pa-risc2w.s

mediastreamer2是开源的网络传输媒体流的库

    FSTD    ht_temp,-24(%sp)          ; -24(sp) = ht
    XMPYU   flt_0,fw_l,lt_temp        ; lt_temp = lt*fw_l
    FSTD    lt_temp,-32(%sp)          ; -32(sp) = lt 

    LDD     -8(%sp),m_0               
    LDD    -16(%sp),m1_0              
    ADD,L   m_0,m1_0,tmp_0            ; tmp_0 = m + m1; 
    LDD     -24(%sp),ht_0             
    LDD     -32(%sp),lt_0             

    CMPCLR,*>>= tmp_0,m1_0,%r0        ; if (m < m1)
    ADD,L   ht_0,top_overflow,ht_0    ; ht += (1<<32)

    EXTRD,U tmp_0,31,32,m_0           ; m>>32  
    DEPD,Z  tmp_0,31,32,m1_0          ; m1 = m<<32 

    ADD,L   ht_0,m_0,ht_0             ; ht+= (m>>32)
    ADD     lt_0,m1_0,lt_0            ; lt= lt+m1;
    ADD,DC  ht_0,%r0,ht_0             ; ht++

    ADD     %ret0,lt_0,lt_0           ; lt = lt + c;
    ADD,DC  ht_0,%r0,ht_0             ; ht++

    COPY    ht_0,%ret0                ; copy carry
    STD     lt_0,0(r_ptr)             ; rp[0] = lt

bn_mul_words_exit
    .EXIT
    LDD     -96(%sp),%r7              ; restore r7  
    LDD     -104(%sp),%r6             ; restore r6  
    LDD     -112(%sp),%r5             ; restore r5  
    LDD     -120(%sp),%r4             ; restore r4  
    BVE     (%rp)
    LDD,MB  -128(%sp),%r3             ; restore r3
	.PROCEND	;in=23,24,25,26,29;out=28;

;----------------------------------------------------------------------------
;
;void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num)
;
; arg0 = rp
; arg1 = ap
; arg2 = num
;

bn_sqr_words
	.proc
	.callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
	.EXPORT	bn_sqr_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
    .entry
	.align 64

    STD     %r3,0(%sp)          ; save r3  
    STD     %r4,8(%sp)          ; save r4  
	NOP
    STD     %r5,16(%sp)         ; save r5  

    CMPIB,>= 0,num,bn_sqr_words_exit
	LDO     128(%sp),%sp       ; bump stack

	;
	; If only 1, the goto straight to cleanup
	;
	CMPIB,= 1,num,bn_sqr_words_single_top
    DEPDI,Z -1,32,33,high_mask   ; Create Mask 0xffffffff80000000L

	;
	; This loop is unrolled 2 times (64-byte aligned as well)
	;

bn_sqr_words_unroll2
    FLDD    0(a_ptr),t_float_0        ; a[0]
    FLDD    8(a_ptr),t_float_1        ; a[1]
    XMPYU   fht_0,flt_0,fm            ; m[0]
    XMPYU   fht_1,flt_1,fm_1          ; m[1]

    FSTD    fm,-24(%sp)               ; store m[0]
    FSTD    fm_1,-56(%sp)             ; store m[1]
    XMPYU   flt_0,flt_0,lt_temp       ; lt[0]
    XMPYU   flt_1,flt_1,lt_temp_1     ; lt[1]

    FSTD    lt_temp,-16(%sp)          ; store lt[0]
    FSTD    lt_temp_1,-48(%sp)        ; store lt[1]
    XMPYU   fht_0,fht_0,ht_temp       ; ht[0]
    XMPYU   fht_1,fht_1,ht_temp_1     ; ht[1]

    FSTD    ht_temp,-8(%sp)           ; store ht[0]
    FSTD    ht_temp_1,-40(%sp)        ; store ht[1]
    LDD     -24(%sp),m_0             
    LDD     -56(%sp),m_1              

    AND     m_0,high_mask,tmp_0       ; m[0] & Mask
    AND     m_1,high_mask,tmp_1       ; m[1] & Mask
    DEPD,Z  m_0,30,31,m_0             ; m[0] << 32+1
    DEPD,Z  m_1,30,31,m_1             ; m[1] << 32+1

