.many b0			// r19 return value
+	;;
+kvm_asm_thash_back1:
+	shr.u r23=r19,61		// get RR number
+	adds r25=VMM_VCPU_VRR0_OFFSET,r21	// get vcpu->arch.vrr[0]'s addr
+	adds r16=VMM_VPD_VPTA_OFFSET,r16	// get vpta
+	;;
+	shladd r27=r23,3,r25	// get vcpu->arch.vrr[r23]'s addr
+	ld8 r17=[r16]		// get PTA
+	mov r26=1
+	;;
+	extr.u r29=r17,2,6		// get pta.size
+	ld8 r25=[r27]		// get vcpu->arch.vrr[r23]'s value
+	;;
+	extr.u r25=r25,2,6		// get rr.ps
+	shl r22=r26,r29		// 1UL << pta.size
+	;;
+	shr.u r23=r19,r25		// vaddr >> rr.ps
+	adds r26=3,r29		// pta.size + 3
+	shl r27=r17,3		// pta << 3
+	;;
+	shl r23=r23,3		// (vaddr >> rr.ps) << 3
+	shr.u r27=r27,r26		// (pta << 3) >> (pta.size+3)
+	movl r16=7<<61
+	;;
+	adds r22=-1,r22		// (1UL << pta.size) - 1
+	shl r27=r27,r29		// ((pta<<3)>>(pta.size+3))<<pta.size
+	and r19=r19,r16		// vaddr & VRN_MASK
+	;;
+	and r22=r22,r23		// vhpt_offset
+	or r19=r19,r27 // (vadr&VRN_MASK)|(((pta<<3)>>(pta.size + 3))<<pta.size)
+	adds r26=asm_mov_to_reg-asm_mov_from_reg,r20
+	;;
+	or r19=r19,r22		// calc pval
+	shladd r17=r18,4,r26
+	adds r30=kvm_resume_to_guest-asm_mov_from_reg,r20
+	;;
+	mov b0=r17
+	br.many b0
+END(kvm_asm_thash)
+
+#define MOV_TO_REG0	\
+{;			\
+	nop.b 0x0;		\
+	nop.b 0x0;		\
+	nop.b 0x0;		\
+	;;			\
+};
+
+
+#define MOV_TO_REG(n)	\
+{;			\
+	mov r##n##=r19;	\
+	mov b0=r30;	\
+	br.sptk.many b0;	\
+	;;			\
+};
+
+
+#define MOV_FROM_REG(n)	\
+{;				\
+	mov r19=r##n##;		\
+	mov b0=r30;		\
+	br.sptk.many b0;		\
+	;;				\
+};
+
+
+#define MOV_TO_BANK0_REG(n)			\
+ENTRY_MIN_ALIGN(asm_mov_to_bank0_reg##n##);	\
+{;						\
+	mov r26=r2;				\
+	mov r2=r19;				\
+	bsw.1;					\
+	;;						\
+};						\
+{;						\
+	mov r##n##=r2;				\
+	nop.b 0x0;					\
+	bsw.0;					\
+	;;						\
+};						\
+{;						\
+	mov r2=r26;				\
+	mov b0=r30;				\
+	br.sptk.many b0;				\
+	;;						\
+};						\
+END(asm_mov_to_bank0_reg##n##)
+
+
+#define MOV_FROM_BANK0_REG(n)			\
+ENTRY_MIN_ALIGN(asm_mov_from_bank0_reg##n##);	\
+{;						\
+	mov r26=r2;				\
+	nop.b 0x0;					\
+	bsw.1;					\
+	;;						\
+};						\
+{;						\
+	mov r2=r##n##;				\
+	nop.b 0x0;					\
+	bsw.0;					\
+	;;						\
+};						\
+{;						\
+	mov r19=r2;				\
+	mov r2=r26;				\
+	mov b0=r30;				\
+};						\
+{;						\
+	nop.b 0x0;					\
+	nop.b 0x0;					\
+	br.sptk.many b0;				\
+	;;						\
+};						\
+END(asm_mov_from_bank0_reg##n##)
+
+
+#define JMP_TO_MOV_TO_BANK0_REG(n)		\
