perf_asm.s

这个linux源代码是很全面的~基本完整了~使用c编译的~由于时间问题我没有亲自测试~但就算用来做参考资料也是非常好的

; 
;   Purpose:
;	This file has the overall purpose of suppyling low-level
;   assembly to program the intrigue portion of the cpu.
; 

#include <linux/config.h>
#include <asm/assembly.h>

#ifdef __LP64__
	.level		2.0w
#endif /* __LP64__ */

#define MTDIAG_1(gr)    .word 0x14201840 + gr*0x10000
#define MTDIAG_2(gr)    .word 0x14401840 + gr*0x10000
#define MFDIAG_1(gr)    .word 0x142008A0 + gr
#define MFDIAG_2(gr)    .word 0x144008A0 + gr
#define STDIAG(dr)      .word 0x14000AA0 + dr*0x200000
#define SFDIAG(dr)      .word 0x14000BA0 + dr*0x200000
#define DR2_SLOW_RET    53


;
; Enable the performance counters
;
; The coprocessor only needs to be enabled when
; starting/stopping the coprocessor with the pmenb/pmdis.
;
	.text
	.align 32

	.export perf_intrigue_enable_perf_counters,code
perf_intrigue_enable_perf_counters:
	.proc
	.callinfo  frame=0,NO_CALLS
	.entry
	
	ldi     0x20,%r25                ; load up perfmon bit
	mfctl   ccr,%r26                 ; get coprocessor register
	or      %r25,%r26,%r26             ; set bit
	mtctl   %r26,ccr                 ; turn on performance coprocessor
	pmenb                           ; enable performance monitor
	ssm     0,0                     ; dummy op to ensure completion
	sync                            ; follow ERS
	andcm   %r26,%r25,%r26             ; clear bit now 
	mtctl   %r26,ccr                 ; turn off performance coprocessor
	nop                             ; NOPs as specified in ERS
	nop
	nop
	nop
	nop
	nop
	nop
	bve    (%r2)
	nop
	.exit
	.procend

	.export perf_intrigue_disable_perf_counters,code
perf_intrigue_disable_perf_counters:
	.proc
	.callinfo  frame=0,NO_CALLS
	.entry
	ldi     0x20,%r25                ; load up perfmon bit
	mfctl   ccr,%r26                 ; get coprocessor register
	or      %r25,%r26,%r26             ; set bit
	mtctl   %r26,ccr                 ; turn on performance coprocessor
	pmdis                           ; disable performance monitor
	ssm     0,0                     ; dummy op to ensure completion
	andcm   %r26,%r25,%r26             ; clear bit now 
	bve    (%r2)
	mtctl   %r26,ccr                 ; turn off performance coprocessor
	.exit
	.procend

;************************************************************************
;*																		*
;* Name: perf_rdr_shift_in_W												*
;*																		*
;* Description:															*
;*	This routine shifts data in from the RDR in arg0 and returns		*
;*	the result in ret0.  If the RDR is <= 64 bits in length, it			*
;*	is shifted shifted backup immediately.  This is to compensate		*
;*	for RDR10 which has bits that preclude PDC stack operations			*
;*	when they are in the wrong state.									*
;*																		*
;* Arguments:															*
;*	arg0 : rdr to be read												*
;*	arg1 : bit length of rdr											*
;*																		*
;* Returns:																*
;*	ret0 = next 64 bits of rdr data from staging register				*
;*																		*
;* Register usage:														*
;*	arg0 : rdr to be read												*
;*	arg1 : bit length of rdr											*
;*	%r24  - original DR2 value											*
;*	%r1   - scratch														*
;*  %r29  - scratch														*
;*																		*
;* Returns:																*
;*	ret0 = RDR data (right justified)									*
;*																		*
;************************************************************************

	.export perf_rdr_shift_in_W,code
perf_rdr_shift_in_W:
	.proc
	.callinfo frame=0,NO_CALLS
	.entry
;
; read(shift in) the RDR.
;

; NOTE: The PCX-W ERS states that DR2_SLOW_RET must be set before any
; shifting is done, from or to, remote diagnose registers.
;

	depdi,z		1,DR2_SLOW_RET,1,%r29
	MFDIAG_2	(24)
	or		    %r24,%r29,%r29
	MTDIAG_2	(29)			; set DR2_SLOW_RET

	nop
	nop
	nop
	nop	

;
; Cacheline start (32-byte cacheline)
;
	nop
	nop
	nop
	extrd,u		arg1,63,6,%r1    ; setup shift amount based on bits to move 

	mtsar		%r1
	shladd		arg0,2,%r0,%r1	; %r1 = 4 * RDR number
	blr  		%r1,%r0		    ; branch to 8-instruction sequence
	nop

;
; Cacheline start (32-byte cacheline)
;

	;
	; RDR 0 sequence
	;
	SFDIAG		(0)		
	ssm		    0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)					; mtdiag %dr1, %r1 
	STDIAG		(0)
	ssm		    0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 1 sequence
	;
	sync			
	ssm		    0,0
	SFDIAG		(1)
	ssm		    0,0
	MFDIAG_1	(28)
	ssm		    0,0
	b,n         perf_rdr_shift_in_W_leave
	nop
	
