head 1.2; access; symbols perseant-exfatfs-base-20250801:1.2 perseant-exfatfs-base-20240630:1.2 cjep_sun2x:1.2.0.44 cjep_sun2x-base:1.2 cjep_staticlib_x-base1:1.2 cjep_staticlib_x:1.2.0.42 cjep_staticlib_x-base:1.2 phil-wifi-20200421:1.2 phil-wifi-20200411:1.2 phil-wifi-20200406:1.2 pgoyette-compat-merge-20190127:1.2 pgoyette-compat-20190127:1.2 pgoyette-compat-20190118:1.2 pgoyette-compat-1226:1.2 pgoyette-compat-1126:1.2 pgoyette-compat-1020:1.2 pgoyette-compat-0930:1.2 pgoyette-compat-0906:1.2 pgoyette-compat-0728:1.2 pgoyette-compat-0625:1.2 pgoyette-compat-0521:1.2 pgoyette-compat-0502:1.2 pgoyette-compat-0422:1.2 pgoyette-compat-0415:1.2 pgoyette-compat-0407:1.2 pgoyette-compat-0330:1.2 pgoyette-compat-0322:1.2 pgoyette-compat-0315:1.2 pgoyette-compat:1.2.0.40 pgoyette-compat-base:1.2 perseant-stdc-iso10646:1.2.0.38 perseant-stdc-iso10646-base:1.2 prg-localcount2-base3:1.2 prg-localcount2-base2:1.2 prg-localcount2-base1:1.2 prg-localcount2:1.2.0.36 prg-localcount2-base:1.2 pgoyette-localcount-20170426:1.2 bouyer-socketcan-base1:1.2 pgoyette-localcount-20170320:1.2 bouyer-socketcan:1.2.0.34 bouyer-socketcan-base:1.2 pgoyette-localcount-20170107:1.2 pgoyette-localcount-20161104:1.2 localcount-20160914:1.2 pgoyette-localcount-20160806:1.2 pgoyette-localcount-20160726:1.2 pgoyette-localcount:1.2.0.32 pgoyette-localcount-base:1.2 netbsd-5-2-3-RELEASE:1.2 netbsd-5-1-5-RELEASE:1.2 yamt-pagecache-base9:1.2 yamt-pagecache-tag8:1.2 tls-earlyentropy:1.2.0.28 tls-earlyentropy-base:1.2 riastradh-xf86-video-intel-2-7-1-pre-2-21-15:1.2 riastradh-drm2-base3:1.2 netbsd-5-2-2-RELEASE:1.2 netbsd-5-1-4-RELEASE:1.2 netbsd-5-2-1-RELEASE:1.2 netbsd-5-1-3-RELEASE:1.2 agc-symver:1.2.0.30 agc-symver-base:1.2 tls-maxphys-base:1.2 yamt-pagecache-base8:1.2 netbsd-5-2:1.2.0.26 yamt-pagecache-base7:1.2 netbsd-5-2-RELEASE:1.2 netbsd-5-2-RC1:1.2 yamt-pagecache-base6:1.2 yamt-pagecache-base5:1.2 yamt-pagecache-base4:1.2 netbsd-5-1-2-RELEASE:1.2 netbsd-5-1-1-RELEASE:1.2 yamt-pagecache-base3:1.2 yamt-pagecache-base2:1.2 yamt-pagecache:1.2.0.24 yamt-pagecache-base:1.2 bouyer-quota2-nbase:1.2 bouyer-quota2:1.2.0.22 bouyer-quota2-base:1.2 matt-nb5-pq3:1.2.0.20 matt-nb5-pq3-base:1.2 netbsd-5-1:1.2.0.18 netbsd-5-1-RELEASE:1.2 netbsd-5-1-RC4:1.2 netbsd-5-1-RC3:1.2 netbsd-5-1-RC2:1.2 netbsd-5-1-RC1:1.2 netbsd-5-0-2-RELEASE:1.2 netbsd-5-0-1-RELEASE:1.2 jym-xensuspend-nbase:1.2 netbsd-5-0:1.2.0.16 netbsd-5-0-RELEASE:1.2 netbsd-5-0-RC4:1.2 netbsd-5-0-RC3:1.2 netbsd-5-0-RC2:1.2 jym-xensuspend:1.2.0.14 jym-xensuspend-base:1.2 netbsd-5-0-RC1:1.2 netbsd-5:1.2.0.12 netbsd-5-base:1.2 mjf-devfs2:1.2.0.10 mjf-devfs2-base:1.2 yamt-pf42-base4:1.2 yamt-pf42-base3:1.2 hpcarm-cleanup-nbase:1.2 yamt-pf42-base2:1.2 yamt-pf42:1.2.0.8 yamt-pf42-base:1.2 keiichi-mipv6:1.2.0.6 keiichi-mipv6-base:1.2 cube-autoconf:1.2.0.4 cube-autoconf-base:1.2 hpcarm-cleanup:1.2.0.2 hpcarm-cleanup-base:1.2 netbsd-1-6-PATCH002-RELEASE:1.1.1.2 netbsd-1-6-PATCH002:1.1.1.2 netbsd-1-6-PATCH002-RC4:1.1.1.2 netbsd-1-6-PATCH002-RC3:1.1.1.2 netbsd-1-6-PATCH002-RC2:1.1.1.2 netbsd-1-6-PATCH002-RC1:1.1.1.2 netbsd-1-6-PATCH001:1.1.1.2 netbsd-1-6-PATCH001-RELEASE:1.1.1.2 netbsd-1-6-PATCH001-RC3:1.1.1.2 netbsd-1-6-PATCH001-RC2:1.1.1.2 netbsd-1-6-PATCH001-RC1:1.1.1.2 