    LDD     -16(%sp),lt_0        
    LDD     -48(%sp),lt_1        
    EXTRD,U tmp_0,32,33,tmp_0         ; tmp_0 = m[0]&Mask >> 32-1
    EXTRD,U tmp_1,32,33,tmp_1         ; tmp_1 = m[1]&Mask >> 32-1

    LDD     -8(%sp),ht_0            
    LDD     -40(%sp),ht_1           
    ADD,L   ht_0,tmp_0,ht_0           ; ht[0] += tmp_0
    ADD,L   ht_1,tmp_1,ht_1           ; ht[1] += tmp_1

    ADD     lt_0,m_0,lt_0             ; lt = lt+m
    ADD,DC  ht_0,%r0,ht_0             ; ht[0]++
    STD     lt_0,0(r_ptr)             ; rp[0] = lt[0]
    STD     ht_0,8(r_ptr)             ; rp[1] = ht[1]

    ADD     lt_1,m_1,lt_1             ; lt = lt+m
    ADD,DC  ht_1,%r0,ht_1             ; ht[1]++
    STD     lt_1,16(r_ptr)            ; rp[2] = lt[1]
    STD     ht_1,24(r_ptr)            ; rp[3] = ht[1]

	LDO    -2(num),num                ; num = num - 2;
    LDO     16(a_ptr),a_ptr           ; ap += 2
	CMPIB,<= 2,num,bn_sqr_words_unroll2
    LDO     32(r_ptr),r_ptr           ; rp += 4

    CMPIB,=,N 0,num,bn_sqr_words_exit ; are we done?

	;
	; Top of loop aligned on 64-byte boundary
	;
bn_sqr_words_single_top
    FLDD    0(a_ptr),t_float_0        ; load up 64-bit value (fr8L) ht(L)/lt(R)

    XMPYU   fht_0,flt_0,fm            ; m
    FSTD    fm,-24(%sp)               ; store m

    XMPYU   flt_0,flt_0,lt_temp       ; lt
    FSTD    lt_temp,-16(%sp)          ; store lt

    XMPYU   fht_0,fht_0,ht_temp       ; ht
    FSTD    ht_temp,-8(%sp)           ; store ht

    LDD     -24(%sp),m_0              ; load m
    AND     m_0,high_mask,tmp_0       ; m & Mask
    DEPD,Z  m_0,30,31,m_0             ; m << 32+1
    LDD     -16(%sp),lt_0             ; lt

    LDD     -8(%sp),ht_0              ; ht
    EXTRD,U tmp_0,32,33,tmp_0         ; tmp_0 = m&Mask >> 32-1
    ADD     m_0,lt_0,lt_0             ; lt = lt+m
    ADD,L   ht_0,tmp_0,ht_0           ; ht += tmp_0
    ADD,DC  ht_0,%r0,ht_0             ; ht++

    STD     lt_0,0(r_ptr)             ; rp[0] = lt
    STD     ht_0,8(r_ptr)             ; rp[1] = ht

bn_sqr_words_exit
    .EXIT
    LDD     -112(%sp),%r5       ; restore r5  
    LDD     -120(%sp),%r4       ; restore r4  
    BVE     (%rp)
    LDD,MB  -128(%sp),%r3 
	.PROCEND	;in=23,24,25,26,29;out=28;


;----------------------------------------------------------------------------
;
;BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
;
; arg0 = rp 
; arg1 = ap
; arg2 = bp 
; arg3 = n

t  .reg %r22
b  .reg %r21
l  .reg %r20

bn_add_words
	.proc
    .entry
	.callinfo
	.EXPORT	bn_add_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
	.align 64

    CMPIB,>= 0,n,bn_add_words_exit
    COPY    %r0,%ret0           ; return 0 by default

	;
	; If 2 or more numbers do the loop
	;
	CMPIB,= 1,n,bn_add_words_single_top
	NOP

