summaryrefslogtreecommitdiff
path: root/libc/sysdeps/linux/sparc/sparcv7/urem.S
diff options
context:
space:
mode:
authorMike Frysinger <vapier@gentoo.org>2005-08-25 23:50:55 +0000
committerMike Frysinger <vapier@gentoo.org>2005-08-25 23:50:55 +0000
commit46b06e9c1c7d57ae0008b3a282110d8ce4345b5f (patch)
tree8bb70d81ee67ecd175ad14778b2513e4496e3b9e /libc/sysdeps/linux/sparc/sparcv7/urem.S
parent4147ffbcc6e2c20ab9bc1b43330244abe5a10b63 (diff)
import different optimized versions of div funcs based upon target sparc arch
Diffstat (limited to 'libc/sysdeps/linux/sparc/sparcv7/urem.S')
-rw-r--r--libc/sysdeps/linux/sparc/sparcv7/urem.S343
1 files changed, 343 insertions, 0 deletions
diff --git a/libc/sysdeps/linux/sparc/sparcv7/urem.S b/libc/sysdeps/linux/sparc/sparcv7/urem.S
new file mode 100644
index 000000000..ecf34672a
--- /dev/null
+++ b/libc/sysdeps/linux/sparc/sparcv7/urem.S
@@ -0,0 +1,343 @@
+ /* 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:
+ * .urem name of function to generate
+ * rem rem=div => %o0 / %o1; rem=rem => %o0 % %o1
+ * false false=true => signed; false=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.
+ */
+
+
+
+ENTRY(.urem)
+
+ ! 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 ST_DIV0
+ retl
+ clr %o0
+
+1:
+ cmp %o3, %o5 ! if %o1 exceeds %o0, done
+ blu LOC(got_result) ! (and algorithm fails otherwise)
+ clr %o2
+ sethi %hi(1 << (32 - 4 - 1)), %g1
+ cmp %o3, %g1
+ blu LOC(not_really_big)
+ 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 LOC(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 LOC(do_single_div)
+ sub %g2, 1, %g2
+
+ LOC(not_too_big):
+ 3: cmp %o5, %o3
+ blu 2b
+ nop
+ be LOC(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
+ ! 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...
+ LOC(do_single_div):
+ subcc %g2, 1, %g2
+ bl LOC(end_regular_divide)
+ nop
+ sub %o3, %o5, %o3
+ mov 1, %o2
+ b LOC(end_single_divloop)
+ nop
+ LOC(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:
+ LOC(end_single_divloop):
+ subcc %g2, 1, %g2
+ bge LOC(single_divloop)
+ tst %o3
+ b,a LOC(end_regular_divide)
+
+LOC(not_really_big):
+1:
+ sll %o5, 4, %o5
+ cmp %o5, %o3
+ bleu 1b
+ addcc %o4, 1, %o4
+ be LOC(got_result)
+ sub %o4, 1, %o4
+
+ tst %o3 ! set up for initial iteration
+LOC(divloop):
+ sll %o2, 4, %o2
+ ! depth 1, accumulated bits 0
+ bl LOC(1.16)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 2, accumulated bits 1
+ bl LOC(2.17)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 3, accumulated bits 3
+ bl LOC(3.19)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 4, accumulated bits 7
+ bl LOC(4.23)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (7*2+1), %o2
+
+LOC(4.23):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (7*2-1), %o2
+
+
+LOC(3.19):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 4, accumulated bits 5
+ bl LOC(4.21)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (5*2+1), %o2
+
+LOC(4.21):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (5*2-1), %o2
+
+
+
+LOC(2.17):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 3, accumulated bits 1
+ bl LOC(3.17)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 4, accumulated bits 3
+ bl LOC(4.19)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (3*2+1), %o2
+
+LOC(4.19):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (3*2-1), %o2
+
+
+LOC(3.17):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 4, accumulated bits 1
+ bl LOC(4.17)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (1*2+1), %o2
+
+LOC(4.17):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (1*2-1), %o2
+
+
+
+
+LOC(1.16):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 2, accumulated bits -1
+ bl LOC(2.15)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 3, accumulated bits -1
+ bl LOC(3.15)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 4, accumulated bits -1
+ bl LOC(4.15)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-1*2+1), %o2
+
+LOC(4.15):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-1*2-1), %o2
+
+
+LOC(3.15):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 4, accumulated bits -3
+ bl LOC(4.13)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-3*2+1), %o2
+
+LOC(4.13):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-3*2-1), %o2
+
+
+
+LOC(2.15):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 3, accumulated bits -3
+ bl LOC(3.13)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ ! depth 4, accumulated bits -5
+ bl LOC(4.11)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-5*2+1), %o2
+
+LOC(4.11):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-5*2-1), %o2
+
+
+LOC(3.13):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ ! depth 4, accumulated bits -7
+ bl LOC(4.9)
+ srl %o5,1,%o5
+ ! remainder is positive
+ subcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-7*2+1), %o2
+
+LOC(4.9):
+ ! remainder is negative
+ addcc %o3,%o5,%o3
+ b 9f
+ add %o2, (-7*2-1), %o2
+
+
+
+
+ 9:
+LOC(end_regular_divide):
+ subcc %o4, 1, %o4
+ bge LOC(divloop)
+ tst %o3
+ bl,a LOC(got_result)
+ ! non-restoring fixup here (one instruction only!)
+ add %o3, %o1, %o3
+
+
+LOC(got_result):
+
+ retl
+ mov %o3, %o0
+
+END(.urem)