netbsd-1-6-RELEASE:1.1.1.2 netbsd-1-6-RC3:1.1.1.2 netbsd-1-6-RC2:1.1.1.2 netbsd-1-6-RC1:1.1.1.2 netbsd-1-6:1.1.1.2.0.10 netbsd-1-6-base:1.1.1.2 netbsd-1-5-PATCH003:1.1.1.2 netbsd-1-5-PATCH002:1.1.1.2 netbsd-1-5-PATCH001:1.1.1.2 netbsd-1-5-RELEASE:1.1.1.2 netbsd-1-5-BETA2:1.1.1.2 netbsd-1-5-BETA:1.1.1.2 netbsd-1-4-PATCH003:1.1.1.2 netbsd-1-5-ALPHA2:1.1.1.2 netbsd-1-5:1.1.1.2.0.8 netbsd-1-5-base:1.1.1.2 netbsd-1-4-PATCH002:1.1.1.2 wrstuden-devbsize-19991221:1.1.1.2 wrstuden-devbsize:1.1.1.2.0.6 wrstuden-devbsize-base:1.1.1.2 comdex-fall-1999:1.1.1.2.0.4 comdex-fall-1999-base:1.1.1.2 netbsd-1-4-PATCH001:1.1.1.2 netbsd-1-4-RELEASE:1.1.1.2 egcs-1-1-2:1.1.1.2 netbsd-1-4:1.1.1.2.0.2 netbsd-1-4-base:1.1.1.2 egcs-1-1-1:1.1.1.2 egcs-1-1-1-prerelease-2:1.1.1.2 egcs-1-1-19981014:1.1.1.2 egcs-1-1b:1.1.1.2 egcs-1-1-19980824:1.1.1.2 egcs-1-1-19980816:1.1.1.2 egcs-1-0-2:1.1.1.1 CYGNUS:1.1.1; locks; strict; comment @;; @; 1.2 date 2002.09.16.16.58.02; author thorpej; state dead; branches; next 1.1; 1.1 date 98.03.29.08.15.42; author mrg; state Exp; branches 1.1.1.1; next ; 1.1.1.1 date 98.03.29.08.15.42; author mrg; state Exp; branches; next 1.1.1.2; 1.1.1.2 date 98.08.16.17.45.41; author tv; state Exp; branches; next ; desc @@ 1.2 log @Remove the old egcs compiler. @ text @/* This is an assembly language implementation of libgcc1.c for the sparc processor. These routines are derived from the Sparc Architecture Manual, version 8, slightly edited to match the desired calling convention, and also to optimize them for our purposes. */ #ifdef L_mulsi3 .text .align 4 .global .umul .proc 4 .umul: or %o0, %o1, %o4 ! logical or of multiplier and multiplicand mov %o0, %y ! multiplier to Y register andncc %o4, 0xfff, %o5 ! mask out lower 12 bits be mul_shortway ! can do it the short way andcc %g0, %g0, %o4 ! zero the partial product and clear NV cc ! ! long multiply ! 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 ! 32nd iteration mulscc %o4, %g0, %o4 ! last iteration only shifts ! the upper 32 bits of product are wrong, but we do not care retl rd %y, %o0 ! ! short multiply ! mul_shortway: 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 ! 12th iteration mulscc %o4, %g0, %o4 ! last iteration only shifts rd %y, %o5 sll %o4, 12, %o4 ! left shift partial product by 12 bits srl %o5, 20, %o5 ! right shift partial product by 20 bits retl or %o5, %o4, %o0 ! merge for true product #endif #ifdef L_divsi3 .text .align 4 .global .udiv .proc 4 .udiv: save %sp, -64, %sp b divide mov 0, %l2 ! result always positive .global .div .proc 4 .div: save %sp, -64, %sp orcc %i1, %i0, %g0 ! is either operand negative bge divide ! if not, skip this junk xor %i1, %i0, %l2 ! record sign of result in sign of %l2 tst %i1 bge 2f tst %i0 ! %i1 < 0 bge divide neg %i1 2: ! %i0 < 0 neg %i0 ! FALL THROUGH divide: ! Compute size of quotient, scale comparand. orcc %i1, %g0, %l1 ! movcc %i1, %l1 te 2 ! if %i1 = 0 mov %i0, %i3 mov 0, %i2 sethi %hi(1<<(32-4-1)), %l3 cmp %i3, %l3 blu not_really_big mov 0, %l0 ! ! Here, the %i0 is >= 2^(31-3) or so. We must be careful here, ! as our usual 3-at-a-shot divide step will cause overflow and havoc. ! The total number of bits in the result here is 3*%l0+%l4, where ! %l4 <= 3. ! Compute %l0 in an unorthodox manner: know we need to Shift %l1 into ! the top decade: so do not even bother to compare to %i3. 