/* This is an assembly language implementation of mulsi3, divsi3, and modsi3
|
/* This is an assembly language implementation of mulsi3, divsi3, and modsi3
|
for the sparc processor.
|
for the sparc processor.
|
|
|
These routines are derived from the SPARC Architecture Manual, version 8,
|
These routines are derived from the SPARC Architecture Manual, version 8,
|
slightly edited to match the desired calling convention, and also to
|
slightly edited to match the desired calling convention, and also to
|
optimize them for our purposes. */
|
optimize them for our purposes. */
|
|
|
#ifdef L_mulsi3
|
#ifdef L_mulsi3
|
.text
|
.text
|
.align 4
|
.align 4
|
.global .umul
|
.global .umul
|
.proc 4
|
.proc 4
|
.umul:
|
.umul:
|
or %o0, %o1, %o4 ! logical or of multiplier and multiplicand
|
or %o0, %o1, %o4 ! logical or of multiplier and multiplicand
|
mov %o0, %y ! multiplier to Y register
|
mov %o0, %y ! multiplier to Y register
|
andncc %o4, 0xfff, %o5 ! mask out lower 12 bits
|
andncc %o4, 0xfff, %o5 ! mask out lower 12 bits
|
be mul_shortway ! can do it the short way
|
be mul_shortway ! can do it the short way
|
andcc %g0, %g0, %o4 ! zero the partial product and clear NV cc
|
andcc %g0, %g0, %o4 ! zero the partial product and clear NV cc
|
!
|
!
|
! long multiply
|
! long multiply
|
!
|
!
|
mulscc %o4, %o1, %o4 ! first iteration of 33
|
mulscc %o4, %o1, %o4 ! first iteration of 33
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4 ! 32nd iteration
|
mulscc %o4, %o1, %o4 ! 32nd iteration
|
mulscc %o4, %g0, %o4 ! last iteration only shifts
|
mulscc %o4, %g0, %o4 ! last iteration only shifts
|
! the upper 32 bits of product are wrong, but we do not care
|
! the upper 32 bits of product are wrong, but we do not care
|
retl
|
retl
|
rd %y, %o0
|
rd %y, %o0
|
!
|
!
|
! short multiply
|
! short multiply
|
!
|
!
|
mul_shortway:
|
mul_shortway:
|
mulscc %o4, %o1, %o4 ! first iteration of 13
|
mulscc %o4, %o1, %o4 ! first iteration of 13
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4
|
mulscc %o4, %o1, %o4 ! 12th iteration
|
mulscc %o4, %o1, %o4 ! 12th iteration
|
mulscc %o4, %g0, %o4 ! last iteration only shifts
|
mulscc %o4, %g0, %o4 ! last iteration only shifts
|
rd %y, %o5
|
rd %y, %o5
|
sll %o4, 12, %o4 ! left shift partial product by 12 bits
|
sll %o4, 12, %o4 ! left shift partial product by 12 bits
|
srl %o5, 20, %o5 ! right shift partial product by 20 bits
|
srl %o5, 20, %o5 ! right shift partial product by 20 bits
|
retl
|
retl
|
or %o5, %o4, %o0 ! merge for true product
|
or %o5, %o4, %o0 ! merge for true product
|
#endif
|
#endif
|
|
|
#ifdef L_divsi3
|
#ifdef L_divsi3
|
/*
|
/*
|
* Division and remainder, from Appendix E of the SPARC Version 8
|
* Division and remainder, from Appendix E of the SPARC Version 8
|
* Architecture Manual, with fixes from Gordon Irlam.
|
* Architecture Manual, with fixes from Gordon Irlam.
|
*/
|
*/
|
|
|
/*
|
/*
|
* Input: dividend and divisor in %o0 and %o1 respectively.
|
* Input: dividend and divisor in %o0 and %o1 respectively.
