Skip to content

R
riscv-gcc-1
Commit 3ec980b1 by Tom Tromey
Options

This commit was generated by cvs2svn to compensate for changes in r104181, 

which included commits to RCS files with non-trunk default branches.

From-SVN: r104182
parent 303ae446
Showing with 3906 additions and 3348 deletions
zlib/as400/compile.clp
......@@ -118,6 +118,6 @@
&MODLIB/INFTREES &MODLIB/TREES +
&MODLIB/UNCOMPR &MODLIB/ZUTIL) +
SRCFILE(&SRCLIB/&CTLFILE) SRCMBR(BNDSRC) +
TEXT('ZLIB 1.2.1') TGTRLS(V4R4M0)
TEXT('ZLIB 1.2.3') TGTRLS(V4R4M0)
ENDPGM
zlib/as400/readme.txt
ZLIB version 1.2.1 for AS400 installation instructions
ZLIB version 1.2.3 for AS400 installation instructions
I) From an AS400 *SAVF file:
......
zlib/as400/zlib.inc
* ZLIB.INC - Interface to the general purpose compression library
*
* ILE RPG400 version by Patrick Monnerat, DATASPHERE.
* Version 1.2.1
* Version 1.2.3
*
*
* WARNING:
......@@ -20,8 +20,12 @@
* Constants
**************************************************************************
*
D ZLIB_VERSION C '1.2.1' Header's version
D ZLIB_VERNUM C X'1210'
* Versioning information.
*
D ZLIB_VERSION C '1.2.3'
D ZLIB_VERNUM C X'1230'
*
* Other equates.
*
D Z_NO_FLUSH C 0
D Z_SYNC_FLUSH C 2
......
zlib/contrib/ada/mtest.adb
......@@ -5,10 +5,10 @@
-- --
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- Continuous test for ZLib multithreading. If the test is fail
-- Wou should provide thread safe allocation routines for the Z_Stream.
-- Continuous test for ZLib multithreading. If the test would fail
-- we should provide thread safe allocation routines for the Z_Stream.
--
-- $Id: mtest.adb,v 1.2 2003/08/12 12:11:05 vagul Exp $
-- $Id: mtest.adb,v 1.4 2004/07/23 07:49:54 vagul Exp $
with ZLib;
with Ada.Streams;
......@@ -148,6 +148,9 @@ procedure MTest is
pragma Unreferenced (Test);
Dummy : Character;
begin
null;
Ada.Text_IO.Get_Immediate (Dummy);
Stop := True;
end MTest;
zlib/contrib/ada/read.adb
......@@ -6,7 +6,7 @@
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: read.adb,v 1.7 2003/08/12 12:12:35 vagul Exp $
-- $Id: read.adb,v 1.8 2004/05/31 10:53:40 vagul Exp $
-- Test/demo program for the generic read interface.
......@@ -68,7 +68,11 @@ procedure Read is
-- ZLib.Read
-- reading data from the File_In.
procedure Read is new ZLib.Read (Read, Read_Buffer, Read_First, Read_Last);
procedure Read is new ZLib.Read
(Read,
Read_Buffer,
Rest_First => Read_First,
Rest_Last => Read_Last);
----------
-- Read --
......@@ -103,6 +107,7 @@ procedure Read is
Pack_Size := 0;
Offset := 1;
Read_First := Read_Buffer'Last + 1;
Read_Last := Read_Buffer'Last;
end Reset;
begin
......
zlib/contrib/ada/readme.txt
ZLib for Ada thick binding (ZLib.Ada)
Release 1.2
ZLib for Ada thick binding (ZLib.Ada)
Release 1.3
ZLib.Ada is a thick binding interface to the popular ZLib data
compression library, available at http://www.gzip.org/zlib/.
It provides Ada-style access to the ZLib C library.
Here are the main changes since ZLib.Ada 1.1:
Here are the main changes since ZLib.Ada 1.2:
- Attension: ZLib.Read generic routine have a initialization requirement
for Read_Last parameter now. It is a bit incompartible with previous version,
but extends functionality, we could use new parameters Allow_Read_Some and
Flush now.
- Added Is_Open routines to ZLib and ZLib.Streams packages.
- The default header type has a name "Default" now. Auto is used only for
automatic GZip/ZLib header detection.
- Add pragma Assert to check Stream_Element is 8 bit.
- Added test for multitasking mtest.adb.
- Fix extraction to buffer with exact known decompressed size. Error reported by
Steve Sangwine.
- Added GNAT project file zlib.gpr.
- Fix definition of ULong (changed to unsigned_long), fix regression on 64 bits
computers. Patch provided by Pascal Obry.
- Add Status_Error exception definition.
How to build ZLib.Ada under GNAT
- Add pragma Assertion that Ada.Streams.Stream_Element size is 8 bit.
How to build ZLib.Ada under GNAT
You should have the ZLib library already build on your computer, before
building ZLib.Ada. Make the directory of ZLib.Ada sources current and
......@@ -30,7 +41,7 @@ Or use the GNAT project file build for GNAT 3.15 or later:
gnatmake -Pzlib.gpr -L<directory where libz.a is>
How to build ZLib.Ada under Aonix ObjectAda for Win32 7.2.2
How to build ZLib.Ada under Aonix ObjectAda for Win32 7.2.2
1. Make a project with all *.ads and *.adb files from the distribution.
2. Build the libz.a library from the ZLib C sources.
......@@ -40,7 +51,7 @@ Or use the GNAT project file build for GNAT 3.15 or later:
6. Build the executable using test.adb as a main procedure.
How to use ZLib.Ada
How to use ZLib.Ada
The source files test.adb and read.adb are small demo programs that show
the main functionality of ZLib.Ada.
......@@ -50,3 +61,5 @@ The routines from the package specifications are commented.
Homepage: http://zlib-ada.sourceforge.net/
Author: Dmitriy Anisimkov <anisimkov@yahoo.com>
Contributors: Pascal Obry <pascal@obry.org>, Steve Sangwine <sjs@essex.ac.uk>
zlib/contrib/ada/zlib-streams.adb
......@@ -6,7 +6,7 @@
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: zlib-streams.adb,v 1.9 2003/08/12 13:15:31 vagul Exp $
-- $Id: zlib-streams.adb,v 1.10 2004/05/31 10:53:40 vagul Exp $
with Ada.Unchecked_Deallocation;
......@@ -90,6 +90,7 @@ package body ZLib.Streams is
Stream.Buffer := new Buffer_Subtype;
Stream.Rest_First := Stream.Buffer'Last + 1;
Stream.Rest_Last := Stream.Buffer'Last;
end if;
end Create;
......@@ -113,6 +114,15 @@ package body ZLib.Streams is
end loop;
end Flush;
-------------
-- Is_Open --
-------------
function Is_Open (Stream : Stream_Type) return Boolean is
begin
return Is_Open (Stream.Reader) or else Is_Open (Stream.Writer);
end Is_Open;
----------
-- Read --
----------
......
zlib/contrib/ada/zlib-streams.ads
......@@ -6,7 +6,7 @@
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: zlib-streams.ads,v 1.11 2003/08/12 13:15:31 vagul Exp $
-- $Id: zlib-streams.ads,v 1.12 2004/05/31 10:53:40 vagul Exp $
package ZLib.Streams is
......@@ -77,6 +77,8 @@ package ZLib.Streams is
-- !!! When the Need_Header is False ZLib-Ada is using undocumented
-- ZLib 1.1.4 functionality to do not create/wait for ZLib headers.
function Is_Open (Stream : Stream_Type) return Boolean;
procedure Close (Stream : in out Stream_Type);
private
......
zlib/contrib/ada/zlib-thin.adb
......@@ -6,12 +6,11 @@
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: zlib-thin.adb,v 1.6 2003/01/21 15:26:37 vagul Exp $
-- $Id: zlib-thin.adb,v 1.8 2003/12/14 18:27:31 vagul Exp $
package body ZLib.Thin is
ZLIB_VERSION : constant Chars_Ptr :=
Interfaces.C.Strings.New_String ("1.1.4");
ZLIB_VERSION : constant Chars_Ptr := zlibVersion;
Z_Stream_Size : constant Int := Z_Stream'Size / System.Storage_Unit;
......@@ -38,14 +37,6 @@ package body ZLib.Thin is
------------------
function Deflate_Init
(strm : in Z_Streamp;
level : in Int := Z_DEFAULT_COMPRESSION)
return Int is
begin
return deflateInit (strm, level, ZLIB_VERSION, Z_Stream_Size);
end Deflate_Init;
function Deflate_Init
(strm : Z_Streamp;
level : Int;
method : Int;
......@@ -69,16 +60,15 @@ package body ZLib.Thin is
-- Inflate_Init --
------------------
function Inflate_Init (strm : Z_Streamp) return Int is
begin
return inflateInit (strm, ZLIB_VERSION, Z_Stream_Size);
end Inflate_Init;
function Inflate_Init (strm : Z_Streamp; windowBits : Int) return Int is
begin
return inflateInit2 (strm, windowBits, ZLIB_VERSION, Z_Stream_Size);
end Inflate_Init;
------------------------
-- Last_Error_Message --
------------------------
function Last_Error_Message (Strm : in Z_Stream) return String is
use Interfaces.C.Strings;
begin
......@@ -89,54 +79,28 @@ package body ZLib.Thin is
end if;
end Last_Error_Message;
-------------
-- Need_In --
-------------
function Need_In (strm : Z_Stream) return Boolean is
begin
return strm.Avail_In = 0;
end Need_In;
--------------
-- Need_Out --
--------------
function Need_Out (strm : Z_Stream) return Boolean is
begin
return strm.Avail_Out = 0;
end Need_Out;
------------
-- Set_In --
------------
procedure Set_In
(Strm : in out Z_Stream;
Buffer : in Byte_Access;
Size : in UInt) is
Buffer : in Voidp;
Size : in UInt) is
begin
Strm.Next_In := Buffer;
Strm.Avail_In := Size;
end Set_In;
procedure Set_In
(Strm : in out Z_Stream;
Buffer : in Voidp;
Size : in UInt) is
begin
Set_In (Strm, Bytes.To_Pointer (Buffer), Size);
end Set_In;
------------------
-- Set_Mem_Func --
------------------
procedure Set_Mem_Func
(Strm : in out Z_Stream;
Opaque : in Voidp;
Alloc : in alloc_func;
Free : in free_func) is
Opaque : in Voidp;
Alloc : in alloc_func;
Free : in free_func) is
begin
Strm.opaque := Opaque;
Strm.zalloc := Alloc;
......@@ -149,21 +113,13 @@ package body ZLib.Thin is
procedure Set_Out
(Strm : in out Z_Stream;
Buffer : in Byte_Access;
Size : in UInt) is
Buffer : in Voidp;
Size : in UInt) is
begin
Strm.Next_Out := Buffer;
Strm.Avail_Out := Size;
end Set_Out;
procedure Set_Out
(Strm : in out Z_Stream;
Buffer : in Voidp;
Size : in UInt) is
begin
Set_Out (Strm, Bytes.To_Pointer (Buffer), Size);
end Set_Out;
--------------
-- Total_In --
--------------
......
zlib/contrib/ada/zlib-thin.ads
......@@ -6,10 +6,11 @@
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: zlib-thin.ads,v 1.8 2003/08/12 13:16:51 vagul Exp $
-- $Id: zlib-thin.ads,v 1.11 2004/07/23 06:33:11 vagul Exp $
with Interfaces.C.Strings;
with System.Address_To_Access_Conversions;
with System;
private package ZLib.Thin is
......@@ -36,18 +37,18 @@ private package ZLib.Thin is
-- zconf.h:216
type Int is new Interfaces.C.int;
type ULong is new Interfaces.C.unsigned; -- 32 bits or more
-- zconf.h:217
type ULong is new Interfaces.C.unsigned_long; -- 32 bits or more
-- zconf.h:217
subtype Chars_Ptr is Interfaces.C.Strings.chars_ptr;
type ULong_Access is access ULong;
type Int_Access is access Int;
subtype Voidp is System.Address; -- zconf.h:232
package Bytes is new System.Address_To_Access_Conversions (Byte);
subtype Voidp is System.Address; -- zconf.h:232
subtype Byte_Access is Bytes.Object_Pointer;
subtype Byte_Access is Voidp;
Nul : constant Voidp := System.Null_Address;
-- end from zconf
Z_NO_FLUSH : constant := 8#0000#; -- zlib.h:125
......@@ -251,12 +252,6 @@ private package ZLib.Thin is
stream_size : Int)
return Int;
function Deflate_Init
(strm : in Z_Streamp;
level : in Int := Z_DEFAULT_COMPRESSION)
return Int;
pragma Inline (Deflate_Init);
function deflateInit2
(strm : Z_Streamp;
level : Int;
......@@ -284,9 +279,6 @@ private package ZLib.Thin is
stream_size : Int)
return Int;
function Inflate_Init (strm : Z_Streamp) return Int;
pragma Inline (Inflate_Init);
function inflateInit2
(strm : in Z_Streamp;
windowBits : in Int;
......@@ -318,20 +310,6 @@ private package ZLib.Thin is
-- has dropped to zero. The application must initialize zalloc, zfree and
-- opaque before calling the init function.
function Need_In (strm : in Z_Stream) return Boolean;
-- return true when we do not need to setup Next_In and Avail_In fields.
pragma Inline (Need_In);
function Need_Out (strm : in Z_Stream) return Boolean;
-- return true when we do not need to setup Next_Out and Avail_Out field.
pragma Inline (Need_Out);
procedure Set_In
(Strm : in out Z_Stream;
Buffer : in Byte_Access;
Size : in UInt);
pragma Inline (Set_In);
procedure Set_In
(Strm : in out Z_Stream;
Buffer : in Voidp;
......@@ -340,12 +318,6 @@ private package ZLib.Thin is
procedure Set_Out
(Strm : in out Z_Stream;
Buffer : in Byte_Access;
Size : in UInt);
pragma Inline (Set_Out);
procedure Set_Out
(Strm : in out Z_Stream;
Buffer : in Voidp;
Size : in UInt);
pragma Inline (Set_Out);
......@@ -388,19 +360,13 @@ private package ZLib.Thin is
function zlibCompileFlags return ULong;
function deflatePrime
(strm : Z_Streamp;
bits : Int;
value : Int)
return Int;
private
type Z_Stream is record -- zlib.h:68
Next_In : Byte_Access; -- next input byte
Next_In : Voidp := Nul; -- next input byte
Avail_In : UInt := 0; -- number of bytes available at next_in
Total_In : ULong := 0; -- total nb of input bytes read so far
Next_Out : Byte_Access; -- next output byte should be put there
Next_Out : Voidp := Nul; -- next output byte should be put there
Avail_Out : UInt := 0; -- remaining free space at next_out
Total_Out : ULong := 0; -- total nb of bytes output so far
msg : Chars_Ptr; -- last error message, NULL if no error
......@@ -460,14 +426,13 @@ private
pragma Import (C, inflateSyncPoint, "inflateSyncPoint");
pragma Import (C, get_crc_table, "get_crc_table");
-- added in zlib 1.2.1:
-- since zlib 1.2.0:
pragma Import (C, inflateCopy, "inflateCopy");
pragma Import (C, compressBound, "compressBound");
pragma Import (C, deflateBound, "deflateBound");
pragma Import (C, gzungetc, "gzungetc");
pragma Import (C, zlibCompileFlags, "zlibCompileFlags");
pragma Import (C, deflatePrime, "deflatePrime");
pragma Import (C, inflateBackInit, "inflateBackInit_");
......
zlib/contrib/ada/zlib.adb
----------------------------------------------------------------
-- ZLib for Ada thick binding. --
-- --
-- Copyright (C) 2002-2003 Dmitriy Anisimkov --
-- Copyright (C) 2002-2004 Dmitriy Anisimkov --
-- --
-- Open source license information is in the zlib.ads file. --
----------------------------------------------------------------
-- $Id: zlib.adb,v 1.19 2003/07/13 16:02:19 vagul Exp $
-- $Id: zlib.adb,v 1.31 2004/09/06 06:53:19 vagul Exp $
with Ada.Exceptions;
with Ada.Unchecked_Conversion;
......@@ -34,7 +34,7 @@ package body ZLib is
VERSION_ERROR);
type Flate_Step_Function is access
function (Strm : Thin.Z_Streamp; flush : Thin.Int) return Thin.Int;
function (Strm : in Thin.Z_Streamp; Flush : in Thin.Int) return Thin.Int;
pragma Convention (C, Flate_Step_Function);
type Flate_End_Function is access
......@@ -82,13 +82,13 @@ package body ZLib is
Flush_Finish : constant array (Boolean) of Flush_Mode
:= (True => Finish, False => No_Flush);
procedure Raise_Error (Stream : Z_Stream);
procedure Raise_Error (Stream : in Z_Stream);
pragma Inline (Raise_Error);
procedure Raise_Error (Message : String);
procedure Raise_Error (Message : in String);
pragma Inline (Raise_Error);
procedure Check_Error (Stream : Z_Stream; Code : Thin.Int);
procedure Check_Error (Stream : in Z_Stream; Code : in Thin.Int);
procedure Free is new Ada.Unchecked_Deallocation
(Z_Stream, Z_Stream_Access);
......@@ -118,7 +118,7 @@ package body ZLib is
-- Check_Error --
-----------------
procedure Check_Error (Stream : Z_Stream; Code : Thin.Int) is
procedure Check_Error (Stream : in Z_Stream; Code : in Thin.Int) is
use type Thin.Int;
begin
if Code /= Thin.Z_OK then
......@@ -138,10 +138,11 @@ package body ZLib is
is
Code : Thin.Int;
begin
Code := Flate (Filter.Compression).Done
(To_Thin_Access (Filter.Strm));
if not Ignore_Error and then not Is_Open (Filter) then
raise Status_Error;
end if;
Filter.Opened := False;
Code := Flate (Filter.Compression).Done (To_Thin_Access (Filter.Strm));
if Ignore_Error or else Code = Thin.Z_OK then
Free (Filter.Strm);
......@@ -154,7 +155,7 @@ package body ZLib is
Ada.Exceptions.Raise_Exception
(ZLib_Error'Identity,
Return_Code_Enum'Image (Return_Code (Code))
& ": " & Error_Message);
& ": " & Error_Message);
end;
end if;
end Close;
......@@ -170,10 +171,9 @@ package body ZLib is
is
use Thin;
begin
return Unsigned_32 (crc32
(ULong (CRC),
Bytes.To_Pointer (Data'Address),
Data'Length));
return Unsigned_32 (crc32 (ULong (CRC),
Data'Address,
Data'Length));
end CRC32;
procedure CRC32
......@@ -192,13 +192,17 @@ package body ZLib is
Level : in Compression_Level := Default_Compression;
Strategy : in Strategy_Type := Default_Strategy;
Method : in Compression_Method := Deflated;
Window_Bits : in Window_Bits_Type := 15;
Memory_Level : in Memory_Level_Type := 8;
Window_Bits : in Window_Bits_Type := Default_Window_Bits;
Memory_Level : in Memory_Level_Type := Default_Memory_Level;
Header : in Header_Type := Default)
is
use type Thin.Int;
Win_Bits : Thin.Int := Thin.Int (Window_Bits);
begin
if Is_Open (Filter) then
raise Status_Error;
end if;
-- We allow ZLib to make header only in case of default header type.
-- Otherwise we would either do header by ourselfs, or do not do
-- header at all.
