Beginnings of experimental support for NASM assembler. This is a free
authorDr. Stephen Henson <steve@openssl.org>
Mon, 12 Jul 1999 12:33:16 +0000 (12:33 +0000)
committerDr. Stephen Henson <steve@openssl.org>
Mon, 12 Jul 1999 12:33:16 +0000 (12:33 +0000)
assembler for various X86 platforms including Win32. It can output object files
that VC++ will tolerate so it could be used to provide assembly language support
to Win32 without the need for MASM.

This is preliminary stuff: it doesn't even work yet.

crypto/perlasm/x86asm.pl
crypto/perlasm/x86nasm.pl [new file with mode: 0644]

index bbbc1cb91a298f48e25f6f1c36248407ddbc0b74..44e330eb8d16d145a88dfb3bc1e8c75608178db8 100644 (file)
@@ -29,6 +29,8 @@ sub main'asm_init
                { $cpp=1; require "x86unix.pl"; }
        elsif ( ($type eq "win32"))
                { $win32=1; require "x86ms.pl"; }
+       elsif ( ($type eq "win32n"))
+               { $win32=1; require "x86nasm.pl"; }
        else
                {
                print STDERR <<"EOF";
@@ -38,6 +40,7 @@ Pick one target type from
        sol     - x86 solaris
        cpp     - format so x86unix.cpp can be used
        win32   - Windows 95/Windows NT
+       win32n  - Windows 95/Windows NT NASM format
 EOF
                exit(1);
                }
diff --git a/crypto/perlasm/x86nasm.pl b/crypto/perlasm/x86nasm.pl
new file mode 100644 (file)
index 0000000..3aab34a
--- /dev/null
@@ -0,0 +1,325 @@
+#!/usr/local/bin/perl
+
+package x86nasm;
+
+$label="L000";
+
+%lb=(  'eax',  'al',
+       'ebx',  'bl',
+       'ecx',  'cl',
+       'edx',  'dl',
+       'ax',   'al',
+       'bx',   'bl',
+       'cx',   'cl',
+       'dx',   'dl',
+       );
+
+%hb=(  'eax',  'ah',
+       'ebx',  'bh',
+       'ecx',  'ch',
+       'edx',  'dh',
+       'ax',   'ah',
+       'bx',   'bh',
+       'cx',   'ch',
+       'dx',   'dh',
+       );
+
+sub main'asm_init_output { @out=(); }
+sub main'asm_get_output { return(@out); }
+sub main'get_labels { return(@labels); }
+sub main'external_label { push(@labels,@_); push(@out, "EXTERN\t$_[0]"); }
+
+sub main'LB
+       {
+       (defined($lb{$_[0]})) || die "$_[0] does not have a 'low byte'\n";
+       return($lb{$_[0]});
+       }
+
+sub main'HB
+       {
+       (defined($hb{$_[0]})) || die "$_[0] does not have a 'high byte'\n";
+       return($hb{$_[0]});
+       }
+
+sub main'BP
+       {
+       &get_mem("BYTE",@_);
+       }
+
+sub main'DWP
+       {
+       &get_mem("DWORD",@_);
+       }
+
+sub main'stack_push
+       {
+       my($num)=@_;
+       $stack+=$num*4;
+       &main'sub("esp",$num*4);
+       }
+
+sub main'stack_pop
+       {
+       my($num)=@_;
+       $stack-=$num*4;
+       &main'add("esp",$num*4);
+       }
+
+sub get_mem
+       {
+       my($size,$addr,$reg1,$reg2,$idx)=@_;
+       my($t,$post);
+       my($ret)="[$size ";
+       $addr =~ s/^\s+//;
+       if ($addr =~ /^(.+)\+(.+)$/)
+               {
+               $reg2=&conv($1);
+               $addr="_$2";
+               }
+       elsif ($addr =~ /^[_a-zA-Z]/)
+               {
+               $addr="_$addr";
+               }
+
+       $reg1="$regs{$reg1}" if defined($regs{$reg1});
+       $reg2="$regs{$reg2}" if defined($regs{$reg2});
+       if (($addr ne "") && ($addr ne 0))
+               {
+               if ($addr !~ /^-/)
+                       { $ret.="${addr}+"; }
+               else    { $post=$addr; }
+               }
+       if ($reg2 ne "")
+               {
+               $t="";
+               $t="*$idx" if ($idx != 0);
+               $reg1="+".$reg1 if ("$reg1$post" ne "");
+               $ret.="$reg2$t$reg1$post]";
+               }
+       else
+               {
+               $ret.="$reg1$post]"
+               }
+       return($ret);
+       }
+
+sub main'mov   { &out2("mov",@_); }
+sub main'movb  { &out2("mov",@_); }
+sub main'and   { &out2("and",@_); }
+sub main'or    { &out2("or",@_); }
+sub main'shl   { &out2("shl",@_); }
+sub main'shr   { &out2("shr",@_); }
+sub main'xor   { &out2("xor",@_); }
+sub main'xorb  { &out2("xor",@_); }
+sub main'add   { &out2("add",@_); }
+sub main'adc   { &out2("adc",@_); }
+sub main'sub   { &out2("sub",@_); }
+sub main'rotl  { &out2("rol",@_); }