	;
	; This loop is unrolled 2 times (64-byte aligned as well)
	;
bn_add_words_unroll2
	LDD     0(a_ptr),t
	LDD     0(b_ptr),b
	ADD     t,%ret0,t                    ; t = t+c;
	ADD,DC  %r0,%r0,%ret0                ; set c to carry
	ADD     t,b,l                        ; l = t + b[0]
	ADD,DC  %ret0,%r0,%ret0              ; c+= carry
	STD     l,0(r_ptr)

	LDD     8(a_ptr),t
	LDD     8(b_ptr),b
	ADD     t,%ret0,t                     ; t = t+c;
	ADD,DC  %r0,%r0,%ret0                 ; set c to carry
	ADD     t,b,l                         ; l = t + b[0]
	ADD,DC  %ret0,%r0,%ret0               ; c+= carry
	STD     l,8(r_ptr)

	LDO     -2(n),n
	LDO     16(a_ptr),a_ptr
	LDO     16(b_ptr),b_ptr

	CMPIB,<= 2,n,bn_add_words_unroll2
	LDO     16(r_ptr),r_ptr

    CMPIB,=,N 0,n,bn_add_words_exit ; are we done?

bn_add_words_single_top
	LDD     0(a_ptr),t
	LDD     0(b_ptr),b

	ADD     t,%ret0,t                 ; t = t+c;
	ADD,DC  %r0,%r0,%ret0             ; set c to carry (could use CMPCLR??)
	ADD     t,b,l                     ; l = t + b[0]
	ADD,DC  %ret0,%r0,%ret0           ; c+= carry
	STD     l,0(r_ptr)

bn_add_words_exit
    .EXIT
    BVE     (%rp)
	NOP
	.PROCEND	;in=23,24,25,26,29;out=28;

;----------------------------------------------------------------------------
;
;BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
;
; arg0 = rp 
; arg1 = ap
; arg2 = bp 
; arg3 = n

t1       .reg %r22
t2       .reg %r21
sub_tmp1 .reg %r20
sub_tmp2 .reg %r19


bn_sub_words
	.proc
	.callinfo 
	.EXPORT	bn_sub_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
    .entry
	.align 64

    CMPIB,>=  0,n,bn_sub_words_exit
    COPY    %r0,%ret0           ; return 0 by default

	;
	; If 2 or more numbers do the loop
	;
	CMPIB,= 1,n,bn_sub_words_single_top
	NOP

	;
	; This loop is unrolled 2 times (64-byte aligned as well)
	;
bn_sub_words_unroll2
	LDD     0(a_ptr),t1
	LDD     0(b_ptr),t2
	SUB     t1,t2,sub_tmp1           ; t3 = t1-t2; 
	SUB     sub_tmp1,%ret0,sub_tmp1  ; t3 = t3- c; 

	CMPCLR,*>> t1,t2,sub_tmp2        ; clear if t1 > t2
	LDO      1(%r0),sub_tmp2
	
	CMPCLR,*= t1,t2,%r0
	COPY    sub_tmp2,%ret0
	STD     sub_tmp1,0(r_ptr)

	LDD     8(a_ptr),t1
	LDD     8(b_ptr),t2
	SUB     t1,t2,sub_tmp1            ; t3 = t1-t2; 
	SUB     sub_tmp1,%ret0,sub_tmp1   ; t3 = t3- c; 
	CMPCLR,*>> t1,t2,sub_tmp2         ; clear if t1 > t2
	LDO      1(%r0),sub_tmp2
	
	CMPCLR,*= t1,t2,%r0
	COPY    sub_tmp2,%ret0
	STD     sub_tmp1,8(r_ptr)

	LDO     -2(n),n
	LDO     16(a_ptr),a_ptr
	LDO     16(b_ptr),b_ptr

	CMPIB,<= 2,n,bn_sub_words_unroll2
	LDO     16(r_ptr),r_ptr

    CMPIB,=,N 0,n,bn_sub_words_exit ; are we done?

bn_sub_words_single_top
	LDD     0(a_ptr),t1
	LDD     0(b_ptr),t2
	SUB     t1,t2,sub_tmp1            ; t3 = t1-t2; 
	SUB     sub_tmp1,%ret0,sub_tmp1   ; t3 = t3- c; 
	CMPCLR,*>> t1,t2,sub_tmp2         ; clear if t1 > t2
	LDO      1(%r0),sub_tmp2
	