......@@ -216,10 +220,9 @@ package body ZLib is
Filter.Offset := Simple_GZip_Header'Last + 1;
end if;
Filter.Strm := new Z_Stream;
Filter.Strm := new Z_Stream;
Filter.Compression := True;
Filter.Stream_End := False;
Filter.Opened := True;
Filter.Header := Header;
if Thin.Deflate_Init
......@@ -255,18 +258,18 @@ package body ZLib is
-----------------------
procedure Generic_Translate
(Filter : in out ZLib.Filter_Type;
In_Buffer_Size : Integer := Default_Buffer_Size;
Out_Buffer_Size : Integer := Default_Buffer_Size)
(Filter : in out ZLib.Filter_Type;
In_Buffer_Size : in Integer := Default_Buffer_Size;
Out_Buffer_Size : in Integer := Default_Buffer_Size)
is
In_Buffer : Stream_Element_Array
(1 .. Stream_Element_Offset (In_Buffer_Size));
In_Buffer : Stream_Element_Array
(1 .. Stream_Element_Offset (In_Buffer_Size));
Out_Buffer : Stream_Element_Array
(1 .. Stream_Element_Offset (Out_Buffer_Size));
Last : Stream_Element_Offset;
In_Last : Stream_Element_Offset;
In_First : Stream_Element_Offset;
Out_Last : Stream_Element_Offset;
(1 .. Stream_Element_Offset (Out_Buffer_Size));
Last : Stream_Element_Offset;
In_Last : Stream_Element_Offset;
In_First : Stream_Element_Offset;
Out_Last : Stream_Element_Offset;
begin
Main : loop
Data_In (In_Buffer, Last);
......@@ -275,18 +278,21 @@ package body ZLib is
loop
Translate
(Filter,
In_Buffer (In_First .. Last),
In_Last,
Out_Buffer,
Out_Last,
Flush_Finish (Last < In_Buffer'First));
(Filter => Filter,
In_Data => In_Buffer (In_First .. Last),
In_Last => In_Last,
Out_Data => Out_Buffer,
Out_Last => Out_Last,
Flush => Flush_Finish (Last < In_Buffer'First));
Data_Out (Out_Buffer (Out_Buffer'First .. Out_Last));
if Out_Buffer'First <= Out_Last then
Data_Out (Out_Buffer (Out_Buffer'First .. Out_Last));
end if;
exit Main when Stream_End (Filter);
-- The end of in buffer.
exit when In_Last = Last;
In_First := In_Last + 1;
......@@ -301,7 +307,7 @@ package body ZLib is
procedure Inflate_Init
(Filter : in out Filter_Type;
Window_Bits : in Window_Bits_Type := 15;
Window_Bits : in Window_Bits_Type := Default_Window_Bits;
Header : in Header_Type := Default)
is
use type Thin.Int;
......@@ -320,6 +326,10 @@ package body ZLib is
end Check_Version;
begin
if Is_Open (Filter) then
raise Status_Error;
end if;
case Header is
when None =>
Check_Version;
......@@ -344,10 +354,9 @@ package body ZLib is
when Default => null;
end case;
Filter.Strm := new Z_Stream;
Filter.Strm := new Z_Stream;
Filter.Compression := False;
Filter.Stream_End := False;
Filter.Opened := True;
Filter.Header := Header;
if Thin.Inflate_Init
......@@ -357,16 +366,25 @@ package body ZLib is
end if;
end Inflate_Init;
-------------
-- Is_Open --
-------------
function Is_Open (Filter : in Filter_Type) return Boolean is
begin
return Filter.Strm /= null;
end Is_Open;
-----------------
-- Raise_Error --
-----------------
procedure Raise_Error (Message : String) is
procedure Raise_Error (Message : in String) is
begin
Ada.Exceptions.Raise_Exception (ZLib_Error'Identity, Message);
end Raise_Error;
procedure Raise_Error (Stream : Z_Stream) is
procedure Raise_Error (Stream : in Z_Stream) is
begin
Raise_Error (Last_Error_Message (Stream));
end Raise_Error;
......@@ -378,21 +396,29 @@ package body ZLib is
procedure Read
(Filter : in out Filter_Type;
Item : out Ada.Streams.Stream_Element_Array;
Last : out Ada.Streams.Stream_Element_Offset)
Last : out Ada.Streams.Stream_Element_Offset;
Flush : in Flush_Mode := No_Flush)
is
In_Last : Stream_Element_Offset;
Item_First : Ada.Streams.Stream_Element_Offset := Item'First;
V_Flush : Flush_Mode := Flush;
begin
pragma Assert (Rest_First in Buffer'First .. Buffer'Last + 1);
pragma Assert (Rest_Last in Buffer'First - 1 .. Buffer'Last);
loop
if Rest_First > Buffer'Last then
if Rest_Last = Buffer'First - 1 then
V_Flush := Finish;
elsif Rest_First > Rest_Last then
Read (Buffer, Rest_Last);
Rest_First := Buffer'First;
end if;
pragma Assert (Rest_Last in Buffer'First - 1 .. Buffer'Last);
if Rest_Last < Buffer'First then
V_Flush := Finish;
end if;
end if;
Translate
(Filter => Filter,
......@@ -400,11 +426,13 @@ package body ZLib is
In_Last => In_Last,
Out_Data => Item (Item_First .. Item'Last),
Out_Last => Last,
Flush => Flush_Finish (Rest_Last < Rest_First));
Flush => V_Flush);
Rest_First := In_Last + 1;
exit when Last = Item'Last or else Stream_End (Filter);
exit when Stream_End (Filter)
or else Last = Item'Last
or else (Last >= Item'First and then Allow_Read_Some);
Item_First := Last + 1;
end loop;
......@@ -489,11 +517,11 @@ package body ZLib is
Code : Thin.Int;
begin
if Filter.Opened = False then
raise ZLib_Error;
if not Is_Open (Filter) then
raise Status_Error;
end if;
if Out_Data'Length = 0 then
if Out_Data'Length = 0 and then In_Data'Length = 0 then
raise Constraint_Error;
end if;
......@@ -514,7 +542,6 @@ package body ZLib is
- Stream_Element_Offset (Avail_In (Filter.Strm.all));
Out_Last := Out_Data'Last
- Stream_Element_Offset (Avail_Out (Filter.Strm.all));
end Translate_Auto;
--------------------
......@@ -529,7 +556,7 @@ package body ZLib is
Out_Last : out Ada.Streams.Stream_Element_Offset;
Flush : in Flush_Mode)
is
Out_First : Stream_Element_Offset;
Out_First : Stream_Element_Offset;
procedure Add_Data (Data : in Stream_Element_Array);
-- Add data to stream from the Filter.Offset till necessary,
......@@ -596,7 +623,7 @@ package body ZLib is
Add_Data (Simple_GZip_Header);
Translate_Auto
(Filter => Filter,
(Filter => Filter,
In_Data => In_Data,
In_Last => In_Last,
Out_Data => Out_Data (Out_First .. Out_Data'Last),
......@@ -604,7 +631,6 @@ package body ZLib is
Flush => Flush);
CRC32 (Filter.CRC, In_Data (In_Data'First .. In_Last));
end if;
if Filter.Stream_End and then Out_Last <= Out_Data'Last then
......@@ -642,10 +668,11 @@ package body ZLib is
procedure Write
(Filter : in out Filter_Type;
Item : in Ada.Streams.Stream_Element_Array;
Flush : in Flush_Mode)
Flush : in Flush_Mode := No_Flush)
is
Buffer : Stream_Element_Array (1 .. Buffer_Size);
In_Last, Out_Last : Stream_Element_Offset;
Buffer : Stream_Element_Array (1 .. Buffer_Size);
In_Last : Stream_Element_Offset;
Out_Last : Stream_Element_Offset;
In_First : Stream_Element_Offset := Item'First;
begin
if Item'Length = 0 and Flush = No_Flush then
......@@ -654,7 +681,7 @@ package body ZLib is
loop
Translate
(Filter => Filter,
(Filter => Filter,
In_Data => Item (In_First .. Item'Last),
In_Last => In_Last,
Out_Data => Buffer,
......
zlib/contrib/ada/zlib.ads
------------------------------------------------------------------------------
-- ZLib for Ada thick binding. --
-- --
-- Copyright (C) 2002-2003 Dmitriy Anisimkov --
-- Copyright (C) 2002-2004 Dmitriy Anisimkov --
-- --
-- This library is free software; you can redistribute it and/or modify --
-- it under the terms of the GNU General Public License as published by --
......@@ -25,7 +25,7 @@
-- covered by the GNU Public License. --
------------------------------------------------------------------------------
-- $Id: zlib.ads,v 1.17 2003/08/12 13:19:07 vagul Exp $
-- $Id: zlib.ads,v 1.26 2004/09/06 06:53:19 vagul Exp $
with Ada.Streams;
......@@ -33,7 +33,8 @@ with Interfaces;
package ZLib is
ZLib_Error : exception;
ZLib_Error : exception;
Status_Error : exception;
type Compression_Level is new Integer range -1 .. 9;
......@@ -55,12 +56,15 @@ package ZLib is
subtype Count is Ada.Streams.Stream_Element_Count;
Default_Memory_Level : constant Memory_Level_Type := 8;
Default_Window_Bits : constant Window_Bits_Type := 15;
----------------------------------
-- Compression method constants --
----------------------------------
Deflated : constant Compression_Method;
-- Only one method allowed in this ZLib version.
-- Only one method allowed in this ZLib version
---------------------------------
-- Compression level constants --
......@@ -79,21 +83,29 @@ package ZLib is
-- Regular way for compression, no flush
Partial_Flush : constant Flush_Mode;
-- will be removed, use Z_SYNC_FLUSH instead
-- Will be removed, use Z_SYNC_FLUSH instead
Sync_Flush : constant Flush_Mode;
-- all pending output is flushed to the output buffer and the output
-- All pending output is flushed to the output buffer and the output
-- is aligned on a byte boundary, so that the decompressor can get all
-- input data available so far. (In particular avail_in is zero after the
-- call if enough output space has been provided before the call.)
-- Flushing may degrade compression for some compression algorithms and so
-- it should be used only when necessary.
Block_Flush : constant Flush_Mode;
-- Z_BLOCK requests that inflate() stop
-- if and when it get to the next deflate block boundary. When decoding the
-- zlib or gzip format, this will cause inflate() to return immediately
-- after the header and before the first block. When doing a raw inflate,
-- inflate() will go ahead and process the first block, and will return
-- when it gets to the end of that block, or when it runs out of data.
Full_Flush : constant Flush_Mode;
-- all output is flushed as with SYNC_FLUSH, and the compression state
-- All output is flushed as with SYNC_FLUSH, and the compression state
-- is reset so that decompression can restart from this point if previous
-- compressed data has been damaged or if random access is desired. Using
-- FULL_FLUSH too often can seriously degrade the compression.
-- Full_Flush too often can seriously degrade the compression.
Finish : constant Flush_Mode;
-- Just for tell the compressor that input data is complete.
......@@ -111,7 +123,7 @@ package ZLib is
Default_Buffer_Size : constant := 4096;
type Filter_Type is limited private;
type Filter_Type is tagged limited private;
-- The filter is for compression and for decompression.
-- The usage of the type is depend of its initialization.
......@@ -124,8 +136,8 @@ package ZLib is
Level : in Compression_Level := Default_Compression;
Strategy : in Strategy_Type := Default_Strategy;
Method : in Compression_Method := Deflated;
Window_Bits : in Window_Bits_Type := 15;
Memory_Level : in Memory_Level_Type := 8;
Window_Bits : in Window_Bits_Type := Default_Window_Bits;
Memory_Level : in Memory_Level_Type := Default_Memory_Level;
Header : in Header_Type := Default);
-- Compressor initialization.
-- When Header parameter is Auto or Default, then default zlib header
......@@ -136,7 +148,7 @@ package ZLib is
procedure Inflate_Init
(Filter : in out Filter_Type;
Window_Bits : in Window_Bits_Type := 15;
Window_Bits : in Window_Bits_Type := Default_Window_Bits;
Header : in Header_Type := Default);
-- Decompressor initialization.
-- Default header type mean that ZLib default header is expecting in the
......@@ -146,10 +158,14 @@ package ZLib is
-- input compressed stream.
-- Auto header type mean that header type (GZip or Native) would be
-- detected automatically in the input stream.
-- Note that header types parameter values None, GZip and Auto is
-- supporting for inflate routine only in ZLib versions 1.2.0.2 and later.
-- Note that header types parameter values None, GZip and Auto are
-- supported for inflate routine only in ZLib versions 1.2.0.2 and later.
-- Deflate_Init is supporting all header types.
function Is_Open (Filter : in Filter_Type) return Boolean;
pragma Inline (Is_Open);
-- Is the filter opened for compression or decompression.
procedure Close
(Filter : in out Filter_Type;
Ignore_Error : in Boolean := False);
......@@ -167,31 +183,31 @@ package ZLib is
(Filter : in out Filter_Type;
In_Buffer_Size : in Integer := Default_Buffer_Size;
Out_Buffer_Size : in Integer := Default_Buffer_Size);
-- Compressing/decompressing data arrived from Data_In routine
-- Compress/decompress data fetch from Data_In routine and pass the result
-- to the Data_Out routine. User should provide Data_In and Data_Out
-- for compression/decompression data flow.
-- Compression or decompression depend on initialization of Filter.
-- Compression or decompression depend on Filter initialization.
function Total_In (Filter : in Filter_Type) return Count;
pragma Inline (Total_In);
-- Return total number of input bytes read so far.
-- Returns total number of input bytes read so far
function Total_Out (Filter : in Filter_Type) return Count;
pragma Inline (Total_Out);
-- Return total number of bytes output so far.
-- Returns total number of bytes output so far
function CRC32
(CRC : in Unsigned_32;
Data : in Ada.Streams.Stream_Element_Array)
return Unsigned_32;
pragma Inline (CRC32);
-- Calculate CRC32, it could be necessary for make gzip format.
-- Compute CRC32, it could be necessary for make gzip format
procedure CRC32
(CRC : in out Unsigned_32;
Data : in Ada.Streams.Stream_Element_Array);
pragma Inline (CRC32);
-- Calculate CRC32, it could be necessary for make gzip format.
-- Compute CRC32, it could be necessary for make gzip format
-------------------------------------------------
-- Below is more complex low level routines. --
......@@ -204,15 +220,11 @@ package ZLib is
Out_Data : out Ada.Streams.Stream_Element_Array;
Out_Last : out Ada.Streams.Stream_Element_Offset;
Flush : in Flush_Mode);
-- Compressing/decompressing the datas from In_Data buffer to the
-- Out_Data buffer.
-- In_Data is incoming data portion,
-- In_Last is the index of last element from In_Data accepted by the
-- Filter.
-- Out_Data is the buffer for output data from the filter.
-- Out_Last is the last element of the received data from Filter.
-- To tell the filter that incoming data is complete put the
-- Flush parameter to FINISH.
-- Compress/decompress the In_Data buffer and place the result into
-- Out_Data. In_Last is the index of last element from In_Data accepted by
-- the Filter. Out_Last is the last element of the received data from
-- Filter. To tell the filter that incoming data are complete put the
-- Flush parameter to Finish.
function Stream_End (Filter : in Filter_Type) return Boolean;
pragma Inline (Stream_End);
......@@ -239,10 +251,9 @@ package ZLib is
procedure Write
(Filter : in out Filter_Type;
Item : in Ada.Streams.Stream_Element_Array;
Flush : in Flush_Mode);
-- Compressing/Decompressing data from Item to the
-- generic parameter procedure Write.
-- Output buffer size could be set in Buffer_Size generic parameter.
Flush : in Flush_Mode := No_Flush);
-- Compress/Decompress data from Item to the generic parameter procedure
-- Write. Output buffer size could be set in Buffer_Size generic parameter.
generic
with procedure Read
......@@ -257,33 +268,41 @@ package ZLib is
Rest_First, Rest_Last : in out Ada.Streams.Stream_Element_Offset;
-- Rest_First have to be initialized to Buffer'Last + 1
-- Rest_Last have to be initialized to Buffer'Last
-- before usage.
Allow_Read_Some : in Boolean := False;
-- Is it allowed to return Last < Item'Last before end of data.
procedure Read
(Filter : in out Filter_Type;
Item : out Ada.Streams.Stream_Element_Array;
Last : out Ada.Streams.Stream_Element_Offset);
-- Compressing/Decompressing data from generic parameter
-- procedure Read to the Item.
-- User should provide Buffer for the operation
-- and Rest_First variable first time initialized to the Buffer'Last + 1.
Last : out Ada.Streams.Stream_Element_Offset;
Flush : in Flush_Mode := No_Flush);
-- Compress/Decompress data from generic parameter procedure Read to the
-- Item. User should provide Buffer and initialized Rest_First, Rest_Last
-- indicators. If Allow_Read_Some is True, Read routines could return
-- Last < Item'Last only at end of stream.
private
use Ada.Streams;
type Flush_Mode is new Integer range 0 .. 4;
pragma Assert (Ada.Streams.Stream_Element'Size = 8);
pragma Assert (Ada.Streams.Stream_Element'Modulus = 2**8);
type Flush_Mode is new Integer range 0 .. 5;
type Compression_Method is new Integer range 8 .. 8;
type Strategy_Type is new Integer range 0 .. 3;
No_Flush : constant Flush_Mode := 0;
Partial_Flush : constant Flush_Mode := 1;
Sync_Flush : constant Flush_Mode := 2;
Full_Flush : constant Flush_Mode := 3;
Finish : constant Flush_Mode := 4;
Partial_Flush : constant Flush_Mode := 1;
-- will be removed, use Z_SYNC_FLUSH instead
Block_Flush : constant Flush_Mode := 5;
Filtered : constant Strategy_Type := 1;
Huffman_Only : constant Strategy_Type := 2;
......@@ -296,7 +315,7 @@ private
type Z_Stream_Access is access all Z_Stream;
type Filter_Type is record
type Filter_Type is tagged limited record
Strm : Z_Stream_Access;
Compression : Boolean;
Stream_End : Boolean;
......@@ -304,8 +323,6 @@ private
CRC : Unsigned_32;
Offset : Stream_Element_Offset;
-- Offset for gzip header/footer output.
Opened : Boolean := False;
end record;
end ZLib;
zlib/contrib/ada/zlib.gpr
......@@ -3,10 +3,10 @@ project Zlib is
for Languages use ("Ada");
for Source_Dirs use (".");
for Object_Dir use ".";
for Main use ("test.adb", "mtest.adb", "read.adb");
for Main use ("test.adb", "mtest.adb", "read.adb", "buffer_demo");
package Compiler is
for Default_Switches ("ada") use ("-gnatwbcfilopru", "-gnatVcdfimorst", "-gnatyabcefhiklmnoprst");
for Default_Switches ("ada") use ("-gnatwcfilopru", "-gnatVcdfimorst", "-gnatyabcefhiklmnoprst");
end Compiler;
package Linker is
......@@ -18,4 +18,3 @@ project Zlib is
end Builder;
end Zlib;
zlib/contrib/delphi/ZLib.pas
......@@ -152,7 +152,7 @@ procedure DecompressToUserBuf(const InBuf: Pointer; InBytes: Integer;
const OutBuf: Pointer; BufSize: Integer);
const
zlib_version = '1.2.1';
zlib_version = '1.2.3';
type
EZlibError = class(Exception);
......@@ -344,7 +344,7 @@ begin
strm.avail_out := OutBytes;
DCheck(inflateInit_(strm, zlib_version, sizeof(strm)));
try
while DCheck(inflate(strm, Z_FINISH)) <> Z_STREAM_END do
while DCheck(inflate(strm, Z_NO_FLUSH)) <> Z_STREAM_END do
begin
P := OutBuf;
Inc(OutBytes, BufInc);
......@@ -510,7 +510,7 @@ begin
Result := Count - FZRec.avail_out;
Exit;
end;
FZRec.next_in := FBuffer;
FZRec.next_in := FBuffer;
FStrmPos := FStrm.Position;
Progress(Self);
end;
......
zlib/contrib/infback9/infback9.c
......@@ -430,6 +430,9 @@ void FAR *out_desc;
}
}
/* handle error breaks in while */
if (mode == BAD) break;
/* build code tables */
state->next = state->codes;
lencode = (code const FAR *)(state->next);
......
zlib/contrib/infback9/infback9.h
......@@ -16,6 +16,10 @@
* zlib.h must be included before this header file.
*/
#ifdef __cplusplus
extern "C" {
#endif
ZEXTERN int ZEXPORT inflateBack9 OF((z_stream FAR *strm,
in_func in, void FAR *in_desc,
out_func out, void FAR *out_desc));
......@@ -27,3 +31,7 @@ ZEXTERN int ZEXPORT inflateBack9Init_ OF((z_stream FAR *strm,
#define inflateBack9Init(strm, window) \
inflateBack9Init_((strm), (window), \
ZLIB_VERSION, sizeof(z_stream))
#ifdef __cplusplus
}
#endif
zlib/contrib/infback9/inftree9.c
/* inftree9.c -- generate Huffman trees for efficient decoding
* Copyright (C) 1995-2003 Mark Adler
* Copyright (C) 1995-2005 Mark Adler
* For conditions of distribution and use, see copyright notice in zlib.h
*/
......@@ -9,7 +9,7 @@
#define MAXBITS 15
const char inflate9_copyright[] =
" inflate9 1.2.1 Copyright 1995-2003 Mark Adler ";
" inflate9 1.2.3 Copyright 1995-2005 Mark Adler ";
/*
If you use the zlib library in a product, an acknowledgment is welcome
in the documentation of your product. If for some reason you cannot
......@@ -64,7 +64,7 @@ unsigned short FAR *work;
static const unsigned short lext[31] = { /* Length codes 257..285 extra */
128, 128, 128, 128, 128, 128, 128, 128, 129, 129, 129, 129,
130, 130, 130, 130, 131, 131, 131, 131, 132, 132, 132, 132,
133, 133, 133, 133, 144, 76, 66};
133, 133, 133, 133, 144, 201, 196};
static const unsigned short dbase[32] = { /* Distance codes 0..31 base */
1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49,
65, 97, 129, 193, 257, 385, 513, 769, 1025, 1537, 2049, 3073,
......@@ -128,7 +128,7 @@ unsigned short FAR *work;
left -= count[len];
if (left < 0) return -1; /* over-subscribed */
}
if (left > 0 && (type == CODES || (codes - count[0] != 1)))
if (left > 0 && (type == CODES || max != 1))
return -1; /* incomplete set */
/* generate offsets into symbol table for each length for sorting */
......
zlib/contrib/infback9/inftree9.h
......@@ -36,12 +36,12 @@ typedef struct {
*/
/* Maximum size of dynamic tree. The maximum found in a long but non-
exhaustive search was 1004 code structures (850 for length/literals
and 154 for distances, the latter actually the result of an
exhaustive search was 1444 code structures (852 for length/literals
and 592 for distances, the latter actually the result of an
exhaustive search). The true maximum is not known, but the value
below is more than safe. */
#define ENOUGH 1440
#define MAXD 154
#define ENOUGH 2048
#define MAXD 592
/* Type of code to build for inftable() */
typedef enum {
......
zlib/contrib/inflate86/inffas86.c
......@@ -7,6 +7,15 @@
* Copyright (C) 2003 Chris Anderson <christop@charm.net>
* Please use the copyright conditions above.
*
* Dec-29-2003 -- I added AMD64 inflate asm support. This version is also
* slightly quicker on x86 systems because, instead of using rep movsb to copy
* data, it uses rep movsw, which moves data in 2-byte chunks instead of single
* bytes. I've tested the AMD64 code on a Fedora Core 1 + the x86_64 updates
* from http://fedora.linux.duke.edu/fc1_x86_64
* which is running on an Athlon 64 3000+ / Gigabyte GA-K8VT800M system with
* 1GB ram. The 64-bit version is about 4% faster than the 32-bit version,
* when decompressing mozilla-source-1.3.tar.gz.
