des_enc.m4

!  des_enc.m4
!  des_enc.S  (generated from des_enc.m4)
!
!  UltraSPARC assembler version of the LibDES/SSLeay/OpenSSL des_enc.c file.
!
!  Version 1.0. 32-bit version.
!
!  June 8, 2000.
!
!  Version 2.0. 32/64-bit, PIC-ification, blended CPU adaptation
!		by Andy Polyakov.
!
!  January 1, 2003.
!
!  Assembler version: Copyright Svend Olaf Mikkelsen.
!
!  Original C code: Copyright Eric A. Young.
!
!  This code can be freely used by LibDES/SSLeay/OpenSSL users.
!
!  The LibDES/SSLeay/OpenSSL copyright notices must be respected.
!
!  This version can be redistributed.
!
!  To expand the m4 macros: m4 -B 8192 des_enc.m4 > des_enc.S
!
!  Global registers 1 to 5 are used. This is the same as done by the
!  cc compiler. The UltraSPARC load/store little endian feature is used.
!
!  Instruction grouping often refers to one CPU cycle.
!
!  Assemble through gcc: gcc -c -mcpu=ultrasparc -o des_enc.o des_enc.S
!
!  Assemble through cc:  cc -c -xarch=v8plusa -o des_enc.o des_enc.S
!
!  Performance improvement according to './apps/openssl speed des'
!
!	32-bit build:
!		23%  faster than cc-5.2 -xarch=v8plus -xO5
!		115% faster than gcc-3.2.1 -m32 -mcpu=ultrasparc -O5
!	64-bit build:
!		50%  faster than cc-5.2 -xarch=v9 -xO5
!		100% faster than gcc-3.2.1 -m64 -mcpu=ultrasparc -O5
!

.ident "des_enc.m4 2.1"
.file  "des_enc-sparc.S"

#if defined(__SUNPRO_C) && defined(__sparcv9)
# define ABI64  /* They've said -xarch=v9 at command line */
#elif defined(__GNUC__) && defined(__arch64__)
# define ABI64  /* They've said -m64 at command line */
#endif

#ifdef ABI64
  .register	%g2,#scratch
  .register	%g3,#scratch
# define	FRAME	-192
# define	BIAS	2047
# define	LDPTR	ldx
# define	STPTR	stx
# define	ARG0	128
# define	ARGSZ	8
# ifndef OPENSSL_SYSNAME_ULTRASPARC
# define OPENSSL_SYSNAME_ULTRASPARC
# endif
#else
# define	FRAME	-96
# define	BIAS	0
# define	LDPTR	ld
# define	STPTR	st
# define	ARG0	68
# define	ARGSZ	4
#endif

#define LOOPS 7

#define global0 %g0
#define global1 %g1
#define global2 %g2
#define global3 %g3
#define global4 %g4
#define global5 %g5

#define local0 %l0
#define local1 %l1
#define local2 %l2
#define local3 %l3
#define local4 %l4
#define local5 %l5
#define local7 %l6
#define local6 %l7

#define in0 %i0
#define in1 %i1
#define in2 %i2
#define in3 %i3
#define in4 %i4
#define in5 %i5
#define in6 %i6
#define in7 %i7

#define out0 %o0
#define out1 %o1
#define out2 %o2
#define out3 %o3
#define out4 %o4
#define out5 %o5
#define out6 %o6
#define out7 %o7

#define stub stb

changequote({,})


! Macro definitions:


! {ip_macro}
!
! The logic used in initial and final permutations is the same as in
! the C code. The permutations are done with a clever shift, xor, and
! technique.
!
! The macro also loads address sbox 1 to 5 to global 1 to 5, address
! sbox 6 to local6, and addres sbox 8 to out3.
!
! Rotates the halfs 3 left to bring the sbox bits in convenient positions.
!
! Loads key first round from address in parameter 5 to out0, out1.
!
! After the the original LibDES initial permutation, the resulting left
! is in the variable initially used for right and vice versa. The macro
! implements the possibility to keep the halfs in the original registers.
!
! parameter 1  left
! parameter 2  right
! parameter 3  result left (modify in first round)
! parameter 4  result right (use in first round)
! parameter 5  key address
! parameter 6  1/2 for include encryption/decryption
! parameter 7  1 for move in1 to in3
! parameter 8  1 for move in3 to in4, 2 for move in4 to in3
! parameter 9  1 for load ks3 and ks2 to in4 and in3

define(ip_macro, {

! {ip_macro}
! $1 $2 $4 $3 $5 $6 $7 $8 $9

	ld	[out2+256], local1
	srl	$2, 4, local4

	xor	local4, $1, local4
	ifelse($7,1,{mov in1, in3},{nop})

	ld	[out2+260], local2
	and	local4, local1, local4
	ifelse($8,1,{mov in3, in4},{})
	ifelse($8,2,{mov in4, in3},{})

	ld	[out2+280], out4          ! loop counter
	sll	local4, 4, local1
	xor	$1, local4, $1

	ld	[out2+264], local3
	srl	$1, 16, local4
	xor	$2, local1, $2

	ifelse($9,1,{LDPTR	KS3, in4},{})
	xor	local4, $2, local4
	nop	!sethi	%hi(DES_SPtrans), global1 ! sbox addr

	ifelse($9,1,{LDPTR	KS2, in3},{})
	and	local4, local2, local4
	nop	!or	global1, %lo(DES_SPtrans), global1   ! sbox addr

	sll	local4, 16, local1
	xor	$2, local4, $2

	srl	$2, 2, local4
	xor	$1, local1, $1

	sethi	%hi(16711680), local5
	xor	local4, $1, local4

	and	local4, local3, local4
	or	local5, 255, local5

	sll	local4, 2, local2
	xor	$1, local4, $1

	srl	$1, 8, local4
	xor	$2, local2, $2

	xor	local4, $2, local4
	add	global1, 768, global4

	and	local4, local5, local4
	add	global1, 1024, global5

	ld	[out2+272], local7
	sll	local4, 8, local1
	xor	$2, local4, $2

	srl	$2, 1, local4
	xor	$1, local1, $1

	ld	[$5], out0                ! key 7531
	xor	local4, $1, local4
	add	global1, 256, global2

	ld	[$5+4], out1              ! key 8642
	and	local4, local7, local4
	add	global1, 512, global3

	sll	local4, 1, local1
	xor	$1, local4, $1

	sll	$1, 3, local3
	xor	$2, local1, $2

	sll	$2, 3, local2
	add	global1, 1280, local6     ! address sbox 8

	srl	$1, 29, local4
	add	global1, 1792, out3       ! address sbox 8

	srl	$2, 29, local1
	or	local4, local3, $4

	or	local2, local1, $3

	ifelse($6, 1, {

		ld	[out2+284], local5     ! 0x0000FC00 used in the rounds
		or	local2, local1, $3
		xor	$4, out0, local1

		call .des_enc.1
		and	local1, 252, local1

	},{})

	ifelse($6, 2, {

		ld	[out2+284], local5     ! 0x0000FC00 used in the rounds
		or	local2, local1, $3
		xor	$4, out0, local1

		call .des_dec.1
		and	local1, 252, local1

	},{})
})


! {rounds_macro}
!
! The logic used in the DES rounds is the same as in the C code,
! except that calculations for sbox 1 and sbox 5 begin before
! the previous round is finished.
!
! In each round one half (work) is modified based on key and the
! other half (use).
!
! In this version we do two rounds in a loop repeated 7 times
! and two rounds seperately.
!
! One half has the bits for the sboxes in the following positions:
!
!	777777xx555555xx333333xx111111xx
!
!	88xx666666xx444444xx222222xx8888
!
! The bits for each sbox are xor-ed with the key bits for that box.
! The above xx bits are cleared, and the result used for lookup in
! the sbox table. Each sbox entry contains the 4 output bits permuted
! into 32 bits according to the P permutation.
!
! In the description of DES, left and right are switched after
! each round, except after last round. In this code the original
! left and right are kept in the same register in all rounds, meaning
! that after the 16 rounds the result for right is in the register
! originally used for left.
!
! parameter 1  first work (left in first round)
! parameter 2  first use (right in first round)
! parameter 3  enc/dec  1/-1
! parameter 4  loop label
! parameter 5  key address register
! parameter 6  optional address for key next encryption/decryption
! parameter 7  not empty for include retl
!
! also compares in2 to 8