1: cmp %l1, %l3 bgeu 3f mov 1, %l4 sll %l1, 3, %l1 b 1b inc %l0 ! ! Now compute %l4 ! 2: addcc %l1, %l1, %l1 bcc not_too_big add %l4, 1, %l4 ! ! We are here if the %i1 overflowed when Shifting. ! This means that %i3 has the high-order bit set. ! Restore %l1 and subtract from %i3. sll %l3, 4, %l3 srl %l1, 1, %l1 add %l1, %l3, %l1 b do_single_div dec %l4 not_too_big: 3: cmp %l1, %i3 blu 2b nop be do_single_div nop ! %l1 > %i3: went too far: back up 1 step ! srl %l1, 1, %l1 ! dec %l4 ! do single-bit divide steps ! ! We have to be careful here. We know that %i3 >= %l1, so we can do the ! first divide step without thinking. BUT, the others are conditional, ! and are only done if %i3 >= 0. Because both %i3 and %l1 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... do_single_div: deccc %l4 bl end_regular_divide nop sub %i3, %l1, %i3 mov 1, %i2 b end_single_divloop nop single_divloop: sll %i2, 1, %i2 bl 1f srl %l1, 1, %l1 ! %i3 >= 0 sub %i3, %l1, %i3 b 2f inc %i2 1: ! %i3 < 0 add %i3, %l1, %i3 dec %i2 end_single_divloop: 2: deccc %l4 bge single_divloop tst %i3 b end_regular_divide nop not_really_big: 1: sll %l1, 3, %l1 cmp %l1, %i3 bleu 1b inccc %l0 be got_result dec %l0 do_regular_divide: ! Do the main division iteration tst %i3 ! Fall through into divide loop divloop: sll %i2, 3, %i2 ! depth 1, accumulated bits 0 bl L.1.8 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 2, accumulated bits 1 bl L.2.9 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 3, accumulated bits 3 bl L.3.11 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (3*2+1), %i2 L.3.11: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (3*2-1), %i2 L.2.9: ! remainder is negative addcc %i3,%l1,%i3 ! depth 3, accumulated bits 1 bl L.3.9 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (1*2+1), %i2 L.3.9: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (1*2-1), %i2 L.1.8: ! remainder is negative addcc %i3,%l1,%i3 ! depth 2, accumulated bits -1 bl L.2.7 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 3, accumulated bits -1 bl L.3.7 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (-1*2+1), %i2 L.3.7: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (-1*2-1), %i2 L.2.7: ! remainder is negative addcc %i3,%l1,%i3 ! depth 3, accumulated bits -3 bl L.3.5 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (-3*2+1), %i2 L.3.5: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (-3*2-1), %i2 end_regular_divide: 9: deccc %l0 bge divloop tst %i3 bge got_result nop ! non-restoring fixup here dec %i2 got_result: tst %l2 bge 1f restore ! answer < 0 retl ! leaf-routine return neg %o2, %o0 ! quotient <- -%i2 1: retl ! leaf-routine return mov %o2, %o0 ! quotient <- %i2 #endif #ifdef L_modsi3 .text .align 4 .global .urem .proc 4 .urem: save %sp, -64, %sp b divide mov 0, %l2 ! result always positive .global .rem .proc 4 .rem: save %sp, -64, %sp orcc %i1, %i0, %g0 ! is either operand negative bge divide ! if not, skip this junk mov %i0, %l2 ! record sign of result in sign of %i2 tst %i1 bge 2f tst %i0 ! %i1 < 0 bge divide neg %i1 2: ! %i0 < 0 neg %i0 ! FALL THROUGH divide: ! Compute size of quotient, scale comparand. orcc %i1, %g0, %l1 ! movcc %i1, %l1 te 2 ! if %i1 = 0 mov %i0, %i3 mov 0, %i2 sethi %hi(1<<(32-4-1)), %l3 cmp %i3, %l3 blu not_really_big mov 0, %l0 ! ! Here, the %i0 is >= 2^(31-3) or so. We must be careful here, ! as our usual 3-at-a-shot divide step will cause overflow and havoc. ! The total number of bits in the result here is 3*%l0+%l4, where ! %l4 <= 3. ! Compute %l0 in an unorthodox manner: know we need to Shift %l1 into ! the top decade: so do not even bother to compare to %i3. 1: cmp %l1, %l3 bgeu 3f mov 1, %l4 sll %l1, 3, %l1 b 1b inc %l0 ! ! Now compute %l4 ! 2: addcc %l1, %l1, %l1 bcc not_too_big add %l4, 1, %l4 ! ! We are here if the %i1 overflowed when Shifting. ! This means that %i3 has the high-order bit set. ! Restore %l1 and subtract from %i3. sll %l3, 4, %l3 srl %l1, 1, %l1 add %l1, %l3, %l1 b do_single_div dec %l4 not_too_big: 3: cmp %l1, %i3 blu 2b nop be do_single_div nop ! %l1 > %i3: went too far: back up 1 step ! srl %l1, 1, %l1 ! dec %l4 ! do single-bit divide steps ! ! We have to be careful here. We know that %i3 >= %l1, so we can do the ! first divide step without thinking. BUT, the others are conditional, ! and are only done if %i3 >= 0. Because both %i3 and %l1 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... do_single_div: deccc %l4 bl end_regular_divide nop sub %i3, %l1, %i3 mov 1, %i2 b end_single_divloop nop single_divloop: sll %i2, 1, %i2 bl 1f srl %l1, 1, %l1 ! %i3 >= 0 sub %i3, %l1, %i3 b 2f inc %i2 1: ! %i3 < 0 add %i3, %l1, %i3 dec %i2 end_single_divloop: 2: deccc %l4 bge single_divloop tst %i3 b end_regular_divide nop not_really_big: 1: sll %l1, 3, %l1 cmp %l1, %i3 bleu 1b inccc %l0 be got_result dec %l0 do_regular_divide: ! Do the main division iteration tst %i3 ! Fall through into divide loop divloop: sll %i2, 3, %i2 ! depth 1, accumulated bits 0 bl L.1.8 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 2, accumulated bits 1 bl L.2.9 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 3, accumulated bits 3 bl L.3.11 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (3*2+1), %i2 L.3.11: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (3*2-1), %i2 L.2.9: ! remainder is negative addcc %i3,%l1,%i3 ! depth 3, accumulated bits 1 bl L.3.9 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (1*2+1), %i2 L.3.9: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (1*2-1), %i2 L.1.8: ! remainder is negative addcc %i3,%l1,%i3 ! depth 2, accumulated bits -1 bl L.2.7 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 ! depth 3, accumulated bits -1 bl L.3.7 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (-1*2+1), %i2 L.3.7: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (-1*2-1), %i2 L.2.7: ! remainder is negative addcc %i3,%l1,%i3 ! depth 3, accumulated bits -3 bl L.3.5 srl %l1,1,%l1 ! remainder is positive subcc %i3,%l1,%i3 b 9f add %i2, (-3*2+1), %i2 L.3.5: ! remainder is negative addcc %i3,%l1,%i3 b 9f add %i2, (-3*2-1), %i2 end_regular_divide: 9: deccc %l0 bge divloop tst %i3 bge got_result nop ! non-restoring fixup here add %i3, %i1, %i3 got_result: tst %l2 bge 1f restore ! answer < 0 retl ! leaf-routine return neg %o3, %o0 ! remainder <- -%i3 1: retl ! leaf-routine