+{;						\
+	nop.b 0x0;					\
+	nop.b 0x0;					\
+	br.sptk.many asm_mov_to_bank0_reg##n##;	\
+	;;						\
+}
+
+
+#define JMP_TO_MOV_FROM_BANK0_REG(n)		\
+{;						\
+	nop.b 0x0;					\
+	nop.b 0x0;					\
+	br.sptk.many asm_mov_from_bank0_reg##n##;	\
+	;;						\
+}
+
+
+MOV_FROM_BANK0_REG(16)
+MOV_FROM_BANK0_REG(17)
+MOV_FROM_BANK0_REG(18)
+MOV_FROM_BANK0_REG(19)
+MOV_FROM_BANK0_REG(20)
+MOV_FROM_BANK0_REG(21)
+MOV_FROM_BANK0_REG(22)
+MOV_FROM_BANK0_REG(23)
+MOV_FROM_BANK0_REG(24)
+MOV_FROM_BANK0_REG(25)
+MOV_FROM_BANK0_REG(26)
+MOV_FROM_BANK0_REG(27)
+MOV_FROM_BANK0_REG(28)
+MOV_FROM_BANK0_REG(29)
+MOV_FROM_BANK0_REG(30)
+MOV_FROM_BANK0_REG(31)
+
+
+// mov from reg table
+ENTRY(asm_mov_from_reg)
+	MOV_FROM_REG(0)
+	MOV_FROM_REG(1)
+	MOV_FROM_REG(2)
+	MOV_FROM_REG(3)
+	MOV_FROM_REG(4)
+	MOV_FROM_REG(5)
+	MOV_FROM_REG(6)
+	MOV_FROM_REG(7)
+	MOV_FROM_REG(8)
+	MOV_FROM_REG(9)
+	MOV_FROM_REG(10)
+	MOV_FROM_REG(11)
+	MOV_FROM_REG(12)
+	MOV_FROM_REG(13)
+	MOV_FROM_REG(14)
+	MOV_FROM_REG(15)
+	JMP_TO_MOV_FROM_BANK0_REG(16)
+	JMP_TO_MOV_FROM_BANK0_REG(17)
+	JMP_TO_MOV_FROM_BANK0_REG(18)
+	JMP_TO_MOV_FROM_BANK0_REG(19)
+	JMP_TO_MOV_FROM_BANK0_REG(20)
+	JMP_TO_MOV_FROM_BANK0_REG(21)
+	JMP_TO_MOV_FROM_BANK0_REG(22)
+	JMP_TO_MOV_FROM_BANK0_REG(23)
+	JMP_TO_MOV_FROM_BANK0_REG(24)
+	JMP_TO_MOV_FROM_BANK0_REG(25)
+	JMP_TO_MOV_FROM_BANK0_REG(26)
+	JMP_TO_MOV_FROM_BANK0_REG(27)
+	JMP_TO_MOV_FROM_BANK0_REG(28)
+	JMP_TO_MOV_FROM_BANK0_REG(29)
+	JMP_TO_MOV_FROM_BANK0_REG(30)
+	JMP_TO_MOV_FROM_BANK0_REG(31)
+	MOV_FROM_REG(32)
+	MOV_FROM_REG(33)
+	MOV_FROM_REG(34)
+	MOV_FROM_REG(35)
+	MOV_FROM_REG(36)
+	MOV_FROM_REG(37)
+	MOV_FROM_REG(38)
+	MOV_FROM_REG(39)
+	MOV_FROM_REG(40)
+	MOV_FROM_REG(41)
+	MOV_FROM_REG(42)
+	MOV_FROM_REG(43)
+	MOV_FROM_REG(44)
+	MOV_FROM_REG(45)
+	MOV_FROM_REG(46)
+	MOV_FROM_REG(47)
+	MOV_FROM_REG(48)
+	MOV_FROM_REG(49)
+	MOV_FROM_REG(50)
+	MOV_FROM_REG(51)
+	MOV_FROM_REG(52)
+	MOV_FROM_REG(53)
+	MOV_FROM_REG(54)
+	MOV_FROM_REG(55)
+	MOV_FROM_REG(56)
+	MOV_FROM_REG(57)
+	MOV_FROM_REG(58)
+	MOV_FROM_REG(59)
+	MOV_FROM_REG(60)
+	MOV_FROM_REG(61)
+	MOV_FROM_REG(62)
+	MOV_FROM_REG(63)
+	MOV_FROM_REG(64)
+	MOV_FROM_REG(65)
+	MOV_FROM_REG(66)
+	MOV_FROM_REG(67)
+	MOV_FROM_REG(68)
+	MOV_FROM_REG(69)
+	MOV_FROM_REG(70)
+	MOV_FROM_REG(71)
+	MOV_FROM_REG(72)
+	MOV_FROM_REG(73)
+	MOV_FROM_REG(74)
+	MOV_FROM_REG(75)
+	MOV_FROM_REG(76)