*
* Mar-13-2003 -- Most of this is derived from inffast.S which is derived from
* the gcc -S output of zlib-1.2.0/inffast.c. Zlib-1.2.0 is in beta release at
* the moment. I have successfully compiled and tested this code with gcc2.96,
......@@ -65,33 +74,44 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
{
struct inflate_state FAR *state;
struct inffast_ar {
void *esp; /* esp save */
unsigned char FAR *in; /* local strm->next_in */
unsigned char FAR *last; /* while in < last, enough input available */
unsigned char FAR *out; /* local strm->next_out */
unsigned char FAR *beg; /* inflate()'s initial strm->next_out */
unsigned char FAR *end; /* while out < end, enough space available */
unsigned wsize; /* window size or zero if not using window */
unsigned write; /* window write index */
unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */
unsigned long hold; /* local strm->hold */
unsigned bits; /* local strm->bits */
code const FAR *lcode; /* local strm->lencode */
code const FAR *dcode; /* local strm->distcode */
unsigned lmask; /* mask for first level of length codes */
unsigned dmask; /* mask for first level of distance codes */
unsigned len; /* match length, unused bytes */
unsigned dist; /* match distance */
unsigned status; /* this is set when state changes */
/* 64 32 x86 x86_64 */
/* ar offset register */
/* 0 0 */ void *esp; /* esp save */
/* 8 4 */ void *ebp; /* ebp save */
/* 16 8 */ unsigned char FAR *in; /* esi rsi local strm->next_in */
/* 24 12 */ unsigned char FAR *last; /* r9 while in < last */
/* 32 16 */ unsigned char FAR *out; /* edi rdi local strm->next_out */
/* 40 20 */ unsigned char FAR *beg; /* inflate()'s init next_out */
/* 48 24 */ unsigned char FAR *end; /* r10 while out < end */
/* 56 28 */ unsigned char FAR *window;/* size of window, wsize!=0 */
/* 64 32 */ code const FAR *lcode; /* ebp rbp local strm->lencode */
/* 72 36 */ code const FAR *dcode; /* r11 local strm->distcode */
/* 80 40 */ unsigned long hold; /* edx rdx local strm->hold */
/* 88 44 */ unsigned bits; /* ebx rbx local strm->bits */
/* 92 48 */ unsigned wsize; /* window size */
/* 96 52 */ unsigned write; /* window write index */
/*100 56 */ unsigned lmask; /* r12 mask for lcode */
/*104 60 */ unsigned dmask; /* r13 mask for dcode */
/*108 64 */ unsigned len; /* r14 match length */
/*112 68 */ unsigned dist; /* r15 match distance */
/*116 72 */ unsigned status; /* set when state chng*/
} ar;
#if defined( __GNUC__ ) && defined( __amd64__ ) && ! defined( __i386 )
#define PAD_AVAIL_IN 6
#define PAD_AVAIL_OUT 258
#else
#define PAD_AVAIL_IN 5
#define PAD_AVAIL_OUT 257
#endif
/* copy state to local variables */
state = (struct inflate_state FAR *)strm->state;
ar.in = strm->next_in;
ar.last = ar.in + (strm->avail_in - 5);
ar.last = ar.in + (strm->avail_in - PAD_AVAIL_IN);
ar.out = strm->next_out;
ar.beg = ar.out - (start - strm->avail_out);
ar.end = ar.out + (strm->avail_out - 257);
ar.end = ar.out + (strm->avail_out - PAD_AVAIL_OUT);
ar.wsize = state->wsize;
ar.write = state->write;
ar.window = state->window;
......@@ -105,32 +125,368 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
/* decode literals and length/distances until end-of-block or not enough
input data or output space */
/* align in on 2 byte boundary */
if (((unsigned long)(void *)ar.in & 0x1) != 0) {
/* align in on 1/2 hold size boundary */
while (((unsigned long)(void *)ar.in & (sizeof(ar.hold) / 2 - 1)) != 0) {
ar.hold += (unsigned long)*ar.in++ << ar.bits;
ar.bits += 8;
}
#if defined( __GNUC__ ) || defined( __ICC )
#if defined( __GNUC__ ) && defined( __amd64__ ) && ! defined( __i386 )
__asm__ __volatile__ (
" leaq %0, %%rax\n"
" movq %%rbp, 8(%%rax)\n" /* save regs rbp and rsp */
" movq %%rsp, (%%rax)\n"
" movq %%rax, %%rsp\n" /* make rsp point to &ar */
" movq 16(%%rsp), %%rsi\n" /* rsi = in */
" movq 32(%%rsp), %%rdi\n" /* rdi = out */
" movq 24(%%rsp), %%r9\n" /* r9 = last */
" movq 48(%%rsp), %%r10\n" /* r10 = end */
" movq 64(%%rsp), %%rbp\n" /* rbp = lcode */
" movq 72(%%rsp), %%r11\n" /* r11 = dcode */
" movq 80(%%rsp), %%rdx\n" /* rdx = hold */
" movl 88(%%rsp), %%ebx\n" /* ebx = bits */
" movl 100(%%rsp), %%r12d\n" /* r12d = lmask */
" movl 104(%%rsp), %%r13d\n" /* r13d = dmask */
/* r14d = len */
/* r15d = dist */
" cld\n"
" cmpq %%rdi, %%r10\n"
" je .L_one_time\n" /* if only one decode left */
" cmpq %%rsi, %%r9\n"
" je .L_one_time\n"
" jmp .L_do_loop\n"
".L_one_time:\n"
" movq %%r12, %%r8\n" /* r8 = lmask */
" cmpb $32, %%bl\n"
" ja .L_get_length_code_one_time\n"
" lodsl\n" /* eax = *(uint *)in++ */
" movb %%bl, %%cl\n" /* cl = bits, needs it for shifting */
" addb $32, %%bl\n" /* bits += 32 */
" shlq %%cl, %%rax\n"
" orq %%rax, %%rdx\n" /* hold |= *((uint *)in)++ << bits */
" jmp .L_get_length_code_one_time\n"
".align 32,0x90\n"
".L_while_test:\n"
" cmpq %%rdi, %%r10\n"
" jbe .L_break_loop\n"
" cmpq %%rsi, %%r9\n"
" jbe .L_break_loop\n"
".L_do_loop:\n"
" movq %%r12, %%r8\n" /* r8 = lmask */
" cmpb $32, %%bl\n"
" ja .L_get_length_code\n" /* if (32 < bits) */
" lodsl\n" /* eax = *(uint *)in++ */
" movb %%bl, %%cl\n" /* cl = bits, needs it for shifting */
" addb $32, %%bl\n" /* bits += 32 */
" shlq %%cl, %%rax\n"
" orq %%rax, %%rdx\n" /* hold |= *((uint *)in)++ << bits */
".L_get_length_code:\n"
" andq %%rdx, %%r8\n" /* r8 &= hold */
" movl (%%rbp,%%r8,4), %%eax\n" /* eax = lcode[hold & lmask] */
" movb %%ah, %%cl\n" /* cl = this.bits */
" subb %%ah, %%bl\n" /* bits -= this.bits */
" shrq %%cl, %%rdx\n" /* hold >>= this.bits */
" testb %%al, %%al\n"
" jnz .L_test_for_length_base\n" /* if (op != 0) 45.7% */
" movq %%r12, %%r8\n" /* r8 = lmask */
" shrl $16, %%eax\n" /* output this.val char */
" stosb\n"
".L_get_length_code_one_time:\n"
" andq %%rdx, %%r8\n" /* r8 &= hold */
" movl (%%rbp,%%r8,4), %%eax\n" /* eax = lcode[hold & lmask] */
".L_dolen:\n"
" movb %%ah, %%cl\n" /* cl = this.bits */
" subb %%ah, %%bl\n" /* bits -= this.bits */
" shrq %%cl, %%rdx\n" /* hold >>= this.bits */
" testb %%al, %%al\n"
" jnz .L_test_for_length_base\n" /* if (op != 0) 45.7% */
" shrl $16, %%eax\n" /* output this.val char */
" stosb\n"
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_test_for_length_base:\n"
" movl %%eax, %%r14d\n" /* len = this */
" shrl $16, %%r14d\n" /* len = this.val */
" movb %%al, %%cl\n"
" testb $16, %%al\n"
" jz .L_test_for_second_level_length\n" /* if ((op & 16) == 0) 8% */
" andb $15, %%cl\n" /* op &= 15 */
" jz .L_decode_distance\n" /* if (!op) */
".L_add_bits_to_len:\n"
" subb %%cl, %%bl\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" shrq %%cl, %%rdx\n"
" addl %%eax, %%r14d\n" /* len += hold & mask[op] */
".L_decode_distance:\n"
" movq %%r13, %%r8\n" /* r8 = dmask */
" cmpb $32, %%bl\n"
" ja .L_get_distance_code\n" /* if (32 < bits) */
" lodsl\n" /* eax = *(uint *)in++ */
" movb %%bl, %%cl\n" /* cl = bits, needs it for shifting */
" addb $32, %%bl\n" /* bits += 32 */
" shlq %%cl, %%rax\n"
" orq %%rax, %%rdx\n" /* hold |= *((uint *)in)++ << bits */
".L_get_distance_code:\n"
" andq %%rdx, %%r8\n" /* r8 &= hold */
" movl (%%r11,%%r8,4), %%eax\n" /* eax = dcode[hold & dmask] */
".L_dodist:\n"
" movl %%eax, %%r15d\n" /* dist = this */
" shrl $16, %%r15d\n" /* dist = this.val */
" movb %%ah, %%cl\n"
" subb %%ah, %%bl\n" /* bits -= this.bits */
" shrq %%cl, %%rdx\n" /* hold >>= this.bits */
" movb %%al, %%cl\n" /* cl = this.op */
" testb $16, %%al\n" /* if ((op & 16) == 0) */
" jz .L_test_for_second_level_dist\n"
" andb $15, %%cl\n" /* op &= 15 */
" jz .L_check_dist_one\n"
".L_add_bits_to_dist:\n"
" subb %%cl, %%bl\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n" /* (1 << op) - 1 */
" andl %%edx, %%eax\n" /* eax &= hold */
" shrq %%cl, %%rdx\n"
" addl %%eax, %%r15d\n" /* dist += hold & ((1 << op) - 1) */
".L_check_window:\n"
" movq %%rsi, %%r8\n" /* save in so from can use it's reg */
" movq %%rdi, %%rax\n"
" subq 40(%%rsp), %%rax\n" /* nbytes = out - beg */
" cmpl %%r15d, %%eax\n"
" jb .L_clip_window\n" /* if (dist > nbytes) 4.2% */
" movl %%r14d, %%ecx\n" /* ecx = len */
" movq %%rdi, %%rsi\n"
" subq %%r15, %%rsi\n" /* from = out - dist */
" sarl %%ecx\n"
" jnc .L_copy_two\n" /* if len % 2 == 0 */
" rep movsw\n"
" movb (%%rsi), %%al\n"
" movb %%al, (%%rdi)\n"
" incq %%rdi\n"
" movq %%r8, %%rsi\n" /* move in back to %rsi, toss from */
" jmp .L_while_test\n"
".L_copy_two:\n"
" rep movsw\n"
" movq %%r8, %%rsi\n" /* move in back to %rsi, toss from */
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_check_dist_one:\n"
" cmpl $1, %%r15d\n" /* if dist 1, is a memset */
" jne .L_check_window\n"
" cmpq %%rdi, 40(%%rsp)\n" /* if out == beg, outside window */
" je .L_check_window\n"
" movl %%r14d, %%ecx\n" /* ecx = len */
" movb -1(%%rdi), %%al\n"
" movb %%al, %%ah\n"
" sarl %%ecx\n"
" jnc .L_set_two\n"
" movb %%al, (%%rdi)\n"
" incq %%rdi\n"
".L_set_two:\n"
" rep stosw\n"
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_test_for_second_level_length:\n"
" testb $64, %%al\n"
" jnz .L_test_for_end_of_block\n" /* if ((op & 64) != 0) */
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" addl %%r14d, %%eax\n" /* eax += len */
" movl (%%rbp,%%rax,4), %%eax\n" /* eax = lcode[val+(hold&mask[op])]*/
" jmp .L_dolen\n"
".align 32,0x90\n"
".L_test_for_second_level_dist:\n"
" testb $64, %%al\n"
" jnz .L_invalid_distance_code\n" /* if ((op & 64) != 0) */
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" addl %%r15d, %%eax\n" /* eax += dist */
" movl (%%r11,%%rax,4), %%eax\n" /* eax = dcode[val+(hold&mask[op])]*/
" jmp .L_dodist\n"
".align 32,0x90\n"
".L_clip_window:\n"
" movl %%eax, %%ecx\n" /* ecx = nbytes */
" movl 92(%%rsp), %%eax\n" /* eax = wsize, prepare for dist cmp */
" negl %%ecx\n" /* nbytes = -nbytes */
" cmpl %%r15d, %%eax\n"
" jb .L_invalid_distance_too_far\n" /* if (dist > wsize) */
" addl %%r15d, %%ecx\n" /* nbytes = dist - nbytes */
" cmpl $0, 96(%%rsp)\n"
" jne .L_wrap_around_window\n" /* if (write != 0) */
" movq 56(%%rsp), %%rsi\n" /* from = window */
" subl %%ecx, %%eax\n" /* eax -= nbytes */
" addq %%rax, %%rsi\n" /* from += wsize - nbytes */
" movl %%r14d, %%eax\n" /* eax = len */
" cmpl %%ecx, %%r14d\n"
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* eax -= nbytes */
" rep movsb\n"
" movq %%rdi, %%rsi\n"
" subq %%r15, %%rsi\n" /* from = &out[ -dist ] */
" jmp .L_do_copy\n"
".align 32,0x90\n"
".L_wrap_around_window:\n"
" movl 96(%%rsp), %%eax\n" /* eax = write */
" cmpl %%eax, %%ecx\n"
" jbe .L_contiguous_in_window\n" /* if (write >= nbytes) */
" movl 92(%%rsp), %%esi\n" /* from = wsize */
" addq 56(%%rsp), %%rsi\n" /* from += window */
" addq %%rax, %%rsi\n" /* from += write */
" subq %%rcx, %%rsi\n" /* from -= nbytes */
" subl %%eax, %%ecx\n" /* nbytes -= write */
" movl %%r14d, %%eax\n" /* eax = len */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movq 56(%%rsp), %%rsi\n" /* from = window */
" movl 96(%%rsp), %%ecx\n" /* nbytes = write */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movq %%rdi, %%rsi\n"
" subq %%r15, %%rsi\n" /* from = out - dist */
" jmp .L_do_copy\n"
".align 32,0x90\n"
".L_contiguous_in_window:\n"
" movq 56(%%rsp), %%rsi\n" /* rsi = window */
" addq %%rax, %%rsi\n"
" subq %%rcx, %%rsi\n" /* from += write - nbytes */
" movl %%r14d, %%eax\n" /* eax = len */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movq %%rdi, %%rsi\n"
" subq %%r15, %%rsi\n" /* from = out - dist */
" jmp .L_do_copy\n" /* if (nbytes >= len) */
".align 32,0x90\n"
".L_do_copy:\n"
" movl %%eax, %%ecx\n" /* ecx = len */
" rep movsb\n"
" movq %%r8, %%rsi\n" /* move in back to %esi, toss from */
" jmp .L_while_test\n"
".L_test_for_end_of_block:\n"
" testb $32, %%al\n"
" jz .L_invalid_literal_length_code\n"
" movl $1, 116(%%rsp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_literal_length_code:\n"
" movl $2, 116(%%rsp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_distance_code:\n"
" movl $3, 116(%%rsp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_distance_too_far:\n"
" movl $4, 116(%%rsp)\n"
" jmp .L_break_loop_with_status\n"
".L_break_loop:\n"
" movl $0, 116(%%rsp)\n"
".L_break_loop_with_status:\n"
/* put in, out, bits, and hold back into ar and pop esp */
" movq %%rsi, 16(%%rsp)\n" /* in */
" movq %%rdi, 32(%%rsp)\n" /* out */
" movl %%ebx, 88(%%rsp)\n" /* bits */
" movq %%rdx, 80(%%rsp)\n" /* hold */
" movq (%%rsp), %%rax\n" /* restore rbp and rsp */
" movq 8(%%rsp), %%rbp\n"
" movq %%rax, %%rsp\n"
:
: "m" (ar)
: "memory", "%rax", "%rbx", "%rcx", "%rdx", "%rsi", "%rdi",
"%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15"
);
#elif ( defined( __GNUC__ ) || defined( __ICC ) ) && defined( __i386 )
__asm__ __volatile__ (
" leal %0, %%eax\n"
" pushf\n"
" pushl %%ebp\n"
" movl %%esp, (%%eax)\n"
" movl %%esp, (%%eax)\n" /* save esp, ebp */
" movl %%ebp, 4(%%eax)\n"
" movl %%eax, %%esp\n"
" movl 4(%%esp), %%esi\n" /* esi = in */
" movl 12(%%esp), %%edi\n" /* edi = out */
" movl 36(%%esp), %%edx\n" /* edx = hold */
" movl 40(%%esp), %%ebx\n" /* ebx = bits */
" movl 44(%%esp), %%ebp\n" /* ebp = lcode */
" movl 8(%%esp), %%esi\n" /* esi = in */
" movl 16(%%esp), %%edi\n" /* edi = out */
" movl 40(%%esp), %%edx\n" /* edx = hold */
" movl 44(%%esp), %%ebx\n" /* ebx = bits */
" movl 32(%%esp), %%ebp\n" /* ebp = lcode */
" cld\n"
" jmp .L_do_loop\n"
".align 32,0x90\n"
".L_while_test:\n"
" cmpl %%edi, 20(%%esp)\n"
" cmpl %%edi, 24(%%esp)\n" /* out < end */
" jbe .L_break_loop\n"
" cmpl %%esi, 8(%%esp)\n"
" cmpl %%esi, 12(%%esp)\n" /* in < last */
" jbe .L_break_loop\n"
".L_do_loop:\n"
......@@ -145,7 +501,7 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
" orl %%eax, %%edx\n" /* hold |= *((ushort *)in)++ << bits */
".L_get_length_code:\n"
" movl 52(%%esp), %%eax\n" /* eax = lmask */
" movl 56(%%esp), %%eax\n" /* eax = lmask */
" andl %%edx, %%eax\n" /* eax &= hold */
" movl (%%ebp,%%eax,4), %%eax\n" /* eax = lcode[hold & lmask] */
......@@ -161,10 +517,11 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
" stosb\n"
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_test_for_length_base:\n"
" movl %%eax, %%ecx\n" /* len = this */
" shrl $16, %%ecx\n" /* len = this.val */
" movl %%ecx, 60(%%esp)\n" /* len = this */
" movl %%ecx, 64(%%esp)\n" /* save len */
" movb %%al, %%cl\n"
" testb $16, %%al\n"
......@@ -184,13 +541,14 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
" movb %%ch, %%cl\n" /* move op back to ecx */
".L_add_bits_to_len:\n"
" movl $1, %%eax\n"
" subb %%cl, %%bl\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" subb %%cl, %%bl\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" shrl %%cl, %%edx\n"
" addl %%eax, 60(%%esp)\n" /* len += hold & mask[op] */
" addl %%eax, 64(%%esp)\n" /* len += hold & mask[op] */
".L_decode_distance:\n"
" cmpb $15, %%bl\n"
......@@ -204,8 +562,8 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
" orl %%eax, %%edx\n" /* hold |= *((ushort *)in)++ << bits */
".L_get_distance_code:\n"
" movl 56(%%esp), %%eax\n" /* eax = dmask */
" movl 48(%%esp), %%ecx\n" /* ecx = dcode */
" movl 60(%%esp), %%eax\n" /* eax = dmask */
" movl 36(%%esp), %%ecx\n" /* ecx = dcode */
" andl %%edx, %%eax\n" /* eax &= hold */
" movl (%%ecx,%%eax,4), %%eax\n"/* eax = dcode[hold & dmask] */
......@@ -234,223 +592,228 @@ unsigned start; /* inflate()'s starting value for strm->avail_out */
" movb %%ch, %%cl\n" /* move op back to ecx */
".L_add_bits_to_dist:\n"
" movl $1, %%eax\n"
" subb %%cl, %%bl\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n" /* (1 << op) - 1 */
" subb %%cl, %%bl\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" shrl %%cl, %%edx\n"
" addl %%eax, %%ebp\n" /* dist += hold & ((1 << op) - 1) */
".L_check_window:\n"
" movl %%esi, 4(%%esp)\n" /* save in so from can use it's reg */
" movl %%esi, 8(%%esp)\n" /* save in so from can use it's reg */
" movl %%edi, %%eax\n"
" subl 16(%%esp), %%eax\n" /* nbytes = out - beg */
" subl 20(%%esp), %%eax\n" /* nbytes = out - beg */
" cmpl %%ebp, %%eax\n"
" jb .L_clip_window\n" /* if (dist > nbytes) 4.2% */
" movl 60(%%esp), %%ecx\n"
" movl 64(%%esp), %%ecx\n" /* ecx = len */
" movl %%edi, %%esi\n"
" subl %%ebp, %%esi\n" /* from = out - dist */
" subl $3, %%ecx\n" /* copy from to out */
" sarl %%ecx\n"
" jnc .L_copy_two\n" /* if len % 2 == 0 */
" rep movsw\n"
" movb (%%esi), %%al\n"
" movb %%al, (%%edi)\n"
" movb 1(%%esi), %%al\n"
" movb 2(%%esi), %%ah\n"
" addl $3, %%esi\n"
" movb %%al, 1(%%edi)\n"
" movb %%ah, 2(%%edi)\n"
" addl $3, %%edi\n"
" rep movsb\n"
" incl %%edi\n"
" movl 4(%%esp), %%esi\n" /* move in back to %esi, toss from */
" movl 44(%%esp), %%ebp\n" /* ebp = lcode */
" movl 8(%%esp), %%esi\n" /* move in back to %esi, toss from */
" movl 32(%%esp), %%ebp\n" /* ebp = lcode */
" jmp .L_while_test\n"
".L_copy_two:\n"
" rep movsw\n"
" movl 8(%%esp), %%esi\n" /* move in back to %esi, toss from */
" movl 32(%%esp), %%ebp\n" /* ebp = lcode */
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_check_dist_one:\n"
" cmpl $1, %%ebp\n" /* if dist 1, is a memset */
" jne .L_check_window\n"
" cmpl %%edi, 16(%%esp)\n"
" je .L_check_window\n"
" cmpl %%edi, 20(%%esp)\n"
" je .L_check_window\n" /* out == beg, if outside window */
" movl 64(%%esp), %%ecx\n" /* ecx = len */
" movb -1(%%edi), %%al\n"
" movb %%al, %%ah\n"
" sarl %%ecx\n"
