1 1.1 is # 2 1.1 is # $NetBSD: ilsp.doc,v 1.1 2000/04/14 20:24:39 is Exp $ 3 1.1 is # 4 1.1 is 5 1.1 is #~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 6 1.1 is # MOTOROLA MICROPROCESSOR & MEMORY TECHNOLOGY GROUP 7 1.1 is # M68000 Hi-Performance Microprocessor Division 8 1.1 is # M68060 Software Package Production Release 9 1.1 is # 10 1.1 is # M68060 Software Package Copyright (C) 1993, 1994, 1995, 1996 Motorola Inc. 11 1.1 is # All rights reserved. 12 1.1 is # 13 1.1 is # THE SOFTWARE is provided on an "AS IS" basis and without warranty. 14 1.1 is # To the maximum extent permitted by applicable law, 15 1.1 is # MOTOROLA DISCLAIMS ALL WARRANTIES WHETHER EXPRESS OR IMPLIED, 16 1.1 is # INCLUDING IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS 17 1.1 is # FOR A PARTICULAR PURPOSE and any warranty against infringement with 18 1.1 is # regard to the SOFTWARE (INCLUDING ANY MODIFIED VERSIONS THEREOF) 19 1.1 is # and any accompanying written materials. 20 1.1 is # 21 1.1 is # To the maximum extent permitted by applicable law, 22 1.1 is # IN NO EVENT SHALL MOTOROLA BE LIABLE FOR ANY DAMAGES WHATSOEVER 23 1.1 is # (INCLUDING WITHOUT LIMITATION, DAMAGES FOR LOSS OF BUSINESS PROFITS, 24 1.1 is # BUSINESS INTERRUPTION, LOSS OF BUSINESS INFORMATION, OR OTHER PECUNIARY LOSS) 25 1.1 is # ARISING OF THE USE OR INABILITY TO USE THE SOFTWARE. 26 1.1 is # 27 1.1 is # Motorola assumes no responsibility for the maintenance and support 28 1.1 is # of the SOFTWARE. 29 1.1 is # 30 1.1 is # You are hereby granted a copyright license to use, modify, and distribute the 31 1.1 is # SOFTWARE so long as this entire notice is retained without alteration 32 1.1 is # in any modified and/or redistributed versions, and that such modified 33 1.1 is # versions are clearly identified as such. 34 1.1 is # No licenses are granted by implication, estoppel or otherwise under any 35 1.1 is # patents or trademarks of Motorola, Inc. 36 1.1 is #~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 37 1.1 is 38 1.1 is 68060 INTEGER SOFTWARE PACKAGE (Library version) 39 1.1 is ------------------------------------------------- 40 1.1 is 41 1.1 is The file ilsp.s contains the "Library version" of the 42 1.1 is 68060 Integer Software Package. Routines included in this 43 1.1 is module can be used to emulate 64-bit divide and multiply, 44 1.1 is and the "cmp2" instruction. These instructions are not 45 1.1 is implemented in hardware on the 68060 and normally take 46 1.1 is exception vector #61 "Unimplemented Integer Instruction". 47 1.1 is 48 1.1 is By re-compiling a program that uses these instructions, and 49 1.1 is making subroutine calls in place of the unimplemented 50 1.1 is instructions, a program can avoid the overhead associated with 51 1.1 is taking the exception. 52 1.1 is 53 1.1 is Release file format: 54 1.1 is -------------------- 55 1.1 is The file ilsp.sa is essentially a hexadecimal image of the 56 1.1 is release package. This is the ONLY format which will be supported. 57 1.1 is The hex image was created by assembling the source code and 58 1.1 is then converting the resulting binary output image into an 59 1.1 is ASCII text file. The hexadecimal numbers are listed 60 1.1 is using the Motorola Assembly Syntax assembler directive "dc.l" 61 1.1 is (define constant longword). The file can be converted to other 62 1.1 is assembly syntaxes by using any word processor with a global 63 1.1 is search and replace function. 64 1.1 is 65 1.1 is To assist in assembling and linking this module with other modules, 66 1.1 is the installer should add a symbolic label to the top of the file. 67 1.1 is This will allow calling routines to access the entry points 68 1.1 is of this package. 69 1.1 is 70 1.1 is The source code ilsp.s has also been included but only for 71 1.1 is documentation purposes. 