define(rounds_macro, {

! {rounds_macro}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	xor	$2, out0, local1

	ld	[out2+284], local5        ! 0x0000FC00
	ba	$4
	and	local1, 252, local1

	.align 32

$4:
	! local6 is address sbox 6
	! out3   is address sbox 8
	! out4   is loop counter

	ld	[global1+local1], local1
	xor	$2, out1, out1            ! 8642
	xor	$2, out0, out0            ! 7531
	! fmovs	%f0, %f0                  ! fxor used for alignment

	srl	out1, 4, local0           ! rotate 4 right
	and	out0, local5, local3      ! 3
	! fmovs	%f0, %f0

	ld	[$5+$3*8], local7         ! key 7531 next round
	srl	local3, 8, local3         ! 3
	and	local0, 252, local2       ! 2
	! fmovs	%f0, %f0

	ld	[global3+local3],local3   ! 3
	sll	out1, 28, out1            ! rotate
	xor	$1, local1, $1            ! 1 finished, local1 now sbox 7

	ld	[global2+local2], local2  ! 2 
	srl	out0, 24, local1          ! 7
	or	out1, local0, out1        ! rotate

	ldub	[out2+local1], local1     ! 7 (and 0xFC)
	srl	out1, 24, local0          ! 8
	and	out1, local5, local4      ! 4

	ldub	[out2+local0], local0     ! 8 (and 0xFC)
	srl	local4, 8, local4         ! 4
	xor	$1, local2, $1            ! 2 finished local2 now sbox 6

	ld	[global4+local4],local4   ! 4
	srl	out1, 16, local2          ! 6
	xor	$1, local3, $1            ! 3 finished local3 now sbox 5

	ld	[out3+local0],local0      ! 8
	and	local2, 252, local2       ! 6
	add	global1, 1536, local5     ! address sbox 7

	ld	[local6+local2], local2   ! 6
	srl	out0, 16, local3          ! 5
	xor	$1, local4, $1            ! 4 finished

	ld	[local5+local1],local1    ! 7
	and	local3, 252, local3       ! 5
	xor	$1, local0, $1            ! 8 finished

	ld	[global5+local3],local3   ! 5
	xor	$1, local2, $1            ! 6 finished
	subcc	out4, 1, out4

	ld	[$5+$3*8+4], out0         ! key 8642 next round
	xor	$1, local7, local2        ! sbox 5 next round
	xor	$1, local1, $1            ! 7 finished

	srl	local2, 16, local2        ! sbox 5 next round
	xor	$1, local3, $1            ! 5 finished

	ld	[$5+$3*16+4], out1        ! key 8642 next round again
	and	local2, 252, local2       ! sbox5 next round
! next round
	xor	$1, local7, local7        ! 7531

	ld	[global5+local2], local2  ! 5
	srl	local7, 24, local3        ! 7
	xor	$1, out0, out0            ! 8642

	ldub	[out2+local3], local3     ! 7 (and 0xFC)
	srl	out0, 4, local0           ! rotate 4 right
	and	local7, 252, local1       ! 1

	sll	out0, 28, out0            ! rotate
	xor	$2, local2, $2            ! 5 finished local2 used

	srl	local0, 8, local4         ! 4
	and	local0, 252, local2       ! 2
	ld	[local5+local3], local3   ! 7

	srl	local0, 16, local5        ! 6
	or	out0, local0, out0        ! rotate
	ld	[global2+local2], local2  ! 2