|
*
|
*
|
* m4 parameters:
|
* m4 parameters:
|
* .div name of function to generate
|
* .div name of function to generate
|
* div div=div => %o0 / %o1; div=rem => %o0 % %o1
|
* div div=div => %o0 / %o1; div=rem => %o0 % %o1
|
* true true=true => signed; true=false => unsigned
|
* true true=true => signed; true=false => unsigned
|
*
|
*
|
* Algorithm parameters:
|
* Algorithm parameters:
|
* N how many bits per iteration we try to get (4)
|
* N how many bits per iteration we try to get (4)
|
* WORDSIZE total number of bits (32)
|
* WORDSIZE total number of bits (32)
|
*
|
*
|
* Derived constants:
|
* Derived constants:
|
* TOPBITS number of bits in the top decade of a number
|
* TOPBITS number of bits in the top decade of a number
|
*
|
*
|
* Important variables:
|
* Important variables:
|
* Q the partial quotient under development (initially 0)
|
* Q the partial quotient under development (initially 0)
|
* R the remainder so far, initially the dividend
|
* R the remainder so far, initially the dividend
|
* ITER number of main division loop iterations required;
|
* ITER number of main division loop iterations required;
|
* equal to ceil(log2(quotient) / N). Note that this
|
* equal to ceil(log2(quotient) / N). Note that this
|
* is the log base (2^N) of the quotient.
|
* is the log base (2^N) of the quotient.
|
* V the current comparand, initially divisor*2^(ITER*N-1)
|
* V the current comparand, initially divisor*2^(ITER*N-1)
|
*
|
*
|
* Cost:
|
* Cost:
|
* Current estimate for non-large dividend is
|
* Current estimate for non-large dividend is
|
* ceil(log2(quotient) / N) * (10 + 7N/2) + C
|
* ceil(log2(quotient) / N) * (10 + 7N/2) + C
|
* A large dividend is one greater than 2^(31-TOPBITS) and takes a
|
* A large dividend is one greater than 2^(31-TOPBITS) and takes a
|
* different path, as the upper bits of the quotient must be developed
|
* different path, as the upper bits of the quotient must be developed
|
* one bit at a time.
|
* one bit at a time.
|
*/
|
*/
|
.global .udiv
|
.global .udiv
|
.align 4
|
.align 4
|
.proc 4
|
.proc 4
|
.text
|
.text
|
.udiv:
|
.udiv:
|
b ready_to_divide
|
b ready_to_divide
|
mov 0, %g3 ! result is always positive
|
mov 0, %g3 ! result is always positive
|
|
|
.global .div
|
.global .div
|
.align 4
|
.align 4
|
.proc 4
|
.proc 4
|
.text
|
.text
|
.div:
|
.div:
|
! compute sign of result; if neither is negative, no problem
|
! compute sign of result; if neither is negative, no problem
|
orcc %o1, %o0, %g0 ! either negative?
|
orcc %o1, %o0, %g0 ! either negative?
|
bge ready_to_divide ! no, go do the divide
|
bge ready_to_divide ! no, go do the divide
|
xor %o1, %o0, %g3 ! compute sign in any case
|
xor %o1, %o0, %g3 ! compute sign in any case
|
tst %o1
|
tst %o1
|
bge 1f
|
bge 1f
|
tst %o0
|
tst %o0
|
! %o1 is definitely negative; %o0 might also be negative
|
! %o1 is definitely negative; %o0 might also be negative
|
bge ready_to_divide ! if %o0 not negative...
|
bge ready_to_divide ! if %o0 not negative...
|
sub %g0, %o1, %o1 ! in any case, make %o1 nonneg
|
sub %g0, %o1, %o1 ! in any case, make %o1 nonneg
|
1: ! %o0 is negative, %o1 is nonnegative
|
1: ! %o0 is negative, %o1 is nonnegative
|
sub %g0, %o0, %o0 ! make %o0 nonnegative
|
sub %g0, %o0, %o0 ! make %o0 nonnegative
|
|
|
|
|
ready_to_divide:
|
ready_to_divide:
|
|
|
! Ready to divide. Compute size of quotient; scale comparand.
|
! Ready to divide. Compute size of quotient; scale comparand.
|
orcc %o1, %g0, %o5
|
orcc %o1, %g0, %o5
|
bne 1f
|
bne 1f
|
mov %o0, %o3
|
mov %o0, %o3
|
|
|
! Divide by zero trap. If it returns, return 0 (about as
|
! Divide by zero trap. If it returns, return 0 (about as
|
! wrong as possible, but that is what SunOS does...).
|
! wrong as possible, but that is what SunOS does...).