" jnc .L_set_two\n"
" movb %%al, (%%edi)\n"
" incl %%edi\n"
" decl %%edi\n"
" movl 60(%%esp), %%ecx\n"
" movb (%%edi), %%al\n"
" subl $3, %%ecx\n"
" movb %%al, 1(%%edi)\n" /* memset out with from[-1] */
" movb %%al, 2(%%edi)\n"
" movb %%al, 3(%%edi)\n"
" addl $4, %%edi\n"
" rep stosb\n"
" movl 44(%%esp), %%ebp\n" /* ebp = lcode */
".L_set_two:\n"
" rep stosw\n"
" movl 32(%%esp), %%ebp\n" /* ebp = lcode */
" jmp .L_while_test\n"
".align 32,0x90\n"
".L_test_for_second_level_length:\n"
" testb $64, %%al\n"
" jnz .L_test_for_end_of_block\n" /* if ((op & 64) != 0) */
" movl $1, %%eax\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" addl 60(%%esp), %%eax\n" /* eax += this.val */
" addl 64(%%esp), %%eax\n" /* eax += len */
" movl (%%ebp,%%eax,4), %%eax\n" /* eax = lcode[val+(hold&mask[op])]*/
" jmp .L_dolen\n"
".align 32,0x90\n"
".L_test_for_second_level_dist:\n"
" testb $64, %%al\n"
" jnz .L_invalid_distance_code\n" /* if ((op & 64) != 0) */
" movl $1, %%eax\n"
" xorl %%eax, %%eax\n"
" incl %%eax\n"
" shll %%cl, %%eax\n"
" decl %%eax\n"
" andl %%edx, %%eax\n" /* eax &= hold */
" addl %%ebp, %%eax\n" /* eax += this.val */
" movl 48(%%esp), %%ecx\n" /* ecx = dcode */
" addl %%ebp, %%eax\n" /* eax += dist */
" movl 36(%%esp), %%ecx\n" /* ecx = dcode */
" movl (%%ecx,%%eax,4), %%eax\n" /* eax = dcode[val+(hold&mask[op])]*/
" jmp .L_dodist\n"
".align 32,0x90\n"
".L_clip_window:\n"
" movl %%eax, %%ecx\n"
" movl 24(%%esp), %%eax\n" /* prepare for dist compare */
" movl 48(%%esp), %%eax\n" /* eax = wsize */
" negl %%ecx\n" /* nbytes = -nbytes */
" movl 32(%%esp), %%esi\n" /* from = window */
" movl 28(%%esp), %%esi\n" /* from = window */
" cmpl %%ebp, %%eax\n"
" jb .L_invalid_distance_too_far\n" /* if (dist > wsize) */
" addl %%ebp, %%ecx\n" /* nbytes = dist - nbytes */
" cmpl $0, 28(%%esp)\n"
" cmpl $0, 52(%%esp)\n"
" jne .L_wrap_around_window\n" /* if (write != 0) */
" subl %%ecx, %%eax\n"
" addl %%eax, %%esi\n" /* from += wsize - nbytes */
" movl 60(%%esp), %%eax\n"
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy1\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movl %%edi, %%esi\n"
" subl %%ebp, %%esi\n" /* from = out - dist */
" jmp .L_do_copy1\n"
" movl 64(%%esp), %%eax\n" /* eax = len */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy1\n" /* if (nbytes >= len) */
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movl %%edi, %%esi\n"
" subl %%ebp, %%esi\n" /* from = out - dist */
" jmp .L_do_copy1\n"
" jmp .L_do_copy\n"
".align 32,0x90\n"
".L_wrap_around_window:\n"
" movl 28(%%esp), %%eax\n"
" movl 52(%%esp), %%eax\n" /* eax = write */
" cmpl %%eax, %%ecx\n"
" jbe .L_contiguous_in_window\n" /* if (write >= nbytes) */
" addl 24(%%esp), %%esi\n"
" addl %%eax, %%esi\n"
" subl %%ecx, %%esi\n" /* from += wsize + write - nbytes */
" addl 48(%%esp), %%esi\n" /* from += wsize */
" addl %%eax, %%esi\n" /* from += write */
" subl %%ecx, %%esi\n" /* from -= nbytes */
" subl %%eax, %%ecx\n" /* nbytes -= write */
" movl 60(%%esp), %%eax\n"
" movl 64(%%esp), %%eax\n" /* eax = len */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy1\n" /* if (nbytes >= len) */
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movl 32(%%esp), %%esi\n" /* from = window */
" movl 28(%%esp), %%ecx\n" /* nbytes = write */
" movl 28(%%esp), %%esi\n" /* from = window */
" movl 52(%%esp), %%ecx\n" /* nbytes = write */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy1\n" /* if (nbytes >= len) */
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movl %%edi, %%esi\n"
" subl %%ebp, %%esi\n" /* from = out - dist */
" jmp .L_do_copy1\n"
" jmp .L_do_copy\n"
".align 32,0x90\n"
".L_contiguous_in_window:\n"
" addl %%eax, %%esi\n"
" subl %%ecx, %%esi\n" /* from += write - nbytes */
" movl 60(%%esp), %%eax\n"
" movl 64(%%esp), %%eax\n" /* eax = len */
" cmpl %%ecx, %%eax\n"
" jbe .L_do_copy1\n" /* if (nbytes >= len) */
" jbe .L_do_copy\n" /* if (nbytes >= len) */
" subl %%ecx, %%eax\n" /* len -= nbytes */
" rep movsb\n"
" movl %%edi, %%esi\n"
" subl %%ebp, %%esi\n" /* from = out - dist */
" jmp .L_do_copy\n" /* if (nbytes >= len) */
".L_do_copy1:\n"
".align 32,0x90\n"
".L_do_copy:\n"
" movl %%eax, %%ecx\n"
" rep movsb\n"
" movl 4(%%esp), %%esi\n" /* move in back to %esi, toss from */
" movl 44(%%esp), %%ebp\n" /* ebp = lcode */
" movl 8(%%esp), %%esi\n" /* move in back to %esi, toss from */
" movl 32(%%esp), %%ebp\n" /* ebp = lcode */
" jmp .L_while_test\n"
".L_test_for_end_of_block:\n"
" testb $32, %%al\n"
" jz .L_invalid_literal_length_code\n"
" movl $1, 68(%%esp)\n"
" movl $1, 72(%%esp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_literal_length_code:\n"
" movl $2, 68(%%esp)\n"
" movl $2, 72(%%esp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_distance_code:\n"
" movl $3, 68(%%esp)\n"
" movl $3, 72(%%esp)\n"
" jmp .L_break_loop_with_status\n"
".L_invalid_distance_too_far:\n"
" movl 4(%%esp), %%esi\n"
" movl $4, 68(%%esp)\n"
" movl 8(%%esp), %%esi\n"
" movl $4, 72(%%esp)\n"
" jmp .L_break_loop_with_status\n"
".L_break_loop:\n"
" movl $0, 68(%%esp)\n"
" movl $0, 72(%%esp)\n"
".L_break_loop_with_status:\n"
/* put in, out, bits, and hold back into ar and pop esp */
" movl %%esi, 4(%%esp)\n"
" movl %%edi, 12(%%esp)\n"
" movl %%ebx, 40(%%esp)\n"
" movl %%edx, 36(%%esp)\n"
" movl %%esi, 8(%%esp)\n" /* save in */
" movl %%edi, 16(%%esp)\n" /* save out */
" movl %%ebx, 44(%%esp)\n" /* save bits */
" movl %%edx, 40(%%esp)\n" /* save hold */
" movl 4(%%esp), %%ebp\n" /* restore esp, ebp */
" movl (%%esp), %%esp\n"
" popl %%ebp\n"
" popf\n"
:
: "m" (ar)
: "memory", "%eax", "%ebx", "%ecx", "%edx", "%esi", "%edi"
);
#elif defined( _MSC_VER )
#elif defined( _MSC_VER ) && ! defined( _M_AMD64 )
__asm {
lea eax, ar
pushfd
push ebp
mov [eax], esp
mov [eax], esp /* save esp, ebp */
mov [eax+4], ebp
mov esp, eax
mov esi, [esp+4] /* esi = in */
mov edi, [esp+12] /* edi = out */
mov edx, [esp+36] /* edx = hold */
mov ebx, [esp+40] /* ebx = bits */
mov ebp, [esp+44] /* ebp = lcode */
mov esi, [esp+8] /* esi = in */
mov edi, [esp+16] /* edi = out */
mov edx, [esp+40] /* edx = hold */
mov ebx, [esp+44] /* ebx = bits */
mov ebp, [esp+32] /* ebp = lcode */
cld
jmp L_do_loop
ALIGN 4
L_while_test:
cmp [esp+20], edi
cmp [esp+24], edi
jbe L_break_loop
cmp [esp+8], esi
cmp [esp+12], esi
jbe L_break_loop
L_do_loop:
......@@ -465,7 +828,7 @@ L_do_loop:
or edx, eax /* hold |= *((ushort *)in)++ << bits */
L_get_length_code:
mov eax, [esp+52] /* eax = lmask */
mov eax, [esp+56] /* eax = lmask */
and eax, edx /* eax &= hold */
mov eax, [ebp+eax*4] /* eax = lcode[hold & lmask] */
......@@ -481,10 +844,11 @@ L_dolen:
stosb
jmp L_while_test
ALIGN 4
L_test_for_length_base:
mov ecx, eax /* len = this */
shr ecx, 16 /* len = this.val */
mov [esp+60], ecx /* len = this */
mov [esp+64], ecx /* save len */
mov cl, al
test al, 16
......@@ -504,13 +868,14 @@ L_test_for_length_base:
mov cl, ch /* move op back to ecx */
L_add_bits_to_len:
mov eax, 1
sub bl, cl
xor eax, eax
inc eax
shl eax, cl
dec eax
sub bl, cl
and eax, edx /* eax &= hold */
shr edx, cl
add [esp+60], eax /* len += hold & mask[op] */
add [esp+64], eax /* len += hold & mask[op] */
L_decode_distance:
cmp bl, 15
......@@ -524,8 +889,8 @@ L_decode_distance:
or edx, eax /* hold |= *((ushort *)in)++ << bits */
L_get_distance_code:
mov eax, [esp+56] /* eax = dmask */
mov ecx, [esp+48] /* ecx = dcode */
mov eax, [esp+60] /* eax = dmask */
mov ecx, [esp+36] /* ecx = dcode */
and eax, edx /* eax &= hold */
mov eax, [ecx+eax*4]/* eax = dcode[hold & dmask] */
......@@ -554,200 +919,207 @@ L_dodist:
mov cl, ch /* move op back to ecx */
L_add_bits_to_dist:
mov eax, 1
sub bl, cl
xor eax, eax
inc eax
shl eax, cl
dec eax /* (1 << op) - 1 */
sub bl, cl
and eax, edx /* eax &= hold */
shr edx, cl
add ebp, eax /* dist += hold & ((1 << op) - 1) */
L_check_window:
mov [esp+4], esi /* save in so from can use it's reg */
mov [esp+8], esi /* save in so from can use it's reg */
mov eax, edi
sub eax, [esp+16] /* nbytes = out - beg */
sub eax, [esp+20] /* nbytes = out - beg */
cmp eax, ebp
jb L_clip_window /* if (dist > nbytes) 4.2% */
mov ecx, [esp+60]
mov ecx, [esp+64] /* ecx = len */
mov esi, edi
sub esi, ebp /* from = out - dist */
sub ecx, 3 /* copy from to out */
sar ecx, 1
jnc L_copy_two
rep movsw
mov al, [esi]
mov [edi], al
mov al, [esi+1]
mov ah, [esi+2]
add esi, 3
mov [edi+1], al
mov [edi+2], ah
add edi, 3
rep movsb
inc edi
mov esi, [esp+8] /* move in back to %esi, toss from */
mov ebp, [esp+32] /* ebp = lcode */
jmp L_while_test
mov esi, [esp+4] /* move in back to %esi, toss from */
mov ebp, [esp+44] /* ebp = lcode */
L_copy_two:
rep movsw
mov esi, [esp+8] /* move in back to %esi, toss from */
mov ebp, [esp+32] /* ebp = lcode */
jmp L_while_test
ALIGN 4
L_check_dist_one:
cmp ebp, 1 /* if dist 1, is a memset */
jne L_check_window
cmp [esp+16], edi
je L_check_window
dec edi
mov ecx, [esp+60]
mov al, [edi]
sub ecx, 3
mov [edi+1], al /* memset out with from[-1] */
mov [edi+2], al
mov [edi+3], al
add edi, 4
rep stosb
mov ebp, [esp+44] /* ebp = lcode */
cmp [esp+20], edi
je L_check_window /* out == beg, if outside window */
mov ecx, [esp+64] /* ecx = len */
mov al, [edi-1]
mov ah, al
sar ecx, 1
jnc L_set_two
mov [edi], al /* memset out with from[-1] */
inc edi
L_set_two:
rep stosw
mov ebp, [esp+32] /* ebp = lcode */
jmp L_while_test
ALIGN 4
L_test_for_second_level_length:
test al, 64
jnz L_test_for_end_of_block /* if ((op & 64) != 0) */
mov eax, 1
xor eax, eax
inc eax
shl eax, cl
dec eax
and eax, edx /* eax &= hold */
add eax, [esp+60] /* eax += this.val */
add eax, [esp+64] /* eax += len */
mov eax, [ebp+eax*4] /* eax = lcode[val+(hold&mask[op])]*/
jmp L_dolen
ALIGN 4
L_test_for_second_level_dist:
test al, 64
jnz L_invalid_distance_code /* if ((op & 64) != 0) */
mov eax, 1
xor eax, eax
inc eax
shl eax, cl
dec eax
and eax, edx /* eax &= hold */
add eax, ebp /* eax += this.val */
mov ecx, [esp+48] /* ecx = dcode */
add eax, ebp /* eax += dist */
mov ecx, [esp+36] /* ecx = dcode */
mov eax, [ecx+eax*4] /* eax = dcode[val+(hold&mask[op])]*/
jmp L_dodist
ALIGN 4
L_clip_window:
mov ecx, eax
mov eax, [esp+24] /* prepare for dist compare */
mov eax, [esp+48] /* eax = wsize */
neg ecx /* nbytes = -nbytes */
mov esi, [esp+32] /* from = window */
mov esi, [esp+28] /* from = window */
cmp eax, ebp
jb L_invalid_distance_too_far /* if (dist > wsize) */
add ecx, ebp /* nbytes = dist - nbytes */
cmp dword ptr [esp+28], 0
cmp dword ptr [esp+52], 0
jne L_wrap_around_window /* if (write != 0) */
sub eax, ecx
add esi, eax /* from += wsize - nbytes */
mov eax, [esp+60]
cmp eax, ecx
jbe L_do_copy1 /* if (nbytes >= len) */
sub eax, ecx /* len -= nbytes */
rep movsb
mov esi, edi
sub esi, ebp /* from = out - dist */
jmp L_do_copy1
mov eax, [esp+64] /* eax = len */
cmp eax, ecx
jbe L_do_copy1 /* if (nbytes >= len) */
jbe L_do_copy /* if (nbytes >= len) */
sub eax, ecx /* len -= nbytes */
rep movsb
mov esi, edi
sub esi, ebp /* from = out - dist */
jmp L_do_copy1
jmp L_do_copy
ALIGN 4
L_wrap_around_window:
mov eax, [esp+28]
mov eax, [esp+52] /* eax = write */
cmp ecx, eax
jbe L_contiguous_in_window /* if (write >= nbytes) */
add esi, [esp+24]
add esi, eax
sub esi, ecx /* from += wsize + write - nbytes */
add esi, [esp+48] /* from += wsize */
add esi, eax /* from += write */
sub esi, ecx /* from -= nbytes */
sub ecx, eax /* nbytes -= write */
mov eax, [esp+60]
mov eax, [esp+64] /* eax = len */
cmp eax, ecx
jbe L_do_copy1 /* if (nbytes >= len) */
jbe L_do_copy /* if (nbytes >= len) */
sub eax, ecx /* len -= nbytes */
rep movsb
mov esi, [esp+32] /* from = window */
mov ecx, [esp+28] /* nbytes = write */
mov esi, [esp+28] /* from = window */
mov ecx, [esp+52] /* nbytes = write */
cmp eax, ecx
jbe L_do_copy1 /* if (nbytes >= len) */
jbe L_do_copy /* if (nbytes >= len) */
sub eax, ecx /* len -= nbytes */
rep movsb
mov esi, edi
sub esi, ebp /* from = out - dist */
jmp L_do_copy1
jmp L_do_copy
ALIGN 4
L_contiguous_in_window:
add esi, eax
sub esi, ecx /* from += write - nbytes */
mov eax, [esp+60]
mov eax, [esp+64] /* eax = len */
cmp eax, ecx
jbe L_do_copy1 /* if (nbytes >= len) */
jbe L_do_copy /* if (nbytes >= len) */
sub eax, ecx /* len -= nbytes */
rep movsb
mov esi, edi
sub esi, ebp /* from = out - dist */
jmp L_do_copy
L_do_copy1:
ALIGN 4
L_do_copy:
mov ecx, eax
rep movsb
mov esi, [esp+4] /* move in back to %esi, toss from */
mov ebp, [esp+44] /* ebp = lcode */
mov esi, [esp+8] /* move in back to %esi, toss from */
mov ebp, [esp+32] /* ebp = lcode */
jmp L_while_test
L_test_for_end_of_block:
test al, 32
jz L_invalid_literal_length_code
mov dword ptr [esp+68], 1
mov dword ptr [esp+72], 1
jmp L_break_loop_with_status
L_invalid_literal_length_code:
mov dword ptr [esp+68], 2
mov dword ptr [esp+72], 2
jmp L_break_loop_with_status
L_invalid_distance_code:
mov dword ptr [esp+68], 3
mov dword ptr [esp+72], 3
jmp L_break_loop_with_status
L_invalid_distance_too_far:
mov esi, [esp+4]
mov dword ptr [esp+68], 4
mov dword ptr [esp+72], 4
jmp L_break_loop_with_status
L_break_loop:
mov dword ptr [esp+68], 0
mov dword ptr [esp+72], 0
L_break_loop_with_status:
/* put in, out, bits, and hold back into ar and pop esp */
mov [esp+4], esi
mov [esp+12], edi
mov [esp+40], ebx
mov [esp+36], edx
mov [esp+8], esi /* save in */
mov [esp+16], edi /* save out */
mov [esp+44], ebx /* save bits */
mov [esp+40], edx /* save hold */
mov ebp, [esp+4] /* restore esp, ebp */
mov esp, [esp]
pop ebp
popfd
}
#else
#error "x86 architecture not defined"
#endif
if (ar.status > 1) {
......@@ -772,10 +1144,12 @@ L_break_loop_with_status:
/* update state and return */
strm->next_in = ar.in;
strm->next_out = ar.out;
strm->avail_in = (unsigned)(ar.in < ar.last ? 5 + (ar.last - ar.in) :
5 - (ar.in - ar.last));
strm->avail_out = (unsigned)(ar.out < ar.end ? 257 + (ar.end - ar.out) :
257 - (ar.out - ar.end));
strm->avail_in = (unsigned)(ar.in < ar.last ?
PAD_AVAIL_IN + (ar.last - ar.in) :
PAD_AVAIL_IN - (ar.in - ar.last));
strm->avail_out = (unsigned)(ar.out < ar.end ?
PAD_AVAIL_OUT + (ar.end - ar.out) :
PAD_AVAIL_OUT - (ar.out - ar.end));
state->hold = ar.hold;
state->bits = ar.bits;
return;
......
zlib/contrib/inflate86/inffast.S
......@@ -188,17 +188,8 @@
/*
* typedef enum inflate_mode consts, in inflate.h
*/
#ifndef NO_GUNZIP
#define GUNZIP
#endif
#ifdef GUNZIP
#define INFLATE_MODE_TYPE 11 /* state->mode flags enum-ed in inflate.h */
#define INFLATE_MODE_BAD 26
#else
#define INFLATE_MODE_TYPE 3
#define INFLATE_MODE_BAD 17
#endif
#if ! defined( USE_MMX ) && ! defined( NO_MMX )
......
zlib/contrib/masm686/match.asm
; match.asm -- Pentium-Pro optimized version of longest_match()
;
; Updated for zlib 1.1.3 and converted to MASM 6.1x
; Copyright (C) 2000 Dan Higdon <hdan@kinesoft.com>
; and Chuck Walbourn <chuckw@kinesoft.com>
; Corrections by Cosmin Truta <cosmint@cs.ubbcluj.ro>
;
; This is free software; you can redistribute it and/or modify it
; under the terms of the GNU General Public License.
; Based on match.S
; Written for zlib 1.1.2
; Copyright (C) 1998 Brian Raiter <breadbox@muppetlabs.com>
.686P
.MODEL FLAT
;===========================================================================
; EQUATES
;===========================================================================
MAX_MATCH EQU 258
MIN_MATCH EQU 3
MIN_LOOKAHEAD EQU (MAX_MATCH + MIN_MATCH + 1)
MAX_MATCH_8 EQU ((MAX_MATCH + 7) AND (NOT 7))
;===========================================================================
; STRUCTURES
;===========================================================================
; This STRUCT assumes a 4-byte alignment
DEFLATE_STATE STRUCT
ds_strm dd ?
ds_status dd ?
ds_pending_buf dd ?
ds_pending_buf_size dd ?
ds_pending_out dd ?
ds_pending dd ?
ds_wrap dd ?
ds_data_type db ?
ds_method db ?
db ? ; padding
db ? ; padding
ds_last_flush dd ?
ds_w_size dd ? ; used
ds_w_bits dd ?
ds_w_mask dd ? ; used
ds_window dd ? ; used
ds_window_size dd ?
ds_prev dd ? ; used
ds_head dd ?
ds_ins_h dd ?
ds_hash_size dd ?
ds_hash_bits dd ?
ds_hash_mask dd ?
ds_hash_shift dd ?
ds_block_start dd ?
ds_match_length dd ? ; used
ds_prev_match dd ? ; used
ds_match_available dd ?
ds_strstart dd ? ; used
ds_match_start dd ? ; used
ds_lookahead dd ? ; used
ds_prev_length dd ? ; used
ds_max_chain_length dd ? ; used
ds_max_laxy_match dd ?
ds_level dd ?
ds_strategy dd ?
ds_good_match dd ? ; used
ds_nice_match dd ? ; used
; Don't need anymore of the struct for match
DEFLATE_STATE ENDS
;===========================================================================
; CODE
;===========================================================================
_TEXT SEGMENT
;---------------------------------------------------------------------------
; match_init
;---------------------------------------------------------------------------
ALIGN 4
PUBLIC _match_init
_match_init PROC
; no initialization needed
ret
_match_init ENDP
;---------------------------------------------------------------------------
; uInt longest_match(deflate_state *deflatestate, IPos curmatch)
;---------------------------------------------------------------------------
ALIGN 4
PUBLIC _longest_match
_longest_match PROC
; Since this code uses EBP for a scratch register, the stack frame must
; be manually constructed and referenced relative to the ESP register.