	srl	out0, 24, local0
	ld	[$5+$3*16], out0          ! key 7531 next round
	and	local4, 252, local4	  ! 4

	and	local5, 252, local5       ! 6
	ld	[global4+local4], local4  ! 4
	xor	$2, local3, $2            ! 7 finished local3 used

	and	local0, 252, local0       ! 8
	ld	[local6+local5], local5   ! 6
	xor	$2, local2, $2            ! 2 finished local2 now sbox 3

	srl	local7, 8, local2         ! 3 start
	ld	[out3+local0], local0     ! 8
	xor	$2, local4, $2            ! 4 finished

	and	local2, 252, local2       ! 3
	ld	[global1+local1], local1  ! 1
	xor	$2, local5, $2            ! 6 finished local5 used

	ld	[global3+local2], local2  ! 3
	xor	$2, local0, $2            ! 8 finished
	add	$5, $3*16, $5             ! enc add 8, dec add -8 to key pointer

	ld	[out2+284], local5        ! 0x0000FC00
	xor	$2, out0, local4          ! sbox 1 next round
	xor	$2, local1, $2            ! 1 finished

	xor	$2, local2, $2            ! 3 finished
#ifdef OPENSSL_SYSNAME_ULTRASPARC
	bne,pt	%icc, $4
#else
	bne	$4
#endif
	and	local4, 252, local1       ! sbox 1 next round

! two rounds more:

	ld	[global1+local1], local1
	xor	$2, out1, out1
	xor	$2, out0, out0

	srl	out1, 4, local0           ! rotate
	and	out0, local5, local3

	ld	[$5+$3*8], local7         ! key 7531
	srl	local3, 8, local3
	and	local0, 252, local2

	ld	[global3+local3],local3
	sll	out1, 28, out1            ! rotate
	xor	$1, local1, $1            ! 1 finished, local1 now sbox 7

	ld	[global2+local2], local2
	srl	out0, 24, local1
	or	out1, local0, out1        ! rotate

	ldub	[out2+local1], local1
	srl	out1, 24, local0
	and	out1, local5, local4

	ldub	[out2+local0], local0
	srl	local4, 8, local4
	xor	$1, local2, $1            ! 2 finished local2 now sbox 6

	ld	[global4+local4],local4
	srl	out1, 16, local2
	xor	$1, local3, $1            ! 3 finished local3 now sbox 5

	ld	[out3+local0],local0
	and	local2, 252, local2
	add	global1, 1536, local5     ! address sbox 7

	ld	[local6+local2], local2
	srl	out0, 16, local3
	xor	$1, local4, $1            ! 4 finished

	ld	[local5+local1],local1
	and	local3, 252, local3
	xor	$1, local0, $1

	ld	[global5+local3],local3
	xor	$1, local2, $1            ! 6 finished
	cmp	in2, 8

	ifelse($6,{}, {}, {ld	[out2+280], out4})  ! loop counter
	xor	$1, local7, local2        ! sbox 5 next round
	xor	$1, local1, $1            ! 7 finished

	ld	[$5+$3*8+4], out0
	srl	local2, 16, local2        ! sbox 5 next round
	xor	$1, local3, $1            ! 5 finished

	and	local2, 252, local2
! next round (two rounds more)
	xor	$1, local7, local7        ! 7531

	ld	[global5+local2], local2
	srl	local7, 24, local3
	xor	$1, out0, out0            ! 8642

	ldub	[out2+local3], local3
	srl	out0, 4, local0           ! rotate
	and	local7, 252, local1

	sll	out0, 28, out0            ! rotate
	xor	$2, local2, $2            ! 5 finished local2 used

	srl	local0, 8, local4
	and	local0, 252, local2
	ld	[local5+local3], local3

	srl	local0, 16, local5
	or	out0, local0, out0        ! rotate
	ld	[global2+local2], local2

	srl	out0, 24, local0
	ifelse($6,{}, {}, {ld	[$6], out0})   ! key next encryption/decryption
	and	local4, 252, local4