72 1.1 is 73 1.1 is Release file structure: 74 1.1 is ----------------------- 75 1.1 is The file ilsp.sa contains an "Entry-Point" section and a 76 1.1 is code section. The ILSP has no "Call-Out" section. The first section 77 1.1 is is the "Entry-Point" section. In order to access a function in the 78 1.1 is package, a program must "bsr" or "jsr" to the location listed 79 1.1 is below in "68060ILSP Entry Points" that corresponds to the desired 80 1.1 is function. A branch instruction located at the selected entry point 81 1.1 is within the package will then enter the correct emulation code routine. 82 1.1 is 83 1.1 is The entry point addresses at the beginning of the package will remain 84 1.1 is fixed so that a program calling the routines will not have to be 85 1.1 is re-compiled with every new 68060ILSP release. 86 1.1 is 87 1.1 is For example, to use a 64-bit multiply instruction, 88 1.1 is do a "bsr" or "jsr" to the entry point defined by 89 1.1 is the 060ILSP entry table. A compiler generated code sequence 90 1.1 is for unsigned multiply could look like: 91 1.1 is 92 1.1 is # mulu.l <ea>,Dh:Dl 93 1.1 is # mulu.l _multiplier,%d1:%d0 94 1.1 is 95 1.1 is subq.l &0x8,%sp # make room for result on stack 96 1.1 is pea (%sp) # pass: result addr on stack 97 1.1 is mov.l %d0,-(%sp) # pass: multiplicand on stack 98 1.1 is mov.l _multiplier,-(%sp) # pass: multiplier on stack 99 1.1 is bsr.l _060LISP_TOP+0x18 # branch to multiply routine 100 1.1 is add.l &0xc,%sp # clear arguments from stack 101 1.1 is mov.l (%sp)+,%d1 # load result[63:32] 102 1.1 is mov.l (%sp)+,%d0 # load result[31:0] 103 1.1 is 104 1.1 is For a divide: 105 1.1 is 106 1.1 is # divu.l <ea>,Dr:Dq 107 1.1 is # divu.l _divisor,%d1:%d0 108 1.1 is 109 1.1 is subq.l &0x8,%sp # make room for result on stack 110 1.1 is pea (%sp) # pass: result addr on stack 111 1.1 is mov.l %d0,-(%sp) # pass: dividend hi on stack 112 1.1 is mov.l %d1,-(%sp) # pass: dividend hi on stack 113 1.1 is mov.l _divisor,-(%sp) # pass: divisor on stack 114 1.1 is bsr.l _060LISP_TOP+0x08 # branch to divide routine 115 1.1 is add.l &0xc,%sp # clear arguments from stack 116 1.1 is mov.l (%sp)+,%d1 # load remainder 117 1.1 is mov.l (%sp)+,%d0 # load quotient 118 1.1 is 119 1.1 is The library routines also return the correct condition code 120 1.1 is register value. If this is important, then the caller of the library 121 1.1 is routine must make sure that the value isn't lost while popping 122 1.1 is other items off of the stack. 123 1.1 is 124 1.1 is An example of using the "cmp2" instruction is as follows: 125 1.1 is 126 1.1 is # cmp2.l <ea>,Rn 127 1.1 is # cmp2.l _bounds,%d0 128 1.1 is 129 1.1 is pea _bounds # pass ptr to bounds 130 1.1 is mov.l %d0,-(%sp) # pass Rn 131 1.1 is bsr.l _060LSP_TOP_+0x48 # branch to "cmp2" routine 132 1.1 is mov.w %cc,_tmp # save off condition codes 133 1.1 is addq.l &0x8,%sp # clear arguments from stack 134 1.1 is 135 1.1 is Exception reporting: 136 1.1 is -------------------- 137 1.1 is If the instruction being emulated is a divide and the source 138 1.1 is operand is a zero, then the library routine, as it's last 139 1.1 is instruction, executes an implemented divide using a zero 140 1.1 is source operand so that an "Integer Divide-by-Zero" exception 141 1.1 is will be taken. Although the exception stack frame will not 142 1.1 is point to the correct instruction, the user will at least be able 143 1.1 is to record that such an event occurred if desired. 144 1.1 is 145 1.1 is 68060ILSP entry points: 146 1.1 is ----------------------- 147 1.1 is _060ILSP_TOP: 148 1.1 is 0x000: _060LSP__idivs64_ 149 1.1 is 0x008: _060LSP__idivu64_ 150 1.1 is 151 1.1 is 0x010: _060LSP__imuls64_ 152 1.1 is 0x018: _060LSP__imulu64_ 153 1.1 is 154 1.1 is 0x020: _060LSP__cmp2_Ab_ 155 1.1 is 0x028: _060LSP__cmp2_Aw_ 156 1.1 is 0x030: _060LSP__cmp2_Al_ 157 1.1 is 0x038: _060LSP__cmp2_Db_ 158 1.1 is 0x040: _060LSP__cmp2_Dw_ 159 1.1 is 0x048: _060LSP__cmp2_Dl_ 160