; Stack image
; Variables
chainlenwmask = 0 ; high word: current chain len
; low word: s->wmask
window = 4 ; local copy of s->window
windowbestlen = 8 ; s->window + bestlen
scanend = 12 ; last two bytes of string
scanstart = 16 ; first two bytes of string
scanalign = 20 ; dword-misalignment of string
nicematch = 24 ; a good enough match size
bestlen = 28 ; size of best match so far
scan = 32 ; ptr to string wanting match
varsize = 36 ; number of bytes (also offset to last saved register)
; Saved Registers (actually pushed into place)
ebx_save = 36
edi_save = 40
esi_save = 44
ebp_save = 48
; Parameters
retaddr = 52
deflatestate = 56
curmatch = 60
; Save registers that the compiler may be using
push ebp
push edi
push esi
push ebx
; Allocate local variable space
sub esp,varsize
; Retrieve the function arguments. ecx will hold cur_match
; throughout the entire function. edx will hold the pointer to the
; deflate_state structure during the function's setup (before
; entering the main loop).
mov edx, [esp+deflatestate]
ASSUME edx:PTR DEFLATE_STATE
mov ecx, [esp+curmatch]
; uInt wmask = s->w_mask;
; unsigned chain_length = s->max_chain_length;
; if (s->prev_length >= s->good_match) {
; chain_length >>= 2;
; }
mov eax, [edx].ds_prev_length
mov ebx, [edx].ds_good_match
cmp eax, ebx
mov eax, [edx].ds_w_mask
mov ebx, [edx].ds_max_chain_length
jl SHORT LastMatchGood
shr ebx, 2
LastMatchGood:
; chainlen is decremented once beforehand so that the function can
; use the sign flag instead of the zero flag for the exit test.
; It is then shifted into the high word, to make room for the wmask
; value, which it will always accompany.
dec ebx
shl ebx, 16
or ebx, eax
mov [esp+chainlenwmask], ebx
; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
mov eax, [edx].ds_nice_match
mov ebx, [edx].ds_lookahead
cmp ebx, eax
jl SHORT LookaheadLess
mov ebx, eax
LookaheadLess:
mov [esp+nicematch], ebx
;/* register Bytef *scan = s->window + s->strstart; */
mov esi, [edx].ds_window
mov [esp+window], esi
mov ebp, [edx].ds_strstart
lea edi, [esi+ebp]
mov [esp+scan],edi
;/* Determine how many bytes the scan ptr is off from being */
;/* dword-aligned. */
mov eax, edi
neg eax
and eax, 3
mov [esp+scanalign], eax
;/* IPos limit = s->strstart > (IPos)MAX_DIST(s) ? */
;/* s->strstart - (IPos)MAX_DIST(s) : NIL; */
mov eax, [edx].ds_w_size
sub eax, MIN_LOOKAHEAD
sub ebp, eax
jg SHORT LimitPositive
xor ebp, ebp
LimitPositive:
;/* int best_len = s->prev_length; */
mov eax, [edx].ds_prev_length
mov [esp+bestlen], eax
;/* Store the sum of s->window + best_len in %esi locally, and in %esi. */
add esi, eax
mov [esp+windowbestlen], esi
;/* register ush scan_start = *(ushf*)scan; */
;/* register ush scan_end = *(ushf*)(scan+best_len-1); */
;/* Posf *prev = s->prev; */
movzx ebx, WORD PTR[edi]
mov [esp+scanstart], ebx
movzx ebx, WORD PTR[eax+edi-1]
mov [esp+scanend], ebx
mov edi, [edx].ds_prev
;/* Jump into the main loop. */
mov edx, [esp+chainlenwmask]
jmp SHORT LoopEntry
;/* do {
; * match = s->window + cur_match;
; * if (*(ushf*)(match+best_len-1) != scan_end ||
; * *(ushf*)match != scan_start) continue;
; * [...]
; * } while ((cur_match = prev[cur_match & wmask]) > limit
; * && --chain_length != 0);
; *
; * Here is the inner loop of the function. The function will spend the
; * majority of its time in this loop, and majority of that time will
; * be spent in the first ten instructions.
; *
; * Within this loop:
; * %ebx = scanend
; * %ecx = curmatch
; * %edx = chainlenwmask - i.e., ((chainlen << 16) | wmask)
; * %esi = windowbestlen - i.e., (window + bestlen)
; * %edi = prev
; * %ebp = limit
; */
ALIGN 4
LookupLoop:
and ecx, edx
movzx ecx, WORD PTR[edi+ecx*2]
cmp ecx, ebp
jbe LeaveNow
sub edx, 000010000H
js LeaveNow
LoopEntry:
movzx eax, WORD PTR[esi+ecx-1]
cmp eax, ebx
jnz SHORT LookupLoop
mov eax, [esp+window]
movzx eax, WORD PTR[eax+ecx]
cmp eax, [esp+scanstart]
jnz SHORT LookupLoop
;/* Store the current value of chainlen. */
mov [esp+chainlenwmask], edx
;/* Point %edi to the string under scrutiny, and %esi to the string we */
;/* are hoping to match it up with. In actuality, %esi and %edi are */
;/* both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and %edx is */
;/* initialized to -(MAX_MATCH_8 - scanalign). */
mov esi, [esp+window]
mov edi, [esp+scan]
add esi, ecx
mov eax, [esp+scanalign]
mov edx, -MAX_MATCH_8
lea edi, [edi+eax+MAX_MATCH_8]
lea esi, [esi+eax+MAX_MATCH_8]
;/* Test the strings for equality, 8 bytes at a time. At the end,
; * adjust %edx so that it is offset to the exact byte that mismatched.
; *
; * We already know at this point that the first three bytes of the
; * strings match each other, and they can be safely passed over before
; * starting the compare loop. So what this code does is skip over 0-3
; * bytes, as much as necessary in order to dword-align the %edi
; * pointer. (%esi will still be misaligned three times out of four.)
; *
; * It should be confessed that this loop usually does not represent
; * much of the total running time. Replacing it with a more
; * straightforward "rep cmpsb" would not drastically degrade
; * performance.
; */
LoopCmps:
mov eax, DWORD PTR[esi+edx]
xor eax, DWORD PTR[edi+edx]
jnz SHORT LeaveLoopCmps
mov eax, DWORD PTR[esi+edx+4]
xor eax, DWORD PTR[edi+edx+4]
jnz SHORT LeaveLoopCmps4
add edx, 8
jnz SHORT LoopCmps
jmp LenMaximum
ALIGN 4
LeaveLoopCmps4:
add edx, 4
LeaveLoopCmps:
test eax, 00000FFFFH
jnz SHORT LenLower
add edx, 2
shr eax, 16
LenLower:
sub al, 1
adc edx, 0
;/* Calculate the length of the match. If it is longer than MAX_MATCH, */
;/* then automatically accept it as the best possible match and leave. */
lea eax, [edi+edx]
mov edi, [esp+scan]
sub eax, edi
cmp eax, MAX_MATCH
jge SHORT LenMaximum
;/* If the length of the match is not longer than the best match we */
;/* have so far, then forget it and return to the lookup loop. */
mov edx, [esp+deflatestate]
mov ebx, [esp+bestlen]
cmp eax, ebx
jg SHORT LongerMatch
mov esi, [esp+windowbestlen]
mov edi, [edx].ds_prev
mov ebx, [esp+scanend]
mov edx, [esp+chainlenwmask]
jmp LookupLoop
ALIGN 4
;/* s->match_start = cur_match; */
;/* best_len = len; */
;/* if (len >= nice_match) break; */
;/* scan_end = *(ushf*)(scan+best_len-1); */
LongerMatch:
mov ebx, [esp+nicematch]
mov [esp+bestlen], eax
mov [edx].ds_match_start, ecx
cmp eax, ebx
jge SHORT LeaveNow
mov esi, [esp+window]
add esi, eax
mov [esp+windowbestlen], esi
movzx ebx, WORD PTR[edi+eax-1]
mov edi, [edx].ds_prev
mov [esp+scanend], ebx
mov edx, [esp+chainlenwmask]
jmp LookupLoop
ALIGN 4
;/* Accept the current string, with the maximum possible length. */
LenMaximum:
mov edx, [esp+deflatestate]
mov DWORD PTR[esp+bestlen], MAX_MATCH
mov [edx].ds_match_start, ecx
;/* if ((uInt)best_len <= s->lookahead) return (uInt)best_len; */
;/* return s->lookahead; */
LeaveNow:
mov edx, [esp+deflatestate]
mov ebx, [esp+bestlen]
mov eax, [edx].ds_lookahead
cmp ebx, eax
jg SHORT LookaheadRet
mov eax, ebx
LookaheadRet:
; Restore the stack and return from whence we came.
add esp, varsize
pop ebx
pop esi
pop edi
pop ebp
ret
_longest_match ENDP
_TEXT ENDS
END
; match.asm -- Pentium-Pro optimized version of longest_match()
;
; Updated for zlib 1.1.3 and converted to MASM 6.1x
; Copyright (C) 2000 Dan Higdon <hdan@kinesoft.com>
; and Chuck Walbourn <chuckw@kinesoft.com>
; Corrections by Cosmin Truta <cosmint@cs.ubbcluj.ro>
;
; This is free software; you can redistribute it and/or modify it
; under the terms of the GNU General Public License.
; Based on match.S
; Written for zlib 1.1.2
; Copyright (C) 1998 Brian Raiter <breadbox@muppetlabs.com>
;
; Modified by Gilles Vollant (2005) for add gzhead and gzindex
.686P
.MODEL FLAT
;===========================================================================
; EQUATES
;===========================================================================
MAX_MATCH EQU 258
MIN_MATCH EQU 3
MIN_LOOKAHEAD EQU (MAX_MATCH + MIN_MATCH + 1)
MAX_MATCH_8 EQU ((MAX_MATCH + 7) AND (NOT 7))
;===========================================================================
; STRUCTURES
;===========================================================================
; This STRUCT assumes a 4-byte alignment
DEFLATE_STATE STRUCT
ds_strm dd ?
ds_status dd ?
ds_pending_buf dd ?
ds_pending_buf_size dd ?
ds_pending_out dd ?
ds_pending dd ?
ds_wrap dd ?
; gzhead and gzindex are added in zlib 1.2.2.2 (see deflate.h)
ds_gzhead dd ?
ds_gzindex dd ?
ds_data_type db ?
ds_method db ?
db ? ; padding
db ? ; padding
ds_last_flush dd ?
ds_w_size dd ? ; used
ds_w_bits dd ?
ds_w_mask dd ? ; used
ds_window dd ? ; used
ds_window_size dd ?
ds_prev dd ? ; used
ds_head dd ?
ds_ins_h dd ?
ds_hash_size dd ?
ds_hash_bits dd ?
ds_hash_mask dd ?
ds_hash_shift dd ?
ds_block_start dd ?
ds_match_length dd ? ; used
ds_prev_match dd ? ; used
ds_match_available dd ?
ds_strstart dd ? ; used
ds_match_start dd ? ; used
ds_lookahead dd ? ; used
ds_prev_length dd ? ; used
ds_max_chain_length dd ? ; used
ds_max_laxy_match dd ?
ds_level dd ?
ds_strategy dd ?
ds_good_match dd ? ; used
ds_nice_match dd ? ; used
; Don't need anymore of the struct for match
DEFLATE_STATE ENDS
;===========================================================================
; CODE
;===========================================================================
_TEXT SEGMENT
;---------------------------------------------------------------------------
; match_init
;---------------------------------------------------------------------------
ALIGN 4
PUBLIC _match_init
_match_init PROC
; no initialization needed
ret
_match_init ENDP
;---------------------------------------------------------------------------
; uInt longest_match(deflate_state *deflatestate, IPos curmatch)
;---------------------------------------------------------------------------
ALIGN 4
PUBLIC _longest_match
_longest_match PROC
; Since this code uses EBP for a scratch register, the stack frame must
; be manually constructed and referenced relative to the ESP register.
; Stack image
; Variables
chainlenwmask = 0 ; high word: current chain len
; low word: s->wmask
window = 4 ; local copy of s->window
windowbestlen = 8 ; s->window + bestlen
scanend = 12 ; last two bytes of string
scanstart = 16 ; first two bytes of string
scanalign = 20 ; dword-misalignment of string
nicematch = 24 ; a good enough match size
bestlen = 28 ; size of best match so far
scan = 32 ; ptr to string wanting match
varsize = 36 ; number of bytes (also offset to last saved register)
; Saved Registers (actually pushed into place)
ebx_save = 36
edi_save = 40
esi_save = 44
ebp_save = 48
; Parameters
retaddr = 52
deflatestate = 56
curmatch = 60
; Save registers that the compiler may be using
push ebp
push edi
push esi
push ebx
; Allocate local variable space
sub esp,varsize
; Retrieve the function arguments. ecx will hold cur_match
; throughout the entire function. edx will hold the pointer to the
; deflate_state structure during the function's setup (before
; entering the main loop).
mov edx, [esp+deflatestate]
ASSUME edx:PTR DEFLATE_STATE
mov ecx, [esp+curmatch]
; uInt wmask = s->w_mask;
; unsigned chain_length = s->max_chain_length;
; if (s->prev_length >= s->good_match) {
; chain_length >>= 2;
; }
mov eax, [edx].ds_prev_length
mov ebx, [edx].ds_good_match
cmp eax, ebx
mov eax, [edx].ds_w_mask
mov ebx, [edx].ds_max_chain_length
jl SHORT LastMatchGood
shr ebx, 2
LastMatchGood:
; chainlen is decremented once beforehand so that the function can
; use the sign flag instead of the zero flag for the exit test.
; It is then shifted into the high word, to make room for the wmask
; value, which it will always accompany.
dec ebx
shl ebx, 16
or ebx, eax
mov [esp+chainlenwmask], ebx
; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
mov eax, [edx].ds_nice_match
mov ebx, [edx].ds_lookahead
cmp ebx, eax
jl SHORT LookaheadLess
mov ebx, eax
LookaheadLess:
mov [esp+nicematch], ebx
;/* register Bytef *scan = s->window + s->strstart; */
mov esi, [edx].ds_window
mov [esp+window], esi
mov ebp, [edx].ds_strstart
lea edi, [esi+ebp]
mov [esp+scan],edi
;/* Determine how many bytes the scan ptr is off from being */
;/* dword-aligned. */
mov eax, edi
neg eax
and eax, 3
mov [esp+scanalign], eax
;/* IPos limit = s->strstart > (IPos)MAX_DIST(s) ? */
;/* s->strstart - (IPos)MAX_DIST(s) : NIL; */
mov eax, [edx].ds_w_size
sub eax, MIN_LOOKAHEAD
sub ebp, eax
jg SHORT LimitPositive
xor ebp, ebp
LimitPositive:
;/* int best_len = s->prev_length; */
mov eax, [edx].ds_prev_length
mov [esp+bestlen], eax
;/* Store the sum of s->window + best_len in %esi locally, and in %esi. */
add esi, eax
mov [esp+windowbestlen], esi
;/* register ush scan_start = *(ushf*)scan; */
;/* register ush scan_end = *(ushf*)(scan+best_len-1); */
;/* Posf *prev = s->prev; */
movzx ebx, WORD PTR[edi]
mov [esp+scanstart], ebx
movzx ebx, WORD PTR[eax+edi-1]
mov [esp+scanend], ebx
mov edi, [edx].ds_prev
;/* Jump into the main loop. */
mov edx, [esp+chainlenwmask]
jmp SHORT LoopEntry
;/* do {
; * match = s->window + cur_match;
; * if (*(ushf*)(match+best_len-1) != scan_end ||
; * *(ushf*)match != scan_start) continue;
; * [...]
; * } while ((cur_match = prev[cur_match & wmask]) > limit
; * && --chain_length != 0);
; *
; * Here is the inner loop of the function. The function will spend the
; * majority of its time in this loop, and majority of that time will
; * be spent in the first ten instructions.
; *
; * Within this loop:
; * %ebx = scanend
; * %ecx = curmatch
; * %edx = chainlenwmask - i.e., ((chainlen << 16) | wmask)
; * %esi = windowbestlen - i.e., (window + bestlen)
; * %edi = prev
; * %ebp = limit
; */
ALIGN 4
LookupLoop:
and ecx, edx
movzx ecx, WORD PTR[edi+ecx*2]
cmp ecx, ebp
jbe LeaveNow
sub edx, 000010000H
js LeaveNow
LoopEntry:
movzx eax, WORD PTR[esi+ecx-1]
cmp eax, ebx
jnz SHORT LookupLoop
mov eax, [esp+window]
movzx eax, WORD PTR[eax+ecx]
cmp eax, [esp+scanstart]
jnz SHORT LookupLoop
;/* Store the current value of chainlen. */
mov [esp+chainlenwmask], edx
;/* Point %edi to the string under scrutiny, and %esi to the string we */
;/* are hoping to match it up with. In actuality, %esi and %edi are */
;/* both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and %edx is */
;/* initialized to -(MAX_MATCH_8 - scanalign). */
mov esi, [esp+window]
mov edi, [esp+scan]
add esi, ecx
mov eax, [esp+scanalign]
mov edx, -MAX_MATCH_8
lea edi, [edi+eax+MAX_MATCH_8]
lea esi, [esi+eax+MAX_MATCH_8]
;/* Test the strings for equality, 8 bytes at a time. At the end,
; * adjust %edx so that it is offset to the exact byte that mismatched.
; *
; * We already know at this point that the first three bytes of the
; * strings match each other, and they can be safely passed over before
; * starting the compare loop. So what this code does is skip over 0-3
; * bytes, as much as necessary in order to dword-align the %edi
; * pointer. (%esi will still be misaligned three times out of four.)
; *
; * It should be confessed that this loop usually does not represent
; * much of the total running time. Replacing it with a more
; * straightforward "rep cmpsb" would not drastically degrade
; * performance.
; */
LoopCmps:
mov eax, DWORD PTR[esi+edx]
xor eax, DWORD PTR[edi+edx]
jnz SHORT LeaveLoopCmps
mov eax, DWORD PTR[esi+edx+4]
xor eax, DWORD PTR[edi+edx+4]
jnz SHORT LeaveLoopCmps4
add edx, 8
jnz SHORT LoopCmps
jmp LenMaximum
ALIGN 4
LeaveLoopCmps4:
add edx, 4
LeaveLoopCmps:
test eax, 00000FFFFH
jnz SHORT LenLower
add edx, 2
shr eax, 16
LenLower:
sub al, 1
adc edx, 0
;/* Calculate the length of the match. If it is longer than MAX_MATCH, */
;/* then automatically accept it as the best possible match and leave. */
lea eax, [edi+edx]
mov edi, [esp+scan]
sub eax, edi
cmp eax, MAX_MATCH
jge SHORT LenMaximum
;/* If the length of the match is not longer than the best match we */
;/* have so far, then forget it and return to the lookup loop. */
mov edx, [esp+deflatestate]
mov ebx, [esp+bestlen]
cmp eax, ebx
jg SHORT LongerMatch
mov esi, [esp+windowbestlen]
mov edi, [edx].ds_prev
mov ebx, [esp+scanend]
mov edx, [esp+chainlenwmask]
jmp LookupLoop
ALIGN 4
;/* s->match_start = cur_match; */
;/* best_len = len; */
;/* if (len >= nice_match) break; */
;/* scan_end = *(ushf*)(scan+best_len-1); */
LongerMatch:
mov ebx, [esp+nicematch]
mov [esp+bestlen], eax
mov [edx].ds_match_start, ecx
cmp eax, ebx
jge SHORT LeaveNow
mov esi, [esp+window]
add esi, eax
mov [esp+windowbestlen], esi
movzx ebx, WORD PTR[edi+eax-1]
mov edi, [edx].ds_prev
mov [esp+scanend], ebx
mov edx, [esp+chainlenwmask]
jmp LookupLoop
ALIGN 4
;/* Accept the current string, with the maximum possible length. */
LenMaximum:
mov edx, [esp+deflatestate]
mov DWORD PTR[esp+bestlen], MAX_MATCH
mov [edx].ds_match_start, ecx
;/* if ((uInt)best_len <= s->lookahead) return (uInt)best_len; */
;/* return s->lookahead; */
LeaveNow:
mov edx, [esp+deflatestate]
mov ebx, [esp+bestlen]
mov eax, [edx].ds_lookahead
cmp ebx, eax
jg SHORT LookaheadRet
mov eax, ebx
LookaheadRet:
; Restore the stack and return from whence we came.
add esp, varsize
pop ebx
pop esi
pop edi
pop ebp
ret
_longest_match ENDP
_TEXT ENDS
END
zlib/contrib/masmx86/gvmat32.asm
;
; gvmat32.asm -- Asm portion of the optimized longest_match for 32 bits x86
; Copyright (C) 1995-1996 Jean-loup Gailly and Gilles Vollant.
; File written by Gilles Vollant, by modifiying the longest_match
; from Jean-loup Gailly in deflate.c
; It need wmask == 0x7fff
; (assembly code is faster with a fixed wmask)
;
; For Visual C++ 4.2 and ML 6.11c (version in directory \MASM611C of Win95 DDK)
; I compile with : "ml /coff /Zi /c gvmat32.asm"
;
;uInt longest_match_7fff(s, cur_match)
; deflate_state *s;
; IPos cur_match; /* current match */
NbStack equ 76
cur_match equ dword ptr[esp+NbStack-0]
str_s equ dword ptr[esp+NbStack-4]
; 5 dword on top (ret,ebp,esi,edi,ebx)
adrret equ dword ptr[esp+NbStack-8]
pushebp equ dword ptr[esp+NbStack-12]
pushedi equ dword ptr[esp+NbStack-16]
pushesi equ dword ptr[esp+NbStack-20]
pushebx equ dword ptr[esp+NbStack-24]
chain_length equ dword ptr [esp+NbStack-28]
limit equ dword ptr [esp+NbStack-32]
best_len equ dword ptr [esp+NbStack-36]
window equ dword ptr [esp+NbStack-40]
prev equ dword ptr [esp+NbStack-44]
scan_start equ word ptr [esp+NbStack-48]
wmask equ dword ptr [esp+NbStack-52]
match_start_ptr equ dword ptr [esp+NbStack-56]
nice_match equ dword ptr [esp+NbStack-60]
scan equ dword ptr [esp+NbStack-64]
windowlen equ dword ptr [esp+NbStack-68]
match_start equ dword ptr [esp+NbStack-72]
strend equ dword ptr [esp+NbStack-76]
NbStackAdd equ (NbStack-24)
.386p
name gvmatch
.MODEL FLAT
; all the +4 offsets are due to the addition of pending_buf_size (in zlib
; in the deflate_state structure since the asm code was first written
; (if you compile with zlib 1.0.4 or older, remove the +4).
