1#! /usr/bin/perl 2 3# Copyright 2000, 2001 Free Software Foundation, Inc. 4# 5# This file is part of the GNU MP Library. 6# 7# The GNU MP Library is free software; you can redistribute it and/or modify 8# it under the terms of either: 9# 10# * the GNU Lesser General Public License as published by the Free 11# Software Foundation; either version 3 of the License, or (at your 12# option) any later version. 13# 14# or 15# 16# * the GNU General Public License as published by the Free Software 17# Foundation; either version 2 of the License, or (at your option) any 18# later version. 19# 20# or both in parallel, as here. 21# 22# The GNU MP Library is distributed in the hope that it will be useful, but 23# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 24# or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 25# for more details. 26# 27# You should have received copies of the GNU General Public License and the 28# GNU Lesser General Public License along with the GNU MP Library. If not, 29# see https://www.gnu.org/licenses/. 30 31 32# Usage: cross.pl [filename.o]... 33# 34# Produce an annotated disassembly of the given object files, indicating 35# certain code alignment and addressing mode problems afflicting K6 chips. 36# "ZZ" is used on all annotations, so this can be searched for. 37# 38# With no arguments, all .o files corresponding to .asm files are processed. 39# This is good in the mpn object directory of a k6*-*-* build. 40# 41# Code alignments of 8 bytes or more are handled. When 32 is used, cache 42# line boundaries will fall in at offsets 0x20,0x40,etc and problems are 43# flagged at those locations. When 16 is used, the line boundaries can also 44# fall at offsets 0x10,0x30,0x50,etc, depending where the file is loaded, so 45# problems are identified there too. Likewise when 8 byte alignment is used 46# problems are flagged additionally at 0x08,0x18,0x28,etc. 47# 48# Usually 32 byte alignment is used for k6 routines, but less is certainly 49# possible if through good luck, or a little tweaking, cache line crossing 50# problems can be avoided at the extra locations. 51# 52# Bugs: 53# 54# Instructions without mod/rm bytes or which are already vector decoded are 55# unaffected by cache line boundary crossing, but not all of these have yet 56# been put in as exceptions. All that occur in practice in GMP are present 57# though. 58# 59# There's no messages for using the vector decoded addressing mode (%esi), 60# but that's easy to avoid when coding. 61# 62# Future: 63# 64# Warn about jump targets that are poorly aligned (less than 2 instructions 65# before a cache line boundary). 66 67use strict; 68 69sub disassemble { 70 my ($file) = @_; 71 my ($addr,$b1,$b2,$b3, $prefix,$opcode,$modrm); 72 my $align; 73 74 open (IN, "objdump -Srfh $file |") 75 || die "Cannot open pipe from objdump\n"; 76 while (<IN>) { 77 print; 78 79 if (/^[ \t]*[0-9]+[ \t]+\.text[ \t]/ && /2\*\*([0-9]+)$/) { 80 $align = 1 << $1; 81 if ($align < 8) { 82 print "ZZ cross.pl cannot handle alignment < 2**3\n"; 83 $align = 8 84 } 85 } 86 87 if (/^[ \t]*([0-9a-f]*):[ \t]*([0-9a-f]+)[ \t]+([0-9a-f]+)[ \t]+([0-9a-f]+)/) { 88 ($addr,$b1,$b2,$b3) = ($1,$2,$3,$4); 89 90 } elsif (/^[ \t]*([0-9a-f]*):[ \t]*([0-9a-f]+)[ \t]+([0-9a-f]+)/) { 91 ($addr,$b1,$b2,$b3) = ($1,$2,$3,''); 92 93 } elsif (/^[ \t]*([0-9a-f]*):[ \t]*([0-9a-f]+)/) { 94 ($addr,$b1,$b2,$b3) = ($1,$2,'',''); 95 96 } else { 97 next; 98 } 99 100 if ($b1 =~ /0f/) { 101 $prefix = $b1; 102 $opcode = $b2; 103 $modrm = $b3; 104 } else { 105 $prefix = ''; 106 $opcode = $b1; 107 $modrm = $b2; 108 } 109 110 # modrm of the form 00-xxx-100 with an 0F prefix is the problem case 111 # for K6 and pre-CXT K6-2 112 if ($prefix =~ /0f/ 113 && $opcode !~ /^8/ # jcond disp32 114 && $modrm =~ /^[0-3][4c]/) { 115 print "ZZ ($file) >3 bytes to determine instruction length [K6]\n"; 116 } 117 118 # with just an opcode, starting 1f mod 20h 119 if (($align==32 && $addr =~ /[13579bdf]f$/ 120 || $align==16 && $addr =~ /f$/ 121 || $align==8 && $addr =~ /[7f]$/) 122 && $prefix !~ /0f/ 123 && $opcode !~ /1[012345]/ # adc 124 && $opcode !~ /1[89abcd]/ # sbb 125 && $opcode !~ /^4/ # inc/dec reg 126 && $opcode !~ /^5/ # push/pop reg 127 && $opcode !~ /68/ # push $imm32 128 && $opcode !~ /^7/ # jcond disp8 129 && $opcode !~ /a[89]/ # test+imm 130 && $opcode !~ /a[a-f]/ # stos/lods/scas 131 && $opcode !~ /b8/ # movl $imm32,%eax 132 && $opcode !~ /d[0123]/ # rcl 133 && $opcode !~ /e[0123]/ # loop/loopz/loopnz/jcxz 134 && $opcode !~ /e8/ # call disp32 135 && $opcode !~ /e[9b]/ # jmp disp32/disp8 136 && $opcode !~ /f[89abcd]/ # clc,stc,cli,sti,cld,std 137 && !($opcode =~ /f[67]/ # grp 1 138 && $modrm =~ /^[2367abef]/) # mul, imul, div, idiv 139 && $modrm !~ /^$/) { 140 print "ZZ ($file) opcode/modrm cross 32-byte boundary\n"; 141 } 142 143 # with an 0F prefix, anything starting at 1f mod 20h 144 if (($align==32 && $addr =~ /[13579bdf][f]$/ 145 || $align==16 && $addr =~ /f$/ 146 || $align==8 && $addr =~ /[7f]$/) 147 && $prefix =~ /0f/ 148 && $opcode !~ /af/ # imul 149 && $opcode !~ /a[45]/ # shldl 150 && $opcode !~ /a[cd]/ # shrdl 151 ) { 152 print "ZZ ($file) prefix/opcode cross 32-byte boundary\n"; 153 } 154 155 # with an 0F prefix, anything with mod/rm starting at 1e mod 20h 156 if (($align==32 && $addr =~ /[13579bdf][e]$/ 157 || $align==16 && $addr =~ /[e]$/ 158 || $align==8 && $addr =~ /[6e]$/) 159 && $prefix =~ /0f/ 160 && $opcode !~ /^8/ # jcond disp32 161 && $opcode !~ /af/ # imull reg,reg 162 && $opcode !~ /a[45]/ # shldl 163 && $opcode !~ /a[cd]/ # shrdl 164 && $modrm !~ /^$/) { 165 print "ZZ ($file) prefix/opcode/modrm cross 32-byte boundary\n"; 166 } 167 } 168 close IN || die "Error from objdump (or objdump not available)\n"; 169} 170 171 172my @files; 173if ($#ARGV >= 0) { 174 @files = @ARGV; 175} else { 176 @files = glob "*.asm"; 177 map {s/.asm/.o/} @files; 178} 179 180foreach (@files) { 181 disassemble($_); 182} 183