|
ta 0x2 ! ST_DIV0
|
ta 0x2 ! ST_DIV0
|
retl
|
retl
|
clr %o0
|
clr %o0
|
|
|
1:
|
1:
|
cmp %o3, %o5 ! if %o1 exceeds %o0, done
|
cmp %o3, %o5 ! if %o1 exceeds %o0, done
|
blu got_result ! (and algorithm fails otherwise)
|
blu got_result ! (and algorithm fails otherwise)
|
clr %o2
|
clr %o2
|
sethi %hi(1 << (32 - 4 - 1)), %g1
|
sethi %hi(1 << (32 - 4 - 1)), %g1
|
cmp %o3, %g1
|
cmp %o3, %g1
|
blu not_really_big
|
blu not_really_big
|
clr %o4
|
clr %o4
|
|
|
! Here the dividend is >= 2**(31-N) or so. We must be careful here,
|
! Here the dividend is >= 2**(31-N) or so. We must be careful here,
|
! as our usual N-at-a-shot divide step will cause overflow and havoc.
|
! as our usual N-at-a-shot divide step will cause overflow and havoc.
|
! The number of bits in the result here is N*ITER+SC, where SC <= N.
|
! The number of bits in the result here is N*ITER+SC, where SC <= N.
|
! Compute ITER in an unorthodox manner: know we need to shift V into
|
! Compute ITER in an unorthodox manner: know we need to shift V into
|
! the top decade: so do not even bother to compare to R.
|
! the top decade: so do not even bother to compare to R.
|
1:
|
1:
|
cmp %o5, %g1
|
cmp %o5, %g1
|
bgeu 3f
|
bgeu 3f
|
mov 1, %g2
|
mov 1, %g2
|
sll %o5, 4, %o5
|
sll %o5, 4, %o5
|
b 1b
|
b 1b
|
add %o4, 1, %o4
|
add %o4, 1, %o4
|
|
|
! Now compute %g2.
|
! Now compute %g2.
|
2: addcc %o5, %o5, %o5
|
2: addcc %o5, %o5, %o5
|
bcc not_too_big
|
bcc not_too_big
|
add %g2, 1, %g2
|
add %g2, 1, %g2
|
|
|
! We get here if the %o1 overflowed while shifting.
|
! We get here if the %o1 overflowed while shifting.
|
! This means that %o3 has the high-order bit set.
|
! This means that %o3 has the high-order bit set.
|
! Restore %o5 and subtract from %o3.
|
! Restore %o5 and subtract from %o3.
|
sll %g1, 4, %g1 ! high order bit
|
sll %g1, 4, %g1 ! high order bit
|
srl %o5, 1, %o5 ! rest of %o5
|
srl %o5, 1, %o5 ! rest of %o5
|
add %o5, %g1, %o5
|
add %o5, %g1, %o5
|
b do_single_div
|
b do_single_div
|
sub %g2, 1, %g2
|
sub %g2, 1, %g2
|
|
|
not_too_big:
|
not_too_big:
|
3: cmp %o5, %o3
|
3: cmp %o5, %o3
|
blu 2b
|
blu 2b
|
nop
|
nop
|
be do_single_div
|
be do_single_div
|
nop
|
nop
|
/* NB: these are commented out in the V8-SPARC manual as well */
|
/* NB: these are commented out in the V8-SPARC manual as well */
|
/* (I do not understand this) */
|
/* (I do not understand this) */
|
! %o5 > %o3: went too far: back up 1 step
|
! %o5 > %o3: went too far: back up 1 step
|
! srl %o5, 1, %o5
|
! srl %o5, 1, %o5
|
! dec %g2
|
! dec %g2
|
! do single-bit divide steps
|
! do single-bit divide steps
|
!
|
!
|
! We have to be careful here. We know that %o3 >= %o5, so we can do the
|
! We have to be careful here. We know that %o3 >= %o5, so we can do the
|
! first divide step without thinking. BUT, the others are conditional,
|
! first divide step without thinking. BUT, the others are conditional,
|
! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-
|
! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-
|
! order bit set in the first step, just falling into the regular
|
! order bit set in the first step, just falling into the regular
|
! division loop will mess up the first time around.
|
! division loop will mess up the first time around.
|
! So we unroll slightly...
|
! So we unroll slightly...
