[BACK]Return to cross.pl CVS log [TXT][DIR] Up to [local] / OpenXM_contrib / gmp / mpn / x86 / k6

Diff for /OpenXM_contrib/gmp/mpn/x86/k6/Attic/cross.pl between version 1.1.1.1 and 1.1.1.2

version 1.1.1.1, 2000/09/09 14:12:42 version 1.1.1.2, 2003/08/25 16:06:27
Line 1 
Line 1 
 #! /usr/bin/perl  #! /usr/bin/perl
   
 # Copyright (C) 2000 Free Software Foundation, Inc.  # Copyright 2000, 2001 Free Software Foundation, Inc.
 #  #
 # This file is part of the GNU MP Library.  # This file is part of the GNU MP Library.
 #  #
Line 29 
Line 29 
 # With no arguments, all .o files corresponding to .asm files are processed.  # With no arguments, all .o files corresponding to .asm files are processed.
 # This is good in the mpn object directory of a k6*-*-* build.  # This is good in the mpn object directory of a k6*-*-* build.
 #  #
 # As far as fixing problems goes, any cache line crossing problems in loops  # Code alignments of 8 bytes or more are handled.  When 32 is used, cache
 # get attention, but as a rule it's too tedious to rearrange code or slip in  # line boundaries will fall in at offsets 0x20,0x40,etc and problems are
 # nops to fix every problem in setup or finishup code.  # flagged at those locations.  When 16 is used, the line boundaries can also
   # fall at offsets 0x10,0x30,0x50,etc, depending where the file is loaded, so
   # problems are identified there too.  Likewise when 8 byte alignment is used
   # problems are flagged additionally at 0x08,0x18,0x28,etc.
 #  #
   # Usually 32 byte alignment is used for k6 routines, but less is certainly
   # possible if through good luck, or a little tweaking, cache line crossing
   # problems can be avoided at the extra locations.
   #
 # Bugs:  # Bugs:
 #  #
 # Instructions without mod/rm bytes or which are already vector decoded are  # Instructions without mod/rm bytes or which are already vector decoded are
Line 41 
Line 48 
 # though.  # though.
 #  #
 # There's no messages for using the vector decoded addressing mode (%esi),  # There's no messages for using the vector decoded addressing mode (%esi),
 # but that mode is easy to avoid when coding.  # but that's easy to avoid when coding.
   #
   # Future:
   #
   # Warn about jump targets that are poorly aligned (less than 2 instructions
   # before a cache line boundary).
   
 use strict;  use strict;
   
 sub disassemble {  sub disassemble {
     my ($file) = @_;      my ($file) = @_;
     my ($addr,$b1,$b2,$b3, $prefix,$opcode,$modrm);      my ($addr,$b1,$b2,$b3, $prefix,$opcode,$modrm);
       my $align;
   
     open (IN, "objdump -Srfh $file |")      open (IN, "objdump -Srfh $file |")
         || die "Cannot open pipe from objdump\n";          || die "Cannot open pipe from objdump\n";
Line 55  sub disassemble {
Line 68  sub disassemble {
         print;          print;
   
         if (/^[ \t]*[0-9]+[ \t]+\.text[ \t]/ && /2\*\*([0-9]+)$/) {          if (/^[ \t]*[0-9]+[ \t]+\.text[ \t]/ && /2\*\*([0-9]+)$/) {
             if ($1 < 5) {              $align = 1 << $1;
                 print "ZZ need at least 2**5 for predictable cache line crossing\n";              if ($align < 8) {
                   print "ZZ cross.pl cannot handle alignment < 2**3\n";
                   $align = 8
             }              }
         }          }
   
Line 88  sub disassemble {
Line 103  sub disassemble {
         if ($prefix =~ /0f/          if ($prefix =~ /0f/
             && $opcode !~ /^8/         # jcond disp32              && $opcode !~ /^8/         # jcond disp32
             && $modrm =~ /^[0-3][4c]/) {              && $modrm =~ /^[0-3][4c]/) {
             print "ZZ ($file) >3 bytes to determine instruction length\n";              print "ZZ ($file) >3 bytes to determine instruction length [K6]\n";
         }          }
   