	and	local5, 252, local5
	ld	[global4+local4], local4
	xor	$2, local3, $2            ! 7 finished local3 used

	and	local0, 252, local0
	ld	[local6+local5], local5
	xor	$2, local2, $2            ! 2 finished local2 now sbox 3

	srl	local7, 8, local2         ! 3 start
	ld	[out3+local0], local0
	xor	$2, local4, $2

	and	local2, 252, local2
	ld	[global1+local1], local1
	xor	$2, local5, $2            ! 6 finished local5 used

	ld	[global3+local2], local2
	srl	$1, 3, local3
	xor	$2, local0, $2

	ifelse($6,{}, {}, {ld	[$6+4], out1}) ! key next encryption/decryption
	sll	$1, 29, local4
	xor	$2, local1, $2

	ifelse($7,{}, {}, {retl})
	xor	$2, local2, $2
})


! {fp_macro}
!
!  parameter 1   right (original left)
!  parameter 2   left (original right)
!  parameter 3   1 for optional store to [in0]
!  parameter 4   1 for load input/output address to local5/7
!
!  The final permutation logic switches the halfes, meaning that
!  left and right ends up the the registers originally used.

define(fp_macro, {

! {fp_macro}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	! initially undo the rotate 3 left done after initial permutation
	! original left is received shifted 3 right and 29 left in local3/4

	sll	$2, 29, local1
	or	local3, local4, $1

	srl	$2, 3, $2
	sethi	%hi(0x55555555), local2

	or	$2, local1, $2
	or	local2, %lo(0x55555555), local2

	srl	$2, 1, local3
	sethi	%hi(0x00ff00ff), local1
	xor	local3, $1, local3
	or	local1, %lo(0x00ff00ff), local1
	and	local3, local2, local3
	sethi	%hi(0x33333333), local4
	sll	local3, 1, local2

	xor	$1, local3, $1

	srl	$1, 8, local3
	xor	$2, local2, $2
	xor	local3, $2, local3
	or	local4, %lo(0x33333333), local4
	and	local3, local1, local3
	sethi	%hi(0x0000ffff), local1
	sll	local3, 8, local2

	xor	$2, local3, $2

	srl	$2, 2, local3
	xor	$1, local2, $1
	xor	local3, $1, local3
	or	local1, %lo(0x0000ffff), local1
	and	local3, local4, local3
	sethi	%hi(0x0f0f0f0f), local4
	sll	local3, 2, local2

	ifelse($4,1, {LDPTR INPUT, local5})
	xor	$1, local3, $1

	ifelse($4,1, {LDPTR OUTPUT, local7})
	srl	$1, 16, local3
	xor	$2, local2, $2
	xor	local3, $2, local3
	or	local4, %lo(0x0f0f0f0f), local4
	and	local3, local1, local3
	sll	local3, 16, local2

	xor	$2, local3, local1

	srl	local1, 4, local3
	xor	$1, local2, $1
	xor	local3, $1, local3
	and	local3, local4, local3
	sll	local3, 4, local2

	xor	$1, local3, $1

	! optional store:

	ifelse($3,1, {st $1, [in0]})

	xor	local1, local2, $2

	ifelse($3,1, {st $2, [in0+4]})

})


! {fp_ip_macro}
!
! Does initial permutation for next block mixed with
! final permutation for current block.
!
! parameter 1   original left
! parameter 2   original right
! parameter 3   left ip
! parameter 4   right ip
! parameter 5   1: load ks1/ks2 to in3/in4, add 120 to in4
!                2: mov in4 to in3
!
! also adds -8 to length in2 and loads loop counter to out4

define(fp_ip_macro, {

! {fp_ip_macro}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	define({temp1},{out4})
	define({temp2},{local3})

	define({ip1},{local1})
	define({ip2},{local2})
	define({ip4},{local4})
	define({ip5},{local5})