|
do_single_div:
|
do_single_div:
|
subcc %g2, 1, %g2
|
subcc %g2, 1, %g2
|
bl end_regular_divide
|
bl end_regular_divide
|
nop
|
nop
|
sub %o3, %o5, %o3
|
sub %o3, %o5, %o3
|
mov 1, %o2
|
mov 1, %o2
|
b end_single_divloop
|
b end_single_divloop
|
nop
|
nop
|
single_divloop:
|
single_divloop:
|
sll %o2, 1, %o2
|
sll %o2, 1, %o2
|
bl 1f
|
bl 1f
|
srl %o5, 1, %o5
|
srl %o5, 1, %o5
|
! %o3 >= 0
|
! %o3 >= 0
|
sub %o3, %o5, %o3
|
sub %o3, %o5, %o3
|
b 2f
|
b 2f
|
add %o2, 1, %o2
|
add %o2, 1, %o2
|
1: ! %o3 < 0
|
1: ! %o3 < 0
|
add %o3, %o5, %o3
|
add %o3, %o5, %o3
|
sub %o2, 1, %o2
|
sub %o2, 1, %o2
|
2:
|
2:
|
end_single_divloop:
|
end_single_divloop:
|
subcc %g2, 1, %g2
|
subcc %g2, 1, %g2
|
bge single_divloop
|
bge single_divloop
|
tst %o3
|
tst %o3
|
b,a end_regular_divide
|
b,a end_regular_divide
|
|
|
not_really_big:
|
not_really_big:
|
1:
|
1:
|
sll %o5, 4, %o5
|
sll %o5, 4, %o5
|
cmp %o5, %o3
|
cmp %o5, %o3
|
bleu 1b
|
bleu 1b
|
addcc %o4, 1, %o4
|
addcc %o4, 1, %o4
|
be got_result
|
be got_result
|
sub %o4, 1, %o4
|
sub %o4, 1, %o4
|
|
|
tst %o3 ! set up for initial iteration
|
tst %o3 ! set up for initial iteration
|
divloop:
|
divloop:
|
sll %o2, 4, %o2
|
sll %o2, 4, %o2
|
! depth 1, accumulated bits 0
|
! depth 1, accumulated bits 0
|
bl L1.16
|
bl L1.16
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 2, accumulated bits 1
|
! depth 2, accumulated bits 1
|
bl L2.17
|
bl L2.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 3, accumulated bits 3
|
! depth 3, accumulated bits 3
|
bl L3.19
|
bl L3.19
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits 7
|
! depth 4, accumulated bits 7
|
bl L4.23
|
bl L4.23
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (7*2+1), %o2
|
add %o2, (7*2+1), %o2
|
|
|
L4.23:
|
L4.23:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (7*2-1), %o2
|
add %o2, (7*2-1), %o2
|
|
|
|
|
L3.19:
|
L3.19:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits 5
|
! depth 4, accumulated bits 5
|
bl L4.21
|
bl L4.21
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (5*2+1), %o2
|
add %o2, (5*2+1), %o2
|
|
|
L4.21:
|
L4.21:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (5*2-1), %o2
|
add %o2, (5*2-1), %o2
|
|
|
L2.17:
|
L2.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 3, accumulated bits 1
|
! depth 3, accumulated bits 1
|
bl L3.17
|
bl L3.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits 3
|
! depth 4, accumulated bits 3
|
bl L4.19
|
bl L4.19
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (3*2+1), %o2
|
add %o2, (3*2+1), %o2
|
|
|
L4.19:
|
L4.19:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (3*2-1), %o2
|
add %o2, (3*2-1), %o2
|
|
|
L3.17:
|
L3.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits 1
|
! depth 4, accumulated bits 1
|
bl L4.17
|
bl L4.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (1*2+1), %o2
|
add %o2, (1*2+1), %o2
|
|
|
L4.17:
|
L4.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (1*2-1), %o2
|
add %o2, (1*2-1), %o2
|
|
|
L1.16:
|
L1.16:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 2, accumulated bits -1
|
! depth 2, accumulated bits -1
|
bl L2.15
|
bl L2.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 3, accumulated bits -1
|
! depth 3, accumulated bits -1
|
bl L3.15
|
bl L3.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits -1
|
! depth 4, accumulated bits -1
|
bl L4.15
|
bl L4.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-1*2+1), %o2
|
add %o2, (-1*2+1), %o2
|
|
|
L4.15:
|
L4.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-1*2-1), %o2
|
add %o2, (-1*2-1), %o2
|
|
|
L3.15:
|
L3.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits -3
|
! depth 4, accumulated bits -3
|
bl L4.13
|
bl L4.13
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-3*2+1), %o2
|
add %o2, (-3*2+1), %o2
|
|
|
L4.13:
|
L4.13:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-3*2-1), %o2
|
add %o2, (-3*2-1), %o2
|
|
|
L2.15:
|
L2.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 3, accumulated bits -3
|
! depth 3, accumulated bits -3
|
bl L3.13
|
bl L3.13
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits -5
|
! depth 4, accumulated bits -5
|
bl L4.11
|
bl L4.11
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-5*2+1), %o2
|
add %o2, (-5*2+1), %o2
|
|
|
L4.11:
|
L4.11:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-5*2-1), %o2
|
add %o2, (-5*2-1), %o2
|
|
|
L3.13:
|
L3.13:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits -7
|
! depth 4, accumulated bits -7
|
bl L4.9
|
bl L4.9
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-7*2+1), %o2
|
add %o2, (-7*2+1), %o2
|
|
|
L4.9:
|
L4.9:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-7*2-1), %o2
|
add %o2, (-7*2-1), %o2
|
|
|
9:
|
9:
|
end_regular_divide:
|
end_regular_divide:
|
subcc %o4, 1, %o4
|
subcc %o4, 1, %o4
|
bge divloop
|
bge divloop
|
tst %o3
|
tst %o3
|
bl,a got_result
|
bl,a got_result
|
! non-restoring fixup here (one instruction only!)