         # with just an opcode, starting 1f mod 20h          # with just an opcode, starting 1f mod 20h
         if ($addr =~ /[13579bdf]f$/          if (($align==32 && $addr =~ /[13579bdf]f$/
                || $align==16 && $addr =~ /f$/
                || $align==8 && $addr =~ /[7f]$/)
             && $prefix !~ /0f/              && $prefix !~ /0f/
             && $opcode !~ /1[012345]/ # adc              && $opcode !~ /1[012345]/ # adc
             && $opcode !~ /1[89abcd]/ # sbb              && $opcode !~ /1[89abcd]/ # sbb
               && $opcode !~ /^4/        # inc/dec reg
               && $opcode !~ /^5/        # push/pop reg
             && $opcode !~ /68/        # push $imm32              && $opcode !~ /68/        # push $imm32
             && $opcode !~ /^7/        # jcond disp8              && $opcode !~ /^7/        # jcond disp8
             && $opcode !~ /a[89]/     # test+imm              && $opcode !~ /a[89]/     # test+imm
             && $opcode !~ /a[a-f]/    # stos/lods/scas              && $opcode !~ /a[a-f]/    # stos/lods/scas
             && $opcode !~ /b8/        # movl $imm32,%eax              && $opcode !~ /b8/        # movl $imm32,%eax
               && $opcode !~ /d[0123]/   # rcl
             && $opcode !~ /e[0123]/   # loop/loopz/loopnz/jcxz              && $opcode !~ /e[0123]/   # loop/loopz/loopnz/jcxz
             && $opcode !~ /e[b9]/     # jmp disp8/disp32              && $opcode !~ /e8/        # call disp32
               && $opcode !~ /e[9b]/     # jmp disp32/disp8
             && $opcode !~ /f[89abcd]/ # clc,stc,cli,sti,cld,std              && $opcode !~ /f[89abcd]/ # clc,stc,cli,sti,cld,std
             && !($opcode =~ /f[67]/          # grp 1              && !($opcode =~ /f[67]/          # grp 1
                  && $modrm =~ /^[2367abef]/) # mul, imul, div, idiv                   && $modrm =~ /^[2367abef]/) # mul, imul, div, idiv
Line 111  sub disassemble {
Line 132  sub disassemble {
         }          }
   
         # with an 0F prefix, anything starting at 1f mod 20h          # with an 0F prefix, anything starting at 1f mod 20h
         if ($addr =~ /[13579bdf][f]$/          if (($align==32 && $addr =~ /[13579bdf][f]$/
             && $prefix =~ /0f/) {               || $align==16 && $addr =~ /f$/
                || $align==8 && $addr =~ /[7f]$/)
               && $prefix =~ /0f/
               && $opcode !~ /af/        # imul
               && $opcode !~ /a[45]/     # shldl
               && $opcode !~ /a[cd]/     # shrdl
               ) {
             print "ZZ ($file) prefix/opcode cross 32-byte boundary\n";              print "ZZ ($file) prefix/opcode cross 32-byte boundary\n";
         }          }
   
         # with an 0F prefix, anything with mod/rm starting at 1e mod 20h          # with an 0F prefix, anything with mod/rm starting at 1e mod 20h
         if ($addr =~ /[13579bdf][e]$/          if (($align==32 && $addr =~ /[13579bdf][e]$/
                || $align==16 && $addr =~ /[e]$/
                || $align==8 && $addr =~ /[6e]$/)
             && $prefix =~ /0f/              && $prefix =~ /0f/
              && $opcode !~ /^8/        # jcond disp32               && $opcode !~ /^8/        # jcond disp32
                && $opcode !~ /af/        # imull reg,reg
                && $opcode !~ /a[45]/     # shldl
                && $opcode !~ /a[cd]/     # shrdl
             && $modrm !~ /^$/) {              && $modrm !~ /^$/) {
             print "ZZ ($file) prefix/opcode/modrm cross 32-byte boundary\n";              print "ZZ ($file) prefix/opcode/modrm cross 32-byte boundary\n";
         }          }
Legend:
Removed from v.1.1.1.1  
changed lines
  Added in v.1.1.1.2
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>