return mov %o3, %o0 ! remainder <- %i3 #endif @ 1.1 log @Initial revision @ text @@ 1.1.1.1 log @initial import of the egcs 1.0.2 compiler @ text @@ 1.1.1.2 log @Import egcs-1.1 gcc, 19980816 snapshot @ text @d83 4 a86 39 /* * 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. */ .global .udiv .align 4 .proc 4 .text d88 5 a92 7 b ready_to_divide mov 0, %g3 ! result is always positive .global .div .align 4 .proc 4 .text d94 21 a114 33 ! compute sign of result; if neither is negative, no problem orcc %o1, %o0, %g0 ! either negative? bge ready_to_divide ! no, go do the divide xor %o1, %o0, %g3 ! compute sign in any case tst %o1 bge 1f tst %o0 ! %o1 is definitely negative; %o0 might also be negative bge ready_to_divide ! if %o0 not negative... sub %g0, %o1, %o1 ! in any case, make %o1 nonneg 1: ! %o0 is negative, %o1 is nonnegative sub %g0, %o0, %o0 ! make %o0 nonnegative ready_to_divide: ! Ready to divide. Compute size of quotient; scale comparand. orcc %o1, %g0, %o5 bne 1f mov %o0, %o3 ! Divide by zero trap. If it returns, return 0 (about as ! wrong as possible, but that is what SunOS does...). ta 0x2 ! ST_DIV0 retl clr %o0 1: cmp %o3, %o5 ! if %o1 exceeds %o0, done blu got_result ! (and algorithm fails otherwise) clr %o2 sethi %hi(1 << (32 - 4 - 1)), %g1 cmp %o3, %g1 d116 38 a153 40 clr %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, %g1 bgeu 3f mov 1, %g2 sll %o5, 4, %o5 b 1b add %o4, 1, %o4 ! Now compute %g2. 2: addcc %o5, %o5, %o5 bcc not_too_big add %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 bit srl %o5, 1, %o5 ! rest of %o5 add %o5, %g1, %o5 b do_single_div sub %g2, 1, %g2 not_too_big: 3: cmp %o5, %o3 blu 2b nop be do_single_div nop /* 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 d156 1 a156 1 ! We have to be careful here. We know that %o3 >= %o5, so we can do the d158 2 a159 2 ! 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 d162 25 a186 26 do_single_div: subcc %g2, 1, %g2 bl end_regular_divide nop sub %o3, %o5, %o3 mov 1, %o2 b end_single_divloop nop single_divloop: sll %o2, 1, %o2 bl 1f srl %o5, 1, %o5 ! %o3 >= 0 sub %o3, %o5, %o3 b 2f add %o2, 1, %o2 1: ! %o3 < 0 add %o3, %o5, %o3 sub %o2, 1, %o2 2: end_single_divloop: subcc %g2, 1, %g2 bge single_divloop tst %o3 b,a end_regular_divide d188 2 a189 3 1: sll %o5, 4, %o5 cmp %o5, %o3 d191 1 a191 1 addcc %o4, 1, %o4 d193 5 a197 3 sub %o4, 1, %o4 tst %o3 ! set up for initial iteration d199 1 a199 1 sll %o2, 4, %o2 d201 2 a202 2 bl L1.16 srl %o5,1,%o5 d204 1 a204 1 subcc %o3,%o5,%o3 d206 2 a207 2 bl L2.17 srl %o5,1,%o5 d209 1 a209 1 subcc %o3,%o5,%o3 d211 2 a212 2 bl L3.19 srl %o5,1,%o5 d214 9 a222 36 subcc %o3,%o5,%o3 ! depth 4, accumulated bits 7 bl L4.23 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (7*2+1), %o2 L4.23: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (7*2-1), %o2 L3.19: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits 5 bl L4.21 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (5*2+1), %o2 L4.21: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (5*2-1), %o2 L2.17: ! remainder is negative addcc %o3,%o5,%o3 d224 2 a225 24 bl L3.17 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 ! depth 4, accumulated bits 3 bl L4.19 