|
! non-restoring fixup here (one instruction only!)
|
sub %o2, 1, %o2
|
sub %o2, 1, %o2
|
|
|
|
|
got_result:
|
got_result:
|
! check to see if answer should be < 0
|
! check to see if answer should be < 0
|
tst %g3
|
tst %g3
|
bl,a 1f
|
bl,a 1f
|
sub %g0, %o2, %o2
|
sub %g0, %o2, %o2
|
1:
|
1:
|
retl
|
retl
|
mov %o2, %o0
|
mov %o2, %o0
|
#endif
|
#endif
|
|
|
#ifdef L_modsi3
|
#ifdef L_modsi3
|
/* This implementation was taken from glibc:
|
/* This implementation was taken from glibc:
|
*
|
*
|
* Input: dividend and divisor in %o0 and %o1 respectively.
|
* Input: dividend and divisor in %o0 and %o1 respectively.
|
*
|
*
|
* Algorithm parameters:
|
* Algorithm parameters:
|
* N how many bits per iteration we try to get (4)
|
* N how many bits per iteration we try to get (4)
|
* WORDSIZE total number of bits (32)
|
* WORDSIZE total number of bits (32)
|
*
|
*
|
* Derived constants:
|
* Derived constants:
|
* TOPBITS number of bits in the top decade of a number
|
* TOPBITS number of bits in the top decade of a number
|
*
|
*
|
* Important variables:
|
* Important variables:
|
* Q the partial quotient under development (initially 0)
|
* Q the partial quotient under development (initially 0)
|
* R the remainder so far, initially the dividend
|
* R the remainder so far, initially the dividend
|
* ITER number of main division loop iterations required;
|
* ITER number of main division loop iterations required;
|
* equal to ceil(log2(quotient) / N). Note that this
|
* equal to ceil(log2(quotient) / N). Note that this
|
* is the log base (2^N) of the quotient.
|
* is the log base (2^N) of the quotient.
|
* V the current comparand, initially divisor*2^(ITER*N-1)
|
* V the current comparand, initially divisor*2^(ITER*N-1)
|
*
|
*
|
* Cost:
|
* Cost:
|
* Current estimate for non-large dividend is
|
* Current estimate for non-large dividend is
|
* ceil(log2(quotient) / N) * (10 + 7N/2) + C
|
* ceil(log2(quotient) / N) * (10 + 7N/2) + C
|
* A large dividend is one greater than 2^(31-TOPBITS) and takes a
|
* A large dividend is one greater than 2^(31-TOPBITS) and takes a
|
* different path, as the upper bits of the quotient must be developed
|
* different path, as the upper bits of the quotient must be developed
|
* one bit at a time.
|
* one bit at a time.
|
*/
|
*/
|
.text
|
.text
|
.align 4
|
.align 4
|
.global .urem
|
.global .urem
|
.proc 4
|
.proc 4
|
.urem:
|
.urem:
|
b divide
|
b divide
|
mov 0, %g3 ! result always positive
|
mov 0, %g3 ! result always positive
|
|
|
.align 4
|
.align 4
|
.global .rem
|
.global .rem
|
.proc 4
|
.proc 4
|
.rem:
|
.rem:
|
! compute sign of result; if neither is negative, no problem
|
! compute sign of result; if neither is negative, no problem
|
orcc %o1, %o0, %g0 ! either negative?