; Note : these value are good with a 8 bytes boundary pack structure
dep_chain_length equ 70h+4
dep_window equ 2ch+4
dep_strstart equ 60h+4
dep_prev_length equ 6ch+4
dep_nice_match equ 84h+4
dep_w_size equ 20h+4
dep_prev equ 34h+4
dep_w_mask equ 28h+4
dep_good_match equ 80h+4
dep_match_start equ 64h+4
dep_lookahead equ 68h+4
_TEXT segment
IFDEF NOUNDERLINE
public longest_match_7fff
public longest_match_686
; public match_init
ELSE
public _longest_match_7fff
public _longest_match_686
; public _match_init
ENDIF
MAX_MATCH equ 258
MIN_MATCH equ 3
MIN_LOOKAHEAD equ (MAX_MATCH+MIN_MATCH+1)
IFDEF NOUNDERLINE
;match_init proc near
; ret
;match_init endp
ELSE
;_match_init proc near
; ret
;_match_init endp
ENDIF
IFDEF NOUNDERLINE
longest_match_7fff proc near
ELSE
_longest_match_7fff proc near
ENDIF
mov edx,[esp+4]
push ebp
push edi
push esi
push ebx
sub esp,NbStackAdd
; initialize or check the variables used in match.asm.
mov ebp,edx
; chain_length = s->max_chain_length
; if (prev_length>=good_match) chain_length >>= 2
mov edx,[ebp+dep_chain_length]
mov ebx,[ebp+dep_prev_length]
cmp [ebp+dep_good_match],ebx
ja noshr
shr edx,2
noshr:
; we increment chain_length because in the asm, the --chain_lenght is in the beginning of the loop
inc edx
mov edi,[ebp+dep_nice_match]
mov chain_length,edx
mov eax,[ebp+dep_lookahead]
cmp eax,edi
; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
jae nolookaheadnicematch
mov edi,eax
nolookaheadnicematch:
; best_len = s->prev_length
mov best_len,ebx
; window = s->window
mov esi,[ebp+dep_window]
mov ecx,[ebp+dep_strstart]
mov window,esi
mov nice_match,edi
; scan = window + strstart
add esi,ecx
mov scan,esi
; dx = *window
mov dx,word ptr [esi]
; bx = *(window+best_len-1)
mov bx,word ptr [esi+ebx-1]
add esi,MAX_MATCH-1
; scan_start = *scan
mov scan_start,dx
; strend = scan + MAX_MATCH-1
mov strend,esi
; bx = scan_end = *(window+best_len-1)
; IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
; s->strstart - (IPos)MAX_DIST(s) : NIL;
mov esi,[ebp+dep_w_size]
sub esi,MIN_LOOKAHEAD
; here esi = MAX_DIST(s)
sub ecx,esi
ja nodist
xor ecx,ecx
nodist:
mov limit,ecx
; prev = s->prev
mov edx,[ebp+dep_prev]
mov prev,edx
;
mov edx,dword ptr [ebp+dep_match_start]
mov bp,scan_start
mov eax,cur_match
mov match_start,edx
mov edx,window
mov edi,edx
add edi,best_len
mov esi,prev
dec edi
; windowlen = window + best_len -1
mov windowlen,edi
jmp beginloop2
align 4
; here, in the loop
; eax = ax = cur_match
; ecx = limit
; bx = scan_end
; bp = scan_start
; edi = windowlen (window + best_len -1)
; esi = prev
;// here; chain_length <=16
normalbeg0add16:
add chain_length,16
jz exitloop
normalbeg0:
cmp word ptr[edi+eax],bx
je normalbeg2noroll
rcontlabnoroll:
; cur_match = prev[cur_match & wmask]
and eax,7fffh
mov ax,word ptr[esi+eax*2]
; if cur_match > limit, go to exitloop
cmp ecx,eax
jnb exitloop
; if --chain_length != 0, go to exitloop
dec chain_length
jnz normalbeg0
jmp exitloop
normalbeg2noroll:
; if (scan_start==*(cur_match+window)) goto normalbeg2
cmp bp,word ptr[edx+eax]
jne rcontlabnoroll
jmp normalbeg2
contloop3:
mov edi,windowlen
; cur_match = prev[cur_match & wmask]
and eax,7fffh
mov ax,word ptr[esi+eax*2]
; if cur_match > limit, go to exitloop
cmp ecx,eax
jnbexitloopshort1:
jnb exitloop
; if --chain_length != 0, go to exitloop
; begin the main loop
beginloop2:
sub chain_length,16+1
; if chain_length <=16, don't use the unrolled loop
jna normalbeg0add16
do16:
cmp word ptr[edi+eax],bx
je normalbeg2dc0
maccn MACRO lab
and eax,7fffh
mov ax,word ptr[esi+eax*2]
cmp ecx,eax
jnb exitloop
cmp word ptr[edi+eax],bx
je lab
ENDM
rcontloop0:
maccn normalbeg2dc1
rcontloop1:
maccn normalbeg2dc2
rcontloop2:
maccn normalbeg2dc3
rcontloop3:
maccn normalbeg2dc4
rcontloop4:
maccn normalbeg2dc5
rcontloop5:
maccn normalbeg2dc6
rcontloop6:
maccn normalbeg2dc7
rcontloop7:
maccn normalbeg2dc8
rcontloop8:
maccn normalbeg2dc9
rcontloop9:
maccn normalbeg2dc10
rcontloop10:
maccn short normalbeg2dc11
rcontloop11:
maccn short normalbeg2dc12
rcontloop12:
maccn short normalbeg2dc13
rcontloop13:
maccn short normalbeg2dc14
rcontloop14:
maccn short normalbeg2dc15
rcontloop15:
and eax,7fffh
mov ax,word ptr[esi+eax*2]
cmp ecx,eax
jnb exitloop
sub chain_length,16
ja do16
jmp normalbeg0add16
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
normbeg MACRO rcontlab,valsub
; if we are here, we know that *(match+best_len-1) == scan_end
cmp bp,word ptr[edx+eax]
; if (match != scan_start) goto rcontlab
jne rcontlab
; calculate the good chain_length, and we'll compare scan and match string
add chain_length,16-valsub
jmp iseq
ENDM
normalbeg2dc11:
normbeg rcontloop11,11
normalbeg2dc12:
normbeg short rcontloop12,12
normalbeg2dc13:
normbeg short rcontloop13,13
normalbeg2dc14:
normbeg short rcontloop14,14
normalbeg2dc15:
normbeg short rcontloop15,15
normalbeg2dc10:
normbeg rcontloop10,10
normalbeg2dc9:
normbeg rcontloop9,9
normalbeg2dc8:
normbeg rcontloop8,8
normalbeg2dc7:
normbeg rcontloop7,7
normalbeg2dc6:
normbeg rcontloop6,6
normalbeg2dc5:
normbeg rcontloop5,5
normalbeg2dc4:
normbeg rcontloop4,4
normalbeg2dc3:
normbeg rcontloop3,3
normalbeg2dc2:
normbeg rcontloop2,2
normalbeg2dc1:
normbeg rcontloop1,1
normalbeg2dc0:
normbeg rcontloop0,0
; we go in normalbeg2 because *(ushf*)(match+best_len-1) == scan_end
normalbeg2:
mov edi,window
cmp bp,word ptr[edi+eax]
jne contloop3 ; if *(ushf*)match != scan_start, continue
iseq:
; if we are here, we know that *(match+best_len-1) == scan_end
; and (match == scan_start)
mov edi,edx
mov esi,scan ; esi = scan
add edi,eax ; edi = window + cur_match = match
mov edx,[esi+3] ; compare manually dword at match+3
xor edx,[edi+3] ; and scan +3
jz begincompare ; if equal, go to long compare
; we will determine the unmatch byte and calculate len (in esi)
or dl,dl
je eq1rr
mov esi,3
jmp trfinval
eq1rr:
or dx,dx
je eq1
mov esi,4
jmp trfinval
eq1:
and edx,0ffffffh
jz eq11
mov esi,5
jmp trfinval
eq11:
mov esi,6
jmp trfinval
begincompare:
; here we now scan and match begin same
add edi,6
add esi,6
mov ecx,(MAX_MATCH-(2+4))/4 ; scan for at most MAX_MATCH bytes
repe cmpsd ; loop until mismatch
je trfin ; go to trfin if not unmatch
; we determine the unmatch byte
sub esi,4
mov edx,[edi-4]
xor edx,[esi]
or dl,dl
jnz trfin
inc esi
or dx,dx
jnz trfin
inc esi
and edx,0ffffffh
jnz trfin
inc esi
trfin:
sub esi,scan ; esi = len
trfinval:
; here we have finised compare, and esi contain len of equal string
cmp esi,best_len ; if len > best_len, go newbestlen
ja short newbestlen
; now we restore edx, ecx and esi, for the big loop
mov esi,prev
mov ecx,limit
mov edx,window
jmp contloop3
newbestlen:
mov best_len,esi ; len become best_len
mov match_start,eax ; save new position as match_start
cmp esi,nice_match ; if best_len >= nice_match, exit
jae exitloop
mov ecx,scan
mov edx,window ; restore edx=window
add ecx,esi
add esi,edx
dec esi
mov windowlen,esi ; windowlen = window + best_len-1
mov bx,[ecx-1] ; bx = *(scan+best_len-1) = scan_end
; now we restore ecx and esi, for the big loop :
mov esi,prev
mov ecx,limit
jmp contloop3
exitloop:
; exit : s->match_start=match_start
mov ebx,match_start
mov ebp,str_s
mov ecx,best_len
mov dword ptr [ebp+dep_match_start],ebx
mov eax,dword ptr [ebp+dep_lookahead]
cmp ecx,eax
ja minexlo
mov eax,ecx
minexlo:
; return min(best_len,s->lookahead)
; restore stack and register ebx,esi,edi,ebp
add esp,NbStackAdd
pop ebx
pop esi
pop edi
pop ebp
ret
InfoAuthor:
; please don't remove this string !
; Your are free use gvmat32 in any fre or commercial apps if you don't remove the string in the binary!
db 0dh,0ah,"GVMat32 optimised assembly code written 1996-98 by Gilles Vollant",0dh,0ah
IFDEF NOUNDERLINE
longest_match_7fff endp
ELSE
_longest_match_7fff endp
ENDIF
IFDEF NOUNDERLINE
cpudetect32 proc near
ELSE
_cpudetect32 proc near
ENDIF
push ebx
pushfd ; push original EFLAGS
pop eax ; get original EFLAGS
mov ecx, eax ; save original EFLAGS
xor eax, 40000h ; flip AC bit in EFLAGS
push eax ; save new EFLAGS value on stack
popfd ; replace current EFLAGS value
pushfd ; get new EFLAGS
pop eax ; store new EFLAGS in EAX
xor eax, ecx ; cant toggle AC bit, processor=80386
jz end_cpu_is_386 ; jump if 80386 processor
push ecx
popfd ; restore AC bit in EFLAGS first
pushfd
pushfd
pop ecx
mov eax, ecx ; get original EFLAGS
xor eax, 200000h ; flip ID bit in EFLAGS
push eax ; save new EFLAGS value on stack
popfd ; replace current EFLAGS value
pushfd ; get new EFLAGS
pop eax ; store new EFLAGS in EAX
popfd ; restore original EFLAGS
xor eax, ecx ; cant toggle ID bit,
je is_old_486 ; processor=old
mov eax,1
db 0fh,0a2h ;CPUID
exitcpudetect:
pop ebx
ret
end_cpu_is_386:
mov eax,0300h
jmp exitcpudetect
is_old_486:
mov eax,0400h
jmp exitcpudetect
IFDEF NOUNDERLINE
cpudetect32 endp
ELSE
_cpudetect32 endp
ENDIF
MAX_MATCH equ 258
MIN_MATCH equ 3
MIN_LOOKAHEAD equ (MAX_MATCH + MIN_MATCH + 1)
MAX_MATCH_8_ equ ((MAX_MATCH + 7) AND 0FFF0h)
;;; stack frame offsets
chainlenwmask equ esp + 0 ; high word: current chain len
; low word: s->wmask
window equ esp + 4 ; local copy of s->window
windowbestlen equ esp + 8 ; s->window + bestlen
scanstart equ esp + 16 ; first two bytes of string
scanend equ esp + 12 ; last two bytes of string
scanalign equ esp + 20 ; dword-misalignment of string
nicematch equ esp + 24 ; a good enough match size
bestlen equ esp + 28 ; size of best match so far
scan equ esp + 32 ; ptr to string wanting match
LocalVarsSize equ 36
; saved ebx byte esp + 36
; saved edi byte esp + 40
; saved esi byte esp + 44
; saved ebp byte esp + 48
; return address byte esp + 52
deflatestate equ esp + 56 ; the function arguments
curmatch equ esp + 60
;;; Offsets for fields in the deflate_state structure. These numbers
;;; are calculated from the definition of deflate_state, with the
;;; assumption that the compiler will dword-align the fields. (Thus,
;;; changing the definition of deflate_state could easily cause this
;;; program to crash horribly, without so much as a warning at
;;; compile time. Sigh.)
dsWSize equ 36
dsWMask equ 44
dsWindow equ 48
dsPrev equ 56
dsMatchLen equ 88
dsPrevMatch equ 92
dsStrStart equ 100
dsMatchStart equ 104
dsLookahead equ 108
dsPrevLen equ 112
dsMaxChainLen equ 116
dsGoodMatch equ 132
dsNiceMatch equ 136
;;; match.asm -- Pentium-Pro-optimized version of longest_match()
;;; Written for zlib 1.1.2
;;; Copyright (C) 1998 Brian Raiter <breadbox@muppetlabs.com>
;;; You can look at http://www.muppetlabs.com/~breadbox/software/assembly.html
;;;
;;; This is free software; you can redistribute it and/or modify it
;;; under the terms of the GNU General Public License.
;GLOBAL _longest_match, _match_init
;SECTION .text
;;; uInt longest_match(deflate_state *deflatestate, IPos curmatch)
;_longest_match:
IFDEF NOUNDERLINE
longest_match_686 proc near
ELSE
_longest_match_686 proc near
ENDIF
;;; Save registers that the compiler may be using, and adjust esp to
;;; make room for our stack frame.
push ebp
push edi
push esi
push ebx
sub esp, LocalVarsSize
;;; Retrieve the function arguments. ecx will hold cur_match
;;; throughout the entire function. edx will hold the pointer to the
;;; deflate_state structure during the function's setup (before
;;; entering the main loop.
mov edx, [deflatestate]
mov ecx, [curmatch]
;;; uInt wmask = s->w_mask;
;;; unsigned chain_length = s->max_chain_length;
;;; if (s->prev_length >= s->good_match) {
;;; chain_length >>= 2;
;;; }
mov eax, [edx + dsPrevLen]
mov ebx, [edx + dsGoodMatch]
cmp eax, ebx
mov eax, [edx + dsWMask]
mov ebx, [edx + dsMaxChainLen]
jl LastMatchGood
shr ebx, 2
LastMatchGood:
;;; chainlen is decremented once beforehand so that the function can
;;; use the sign flag instead of the zero flag for the exit test.
;;; It is then shifted into the high word, to make room for the wmask
;;; value, which it will always accompany.
dec ebx
shl ebx, 16
or ebx, eax
mov [chainlenwmask], ebx
;;; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
mov eax, [edx + dsNiceMatch]
mov ebx, [edx + dsLookahead]
cmp ebx, eax
jl LookaheadLess
mov ebx, eax
LookaheadLess: mov [nicematch], ebx
;;; register Bytef *scan = s->window + s->strstart;
mov esi, [edx + dsWindow]
mov [window], esi
mov ebp, [edx + dsStrStart]
lea edi, [esi + ebp]
mov [scan], edi
;;; Determine how many bytes the scan ptr is off from being
;;; dword-aligned.
mov eax, edi
neg eax
and eax, 3
mov [scanalign], eax
;;; IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
;;; s->strstart - (IPos)MAX_DIST(s) : NIL;
mov eax, [edx + dsWSize]
sub eax, MIN_LOOKAHEAD
sub ebp, eax
jg LimitPositive
xor ebp, ebp
LimitPositive:
;;; int best_len = s->prev_length;
mov eax, [edx + dsPrevLen]
mov [bestlen], eax
;;; Store the sum of s->window + best_len in esi locally, and in esi.
add esi, eax
mov [windowbestlen], esi
;;; register ush scan_start = *(ushf*)scan;
;;; register ush scan_end = *(ushf*)(scan+best_len-1);
;;; Posf *prev = s->prev;
movzx ebx, word ptr [edi]
mov [scanstart], ebx
movzx ebx, word ptr [edi + eax - 1]
mov [scanend], ebx
mov edi, [edx + dsPrev]
;;; Jump into the main loop.
mov edx, [chainlenwmask]
jmp short LoopEntry
align 4
;;; do {
;;; match = s->window + cur_match;
;;; if (*(ushf*)(match+best_len-1) != scan_end ||
;;; *(ushf*)match != scan_start) continue;
;;; [...]
;;; } while ((cur_match = prev[cur_match & wmask]) > limit
;;; && --chain_length != 0);
;;;
;;; Here is the inner loop of the function. The function will spend the
;;; majority of its time in this loop, and majority of that time will
;;; be spent in the first ten instructions.
;;;
;;; Within this loop:
;;; ebx = scanend
;;; ecx = curmatch
;;; edx = chainlenwmask - i.e., ((chainlen << 16) | wmask)
;;; esi = windowbestlen - i.e., (window + bestlen)
;;; edi = prev
;;; ebp = limit
LookupLoop:
and ecx, edx
movzx ecx, word ptr [edi + ecx*2]
cmp ecx, ebp
jbe LeaveNow
sub edx, 00010000h
js LeaveNow
LoopEntry: movzx eax, word ptr [esi + ecx - 1]
cmp eax, ebx
jnz LookupLoop
mov eax, [window]
movzx eax, word ptr [eax + ecx]
cmp eax, [scanstart]
jnz LookupLoop
;;; Store the current value of chainlen.
mov [chainlenwmask], edx
;;; Point edi to the string under scrutiny, and esi to the string we
;;; are hoping to match it up with. In actuality, esi and edi are
;;; both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and edx is
;;; initialized to -(MAX_MATCH_8 - scanalign).
mov esi, [window]
mov edi, [scan]
add esi, ecx
mov eax, [scanalign]
mov edx, 0fffffef8h; -(MAX_MATCH_8)
lea edi, [edi + eax + 0108h] ;MAX_MATCH_8]
lea esi, [esi + eax + 0108h] ;MAX_MATCH_8]
;;; Test the strings for equality, 8 bytes at a time. At the end,
;;; adjust edx so that it is offset to the exact byte that mismatched.
;;;
;;; We already know at this point that the first three bytes of the
;;; strings match each other, and they can be safely passed over before
;;; starting the compare loop. So what this code does is skip over 0-3
;;; bytes, as much as necessary in order to dword-align the edi
;;; pointer. (esi will still be misaligned three times out of four.)
;;;
;;; It should be confessed that this loop usually does not represent
;;; much of the total running time. Replacing it with a more
;;; straightforward "rep cmpsb" would not drastically degrade
;;; performance.
LoopCmps:
mov eax, [esi + edx]
xor eax, [edi + edx]
jnz LeaveLoopCmps
mov eax, [esi + edx + 4]
xor eax, [edi + edx + 4]
jnz LeaveLoopCmps4
add edx, 8
jnz LoopCmps
jmp short LenMaximum
LeaveLoopCmps4: add edx, 4
LeaveLoopCmps: test eax, 0000FFFFh
jnz LenLower
add edx, 2
shr eax, 16
LenLower: sub al, 1
adc edx, 0
;;; Calculate the length of the match. If it is longer than MAX_MATCH,
;;; then automatically accept it as the best possible match and leave.
lea eax, [edi + edx]
mov edi, [scan]
sub eax, edi
cmp eax, MAX_MATCH
jge LenMaximum
;;; If the length of the match is not longer than the best match we
;;; have so far, then forget it and return to the lookup loop.
mov edx, [deflatestate]
mov ebx, [bestlen]
cmp eax, ebx
jg LongerMatch
mov esi, [windowbestlen]
mov edi, [edx + dsPrev]
mov ebx, [scanend]
mov edx, [chainlenwmask]
jmp LookupLoop
;;; s->match_start = cur_match;
;;; best_len = len;
;;; if (len >= nice_match) break;
;;; scan_end = *(ushf*)(scan+best_len-1);
LongerMatch: mov ebx, [nicematch]
mov [bestlen], eax
mov [edx + dsMatchStart], ecx
cmp eax, ebx
jge LeaveNow
mov esi, [window]
add esi, eax
mov [windowbestlen], esi
movzx ebx, word ptr [edi + eax - 1]
mov edi, [edx + dsPrev]
mov [scanend], ebx
mov edx, [chainlenwmask]
jmp LookupLoop
;;; Accept the current string, with the maximum possible length.
LenMaximum: mov edx, [deflatestate]
mov dword ptr [bestlen], MAX_MATCH
mov [edx + dsMatchStart], ecx
;;; if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
;;; return s->lookahead;
LeaveNow:
mov edx, [deflatestate]
mov ebx, [bestlen]
mov eax, [edx + dsLookahead]
cmp ebx, eax
jg LookaheadRet
mov eax, ebx
LookaheadRet:
;;; Restore the stack and return from whence we came.
add esp, LocalVarsSize
pop ebx
pop esi
pop edi
pop ebp
ret
; please don't remove this string !