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (3*2+1), %o2 L4.19: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (3*2-1), %o2 L3.17: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits 1 bl L4.17 srl %o5,1,%o5 d227 1 a227 1 subcc %o3,%o5,%o3 d229 7 a235 11 add %o2, (1*2+1), %o2 L4.17: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (1*2-1), %o2 L1.16: ! remainder is negative addcc %o3,%o5,%o3 d237 2 a238 2 bl L2.15 srl %o5,1,%o5 d240 1 a240 1 subcc %o3,%o5,%o3 d242 2 a243 2 bl L3.15 srl %o5,1,%o5 d245 9 a253 35 subcc %o3,%o5,%o3 ! depth 4, accumulated bits -1 bl L4.15 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-1*2+1), %o2 L4.15: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-1*2-1), %o2 L3.15: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits -3 bl L4.13 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-3*2+1), %o2 L4.13: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-3*2-1), %o2 L2.15: ! remainder is negative addcc %o3,%o5,%o3 d255 2 a256 2 bl L3.13 srl %o5,1,%o5 d258 5 a262 23 subcc %o3,%o5,%o3 ! depth 4, accumulated bits -5 bl L4.11 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-5*2+1), %o2 L4.11: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-5*2-1), %o2 L3.13: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits -7 bl L4.9 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 d264 1 a264 9 add %o2, (-7*2+1), %o2 L4.9: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-7*2-1), %o2 9: d266 1 a266 1 subcc %o4, 1, %o4 d268 5 a272 6 tst %o3 bl,a got_result ! non-restoring fixup here (one instruction only!) sub %o2, 1, %o2 d274 8 a281 7 ! check to see if answer should be < 0 tst %g3 bl,a 1f sub %g0, %o2, %o2 1: retl mov %o2, %o0 a284 26 /* This implementation was taken from glibc: * * Input: dividend and divisor in %o0 and %o1 respectively. * * 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. */ d290 1 d292 1 a292 3 mov 0, %g3 ! result always positive .align 4 d296 13 a308 15 ! compute sign of result; if neither is negative, no problem orcc %o1, %o0, %g0 ! either negative? bge 2f ! no, go do the divide mov %o0, %g3 ! sign of remainder matches %o0 tst %o1 bge 1f tst %o0 ! %o1 is definitely negative; %o0 might also be negative bge 2f ! if %o0 not negative... sub %g0, %o1, %o1 ! in any case, make %o1 nonneg 1: ! %o0 is negative, %o1 is nonnegative sub %g0, %o0, %o0 ! make %o0 nonnegative 2: ! Ready to divide. Compute size of quotient; scale comparand. d310 7 a316 16 orcc %o1, %g0, %o5 bne 1f mov %o0, %o3 ! Divide by zero trap. If it returns, return 0 (about as ! wrong as possible, but that is what SunOS does...). ta 0x2 !ST_DIV0 retl clr %o0 1: cmp %o3, %o5 ! if %o1 exceeds %o0, done blu got_result ! (and algorithm fails otherwise) clr %o2 sethi %hi(1 << (32 - 4 - 1)), %g1 cmp %o3, %g1 d318 38 a355 40 clr %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, %g1 bgeu 3f mov 1, %g2 sll %o5, 4, %o5 b 1b add %o4, 1, %o4 ! Now compute %g2. 2: addcc %o5, %o5, %o5 bcc not_too_big add %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 bit srl %o5, 1, %o5 ! rest of %o5 add %o5, %g1, %o5 b do_single_div sub %g2, 1, %g2 not_too_big: 3: cmp %o5, %o3 blu 2b nop be do_single_div nop /* 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 d358 1 a358 1 ! We have to be careful here. We know that %o3 >= %o5, so we can do the d360 2 a361 2 ! 