|
orcc %o1, %o0, %g0 ! either negative?
|
bge 2f ! no, go do the divide
|
bge 2f ! no, go do the divide
|
mov %o0, %g3 ! sign of remainder matches %o0
|
mov %o0, %g3 ! sign of remainder matches %o0
|
tst %o1
|
tst %o1
|
bge 1f
|
bge 1f
|
tst %o0
|
tst %o0
|
! %o1 is definitely negative; %o0 might also be negative
|
! %o1 is definitely negative; %o0 might also be negative
|
bge 2f ! if %o0 not negative...
|
bge 2f ! if %o0 not negative...
|
sub %g0, %o1, %o1 ! in any case, make %o1 nonneg
|
sub %g0, %o1, %o1 ! in any case, make %o1 nonneg
|
1: ! %o0 is negative, %o1 is nonnegative
|
1: ! %o0 is negative, %o1 is nonnegative
|
sub %g0, %o0, %o0 ! make %o0 nonnegative
|
sub %g0, %o0, %o0 ! make %o0 nonnegative
|
2:
|
2:
|
|
|
! Ready to divide. Compute size of quotient; scale comparand.
|
! Ready to divide. Compute size of quotient; scale comparand.
|
divide:
|
divide:
|
orcc %o1, %g0, %o5
|
orcc %o1, %g0, %o5
|
bne 1f
|
bne 1f
|
mov %o0, %o3
|
mov %o0, %o3
|
|
|
! Divide by zero trap. If it returns, return 0 (about as
|
! Divide by zero trap. If it returns, return 0 (about as
|
! wrong as possible, but that is what SunOS does...).
|
! wrong as possible, but that is what SunOS does...).
|
ta 0x2 !ST_DIV0
|
ta 0x2 !ST_DIV0
|
retl
|
retl
|
clr %o0
|
clr %o0
|
|
|
1:
|
1:
|
cmp %o3, %o5 ! if %o1 exceeds %o0, done
|
cmp %o3, %o5 ! if %o1 exceeds %o0, done
|
blu got_result ! (and algorithm fails otherwise)
|
blu got_result ! (and algorithm fails otherwise)
|
clr %o2
|
clr %o2
|
sethi %hi(1 << (32 - 4 - 1)), %g1
|
sethi %hi(1 << (32 - 4 - 1)), %g1
|
cmp %o3, %g1
|
cmp %o3, %g1
|
blu not_really_big
|
blu not_really_big
|
clr %o4
|
clr %o4
|
|
|
! Here the dividend is >= 2**(31-N) or so. We must be careful here,
|
! Here the dividend is >= 2**(31-N) or so. We must be careful here,
|
! as our usual N-at-a-shot divide step will cause overflow and havoc.
|
! as our usual N-at-a-shot divide step will cause overflow and havoc.
|
! The number of bits in the result here is N*ITER+SC, where SC <= N.
|
! The number of bits in the result here is N*ITER+SC, where SC <= N.
|
! Compute ITER in an unorthodox manner: know we need to shift V into
|
! Compute ITER in an unorthodox manner: know we need to shift V into
|
! the top decade: so do not even bother to compare to R.
|
! the top decade: so do not even bother to compare to R.
|
1:
|
1:
|
cmp %o5, %g1
|
cmp %o5, %g1
|
bgeu 3f
|
bgeu 3f
|
mov 1, %g2
|
mov 1, %g2
|
sll %o5, 4, %o5
|
sll %o5, 4, %o5
|
b 1b
|
b 1b
|
add %o4, 1, %o4
|
add %o4, 1, %o4
|
|
|
! Now compute %g2.
|
! Now compute %g2.
|
2: addcc %o5, %o5, %o5
|
2: addcc %o5, %o5, %o5
|
bcc not_too_big
|
bcc not_too_big
|
add %g2, 1, %g2
|
add %g2, 1, %g2
|
|
|
! We get here if the %o1 overflowed while shifting.
|
! We get here if the %o1 overflowed while shifting.
|
! This means that %o3 has the high-order bit set.
|
! This means that %o3 has the high-order bit set.
|
! Restore %o5 and subtract from %o3.
|
! Restore %o5 and subtract from %o3.
