summaryrefslogtreecommitdiff
path: root/arch/c6x/lib/divu.S
diff options
context:
space:
mode:
Diffstat (limited to 'arch/c6x/lib/divu.S')
-rw-r--r--arch/c6x/lib/divu.S98
1 files changed, 98 insertions, 0 deletions
diff --git a/arch/c6x/lib/divu.S b/arch/c6x/lib/divu.S
new file mode 100644
index 000000000..64af3c006
--- /dev/null
+++ b/arch/c6x/lib/divu.S
@@ -0,0 +1,98 @@
+;; Copyright 2010 Free Software Foundation, Inc.
+;; Contributed by Bernd Schmidt <bernds@codesourcery.com>.
+;;
+;; This program is free software; you can redistribute it and/or modify
+;; it under the terms of the GNU General Public License as published by
+;; the Free Software Foundation; either version 2 of the License, or
+;; (at your option) any later version.
+;;
+;; This program is distributed in the hope that it will be useful,
+;; but WITHOUT ANY WARRANTY; without even the implied warranty of
+;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+;; GNU General Public License for more details.
+;;
+;; You should have received a copy of the GNU General Public License
+;; along with this program; if not, write to the Free Software
+;; Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+#include <linux/linkage.h>
+
+ ;; ABI considerations for the divide functions
+ ;; The following registers are call-used:
+ ;; __c6xabi_divi A0,A1,A2,A4,A6,B0,B1,B2,B4,B5
+ ;; __c6xabi_divu A0,A1,A2,A4,A6,B0,B1,B2,B4
+ ;; __c6xabi_remi A1,A2,A4,A5,A6,B0,B1,B2,B4
+ ;; __c6xabi_remu A1,A4,A5,A7,B0,B1,B2,B4
+ ;;
+ ;; In our implementation, divu and remu are leaf functions,
+ ;; while both divi and remi call into divu.
+ ;; A0 is not clobbered by any of the functions.
+ ;; divu does not clobber B2 either, which is taken advantage of
+ ;; in remi.
+ ;; divi uses B5 to hold the original return address during
+ ;; the call to divu.
+ ;; remi uses B2 and A5 to hold the input values during the
+ ;; call to divu. It stores B3 in on the stack.
+
+ .text
+ENTRY(__c6xabi_divu)
+ ;; We use a series of up to 31 subc instructions. First, we find
+ ;; out how many leading zero bits there are in the divisor. This
+ ;; gives us both a shift count for aligning (shifting) the divisor
+ ;; to the, and the number of times we have to execute subc.
+
+ ;; At the end, we have both the remainder and most of the quotient
+ ;; in A4. The top bit of the quotient is computed first and is
+ ;; placed in A2.
+
+ ;; Return immediately if the dividend is zero.
+ mv .s2x A4, B1
+ [B1] lmbd .l2 1, B4, B1
+|| [!B1] b .s2 B3 ; RETURN A
+|| [!B1] mvk .d2 1, B4
+ mv .l1x B1, A6
+|| shl .s2 B4, B1, B4
+
+ ;; The loop performs a maximum of 28 steps, so we do the
+ ;; first 3 here.
+ cmpltu .l1x A4, B4, A2
+ [!A2] sub .l1x A4, B4, A4
+|| shru .s2 B4, 1, B4
+|| xor .s1 1, A2, A2
+
+ shl .s1 A2, 31, A2
+|| [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+
+ ;; RETURN A may happen here (note: must happen before the next branch)
+_divu_loop:
+ cmpgt .l2 B1, 7, B0
+|| [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+|| [B0] b .s1 _divu_loop
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ [B1] subc .l1x A4,B4,A4
+|| [B1] add .s2 -1, B1, B1
+ ;; loop backwards branch happens here
+
+ ret .s2 B3
+|| mvk .s1 32, A1
+ sub .l1 A1, A6, A6
+ shl .s1 A4, A6, A4
+ shru .s1 A4, 1, A4
+|| sub .l1 A6, 1, A6
+ or .l1 A2, A4, A4
+ shru .s1 A4, A6, A4
+ nop
+ENDPROC(__c6xabi_divu)