; Your can freely use gvmat32 in any free or commercial app if you don't remove the string in the binary!
db 0dh,0ah,"asm686 with masm, optimised assembly code from Brian Raiter, written 1998",0dh,0ah
IFDEF NOUNDERLINE
longest_match_686 endp
ELSE
_longest_match_686 endp
ENDIF
_TEXT ends
end
; gvmat32.asm -- Asm portion of the optimized longest_match for 32 bits x86
; Copyright (C) 1995-1996 Jean-loup Gailly and Gilles Vollant.
; File written by Gilles Vollant, by modifiying the longest_match
; from Jean-loup Gailly in deflate.c
;
; http://www.zlib.net
; http://www.winimage.com/zLibDll
; http://www.muppetlabs.com/~breadbox/software/assembly.html
;
; For Visual C++ 4.x and higher and ML 6.x and higher
; ml.exe is in directory \MASM611C of Win95 DDK
; ml.exe is also distributed in http://www.masm32.com/masmdl.htm
; and in VC++2003 toolkit at http://msdn.microsoft.com/visualc/vctoolkit2003/
;
; this file contain two implementation of longest_match
;
; longest_match_7fff : written 1996 by Gilles Vollant optimized for
; first Pentium. Assume s->w_mask == 0x7fff
; longest_match_686 : written by Brian raiter (1998), optimized for Pentium Pro
;
; for using an seembly version of longest_match, you need define ASMV in project
; There is two way in using gvmat32.asm
;
; A) Suggested method
; if you want include both longest_match_7fff and longest_match_686
; compile the asm file running
; ml /coff /Zi /Flgvmat32.lst /c gvmat32.asm
; and include gvmat32c.c in your project
; if you have an old cpu (386,486 or first Pentium) and s->w_mask==0x7fff,
; longest_match_7fff will be used
; if you have a more modern CPU (Pentium Pro, II and higher)
; longest_match_686 will be used
; on old cpu with s->w_mask!=0x7fff, longest_match_686 will be used,
; but this is not a sitation you'll find often
;
; B) Alternative
; if you are not interresed in old cpu performance and want the smaller
; binaries possible
;
; compile the asm file running
; ml /coff /Zi /c /Flgvmat32.lst /DNOOLDPENTIUMCODE gvmat32.asm
; and do not include gvmat32c.c in your project (ou define also
; NOOLDPENTIUMCODE)
;
; note : as I known, longest_match_686 is very faster than longest_match_7fff
; on pentium Pro/II/III, faster (but less) in P4, but it seem
; longest_match_7fff can be faster (very very litte) on AMD Athlon64/K8
;
; see below : zlib1222add must be adjuster if you use a zlib version < 1.2.2.2
;uInt longest_match_7fff(s, cur_match)
; deflate_state *s;
; IPos cur_match; /* current match */
NbStack equ 76
cur_match equ dword ptr[esp+NbStack-0]
str_s equ dword ptr[esp+NbStack-4]
; 5 dword on top (ret,ebp,esi,edi,ebx)
adrret equ dword ptr[esp+NbStack-8]
pushebp equ dword ptr[esp+NbStack-12]
pushedi equ dword ptr[esp+NbStack-16]
pushesi equ dword ptr[esp+NbStack-20]
pushebx equ dword ptr[esp+NbStack-24]
chain_length equ dword ptr [esp+NbStack-28]
limit equ dword ptr [esp+NbStack-32]
best_len equ dword ptr [esp+NbStack-36]
window equ dword ptr [esp+NbStack-40]
prev equ dword ptr [esp+NbStack-44]
scan_start equ word ptr [esp+NbStack-48]
wmask equ dword ptr [esp+NbStack-52]
match_start_ptr equ dword ptr [esp+NbStack-56]
nice_match equ dword ptr [esp+NbStack-60]
scan equ dword ptr [esp+NbStack-64]
windowlen equ dword ptr [esp+NbStack-68]
match_start equ dword ptr [esp+NbStack-72]
strend equ dword ptr [esp+NbStack-76]
NbStackAdd equ (NbStack-24)
.386p
name gvmatch
.MODEL FLAT
; all the +zlib1222add offsets are due to the addition of fields
; in zlib in the deflate_state structure since the asm code was first written
; (if you compile with zlib 1.0.4 or older, use "zlib1222add equ (-4)").
; (if you compile with zlib between 1.0.5 and 1.2.2.1, use "zlib1222add equ 0").
; if you compile with zlib 1.2.2.2 or later , use "zlib1222add equ 8").
zlib1222add equ 8
; Note : these value are good with a 8 bytes boundary pack structure
dep_chain_length equ 74h+zlib1222add
dep_window equ 30h+zlib1222add
dep_strstart equ 64h+zlib1222add
dep_prev_length equ 70h+zlib1222add
dep_nice_match equ 88h+zlib1222add
dep_w_size equ 24h+zlib1222add
dep_prev equ 38h+zlib1222add
dep_w_mask equ 2ch+zlib1222add
dep_good_match equ 84h+zlib1222add
dep_match_start equ 68h+zlib1222add
dep_lookahead equ 6ch+zlib1222add
_TEXT segment
IFDEF NOUNDERLINE
IFDEF NOOLDPENTIUMCODE
public longest_match
public match_init
ELSE
public longest_match_7fff
public cpudetect32
public longest_match_686
ENDIF
ELSE
IFDEF NOOLDPENTIUMCODE
public _longest_match
public _match_init
ELSE
public _longest_match_7fff
public _cpudetect32
public _longest_match_686
ENDIF
ENDIF
MAX_MATCH equ 258
MIN_MATCH equ 3
MIN_LOOKAHEAD equ (MAX_MATCH+MIN_MATCH+1)
IFNDEF NOOLDPENTIUMCODE
IFDEF NOUNDERLINE
longest_match_7fff proc near
ELSE
_longest_match_7fff proc near
ENDIF
mov edx,[esp+4]
push ebp
push edi
push esi
push ebx
sub esp,NbStackAdd
; initialize or check the variables used in match.asm.
mov ebp,edx
; chain_length = s->max_chain_length
; if (prev_length>=good_match) chain_length >>= 2
mov edx,[ebp+dep_chain_length]
mov ebx,[ebp+dep_prev_length]
cmp [ebp+dep_good_match],ebx
ja noshr
shr edx,2
noshr:
; we increment chain_length because in the asm, the --chain_lenght is in the beginning of the loop
inc edx
mov edi,[ebp+dep_nice_match]
mov chain_length,edx
mov eax,[ebp+dep_lookahead]
cmp eax,edi
; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
jae nolookaheadnicematch
mov edi,eax
nolookaheadnicematch:
; best_len = s->prev_length
mov best_len,ebx
; window = s->window
mov esi,[ebp+dep_window]
mov ecx,[ebp+dep_strstart]
mov window,esi
mov nice_match,edi
; scan = window + strstart
add esi,ecx
mov scan,esi
; dx = *window
mov dx,word ptr [esi]
; bx = *(window+best_len-1)
mov bx,word ptr [esi+ebx-1]
add esi,MAX_MATCH-1
; scan_start = *scan
mov scan_start,dx
; strend = scan + MAX_MATCH-1
mov strend,esi
; bx = scan_end = *(window+best_len-1)
; IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
; s->strstart - (IPos)MAX_DIST(s) : NIL;
mov esi,[ebp+dep_w_size]
sub esi,MIN_LOOKAHEAD
; here esi = MAX_DIST(s)
sub ecx,esi
ja nodist
xor ecx,ecx
nodist:
mov limit,ecx
; prev = s->prev
mov edx,[ebp+dep_prev]
mov prev,edx
;
mov edx,dword ptr [ebp+dep_match_start]
mov bp,scan_start
mov eax,cur_match
mov match_start,edx
mov edx,window
mov edi,edx
add edi,best_len
mov esi,prev
dec edi
; windowlen = window + best_len -1
mov windowlen,edi
jmp beginloop2
align 4
; here, in the loop
; eax = ax = cur_match
; ecx = limit
; bx = scan_end
; bp = scan_start
; edi = windowlen (window + best_len -1)
; esi = prev
;// here; chain_length <=16
normalbeg0add16:
add chain_length,16
jz exitloop
normalbeg0:
cmp word ptr[edi+eax],bx
je normalbeg2noroll
rcontlabnoroll:
; cur_match = prev[cur_match & wmask]
and eax,7fffh
mov ax,word ptr[esi+eax*2]
; if cur_match > limit, go to exitloop
cmp ecx,eax
jnb exitloop
; if --chain_length != 0, go to exitloop
dec chain_length
jnz normalbeg0
jmp exitloop
normalbeg2noroll:
; if (scan_start==*(cur_match+window)) goto normalbeg2
cmp bp,word ptr[edx+eax]
jne rcontlabnoroll
jmp normalbeg2
contloop3:
mov edi,windowlen
; cur_match = prev[cur_match & wmask]
and eax,7fffh
mov ax,word ptr[esi+eax*2]
; if cur_match > limit, go to exitloop
cmp ecx,eax
jnbexitloopshort1:
jnb exitloop
; if --chain_length != 0, go to exitloop
; begin the main loop
beginloop2:
sub chain_length,16+1
; if chain_length <=16, don't use the unrolled loop
jna normalbeg0add16
do16:
cmp word ptr[edi+eax],bx
je normalbeg2dc0
maccn MACRO lab
and eax,7fffh
mov ax,word ptr[esi+eax*2]
cmp ecx,eax
jnb exitloop
cmp word ptr[edi+eax],bx
je lab
ENDM
rcontloop0:
maccn normalbeg2dc1
rcontloop1:
maccn normalbeg2dc2
rcontloop2:
maccn normalbeg2dc3
rcontloop3:
maccn normalbeg2dc4
rcontloop4:
maccn normalbeg2dc5
rcontloop5:
maccn normalbeg2dc6
rcontloop6:
maccn normalbeg2dc7
rcontloop7:
maccn normalbeg2dc8
rcontloop8:
maccn normalbeg2dc9
rcontloop9:
maccn normalbeg2dc10
rcontloop10:
maccn short normalbeg2dc11
rcontloop11:
maccn short normalbeg2dc12
rcontloop12:
maccn short normalbeg2dc13
rcontloop13:
maccn short normalbeg2dc14
rcontloop14:
maccn short normalbeg2dc15
rcontloop15:
and eax,7fffh
mov ax,word ptr[esi+eax*2]
cmp ecx,eax
jnb exitloop
sub chain_length,16
ja do16
jmp normalbeg0add16
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
normbeg MACRO rcontlab,valsub
; if we are here, we know that *(match+best_len-1) == scan_end
cmp bp,word ptr[edx+eax]
; if (match != scan_start) goto rcontlab
jne rcontlab
; calculate the good chain_length, and we'll compare scan and match string
add chain_length,16-valsub
jmp iseq
ENDM
normalbeg2dc11:
normbeg rcontloop11,11
normalbeg2dc12:
normbeg short rcontloop12,12
normalbeg2dc13:
normbeg short rcontloop13,13
normalbeg2dc14:
normbeg short rcontloop14,14
normalbeg2dc15:
normbeg short rcontloop15,15
normalbeg2dc10:
normbeg rcontloop10,10
normalbeg2dc9:
normbeg rcontloop9,9
normalbeg2dc8:
normbeg rcontloop8,8
normalbeg2dc7:
normbeg rcontloop7,7
normalbeg2dc6:
normbeg rcontloop6,6
normalbeg2dc5:
normbeg rcontloop5,5
normalbeg2dc4:
normbeg rcontloop4,4
normalbeg2dc3:
normbeg rcontloop3,3
normalbeg2dc2:
normbeg rcontloop2,2
normalbeg2dc1:
normbeg rcontloop1,1
normalbeg2dc0:
normbeg rcontloop0,0
; we go in normalbeg2 because *(ushf*)(match+best_len-1) == scan_end
normalbeg2:
mov edi,window
cmp bp,word ptr[edi+eax]
jne contloop3 ; if *(ushf*)match != scan_start, continue
iseq:
; if we are here, we know that *(match+best_len-1) == scan_end
; and (match == scan_start)
mov edi,edx
mov esi,scan ; esi = scan
add edi,eax ; edi = window + cur_match = match
mov edx,[esi+3] ; compare manually dword at match+3
xor edx,[edi+3] ; and scan +3
jz begincompare ; if equal, go to long compare
; we will determine the unmatch byte and calculate len (in esi)
or dl,dl
je eq1rr
mov esi,3
jmp trfinval
eq1rr:
or dx,dx
je eq1
mov esi,4
jmp trfinval
eq1:
and edx,0ffffffh
jz eq11
mov esi,5
jmp trfinval
eq11:
mov esi,6
jmp trfinval
begincompare:
; here we now scan and match begin same
add edi,6
add esi,6
mov ecx,(MAX_MATCH-(2+4))/4 ; scan for at most MAX_MATCH bytes
repe cmpsd ; loop until mismatch
je trfin ; go to trfin if not unmatch
; we determine the unmatch byte
sub esi,4
mov edx,[edi-4]
xor edx,[esi]
or dl,dl
jnz trfin
inc esi
or dx,dx
jnz trfin
inc esi
and edx,0ffffffh
jnz trfin
inc esi
trfin:
sub esi,scan ; esi = len
trfinval:
; here we have finised compare, and esi contain len of equal string
cmp esi,best_len ; if len > best_len, go newbestlen
ja short newbestlen
; now we restore edx, ecx and esi, for the big loop
mov esi,prev
mov ecx,limit
mov edx,window
jmp contloop3
newbestlen:
mov best_len,esi ; len become best_len
mov match_start,eax ; save new position as match_start
cmp esi,nice_match ; if best_len >= nice_match, exit
jae exitloop
mov ecx,scan
mov edx,window ; restore edx=window
add ecx,esi
add esi,edx
dec esi
mov windowlen,esi ; windowlen = window + best_len-1
mov bx,[ecx-1] ; bx = *(scan+best_len-1) = scan_end
; now we restore ecx and esi, for the big loop :
mov esi,prev
mov ecx,limit
jmp contloop3
exitloop:
; exit : s->match_start=match_start
mov ebx,match_start
mov ebp,str_s
mov ecx,best_len
mov dword ptr [ebp+dep_match_start],ebx
mov eax,dword ptr [ebp+dep_lookahead]
cmp ecx,eax
ja minexlo
mov eax,ecx
minexlo:
; return min(best_len,s->lookahead)
; restore stack and register ebx,esi,edi,ebp
add esp,NbStackAdd
pop ebx
pop esi
pop edi
pop ebp
ret
InfoAuthor:
; please don't remove this string !
; Your are free use gvmat32 in any fre or commercial apps if you don't remove the string in the binary!
db 0dh,0ah,"GVMat32 optimised assembly code written 1996-98 by Gilles Vollant",0dh,0ah
IFDEF NOUNDERLINE
longest_match_7fff endp
ELSE
_longest_match_7fff endp
ENDIF
IFDEF NOUNDERLINE
cpudetect32 proc near
ELSE
_cpudetect32 proc near
ENDIF
push ebx
pushfd ; push original EFLAGS
pop eax ; get original EFLAGS
mov ecx, eax ; save original EFLAGS
xor eax, 40000h ; flip AC bit in EFLAGS
push eax ; save new EFLAGS value on stack
popfd ; replace current EFLAGS value
pushfd ; get new EFLAGS
pop eax ; store new EFLAGS in EAX
xor eax, ecx ; cant toggle AC bit, processor=80386
jz end_cpu_is_386 ; jump if 80386 processor
push ecx
popfd ; restore AC bit in EFLAGS first
pushfd
pushfd
pop ecx
mov eax, ecx ; get original EFLAGS
xor eax, 200000h ; flip ID bit in EFLAGS
push eax ; save new EFLAGS value on stack
popfd ; replace current EFLAGS value
pushfd ; get new EFLAGS
pop eax ; store new EFLAGS in EAX
popfd ; restore original EFLAGS
xor eax, ecx ; cant toggle ID bit,
je is_old_486 ; processor=old
mov eax,1
db 0fh,0a2h ;CPUID
exitcpudetect:
pop ebx
ret
end_cpu_is_386:
mov eax,0300h
jmp exitcpudetect
is_old_486:
mov eax,0400h
jmp exitcpudetect
IFDEF NOUNDERLINE
cpudetect32 endp
ELSE
_cpudetect32 endp
ENDIF
ENDIF
MAX_MATCH equ 258
MIN_MATCH equ 3
MIN_LOOKAHEAD equ (MAX_MATCH + MIN_MATCH + 1)
MAX_MATCH_8_ equ ((MAX_MATCH + 7) AND 0FFF0h)
;;; stack frame offsets
chainlenwmask equ esp + 0 ; high word: current chain len
; low word: s->wmask
window equ esp + 4 ; local copy of s->window
windowbestlen equ esp + 8 ; s->window + bestlen
scanstart equ esp + 16 ; first two bytes of string
scanend equ esp + 12 ; last two bytes of string
scanalign equ esp + 20 ; dword-misalignment of string
nicematch equ esp + 24 ; a good enough match size
bestlen equ esp + 28 ; size of best match so far
scan equ esp + 32 ; ptr to string wanting match
LocalVarsSize equ 36
; saved ebx byte esp + 36
; saved edi byte esp + 40
; saved esi byte esp + 44
; saved ebp byte esp + 48
; return address byte esp + 52
deflatestate equ esp + 56 ; the function arguments
curmatch equ esp + 60
;;; Offsets for fields in the deflate_state structure. These numbers
;;; are calculated from the definition of deflate_state, with the
;;; assumption that the compiler will dword-align the fields. (Thus,
;;; changing the definition of deflate_state could easily cause this
;;; program to crash horribly, without so much as a warning at
;;; compile time. Sigh.)
dsWSize equ 36+zlib1222add
dsWMask equ 44+zlib1222add
dsWindow equ 48+zlib1222add
dsPrev equ 56+zlib1222add
dsMatchLen equ 88+zlib1222add
dsPrevMatch equ 92+zlib1222add
dsStrStart equ 100+zlib1222add
dsMatchStart equ 104+zlib1222add
dsLookahead equ 108+zlib1222add
dsPrevLen equ 112+zlib1222add
dsMaxChainLen equ 116+zlib1222add
dsGoodMatch equ 132+zlib1222add
dsNiceMatch equ 136+zlib1222add
;;; match.asm -- Pentium-Pro-optimized version of longest_match()
;;; Written for zlib 1.1.2
;;; Copyright (C) 1998 Brian Raiter <breadbox@muppetlabs.com>
;;; You can look at http://www.muppetlabs.com/~breadbox/software/assembly.html
;;;
;;; This is free software; you can redistribute it and/or modify it
;;; under the terms of the GNU General Public License.
;GLOBAL _longest_match, _match_init
;SECTION .text
;;; uInt longest_match(deflate_state *deflatestate, IPos curmatch)
;_longest_match:
IFDEF NOOLDPENTIUMCODE
IFDEF NOUNDERLINE
longest_match proc near
ELSE
_longest_match proc near
ENDIF
ELSE
IFDEF NOUNDERLINE
longest_match_686 proc near
ELSE
_longest_match_686 proc near
ENDIF
ENDIF
;;; Save registers that the compiler may be using, and adjust esp to
;;; make room for our stack frame.
push ebp
push edi
push esi
push ebx
sub esp, LocalVarsSize
;;; Retrieve the function arguments. ecx will hold cur_match
;;; throughout the entire function. edx will hold the pointer to the
;;; deflate_state structure during the function's setup (before
;;; entering the main loop.
mov edx, [deflatestate]
mov ecx, [curmatch]
;;; uInt wmask = s->w_mask;
;;; unsigned chain_length = s->max_chain_length;
;;; if (s->prev_length >= s->good_match) {
;;; chain_length >>= 2;
;;; }
mov eax, [edx + dsPrevLen]
mov ebx, [edx + dsGoodMatch]
cmp eax, ebx
mov eax, [edx + dsWMask]
mov ebx, [edx + dsMaxChainLen]
jl LastMatchGood
shr ebx, 2
LastMatchGood:
;;; chainlen is decremented once beforehand so that the function can
;;; use the sign flag instead of the zero flag for the exit test.
;;; It is then shifted into the high word, to make room for the wmask
;;; value, which it will always accompany.
dec ebx
shl ebx, 16
or ebx, eax
mov [chainlenwmask], ebx
;;; if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
mov eax, [edx + dsNiceMatch]
mov ebx, [edx + dsLookahead]
cmp ebx, eax
jl LookaheadLess
mov ebx, eax
LookaheadLess: mov [nicematch], ebx
;;; register Bytef *scan = s->window + s->strstart;
mov esi, [edx + dsWindow]
mov [window], esi
mov ebp, [edx + dsStrStart]
lea edi, [esi + ebp]
mov [scan], edi
;;; Determine how many bytes the scan ptr is off from being
;;; dword-aligned.
mov eax, edi
neg eax
and eax, 3
mov [scanalign], eax
;;; IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
;;; s->strstart - (IPos)MAX_DIST(s) : NIL;
mov eax, [edx + dsWSize]
sub eax, MIN_LOOKAHEAD
sub ebp, eax
jg LimitPositive
xor ebp, ebp
LimitPositive:
;;; int best_len = s->prev_length;
mov eax, [edx + dsPrevLen]
mov [bestlen], eax
;;; Store the sum of s->window + best_len in esi locally, and in esi.
add esi, eax
mov [windowbestlen], esi
;;; register ush scan_start = *(ushf*)scan;
;;; register ush scan_end = *(ushf*)(scan+best_len-1);
;;; Posf *prev = s->prev;
movzx ebx, word ptr [edi]
mov [scanstart], ebx
movzx ebx, word ptr [edi + eax - 1]
mov [scanend], ebx
mov edi, [edx + dsPrev]
;;; Jump into the main loop.
mov edx, [chainlenwmask]
jmp short LoopEntry
align 4
;;; do {
;;; match = s->window + cur_match;
;;; if (*(ushf*)(match+best_len-1) != scan_end ||
;;; *(ushf*)match != scan_start) continue;
;;; [...]
;;; } while ((cur_match = prev[cur_match & wmask]) > limit
;;; && --chain_length != 0);
;;;
;;; Here is the inner loop of the function. The function will spend the
;;; majority of its time in this loop, and majority of that time will
;;; be spent in the first ten instructions.