|
sll %g1, 4, %g1 ! high order bit
|
sll %g1, 4, %g1 ! high order bit
|
srl %o5, 1, %o5 ! rest of %o5
|
srl %o5, 1, %o5 ! rest of %o5
|
add %o5, %g1, %o5
|
add %o5, %g1, %o5
|
b do_single_div
|
b do_single_div
|
sub %g2, 1, %g2
|
sub %g2, 1, %g2
|
|
|
not_too_big:
|
not_too_big:
|
3: cmp %o5, %o3
|
3: cmp %o5, %o3
|
blu 2b
|
blu 2b
|
nop
|
nop
|
be do_single_div
|
be do_single_div
|
nop
|
nop
|
/* NB: these are commented out in the V8-SPARC manual as well */
|
/* NB: these are commented out in the V8-SPARC manual as well */
|
/* (I do not understand this) */
|
/* (I do not understand this) */
|
! %o5 > %o3: went too far: back up 1 step
|
! %o5 > %o3: went too far: back up 1 step
|
! srl %o5, 1, %o5
|
! srl %o5, 1, %o5
|
! dec %g2
|
! dec %g2
|
! do single-bit divide steps
|
! do single-bit divide steps
|
!
|
!
|
! We have to be careful here. We know that %o3 >= %o5, so we can do the
|
! We have to be careful here. We know that %o3 >= %o5, so we can do the
|
! first divide step without thinking. BUT, the others are conditional,
|
! first divide step without thinking. BUT, the others are conditional,
|
! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-
|
! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-
|
! order bit set in the first step, just falling into the regular
|
! order bit set in the first step, just falling into the regular
|
! division loop will mess up the first time around.
|
! division loop will mess up the first time around.
|
! So we unroll slightly...
|
! So we unroll slightly...
|
do_single_div:
|
do_single_div:
|
subcc %g2, 1, %g2
|
subcc %g2, 1, %g2
|
bl end_regular_divide
|
bl end_regular_divide
|
nop
|
nop
|
sub %o3, %o5, %o3
|
sub %o3, %o5, %o3
|
mov 1, %o2
|
mov 1, %o2
|
b end_single_divloop
|
b end_single_divloop
|
nop
|
nop
|
single_divloop:
|
single_divloop:
|
sll %o2, 1, %o2
|
sll %o2, 1, %o2
|
bl 1f
|
bl 1f
|
srl %o5, 1, %o5
|
srl %o5, 1, %o5
|
! %o3 >= 0
|
! %o3 >= 0
|
sub %o3, %o5, %o3
|
sub %o3, %o5, %o3
|
b 2f
|
b 2f
|
add %o2, 1, %o2
|
add %o2, 1, %o2
|
1: ! %o3 < 0
|
1: ! %o3 < 0
|
add %o3, %o5, %o3
|
add %o3, %o5, %o3
|
sub %o2, 1, %o2
|
sub %o2, 1, %o2
|
2:
|
2:
|
end_single_divloop:
|
end_single_divloop:
|
subcc %g2, 1, %g2
|
subcc %g2, 1, %g2
|
bge single_divloop
|
bge single_divloop
|
tst %o3
|
tst %o3
|
b,a end_regular_divide
|
b,a end_regular_divide
|
|
|
not_really_big:
|
not_really_big:
|
1:
|
1:
|
sll %o5, 4, %o5
|
sll %o5, 4, %o5
|
cmp %o5, %o3
|
cmp %o5, %o3
|
bleu 1b
|
bleu 1b
|
addcc %o4, 1, %o4
|
addcc %o4, 1, %o4
|
be got_result
|
be got_result
|
sub %o4, 1, %o4
|
sub %o4, 1, %o4
|
|
|
tst %o3 ! set up for initial iteration
|
tst %o3 ! set up for initial iteration
|
divloop:
|
divloop:
|
sll %o2, 4, %o2
|
sll %o2, 4, %o2
|
! depth 1, accumulated bits 0
|
! depth 1, accumulated bits 0
|
bl L1.16
|
bl L1.16
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 2, accumulated bits 1
|
! depth 2, accumulated bits 1
|
bl L2.17
|
bl L2.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 3, accumulated bits 3
|
! depth 3, accumulated bits 3
|
bl L3.19
|
bl L3.19
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits 7
|
! depth 4, accumulated bits 7
|
bl L4.23
|
bl L4.23
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (7*2+1), %o2
|
add %o2, (7*2+1), %o2
|
L4.23:
|
L4.23:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (7*2-1), %o2
|
add %o2, (7*2-1), %o2
|
|
|
L3.19:
|
L3.19:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits 5
|
! depth 4, accumulated bits 5
|
bl L4.21
|
bl L4.21