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 d364 25 a388 26 do_single_div: subcc %g2, 1, %g2 bl end_regular_divide nop sub %o3, %o5, %o3 mov 1, %o2 b end_single_divloop nop single_divloop: sll %o2, 1, %o2 bl 1f srl %o5, 1, %o5 ! %o3 >= 0 sub %o3, %o5, %o3 b 2f add %o2, 1, %o2 1: ! %o3 < 0 add %o3, %o5, %o3 sub %o2, 1, %o2 2: end_single_divloop: subcc %g2, 1, %g2 bge single_divloop tst %o3 b,a end_regular_divide d390 2 a391 3 1: sll %o5, 4, %o5 cmp %o5, %o3 d393 1 a393 1 addcc %o4, 1, %o4 d395 5 a399 3 sub %o4, 1, %o4 tst %o3 ! set up for initial iteration d401 4 a404 4 sll %o2, 4, %o2 ! depth 1, accumulated bits 0 bl L1.16 srl %o5,1,%o5 d406 1 a406 1 subcc %o3,%o5,%o3 d408 2 a409 2 bl L2.17 srl %o5,1,%o5 d411 1 a411 1 subcc %o3,%o5,%o3 d413 2 a414 2 bl L3.19 srl %o5,1,%o5 d416 9 a424 34 subcc %o3,%o5,%o3 ! depth 4, accumulated bits 7 bl L4.23 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (7*2+1), %o2 L4.23: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (7*2-1), %o2 L3.19: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits 5 bl L4.21 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (5*2+1), %o2 L4.21: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (5*2-1), %o2 L2.17: ! remainder is negative addcc %o3,%o5,%o3 d426 2 a427 2 bl L3.17 srl %o5,1,%o5 d429 9 a437 35 subcc %o3,%o5,%o3 ! depth 4, accumulated bits 3 bl L4.19 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (3*2+1), %o2 L4.19: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (3*2-1), %o2 L3.17: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits 1 bl L4.17 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (1*2+1), %o2 L4.17: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (1*2-1), %o2 L1.16: ! remainder is negative addcc %o3,%o5,%o3 d439 2 a440 2 bl L2.15 srl %o5,1,%o5 d442 1 a442 1 subcc %o3,%o5,%o3 d444 2 a445 2 bl L3.15 srl %o5,1,%o5 d447 9 a455 35 subcc %o3,%o5,%o3 ! depth 4, accumulated bits -1 bl L4.15 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-1*2+1), %o2 L4.15: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-1*2-1), %o2 L3.15: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits -3 bl L4.13 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-3*2+1), %o2 L4.13: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-3*2-1), %o2 L2.15: ! remainder is negative addcc %o3,%o5,%o3 d457 2 a458 2 bl L3.13 srl %o5,1,%o5 d460 7 a466 33 subcc %o3,%o5,%o3 ! depth 4, accumulated bits -5 bl L4.11 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-5*2+1), %o2 L4.11: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-5*2-1), %o2 L3.13: ! remainder is negative addcc %o3,%o5,%o3 ! depth 4, accumulated bits -7 bl L4.9 srl %o5,1,%o5 ! remainder is positive subcc %o3,%o5,%o3 b 9f add %o2, (-7*2+1), %o2 L4.9: ! remainder is negative addcc %o3,%o5,%o3 b 9f add %o2, (-7*2-1), %o2 9: d468 1 a468 1 subcc %o4, 1, %o4 d470 5 a474 5 tst %o3 bl,a got_result ! non-restoring fixup here (one instruction only!) add %o3, %o1, %o3 d476 9 a484 7 ! check to see if answer should be < 0 tst %g3 bl,a 1f sub %g0, %o3, %o3 1: retl mov %o3, %o0 a485 1 #endif @