+	MOV_FROM_REG(77)
+	MOV_FROM_REG(78)
+	MOV_FROM_REG(79)
+	MOV_FROM_REG(80)
+	MOV_FROM_REG(81)
+	MOV_FROM_REG(82)
+	MOV_FROM_REG(83)
+	MOV_FROM_REG(84)
+	MOV_FROM_REG(85)
+	MOV_FROM_REG(86)
+	MOV_FROM_REG(87)
+	MOV_FROM_REG(88)
+	MOV_FROM_REG(89)
+	MOV_FROM_REG(90)
+	MOV_FROM_REG(91)
+	MOV_FROM_REG(92)
+	MOV_FROM_REG(93)
+	MOV_FROM_REG(94)
+	MOV_FROM_REG(95)
+	MOV_FROM_REG(96)
+	MOV_FROM_REG(97)
+	MOV_FROM_REG(98)
+	MOV_FROM_REG(99)
+	MOV_FROM_REG(100)
+	MOV_FROM_REG(101)
+	MOV_FROM_REG(102)
+	MOV_FROM_REG(103)
+	MOV_FROM_REG(104)
+	MOV_FROM_REG(105)
+	MOV_FROM_REG(106)
+	MOV_FROM_REG(107)
+	MOV_FROM_REG(108)
+	MOV_FROM_REG(109)
+	MOV_FROM_REG(110)
+	MOV_FROM_REG(111)
+	MOV_FROM_REG(112)
+	MOV_FROM_REG(113)
+	MOV_FROM_REG(114)
+	MOV_FROM_REG(115)
+	MOV_FROM_REG(116)
+	MOV_FROM_REG(117)
+	MOV_FROM_REG(118)
+	MOV_FROM_REG(119)
+	MOV_FROM_REG(120)
+	MOV_FROM_REG(121)
+	MOV_FROM_REG(122)
+	MOV_FROM_REG(123)
+	MOV_FROM_REG(124)
+	MOV_FROM_REG(125)
+	MOV_FROM_REG(126)
+	MOV_FROM_REG(127)
+END(asm_mov_from_reg)
+
+
+/* must be in bank 0
+ * parameter:
+ * r31: pr
+ * r24: b0
+ */
+ENTRY(kvm_resume_to_guest)
+	adds r16 = VMM_VCPU_SAVED_GP_OFFSET,r21
+	;;
+	ld8 r1 =[r16]
+	adds r20 = VMM_VCPU_VSA_BASE_OFFSET,r21
+	;;
+	mov r16=cr.ipsr
+	;;
+	ld8 r20 = [r20]
+	adds r19=VMM_VPD_BASE_OFFSET,r21
+	;;
+	ld8 r25=[r19]
+	extr.u r17=r16,IA64_PSR_RI_BIT,2
+	tbit.nz p6,p7=r16,IA64_PSR_RI_BIT+1
+	;;
+	(p6) mov r18=cr.iip
+	(p6) mov r17=r0
+	;;
+	(p6) add r18=0x10,r18
+	(p7) add r17=1,r17
+	;;
+	(p6) mov cr.iip=r18
+	dep r16=r17,r16,IA64_PSR_RI_BIT,2
+	;;
+	mov cr.ipsr=r16
+	adds r19= VPD_VPSR_START_OFFSET,r25
+	add r28=PAL_VPS_RESUME_NORMAL,r20
+	add r29=PAL_VPS_RESUME_HANDLER,r20
+	;;
+	ld8 r19=[r19]
+	mov b0=r29
+	cmp.ne p6,p7 = r0,r0
+	;;
+	tbit.z p6,p7 = r19,IA64_PSR_IC_BIT		// p1=vpsr.ic
+	;;
+	(p6) ld8 r26=[r25]
+	(p7) mov b0=r28
+	mov pr=r31,-2
+	br.sptk.many b0             // call pal service
+	;;
+END(kvm_resume_to_guest)
+
+
+MOV_TO_BANK0_REG(16)
+MOV_TO_BANK0_REG(17)
+MOV_TO_BANK0_REG(18)
+MOV_TO_BANK0_REG(19)
+MOV_TO_BANK0_REG(20)
+MOV_TO_BANK0_REG(21)
+MOV_TO_BANK0_REG(22)
+MOV_TO_BANK0_REG(23)
+MOV_TO_BANK0_REG(24)
+MOV_TO_BANK0_REG(25)
+MOV_TO_BANK0_REG(26)
+MOV_TO_BANK0_REG(27)
+MOV_TO_BANK0_REG(28)
+MOV_TO_BANK0_REG(29)
+MOV_TO_BANK0_REG(30)
+MOV_TO_BANK0_REG(31)
+
+
+// mov to reg table
+ENTRY(asm_mov_to_reg)
+	MOV_TO_REG0
+	MOV_TO_REG(1)