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (5*2+1), %o2
|
add %o2, (5*2+1), %o2
|
|
|
L4.21:
|
L4.21:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (5*2-1), %o2
|
add %o2, (5*2-1), %o2
|
|
|
L2.17:
|
L2.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 3, accumulated bits 1
|
! depth 3, accumulated bits 1
|
bl L3.17
|
bl L3.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits 3
|
! depth 4, accumulated bits 3
|
bl L4.19
|
bl L4.19
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (3*2+1), %o2
|
add %o2, (3*2+1), %o2
|
|
|
L4.19:
|
L4.19:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (3*2-1), %o2
|
add %o2, (3*2-1), %o2
|
|
|
L3.17:
|
L3.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits 1
|
! depth 4, accumulated bits 1
|
bl L4.17
|
bl L4.17
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (1*2+1), %o2
|
add %o2, (1*2+1), %o2
|
|
|
L4.17:
|
L4.17:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (1*2-1), %o2
|
add %o2, (1*2-1), %o2
|
|
|
L1.16:
|
L1.16:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 2, accumulated bits -1
|
! depth 2, accumulated bits -1
|
bl L2.15
|
bl L2.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 3, accumulated bits -1
|
! depth 3, accumulated bits -1
|
bl L3.15
|
bl L3.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits -1
|
! depth 4, accumulated bits -1
|
bl L4.15
|
bl L4.15
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-1*2+1), %o2
|
add %o2, (-1*2+1), %o2
|
|
|
L4.15:
|
L4.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-1*2-1), %o2
|
add %o2, (-1*2-1), %o2
|
|
|
L3.15:
|
L3.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits -3
|
! depth 4, accumulated bits -3
|
bl L4.13
|
bl L4.13
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-3*2+1), %o2
|
add %o2, (-3*2+1), %o2
|
|
|
L4.13:
|
L4.13:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-3*2-1), %o2
|
add %o2, (-3*2-1), %o2
|
|
|
L2.15:
|
L2.15:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 3, accumulated bits -3
|
! depth 3, accumulated bits -3
|
bl L3.13
|
bl L3.13
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
! depth 4, accumulated bits -5
|
! depth 4, accumulated bits -5
|
bl L4.11
|
bl L4.11
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-5*2+1), %o2
|
add %o2, (-5*2+1), %o2
|
|
|
L4.11:
|
L4.11:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-5*2-1), %o2
|
add %o2, (-5*2-1), %o2
|
|
|
L3.13:
|
L3.13:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
! depth 4, accumulated bits -7
|
! depth 4, accumulated bits -7
|
bl L4.9
|
bl L4.9
|
srl %o5,1,%o5
|
srl %o5,1,%o5
|
! remainder is positive
|
! remainder is positive
|
subcc %o3,%o5,%o3
|
subcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-7*2+1), %o2
|
add %o2, (-7*2+1), %o2
|
|
|
L4.9:
|
L4.9:
|
! remainder is negative
|
! remainder is negative
|
addcc %o3,%o5,%o3
|
addcc %o3,%o5,%o3
|
b 9f
|
b 9f
|
add %o2, (-7*2-1), %o2
|
add %o2, (-7*2-1), %o2
|
|
|
9:
|
9:
|
end_regular_divide:
|
end_regular_divide:
|
subcc %o4, 1, %o4
|
subcc %o4, 1, %o4
|
bge divloop
|
bge divloop
|
tst %o3
|
tst %o3
|
bl,a got_result
|
bl,a got_result
|
! non-restoring fixup here (one instruction only!)
|
! non-restoring fixup here (one instruction only!)
|
add %o3, %o1, %o3
|
add %o3, %o1, %o3
|
|
|
got_result:
|
got_result:
|
! check to see if answer should be < 0
|
! check to see if answer should be < 0
|
tst %g3
|
tst %g3
|
bl,a 1f
|
bl,a 1f
|
sub %g0, %o3, %o3
|
sub %g0, %o3, %o3
|
1:
|
1:
|
retl
|
retl
|
mov %o3, %o0
|
mov %o3, %o0
|
|
|
#endif
|
#endif
|
|
|
|
|
