Home | History | Annotate | Line # | Download | only in gen
      1 /*	$NetBSD: muldi3.S,v 1.2 2020/05/31 12:37:07 rin Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 2020 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Rin Okuyama.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 #include <machine/asm.h>
     33 
     34 RCSID("$NetBSD: muldi3.S,v 1.2 2020/05/31 12:37:07 rin Exp $")
     35 
     36 | int64_t __muldi3(int64_t X, int64_t Y);
     37 |
     38 | * Return lower 64bit of (X * Y) into %d0:%d1.
     39 |
     40 | * Intended for 68060:
     41 |   - GCC does not emit __muldi3() for 68020-40, that have 32 * 32 --> 64 mulul.
     42 |   - mulsl (and moveml) are not implemented for 68010.
     43 |
     44 | * Notation:
     45 |   - H32:L32 --> higher:lower 32bit of variable
     46 |   - H:L     --> higher:lower 16bit of variable/register
     47 
     48 #ifdef __mc68010__
     49 #error "not for 68010"
     50 #endif
     51 
     52 #define X_H32 (4 * 4)
     53 #define X_L32 (X_H32 + 4)
     54 #define Y_H32 (X_L32 + 4)
     55 #define Y_L32 (Y_H32 + 4)
     56 
     57 ENTRY(__muldi3)
     58 	moveml	%d2-%d4, -(%sp)	| push %d2-%d4
     59 
     60 | First, calculate (X_L32 * Y_L32) as a 64bit integer.
     61 
     62 	movel	X_L32(%sp), %a0	| save X_L32
     63 	movel	Y_L32(%sp), %a1	| save Y_L32
     64 
     65 	movel	%a0, %d2	| prepare for X_L32(H) in L
     66 	movel	%a1, %d3	| prepare for Y_L32(H) in L
     67 
     68 	movel	%a0, %d4	| X_L32(L) in L
     69 	movel	%a1, %d1	| Y_L32(L) in L
     70 	movel	%a0, %d0	| X_L32(L) in L
     71 
     72 	swap	%d2		| X_L32(H) in L
     73 	swap	%d3 		| Y_L32(H) in L
     74 
     75 	muluw	%d1, %d4	| A = X_L32(L) * Y_L32(L)
     76 	muluw	%d2, %d1	| B = X_L32(H) * Y_L32(L)
     77 	muluw	%d3, %d2	| C = X_L32(H) * Y_L32(H)
     78 	muluw	%d0, %d3	| D = X_L32(L) * Y_L32(H)
     79 
     80 	movel	%d4, %d0	| extract A(H)
     81 	clrw	%d0
     82 	swap	%d0
     83 
     84 	addl	%d0, %d1	| B += A(H) (no carry; max 0xffff0000)
     85 
     86 	addl	%d3, %d1	| B += D
     87 	bccs	1f		| if (carry)
     88 	addil	#0x10000, %d2	| 	C += 0x10000
     89 
     90 1:	swap	%d1		| B(H) <--> B(L)
     91 
     92 | (%d0), (%d1), %d2 = C, %d3 = free, %d4 = A
     93 
     94 	clrl	%d3		| extract B(H)
     95 	movew	%d1, %d3
     96 
     97 	movew	%d4, %d1	| %d1 = (B(L) << 16) + A(L)
     98 
     99 	addl	%d3, %d2	| C += B(H)
    100 
    101 | We have (X_L32 * Y_L32) in %d2:%d1. Lower 32bit was completed.
    102 | Add (X_L32 * Y_H32 + X_H32 * Y_L32) to higher 32bit.
    103 |
    104 | (%d0), (%d1), %d2 = C, %d3 = free, %d4 = free
    105 
    106 	movel	%a0, %d0	| restore X_L32
    107 	movel	%a1, %d3	| restore Y_L32
    108 	mulsl	Y_H32(%sp), %d0	| E = X_L32 * Y_H32
    109 	mulsl	X_H32(%sp), %d3 | F = X_H32 * Y_L32
    110 	addl	%d2, %d0	| E += C
    111 	addl	%d3, %d0	| %d0 = E + F
    112 
    113 	moveml	(%sp)+, %d2-%d4	| pop %d2-%d4
    114 	rts
    115 END(__muldi3)
    116