URL
https://opencores.org/ocsvn/or1k/or1k/trunk
Subversion Repositories or1k
[/] [or1k/] [trunk/] [linux/] [uClibc/] [libc/] [sysdeps/] [linux/] [sparc/] [sdiv.S] - Rev 1765
Compare with Previous | Blame | View Log
/* This file is generated from divrem.m4; DO NOT EDIT! *//** Division and remainder, from Appendix E of the Sparc Version 8* Architecture Manual, with fixes from Gordon Irlam.*//** Input: dividend and divisor in %o0 and %o1 respectively.** m4 parameters:* .div name of function to generate* div div=div => %o0 / %o1; div=rem => %o0 % %o1* true true=true => signed; true=false => unsigned** Algorithm parameters:* N how many bits per iteration we try to get (4)* WORDSIZE total number of bits (32)** Derived constants:* TOPBITS number of bits in the top decade of a number** Important variables:* Q the partial quotient under development (initially 0)* R the remainder so far, initially the dividend* ITER number of main division loop iterations required;* equal to ceil(log2(quotient) / N). Note that this* is the log base (2^N) of the quotient.* V the current comparand, initially divisor*2^(ITER*N-1)** Cost:* Current estimate for non-large dividend is* ceil(log2(quotient) / N) * (10 + 7N/2) + C* 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* one bit at a time.*/#include <sys/syscall.h>.global .div;.align 4;.type .div ,@function;.div:! compute sign of result; if neither is negative, no problemorcc %o1, %o0, %g0 ! either negative?bge 2f ! no, go do the dividexor %o1, %o0, %g3 ! compute sign in any casetst %o1bge 1ftst %o0! %o1 is definitely negative; %o0 might also be negativebge 2f ! if %o0 not negative...sub %g0, %o1, %o1 ! in any case, make %o1 nonneg1: ! %o0 is negative, %o1 is nonnegativesub %g0, %o0, %o0 ! make %o0 nonnegative2:! Ready to divide. Compute size of quotient; scale comparand.orcc %o1, %g0, %o5bne 1fmov %o0, %o3! Divide by zero trap. If it returns, return 0 (about as! wrong as possible, but that is what SunOS does...).ta 0x02retlclr %o01:cmp %o3, %o5 ! if %o1 exceeds %o0, doneblu .Lgot_result ! (and algorithm fails otherwise)clr %o2sethi %hi(1 << (32 - 4 - 1)), %g1cmp %o3, %g1blu .Lnot_really_bigclr %o4! 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.! 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! the top decade: so do not even bother to compare to R.1:cmp %o5, %g1bgeu 3fmov 1, %g2sll %o5, 4, %o5b 1badd %o4, 1, %o4! Now compute %g2.2: addcc %o5, %o5, %o5bcc .Lnot_too_bigadd %g2, 1, %g2! We get here if the %o1 overflowed while shifting.! This means that %o3 has the high-order bit set.! Restore %o5 and subtract from %o3.sll %g1, 4, %g1 ! high order bitsrl %o5, 1, %o5 ! rest of %o5add %o5, %g1, %o5b .Ldo_single_divsub %g2, 1, %g2.Lnot_too_big:3: cmp %o5, %o3blu 2bnopbe .Ldo_single_divnop/* NB: these are commented out in the V8-Sparc manual as well *//* (I do not understand this) */! %o5 > %o3: went too far: back up 1 step! srl %o5, 1, %o5! dec %g2! do single-bit divide steps!! 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,! 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! division loop will mess up the first time around.! So we unroll slightly....Ldo_single_div:subcc %g2, 1, %g2bl .Lend_regular_dividenopsub %o3, %o5, %o3mov 1, %o2b .Lend_single_divloopnop.Lsingle_divloop:sll %o2, 1, %o2bl 1fsrl %o5, 1, %o5! %o3 >= 0sub %o3, %o5, %o3b 2fadd %o2, 1, %o21: ! %o3 < 0add %o3, %o5, %o3sub %o2, 1, %o22:.Lend_single_divloop:subcc %g2, 1, %g2bge .Lsingle_divlooptst %o3b,a .Lend_regular_divide.Lnot_really_big:1:sll %o5, 4, %o5cmp %o5, %o3bleu 1baddcc %o4, 1, %o4be .Lgot_resultsub %o4, 1, %o4tst %o3 ! set up for initial iteration.Ldivloop:sll %o2, 4, %o2! depth 1, accumulated bits 0bl .L1.16srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 2, accumulated bits 1bl .L2.17srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 3, accumulated bits 3bl .L3.19srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 4, accumulated bits 7bl .L4.23srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (7*2+1), %o2.L4.23:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (7*2-1), %o2.L3.19:! remainder is negativeaddcc %o3,%o5,%o3! depth 4, accumulated bits 5bl .L4.21srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (5*2+1), %o2.L4.21:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (5*2-1), %o2.L2.17:! remainder is negativeaddcc %o3,%o5,%o3! depth 3, accumulated bits 1bl .L3.17srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 4, accumulated bits 3bl .L4.19srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (3*2+1), %o2.L4.19:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (3*2-1), %o2.L3.17:! remainder is negativeaddcc %o3,%o5,%o3! depth 4, accumulated bits 1bl .L4.17srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (1*2+1), %o2.L4.17:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (1*2-1), %o2.L1.16:! remainder is negativeaddcc %o3,%o5,%o3! depth 2, accumulated bits -1bl .L2.15srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 3, accumulated bits -1bl .L3.15srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 4, accumulated bits -1bl .L4.15srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (-1*2+1), %o2.L4.15:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (-1*2-1), %o2.L3.15:! remainder is negativeaddcc %o3,%o5,%o3! depth 4, accumulated bits -3bl .L4.13srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (-3*2+1), %o2.L4.13:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (-3*2-1), %o2.L2.15:! remainder is negativeaddcc %o3,%o5,%o3! depth 3, accumulated bits -3bl .L3.13srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3! depth 4, accumulated bits -5bl .L4.11srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (-5*2+1), %o2.L4.11:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (-5*2-1), %o2.L3.13:! remainder is negativeaddcc %o3,%o5,%o3! depth 4, accumulated bits -7bl .L4.9srl %o5,1,%o5! remainder is positivesubcc %o3,%o5,%o3b 9fadd %o2, (-7*2+1), %o2.L4.9:! remainder is negativeaddcc %o3,%o5,%o3b 9fadd %o2, (-7*2-1), %o29:.Lend_regular_divide:subcc %o4, 1, %o4bge .Ldivlooptst %o3bl,a .Lgot_result! non-restoring fixup here (one instruction only!)sub %o2, 1, %o2.Lgot_result:! check to see if answer should be < 0tst %g3bl,a 1fsub %g0, %o2, %o21:retlmov %o2, %o0.size .div,.-.div;