	! $1 in local3, local4

	ld	[out2+256], ip1
	sll	out5, 29, temp1
	or	local3, local4, $1

	srl	out5, 3, $2
	ifelse($5,2,{mov in4, in3})

	ld	[out2+272], ip5
	srl	$4, 4, local0
	or	$2, temp1, $2

	srl	$2, 1, temp1
	xor	temp1, $1, temp1

	and	temp1, ip5, temp1
	xor	local0, $3, local0

	sll	temp1, 1, temp2
	xor	$1, temp1, $1

	and	local0, ip1, local0
	add	in2, -8, in2

	sll	local0, 4, local7
	xor	$3, local0, $3

	ld	[out2+268], ip4
	srl	$1, 8, temp1
	xor	$2, temp2, $2
	ld	[out2+260], ip2
	srl	$3, 16, local0
	xor	$4, local7, $4
	xor	temp1, $2, temp1
	xor	local0, $4, local0
	and	temp1, ip4, temp1
	and	local0, ip2, local0
	sll	temp1, 8, temp2
	xor	$2, temp1, $2
	sll	local0, 16, local7
	xor	$4, local0, $4

	srl	$2, 2, temp1
	xor	$1, temp2, $1

	ld	[out2+264], temp2         ! ip3
	srl	$4, 2, local0
	xor	$3, local7, $3
	xor	temp1, $1, temp1
	xor	local0, $3, local0
	and	temp1, temp2, temp1
	and	local0, temp2, local0
	sll	temp1, 2, temp2
	xor	$1, temp1, $1
	sll	local0, 2, local7
	xor	$3, local0, $3

	srl	$1, 16, temp1
	xor	$2, temp2, $2
	srl	$3, 8, local0
	xor	$4, local7, $4
	xor	temp1, $2, temp1
	xor	local0, $4, local0
	and	temp1, ip2, temp1
	and	local0, ip4, local0
	sll	temp1, 16, temp2
	xor	$2, temp1, local4
	sll	local0, 8, local7
	xor	$4, local0, $4

	srl	$4, 1, local0
	xor	$3, local7, $3

	srl	local4, 4, temp1
	xor	local0, $3, local0

	xor	$1, temp2, $1
	and	local0, ip5, local0

	sll	local0, 1, local7
	xor	temp1, $1, temp1

	xor	$3, local0, $3
	xor	$4, local7, $4

	sll	$3, 3, local5
	and	temp1, ip1, temp1

	sll	temp1, 4, temp2
	xor	$1, temp1, $1

	ifelse($5,1,{LDPTR	KS2, in4})
	sll	$4, 3, local2
	xor	local4, temp2, $2

	! reload since used as temporar:

	ld	[out2+280], out4          ! loop counter

	srl	$3, 29, local0
	ifelse($5,1,{add in4, 120, in4})

	ifelse($5,1,{LDPTR	KS1, in3})
	srl	$4, 29, local7

	or	local0, local5, $4
	or	local2, local7, $3

})



! {load_little_endian}
!
! parameter 1  address
! parameter 2  destination left
! parameter 3  destination right
! parameter 4  temporar
! parameter 5  label

define(load_little_endian, {

! {load_little_endian}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	! first in memory to rightmost in register

#ifdef OPENSSL_SYSNAME_ULTRASPARC
	andcc	$1, 3, global0
	bne,pn	%icc, $5
	nop

	lda	[$1] 0x88, $2
	add	$1, 4, $4

	ba,pt	%icc, $5a
	lda	[$4] 0x88, $3
#endif

$5:
	ldub	[$1+3], $2

	ldub	[$1+2], $4
	sll	$2, 8, $2
	or	$2, $4, $2

	ldub	[$1+1], $4
	sll	$2, 8, $2
	or	$2, $4, $2

	ldub	[$1+0], $4
	sll	$2, 8, $2
	or	$2, $4, $2


	ldub	[$1+3+4], $3

	ldub	[$1+2+4], $4
	sll	$3, 8, $3
	or	$3, $4, $3

	ldub	[$1+1+4], $4
	sll	$3, 8, $3
	or	$3, $4, $3

	ldub	[$1+0+4], $4
	sll	$3, 8, $3
	or	$3, $4, $3
$5a:

})


! {load_little_endian_inc}
!
! parameter 1  address
! parameter 2  destination left
! parameter 3  destination right
! parameter 4  temporar
! parameter 4  label
!
! adds 8 to address

define(load_little_endian_inc, {

! {load_little_endian_inc}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	! first in memory to rightmost in register

#ifdef OPENSSL_SYSNAME_ULTRASPARC
	andcc	$1, 3, global0
	bne,pn	%icc, $5
	nop

	lda	[$1] 0x88, $2
	add	$1, 4, $1

	lda	[$1] 0x88, $3
	ba,pt	%icc, $5a
	add	$1, 4, $1
#endif

$5:
	ldub	[$1+3], $2

	ldub	[$1+2], $4
	sll	$2, 8, $2
	or	$2, $4, $2

	ldub	[$1+1], $4
	sll	$2, 8, $2
	or	$2, $4, $2

	ldub	[$1+0], $4
	sll	$2, 8, $2
	or	$2, $4, $2

	ldub	[$1+3+4], $3
	add	$1, 8, $1

	ldub	[$1+2+4-8], $4
	sll	$3, 8, $3
	or	$3, $4, $3

	ldub	[$1+1+4-8], $4
	sll	$3, 8, $3
	or	$3, $4, $3

	ldub	[$1+0+4-8], $4
	sll	$3, 8, $3
	or	$3, $4, $3
$5a:

})


! {load_n_bytes}
!
! Loads 1 to 7 bytes little endian
! Remaining bytes are zeroed.
!
! parameter 1  address
! parameter 2  length
! parameter 3  destination register left
! parameter 4  destination register right
! parameter 5  temp
! parameter 6  temp2
! parameter 7  label
! parameter 8  return label

define(load_n_bytes, {

! {load_n_bytes}
! $1 $2 $5 $6 $7 $8 $7 $8 $9

$7.0:	call	.+8
	sll	$2, 2, $6

	add	%o7,$7.jmp.table-$7.0,$5

	add	$5, $6, $5
	mov	0, $4

	ld	[$5], $5

	jmp	%o7+$5
	mov	0, $3

$7.7:
	ldub	[$1+6], $5
	sll	$5, 16, $5
	or	$3, $5, $3
$7.6:
	ldub	[$1+5], $5
	sll	$5, 8, $5
	or	$3, $5, $3
$7.5:
	ldub	[$1+4], $5
	or	$3, $5, $3
$7.4:
	ldub	[$1+3], $5
	sll	$5, 24, $5
	or	$4, $5, $4
$7.3:
	ldub	[$1+2], $5
	sll	$5, 16, $5
	or	$4, $5, $4
$7.2:
	ldub	[$1+1], $5
	sll	$5, 8, $5
	or	$4, $5, $4
$7.1:
	ldub	[$1+0], $5
	ba	$8
	or	$4, $5, $4

	.align 4

$7.jmp.table:
	.word	0
	.word	$7.1-$7.0
	.word	$7.2-$7.0
	.word	$7.3-$7.0
	.word	$7.4-$7.0
	.word	$7.5-$7.0
	.word	$7.6-$7.0
	.word	$7.7-$7.0
})


! {store_little_endian}
!
! parameter 1  address
! parameter 2  source left
! parameter 3  source right
! parameter 4  temporar

define(store_little_endian, {

! {store_little_endian}
! $1 $2 $3 $4 $5 $6 $7 $8 $9

	! rightmost in register to first in memory

#ifdef OPENSSL_SYSNAME_ULTRASPARC
	andcc	$1, 3, global0
	bne,pn	%icc, $5
	nop

	sta	$2, [$1] 0x88
	add	$1, 4, $4

	ba,pt	%icc, $5a
	sta	$3, [$4] 0x88
#endif

$5:
	and	$2, 255, $4
	stub	$4, [$1+0]