+sub main'rotr  { &out2("ror",@_); }
+sub main'exch  { &out2("xchg",@_); }
+sub main'cmp   { &out2("cmp",@_); }
+sub main'lea   { &out2("lea",@_); }
+sub main'mul   { &out1("mul",@_); }
+sub main'div   { &out1("div",@_); }
+sub main'dec   { &out1("dec",@_); }
+sub main'inc   { &out1("inc",@_); }
+sub main'jmp   { &out1("jmp",@_); }
+sub main'jmp_ptr { &out1p("jmp",@_); }
+
+# This is a bit of a kludge: declare all branches as NEAR.
+sub main'je    { &out1("je NEAR",@_); }
+sub main'jle   { &out1("jle NEAR",@_); }
+sub main'jz    { &out1("jz NEAR",@_); }
+sub main'jge   { &out1("jge NEAR",@_); }
+sub main'jl    { &out1("jl NEAR",@_); }
+sub main'jb    { &out1("jb NEAR",@_); }
+sub main'jc    { &out1("jc NEAR",@_); }
+sub main'jnc   { &out1("jnc NEAR",@_); }
+sub main'jnz   { &out1("jnz NEAR",@_); }
+sub main'jne   { &out1("jne NEAR",@_); }
+sub main'jno   { &out1("jno NEAR",@_); }
+
+sub main'push  { &out1("push",@_); $stack+=4; }
+sub main'pop   { &out1("pop",@_); $stack-=4; }
+sub main'bswap { &out1("bswap",@_); &using486(); }
+sub main'not   { &out1("not",@_); }
+sub main'call  { &out1("call",'_'.$_[0]); }
+sub main'ret   { &out0("ret"); }
+sub main'nop   { &out0("nop"); }
+
+sub out2
+       {
+       my($name,$p1,$p2)=@_;
+       my($l,$t);
+
+       push(@out,"\t$name\t");
+       $t=&conv($p1).",";
+       $l=length($t);
+       push(@out,$t);
+       $l=4-($l+9)/8;
+       push(@out,"\t" x $l);
+       push(@out,&conv($p2));
+       push(@out,"\n");
+       }
+
+sub out0
+       {
+       my($name)=@_;
+
+       push(@out,"\t$name\n");
+       }
+
+sub out1
+       {
+       my($name,$p1)=@_;
+       my($l,$t);
+       push(@out,"\t$name\t".&conv($p1)."\n");
+       }
+
+sub conv
+       {
+       my($p)=@_;
+       $p =~ s/0x([0-9A-Fa-f]+)/0$1h/;
+       return $p;
+       }
+
+sub using486
+       {
+       return if $using486;
+       $using486++;
+       grep(s/\.386/\.486/,@out);
+       }
+
+sub main'file
+       {
+       push(@out, "segment .text");
+       }
+
+sub main'function_begin
+       {
+       my($func,$extra)=@_;
+
+       push(@labels,$func);
+       my($tmp)=<<"EOF";
+GLOBAL _$func
+_$func:
+       push    ebp
+       push    ebx
+       push    esi
+       push    edi
+EOF
+       push(@out,$tmp);
+       $stack=20;
+       }
+
+sub main'function_begin_B
+       {
+       my($func,$extra)=@_;
+       my($tmp)=<<"EOF";
+GLOBAL _$func
+_$func:
+EOF
+       push(@out,$tmp);
+       $stack=4;
+       }
+
+sub main'function_end
+       {
+       my($func)=@_;
+
+       my($tmp)=<<"EOF";
+       pop     edi
+       pop     esi
+       pop     ebx
+       pop     ebp
+       ret
+EOF
+       push(@out,$tmp);
+       $stack=0;
+       %label=();
+       }
+
+sub main'function_end_B
+       {
+       $stack=0;
+       %label=();
+       }
+
+sub main'function_end_A
+       {
+       my($func)=@_;
+
+       my($tmp)=<<"EOF";
+       pop     edi
+       pop     esi
+       pop     ebx
+       pop     ebp
+       ret
+EOF
+       push(@out,$tmp);
+       }
+
+sub main'file_end
+       {
+       }
+
+sub main'wparam
+       {
+       my($num)=@_;
+
+       return(&main'DWP($stack+$num*4,"esp","",0));
+       }
+
+sub main'swtmp
+       {
+       return(&main'DWP($_[0]*4,"esp","",0));
+       }
+
+# Should use swtmp, which is above esp.  Linix can trash the stack above esp
+#sub main'wtmp
+#      {
+#      my($num)=@_;
+#
+#      return(&main'DWP(-(($num+1)*4),"esp","",0));
+#      }
+
+sub main'comment
+       {
+       foreach (@_)
+               {
+               push(@out,"\t; $_\n");
+               }
+       }
+
+sub main'label
+       {
+       if (!defined($label{$_[0]}))
+               {
+               $label{$_[0]}="\$${label}${_[0]}";
+               $label++;
+               }
+       return($label{$_[0]});
+       }
+
+sub main'set_label
+       {
+       if (!defined($label{$_[0]}))
+               {
+               $label{$_[0]}="${label}${_[0]}";
+               $label++;
+               }
+       push(@out,"$label{$_[0]}:\n");
+       }
+
+sub main'data_word
+       {
+       push(@out,"\tDD\t$_[0]\n");
+       }
+
+sub out1p
+       {
+       my($name,$p1)=@_;
+       my($l,$t);
+
+       push(@out,"\t$name\t ".&conv($p1)."\n");
+       }