	CMPCLR,*= t1,t2,%r0
	COPY    sub_tmp2,%ret0

	STD     sub_tmp1,0(r_ptr)

bn_sub_words_exit
    .EXIT
    BVE     (%rp)
	NOP
	.PROCEND	;in=23,24,25,26,29;out=28;

;------------------------------------------------------------------------------
;
; unsigned long bn_div_words(unsigned long h, unsigned long l, unsigned long d)
;
; arg0 = h
; arg1 = l
; arg2 = d
;
; This is mainly just modified assembly from the compiler, thus the
; lack of variable names.
;
;------------------------------------------------------------------------------
bn_div_words
	.proc
	.callinfo CALLER,FRAME=272,ENTRY_GR=%r10,SAVE_RP,ARGS_SAVED,ORDERING_AWARE
	.EXPORT	bn_div_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
	.IMPORT	BN_num_bits_word,CODE,NO_RELOCATION
	.IMPORT	__iob,DATA
	.IMPORT	fprintf,CODE,NO_RELOCATION
	.IMPORT	abort,CODE,NO_RELOCATION
	.IMPORT	$$div2U,MILLICODE
    .entry
    STD     %r2,-16(%r30)   
    STD,MA  %r3,352(%r30)   
    STD     %r4,-344(%r30)  
    STD     %r5,-336(%r30)  
    STD     %r6,-328(%r30)  
    STD     %r7,-320(%r30)  
    STD     %r8,-312(%r30)  
    STD     %r9,-304(%r30)  
    STD     %r10,-296(%r30)

    STD     %r27,-288(%r30)             ; save gp

    COPY    %r24,%r3           ; save d 
    COPY    %r26,%r4           ; save h (high 64-bits)
    LDO      -1(%r0),%ret0     ; return -1 by default	

    CMPB,*=  %r0,%arg2,$D3     ; if (d == 0)
    COPY    %r25,%r5           ; save l (low 64-bits)

    LDO     -48(%r30),%r29     ; create ap 
    .CALL   ;in=26,29;out=28;
    B,L     BN_num_bits_word,%r2 
    COPY    %r3,%r26        
    LDD     -288(%r30),%r27    ; restore gp 
    LDI     64,%r21 

    CMPB,=  %r21,%ret0,$00000012   ;if (i == 64) (forward) 
    COPY    %ret0,%r24             ; i   
    MTSARCM %r24    
    DEPDI,Z -1,%sar,1,%r29  
    CMPB,*<<,N %r29,%r4,bn_div_err_case ; if (h > 1<<i) (forward) 

$00000012
    SUBI    64,%r24,%r31                       ; i = 64 - i;
    CMPCLR,*<< %r4,%r3,%r0                     ; if (h >= d)
    SUB     %r4,%r3,%r4                        ; h -= d
    CMPB,=  %r31,%r0,$0000001A                 ; if (i)
    COPY    %r0,%r10                           ; ret = 0
    MTSARCM %r31                               ; i to shift
    DEPD,Z  %r3,%sar,64,%r3                    ; d <<= i;
    SUBI    64,%r31,%r19                       ; 64 - i; redundent
    MTSAR   %r19                               ; (64 -i) to shift
    SHRPD   %r4,%r5,%sar,%r4                   ; l>> (64-i)
    MTSARCM %r31                               ; i to shift
    DEPD,Z  %r5,%sar,64,%r5                    ; l <<= i;

$0000001A
    DEPDI,Z -1,31,32,%r19                      
    EXTRD,U %r3,31,32,%r6                      ; dh=(d&0xfff)>>32
    EXTRD,U %r3,63,32,%r8                      ; dl = d&0xffffff
    LDO     2(%r0),%r9
    STD    %r3,-280(%r30)                      ; "d" to stack

$0000001C
    DEPDI,Z -1,63,32,%r29                      ; 
    EXTRD,U %r4,31,32,%r31                     ; h >> 32
    CMPB,*=,N  %r31,%r6,$D2     	       ; if ((h>>32) != dh)(forward) div
    COPY    %r4,%r26       
    EXTRD,U %r4,31,32,%r25