	;
	; RDR 2 read sequence
	;
	SFDIAG		(2)		
	ssm		    0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(2)
	ssm		    0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 3 read sequence
	;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 4 read sequence
	;
	sync				
	ssm			0,0
	SFDIAG		(4)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	; 
	; RDR 5 read sequence
	;
	sync	
	ssm			0,0
	SFDIAG		(5)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 6 read sequence
	;
	sync	
	ssm			0,0
	SFDIAG		(6)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 7 read sequence
	;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 8 read sequence
	;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 9 read sequence
	;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 10 read sequence
	;
	SFDIAG		(10)
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(10)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 11 read sequence
	;
	SFDIAG		(11)
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(11)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 12 read sequence
	;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 13 read sequence
	;
	sync
	ssm			0,0
	SFDIAG		(13)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 14 read sequence
	;
	SFDIAG		(14)	
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(14)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 15 read sequence
	;
	sync				; RDR 15 read sequence
	ssm			0,0
	SFDIAG		(15)
	ssm			0,0
	MFDIAG_1	(28)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	nop
	
	;
	; RDR 16 read sequence
	;
	sync				; RDR 16 read sequence
	ssm			0,0
	SFDIAG		(16)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 17 read sequence
	;
	SFDIAG		(17)	
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(17)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 18 read sequence
	;
	SFDIAG		(18)	
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(18)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
;
; RDR 19 read sequence
;
	b,n         perf_rdr_shift_in_W_leave
	nop
	nop
	nop
	nop
	nop
	nop
	nop

	;
	; RDR 20 read sequence
	;
	sync
	ssm			0,0
	SFDIAG		(20)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 21 read sequence
	;
	sync
	ssm			0,0
	SFDIAG		(21)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 22 read sequence
	;
	sync
	ssm			0,0
	SFDIAG		(22)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 23 read sequence
	;
	sync		
	ssm			0,0
	SFDIAG		(23)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 24 read sequence
	;
	sync	
	ssm			0,0
	SFDIAG		(24)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 25 read sequence
	;
	sync
	ssm			0,0
	SFDIAG		(25)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 26 read sequence
	;
	SFDIAG		(26)		
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(26)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 27 read sequence
	;
	SFDIAG		(27)
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(27)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 28 read sequence
	;
	sync				
	ssm			0,0
	SFDIAG		(28)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 29 read sequence
	;
	sync		
	ssm			0,0
	SFDIAG		(29)
	ssm			0,0
	MFDIAG_1	(28)
	b,n         perf_rdr_shift_in_W_leave
	ssm			0,0
	nop
	
	;
	; RDR 30 read sequence
	;
	SFDIAG		(30)
	ssm			0,0
	MFDIAG_1	(28)
	shrpd		ret0,%r0,%sar,%r1
	MTDIAG_1	(1)
	STDIAG		(30)
	ssm			0,0
	b,n         perf_rdr_shift_in_W_leave
	
	;
	; RDR 31 read sequence
	;
	sync		
	ssm			0,0
	SFDIAG		(31)
	ssm			0,0
	MFDIAG_1	(28)
	nop
	ssm			0,0
	nop

	;
	; Fallthrough
	;

perf_rdr_shift_in_W_leave:
	bve		    (%r2)
	.exit
	MTDIAG_2	(24)			; restore DR2
	.procend


;************************************************************************
;*																		*
;* Name: perf_rdr_shift_out_W												*
;*																		*
;* Description:															*
;*	This routine moves data to the RDR's.  The double-word that			*
;*	arg1 points to is loaded and moved into the staging register.		*
;*	Then the STDIAG instruction for the RDR # in arg0 is called			*
;*	to move the data to the RDR.										*
;*																		*
;* Arguments:															*
;*	arg0 = rdr number													*
;*	arg1 = 64-bit value to write										*
;*	%r24 - DR2 | DR2_SLOW_RET											*
;*	%r23 - original DR2 value											*
;*																		*
;* Returns:																*
;*	None																*
;*																		*
;* Register usage:														*
;*																		*
;************************************************************************

	.export perf_rdr_shift_out_W,code
perf_rdr_shift_out_W:
	.proc
	.callinfo frame=0,NO_CALLS
	.entry
;
; NOTE: The PCX-W ERS states that DR2_SLOW_RET must be set before any
; shifting is done, from or to, the remote diagnose registers.
;

	depdi,z		1,DR2_SLOW_RET,1,%r24
	MFDIAG_2	(23)
	or		     %r24,%r23,%r24
	MTDIAG_2	(24)			; set DR2_SLOW_RET

	MTDIAG_1	(25)			; data to the staging register
	shladd		arg0,2,%r0,%r1	; %r1 = 4 * RDR number
	blr		    %r1,%r0		    ; branch to 8-instruction sequence
	nop

	;
	; RDR 0 write sequence
	;
	sync				; RDR 0 write sequence
	ssm			0,0
	STDIAG		(0)
	ssm			0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm			0,0
	nop

	;
	; RDR 1 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(1)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 2 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(2)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 3 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(3)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 4 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(4)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 5 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(5)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 6 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(6)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 7 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(7)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 8 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(8)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 9 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(9)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 10 write sequence
	;
	sync	
	ssm		0,0
	STDIAG		(10)
	STDIAG		(26)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	ssm		0,0
	nop

	;
	; RDR 11 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(11)
	STDIAG		(27)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	ssm		0,0
	nop

	;
	; RDR 12 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(12)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 13 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(13)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 14 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(14)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 15 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(15)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 16 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(16)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 17 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(17)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 18 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(18)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 19 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(19)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 20 write sequence
	;
	sync
	ssm		0,0
	STDIAG		(20)
	ssm		0,0
	b,n         perf_rdr_shift_out_W_leave
	nop
	ssm		0,0
	nop

	;
	; RDR 21 write sequence
	;
	sync
	ssm		0,0