+	MOV_TO_REG(2)
+	MOV_TO_REG(3)
+	MOV_TO_REG(4)
+	MOV_TO_REG(5)
+	MOV_TO_REG(6)
+	MOV_TO_REG(7)
+	MOV_TO_REG(8)
+	MOV_TO_REG(9)
+	MOV_TO_REG(10)
+	MOV_TO_REG(11)
+	MOV_TO_REG(12)
+	MOV_TO_REG(13)
+	MOV_TO_REG(14)
+	MOV_TO_REG(15)
+	JMP_TO_MOV_TO_BANK0_REG(16)
+	JMP_TO_MOV_TO_BANK0_REG(17)
+	JMP_TO_MOV_TO_BANK0_REG(18)
+	JMP_TO_MOV_TO_BANK0_REG(19)
+	JMP_TO_MOV_TO_BANK0_REG(20)
+	JMP_TO_MOV_TO_BANK0_REG(21)
+	JMP_TO_MOV_TO_BANK0_REG(22)
+	JMP_TO_MOV_TO_BANK0_REG(23)
+	JMP_TO_MOV_TO_BANK0_REG(24)
+	JMP_TO_MOV_TO_BANK0_REG(25)
+	JMP_TO_MOV_TO_BANK0_REG(26)
+	JMP_TO_MOV_TO_BANK0_REG(27)
+	JMP_TO_MOV_TO_BANK0_REG(28)
+	JMP_TO_MOV_TO_BANK0_REG(29)
+	JMP_TO_MOV_TO_BANK0_REG(30)
+	JMP_TO_MOV_TO_BANK0_REG(31)
+	MOV_TO_REG(32)
+	MOV_TO_REG(33)
+	MOV_TO_REG(34)
+	MOV_TO_REG(35)
+	MOV_TO_REG(36)
+	MOV_TO_REG(37)
+	MOV_TO_REG(38)
+	MOV_TO_REG(39)
+	MOV_TO_REG(40)
+	MOV_TO_REG(41)
+	MOV_TO_REG(42)
+	MOV_TO_REG(43)
+	MOV_TO_REG(44)
+	MOV_TO_REG(45)
+	MOV_TO_REG(46)
+	MOV_TO_REG(47)
+	MOV_TO_REG(48)
+	MOV_TO_REG(49)
+	MOV_TO_REG(50)
+	MOV_TO_REG(51)
+	MOV_TO_REG(52)
+	MOV_TO_REG(53)
+	MOV_TO_REG(54)
+	MOV_TO_REG(55)
+	MOV_TO_REG(56)
+	MOV_TO_REG(57)
+	MOV_TO_REG(58)
+	MOV_TO_REG(59)
+	MOV_TO_REG(60)
+	MOV_TO_REG(61)
+	MOV_TO_REG(62)
+	MOV_TO_REG(63)
+	MOV_TO_REG(64)
+	MOV_TO_REG(65)
+	MOV_TO_REG(66)
+	MOV_TO_REG(67)
+	MOV_TO_REG(68)
+	MOV_TO_REG(69)
+	MOV_TO_REG(70)
+	MOV_TO_REG(71)
+	MOV_TO_REG(72)
+	MOV_TO_REG(73)
+	MOV_TO_REG(74)
+	MOV_TO_REG(75)
+	MOV_TO_REG(76)
+	MOV_TO_REG(77)
+	MOV_TO_REG(78)
+	MOV_TO_REG(79)
+	MOV_TO_REG(80)
+	MOV_TO_REG(81)
+	MOV_TO_REG(82)
+	MOV_TO_REG(83)
+	MOV_TO_REG(84)
+	MOV_TO_REG(85)
+	MOV_TO_REG(86)
+	MOV_TO_REG(87)
+	MOV_TO_REG(88)
+	MOV_TO_REG(89)
+	MOV_TO_REG(90)
+	MOV_TO_REG(91)
+	MOV_TO_REG(92)
+	MOV_TO_REG(93)
+	MOV_TO_REG(94)
+	MOV_TO_REG(95)
+	MOV_TO_REG(96)
+	MOV_TO_REG(97)
+	MOV_TO_REG(98)
+	MOV_TO_REG(99)
+	MOV_TO_REG(100)
+	MOV_TO_REG(101)
+	MOV_TO_REG(102)
+	MOV_TO_REG(103)
+	MOV_TO_REG(104)
+	MOV_TO_REG(105)
+	MOV_TO_REG(106)
+	MOV_TO_REG(107)
+	MOV_TO_REG(108)
+	MOV_TO_REG(109)
+	MOV_TO_REG(110)
+	MOV_TO_REG(111)
+	MOV_TO_REG(112)
+	MOV_TO_REG(113)
+	MOV_TO_REG(114)
+	MOV_TO_REG(115)
+	MOV_TO_REG(116)
+	MOV_TO_REG(117)
+	MOV_TO_REG(118)
+	MOV_TO_REG(119)
+	MOV_TO_REG(120)
+	MOV_TO_REG(121)
+	MOV_TO_REG(122)
+	MOV_TO_REG(123)
+	MOV_TO_REG(124)
+	MOV_TO_REG(125)
+	MOV_TO_REG(126)
+	MOV_TO_REG(127)
+END(asm_mov_to_reg)