;;;
;;; Within this loop:
;;; ebx = scanend
;;; ecx = curmatch
;;; edx = chainlenwmask - i.e., ((chainlen << 16) | wmask)
;;; esi = windowbestlen - i.e., (window + bestlen)
;;; edi = prev
;;; ebp = limit
LookupLoop:
and ecx, edx
movzx ecx, word ptr [edi + ecx*2]
cmp ecx, ebp
jbe LeaveNow
sub edx, 00010000h
js LeaveNow
LoopEntry: movzx eax, word ptr [esi + ecx - 1]
cmp eax, ebx
jnz LookupLoop
mov eax, [window]
movzx eax, word ptr [eax + ecx]
cmp eax, [scanstart]
jnz LookupLoop
;;; Store the current value of chainlen.
mov [chainlenwmask], edx
;;; Point edi to the string under scrutiny, and esi to the string we
;;; are hoping to match it up with. In actuality, esi and edi are
;;; both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and edx is
;;; initialized to -(MAX_MATCH_8 - scanalign).
mov esi, [window]
mov edi, [scan]
add esi, ecx
mov eax, [scanalign]
mov edx, 0fffffef8h; -(MAX_MATCH_8)
lea edi, [edi + eax + 0108h] ;MAX_MATCH_8]
lea esi, [esi + eax + 0108h] ;MAX_MATCH_8]
;;; Test the strings for equality, 8 bytes at a time. At the end,
;;; adjust edx so that it is offset to the exact byte that mismatched.
;;;
;;; We already know at this point that the first three bytes of the
;;; strings match each other, and they can be safely passed over before
;;; starting the compare loop. So what this code does is skip over 0-3
;;; bytes, as much as necessary in order to dword-align the edi
;;; pointer. (esi will still be misaligned three times out of four.)
;;;
;;; It should be confessed that this loop usually does not represent
;;; much of the total running time. Replacing it with a more
;;; straightforward "rep cmpsb" would not drastically degrade
;;; performance.
LoopCmps:
mov eax, [esi + edx]
xor eax, [edi + edx]
jnz LeaveLoopCmps
mov eax, [esi + edx + 4]
xor eax, [edi + edx + 4]
jnz LeaveLoopCmps4
add edx, 8
jnz LoopCmps
jmp short LenMaximum
LeaveLoopCmps4: add edx, 4
LeaveLoopCmps: test eax, 0000FFFFh
jnz LenLower
add edx, 2
shr eax, 16
LenLower: sub al, 1
adc edx, 0
;;; Calculate the length of the match. If it is longer than MAX_MATCH,
;;; then automatically accept it as the best possible match and leave.
lea eax, [edi + edx]
mov edi, [scan]
sub eax, edi
cmp eax, MAX_MATCH
jge LenMaximum
;;; If the length of the match is not longer than the best match we
;;; have so far, then forget it and return to the lookup loop.
mov edx, [deflatestate]
mov ebx, [bestlen]
cmp eax, ebx
jg LongerMatch
mov esi, [windowbestlen]
mov edi, [edx + dsPrev]
mov ebx, [scanend]
mov edx, [chainlenwmask]
jmp LookupLoop
;;; s->match_start = cur_match;
;;; best_len = len;
;;; if (len >= nice_match) break;
;;; scan_end = *(ushf*)(scan+best_len-1);
LongerMatch: mov ebx, [nicematch]
mov [bestlen], eax
mov [edx + dsMatchStart], ecx
cmp eax, ebx
jge LeaveNow
mov esi, [window]
add esi, eax
mov [windowbestlen], esi
movzx ebx, word ptr [edi + eax - 1]
mov edi, [edx + dsPrev]
mov [scanend], ebx
mov edx, [chainlenwmask]
jmp LookupLoop
;;; Accept the current string, with the maximum possible length.
LenMaximum: mov edx, [deflatestate]
mov dword ptr [bestlen], MAX_MATCH
mov [edx + dsMatchStart], ecx
;;; if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
;;; return s->lookahead;
LeaveNow:
mov edx, [deflatestate]
mov ebx, [bestlen]
mov eax, [edx + dsLookahead]
cmp ebx, eax
jg LookaheadRet
mov eax, ebx
LookaheadRet:
;;; Restore the stack and return from whence we came.
add esp, LocalVarsSize
pop ebx
pop esi
pop edi
pop ebp
ret
; please don't remove this string !
; Your can freely use gvmat32 in any free or commercial app if you don't remove the string in the binary!
db 0dh,0ah,"asm686 with masm, optimised assembly code from Brian Raiter, written 1998",0dh,0ah
IFDEF NOOLDPENTIUMCODE
IFDEF NOUNDERLINE
longest_match endp
ELSE
_longest_match endp
ENDIF
IFDEF NOUNDERLINE
match_init proc near
ret
match_init endp
ELSE
_match_init proc near
ret
_match_init endp
ENDIF
ELSE
IFDEF NOUNDERLINE
longest_match_686 endp
ELSE
_longest_match_686 endp
ENDIF
ENDIF
_TEXT ends
end
zlib/contrib/masmx86/gvmat32c.c
/* gvmat32.c -- C portion of the optimized longest_match for 32 bits x86
* Copyright (C) 1995-1996 Jean-loup Gailly and Gilles Vollant.
* File written by Gilles Vollant, by modifiying the longest_match
* from Jean-loup Gailly in deflate.c
* it prepare all parameters and call the assembly longest_match_gvasm
* longest_match execute standard C code is wmask != 0x7fff
* (assembly code is faster with a fixed wmask)
*
*/
#include "deflate.h"
#ifdef ASMV
#define NIL 0
#define UNALIGNED_OK
/* if your C compiler don't add underline before function name,
define ADD_UNDERLINE_ASMFUNC */
#ifdef ADD_UNDERLINE_ASMFUNC
#define longest_match_7fff _longest_match_7fff
#define longest_match_686 _longest_match_686
#define cpudetect32 _cpudetect32
#endif
void match_init()
{
}
unsigned long cpudetect32();
uInt longest_match_c(
deflate_state *s,
IPos cur_match); /* current match */
uInt longest_match_7fff(
deflate_state *s,
IPos cur_match); /* current match */
uInt longest_match_686(
deflate_state *s,
IPos cur_match); /* current match */
uInt longest_match(
deflate_state *s,
IPos cur_match) /* current match */
{
static uInt iIsPPro=2;
if ((s->w_mask == 0x7fff) && (iIsPPro==0))
return longest_match_7fff(s,cur_match);
if (iIsPPro==1)
return longest_match_686(s,cur_match);
if (iIsPPro==2)
iIsPPro = (((cpudetect32()/0x100)&0xf)>=6) ? 1 : 0;
return longest_match_c(s,cur_match);
}
uInt longest_match_c(s, cur_match)
deflate_state *s;
IPos cur_match; /* current match */
{
unsigned chain_length = s->max_chain_length;/* max hash chain length */
register Bytef *scan = s->window + s->strstart; /* current string */
register Bytef *match; /* matched string */
register int len; /* length of current match */
int best_len = s->prev_length; /* best match length so far */
int nice_match = s->nice_match; /* stop if match long enough */
IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
s->strstart - (IPos)MAX_DIST(s) : NIL;
/* Stop when cur_match becomes <= limit. To simplify the code,
* we prevent matches with the string of window index 0.
*/
Posf *prev = s->prev;
uInt wmask = s->w_mask;
#ifdef UNALIGNED_OK
/* Compare two bytes at a time. Note: this is not always beneficial.
* Try with and without -DUNALIGNED_OK to check.
*/
register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
register ush scan_start = *(ushf*)scan;
register ush scan_end = *(ushf*)(scan+best_len-1);
#else
register Bytef *strend = s->window + s->strstart + MAX_MATCH;
register Byte scan_end1 = scan[best_len-1];
register Byte scan_end = scan[best_len];
#endif
/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
* It is easy to get rid of this optimization if necessary.
*/
Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
/* Do not waste too much time if we already have a good match: */
if (s->prev_length >= s->good_match) {
chain_length >>= 2;
}
/* Do not look for matches beyond the end of the input. This is necessary
* to make deflate deterministic.
*/
if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
do {
Assert(cur_match < s->strstart, "no future");
match = s->window + cur_match;
/* Skip to next match if the match length cannot increase
* or if the match length is less than 2:
*/
#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
/* This code assumes sizeof(unsigned short) == 2. Do not use
* UNALIGNED_OK if your compiler uses a different size.
*/
if (*(ushf*)(match+best_len-1) != scan_end ||
*(ushf*)match != scan_start) continue;
/* It is not necessary to compare scan[2] and match[2] since they are
* always equal when the other bytes match, given that the hash keys
* are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
* strstart+3, +5, ... up to strstart+257. We check for insufficient
* lookahead only every 4th comparison; the 128th check will be made
* at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
* necessary to put more guard bytes at the end of the window, or
* to check more often for insufficient lookahead.
*/
Assert(scan[2] == match[2], "scan[2]?");
scan++, match++;
do {
} while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
scan < strend);
/* The funny "do {}" generates better code on most compilers */
/* Here, scan <= window+strstart+257 */
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
if (*scan == *match) scan++;
len = (MAX_MATCH - 1) - (int)(strend-scan);
scan = strend - (MAX_MATCH-1);
#else /* UNALIGNED_OK */
if (match[best_len] != scan_end ||
match[best_len-1] != scan_end1 ||
*match != *scan ||
*++match != scan[1]) continue;
/* The check at best_len-1 can be removed because it will be made
* again later. (This heuristic is not always a win.)
* It is not necessary to compare scan[2] and match[2] since they
* are always equal when the other bytes match, given that
* the hash keys are equal and that HASH_BITS >= 8.
*/
scan += 2, match++;
Assert(*scan == *match, "match[2]?");
/* We check for insufficient lookahead only every 8th comparison;
* the 256th check will be made at strstart+258.
*/
do {
} while (*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
scan < strend);
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
len = MAX_MATCH - (int)(strend - scan);
scan = strend - MAX_MATCH;
#endif /* UNALIGNED_OK */
if (len > best_len) {
s->match_start = cur_match;
best_len = len;
if (len >= nice_match) break;
#ifdef UNALIGNED_OK
scan_end = *(ushf*)(scan+best_len-1);
#else
scan_end1 = scan[best_len-1];
scan_end = scan[best_len];
#endif
}
} while ((cur_match = prev[cur_match & wmask]) > limit
&& --chain_length != 0);
if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
return s->lookahead;
}
#endif /* ASMV */
/* gvmat32.c -- C portion of the optimized longest_match for 32 bits x86
* Copyright (C) 1995-1996 Jean-loup Gailly and Gilles Vollant.
* File written by Gilles Vollant, by modifiying the longest_match
* from Jean-loup Gailly in deflate.c
* it prepare all parameters and call the assembly longest_match_gvasm
* longest_match execute standard C code is wmask != 0x7fff
* (assembly code is faster with a fixed wmask)
*
* Read comment at beginning of gvmat32.asm for more information
*/
#if defined(ASMV) && (!defined(NOOLDPENTIUMCODE))
#include "deflate.h"
/* if your C compiler don't add underline before function name,
define ADD_UNDERLINE_ASMFUNC */
#ifdef ADD_UNDERLINE_ASMFUNC
#define longest_match_7fff _longest_match_7fff
#define longest_match_686 _longest_match_686
#define cpudetect32 _cpudetect32
#endif
unsigned long cpudetect32();
uInt longest_match_c(
deflate_state *s,
IPos cur_match); /* current match */
uInt longest_match_7fff(
deflate_state *s,
IPos cur_match); /* current match */
uInt longest_match_686(
deflate_state *s,
IPos cur_match); /* current match */
static uInt iIsPPro=2;
void match_init ()
{
iIsPPro = (((cpudetect32()/0x100)&0xf)>=6) ? 1 : 0;
}
uInt longest_match(
deflate_state *s,
IPos cur_match) /* current match */
{
if (iIsPPro!=0)
return longest_match_686(s,cur_match);
if (s->w_mask != 0x7fff)
return longest_match_686(s,cur_match);
/* now ((s->w_mask == 0x7fff) && (iIsPPro==0)) */
return longest_match_7fff(s,cur_match);
}
#endif /* defined(ASMV) && (!defined(NOOLDPENTIUMCODE)) */
zlib/contrib/masmx86/inffas32.asm
; 75 "inffast.S"
;FILE "inffast.S"
;;;GLOBAL _inflate_fast
;;;SECTION .text
.586p
.mmx
name inflate_fast_x86
.MODEL FLAT
_DATA segment
inflate_fast_use_mmx:
dd 1
_TEXT segment
PUBLIC _inflate_fast
ALIGN 4
_inflate_fast:
jmp inflate_fast_entry
ALIGN 4
db 'Fast decoding Code from Chris Anderson'
db 0
ALIGN 4
invalid_literal_length_code_msg:
db 'invalid literal/length code'
db 0
ALIGN 4
invalid_distance_code_msg:
db 'invalid distance code'
db 0
ALIGN 4
invalid_distance_too_far_msg:
db 'invalid distance too far back'
db 0
ALIGN 4
inflate_fast_mask:
dd 0
dd 1
dd 3
dd 7
dd 15
dd 31
dd 63
dd 127
dd 255
dd 511
dd 1023
dd 2047
dd 4095
dd 8191
dd 16383
dd 32767
dd 65535
dd 131071
dd 262143
dd 524287
dd 1048575
dd 2097151
dd 4194303
dd 8388607
dd 16777215
dd 33554431
dd 67108863
dd 134217727
dd 268435455
dd 536870911
dd 1073741823
dd 2147483647
dd 4294967295
mode_state equ 0 ;/* state->mode */
wsize_state equ 32 ;/* state->wsize */
write_state equ (36+4) ;/* state->write */
window_state equ (40+4) ;/* state->window */
hold_state equ (44+4) ;/* state->hold */
bits_state equ (48+4) ;/* state->bits */
lencode_state equ (64+4) ;/* state->lencode */
distcode_state equ (68+4) ;/* state->distcode */
lenbits_state equ (72+4) ;/* state->lenbits */
distbits_state equ (76+4) ;/* state->distbits */
;;SECTION .text
; 205 "inffast.S"
;GLOBAL inflate_fast_use_mmx
;SECTION .data
; GLOBAL inflate_fast_use_mmx:object
;.size inflate_fast_use_mmx, 4
; 226 "inffast.S"
;SECTION .text
ALIGN 4
inflate_fast_entry:
push edi
push esi
push ebp
push ebx
pushfd
sub esp,64
cld
mov esi, [esp+88]
mov edi, [esi+28]
mov edx, [esi+4]
mov eax, [esi+0]
add edx,eax
sub edx,11
mov [esp+44],eax
mov [esp+20],edx
mov ebp, [esp+92]
mov ecx, [esi+16]
mov ebx, [esi+12]
sub ebp,ecx
neg ebp
add ebp,ebx
sub ecx,257
add ecx,ebx
mov [esp+60],ebx
mov [esp+40],ebp
mov [esp+16],ecx
; 285 "inffast.S"
mov eax, [edi+lencode_state]
mov ecx, [edi+distcode_state]
mov [esp+8],eax
mov [esp+12],ecx
mov eax,1
mov ecx, [edi+lenbits_state]
shl eax,cl
dec eax
mov [esp+0],eax
mov eax,1
mov ecx, [edi+distbits_state]
shl eax,cl
dec eax
mov [esp+4],eax
mov eax, [edi+wsize_state]
mov ecx, [edi+write_state]
mov edx, [edi+window_state]
mov [esp+52],eax
mov [esp+48],ecx
mov [esp+56],edx
mov ebp, [edi+hold_state]
mov ebx, [edi+bits_state]
; 321 "inffast.S"
mov esi, [esp+44]
mov ecx, [esp+20]
cmp ecx,esi
ja L_align_long
add ecx,11
sub ecx,esi
mov eax,12
sub eax,ecx
lea edi, [esp+28]
rep movsb
mov ecx,eax
xor eax,eax
rep stosb
lea esi, [esp+28]
mov [esp+20],esi
jmp L_is_aligned
L_align_long:
test esi,3
jz L_is_aligned
xor eax,eax
mov al, [esi]
inc esi
mov ecx,ebx
add ebx,8
shl eax,cl
or ebp,eax
jmp L_align_long
L_is_aligned:
mov edi, [esp+60]
; 366 "inffast.S"
L_check_mmx:
cmp dword ptr [inflate_fast_use_mmx],2
je L_init_mmx
ja L_do_loop
push eax
push ebx
push ecx
push edx
pushfd
mov eax, [esp]
xor dword ptr [esp],0200000h
popfd
pushfd
pop edx
xor edx,eax
jz L_dont_use_mmx
xor eax,eax
cpuid
cmp ebx,0756e6547h
jne L_dont_use_mmx
cmp ecx,06c65746eh
jne L_dont_use_mmx
cmp edx,049656e69h
jne L_dont_use_mmx
mov eax,1
cpuid
shr eax,8
and eax,15
cmp eax,6
jne L_dont_use_mmx
test edx,0800000h
jnz L_use_mmx
jmp L_dont_use_mmx
L_use_mmx:
mov dword ptr [inflate_fast_use_mmx],2
jmp L_check_mmx_pop
L_dont_use_mmx:
mov dword ptr [inflate_fast_use_mmx],3
L_check_mmx_pop:
pop edx
pop ecx
pop ebx
pop eax
jmp L_check_mmx
; 426 "inffast.S"
ALIGN 4
L_do_loop:
; 437 "inffast.S"
cmp bl,15
ja L_get_length_code
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
L_get_length_code:
mov edx, [esp+0]
mov ecx, [esp+8]
and edx,ebp
mov eax, [ecx+edx*4]
L_dolen:
mov cl,ah
sub bl,ah
shr ebp,cl
test al,al
jnz L_test_for_length_base
shr eax,16
stosb
L_while_test:
cmp [esp+16],edi
jbe L_break_loop
cmp [esp+20],esi
ja L_do_loop
jmp L_break_loop
L_test_for_length_base:
; 502 "inffast.S"
mov edx,eax
shr edx,16
mov cl,al
test al,16
jz L_test_for_second_level_length
and cl,15
jz L_save_len
cmp bl,cl
jae L_add_bits_to_len
mov ch,cl
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
mov cl,ch
L_add_bits_to_len:
mov eax,1
shl eax,cl
dec eax
sub bl,cl
and eax,ebp
shr ebp,cl
add edx,eax
L_save_len:
mov [esp+24],edx
L_decode_distance:
; 549 "inffast.S"
cmp bl,15
ja L_get_distance_code
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
L_get_distance_code:
mov edx, [esp+4]
mov ecx, [esp+12]
and edx,ebp
mov eax, [ecx+edx*4]
L_dodist:
mov edx,eax
shr edx,16
mov cl,ah
sub bl,ah
shr ebp,cl
; 584 "inffast.S"
mov cl,al
test al,16
jz L_test_for_second_level_dist
and cl,15
jz L_check_dist_one
cmp bl,cl
jae L_add_bits_to_dist
mov ch,cl
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
mov cl,ch
L_add_bits_to_dist:
mov eax,1
shl eax,cl
dec eax
sub bl,cl
and eax,ebp
shr ebp,cl
add edx,eax
jmp L_check_window
L_check_window:
; 625 "inffast.S"
mov [esp+44],esi
mov eax,edi
sub eax, [esp+40]
cmp eax,edx
jb L_clip_window
mov ecx, [esp+24]
mov esi,edi
sub esi,edx
sub ecx,3
mov al, [esi]
mov [edi],al
mov al, [esi+1]
mov dl, [esi+2]
add esi,3
mov [edi+1],al
mov [edi+2],dl
add edi,3
rep movsb
mov esi, [esp+44]
jmp L_while_test
ALIGN 4
L_check_dist_one:
cmp edx,1
jne L_check_window
cmp [esp+40],edi
je L_check_window
dec edi
mov ecx, [esp+24]
mov al, [edi]
sub ecx,3
mov [edi+1],al
mov [edi+2],al
mov [edi+3],al
add edi,4
rep stosb
jmp L_while_test
ALIGN 4
L_test_for_second_level_length:
test al,64
jnz L_test_for_end_of_block
mov eax,1
shl eax,cl
dec eax
and eax,ebp
add eax,edx
mov edx, [esp+8]
mov eax, [edx+eax*4]
jmp L_dolen
ALIGN 4
L_test_for_second_level_dist:
test al,64
jnz L_invalid_distance_code
mov eax,1
shl eax,cl
dec eax
and eax,ebp
add eax,edx
mov edx, [esp+12]
mov eax, [edx+eax*4]
jmp L_dodist
ALIGN 4
L_clip_window:
; 721 "inffast.S"
mov ecx,eax
mov eax, [esp+52]
neg ecx
mov esi, [esp+56]
cmp eax,edx
jb L_invalid_distance_too_far
add ecx,edx
cmp dword ptr [esp+48],0
jne L_wrap_around_window
sub eax,ecx
add esi,eax
; 749 "inffast.S"
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
L_wrap_around_window:
; 793 "inffast.S"
mov eax, [esp+48]
cmp ecx,eax
jbe L_contiguous_in_window
add esi, [esp+52]
add esi,eax
sub esi,ecx
sub ecx,eax
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi, [esp+56]
mov ecx, [esp+48]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
L_contiguous_in_window:
; 836 "inffast.S"
add esi,eax
sub esi,ecx
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
L_do_copy1:
; 862 "inffast.S"
mov ecx,eax
rep movsb
mov esi, [esp+44]
jmp L_while_test
; 878 "inffast.S"
ALIGN 4
L_init_mmx:
emms
movd mm0,ebp
mov ebp,ebx
; 896 "inffast.S"
movd mm4,[esp+0]
movq mm3,mm4
movd mm5,[esp+4]
movq mm2,mm5
pxor mm1,mm1
mov ebx, [esp+8]
jmp L_do_loop_mmx
ALIGN 4
L_do_loop_mmx:
psrlq mm0,mm1
cmp ebp,32
ja L_get_length_code_mmx
movd mm6,ebp
movd mm7,[esi]
add esi,4
psllq mm7,mm6
add ebp,32
por mm0,mm7
L_get_length_code_mmx:
pand mm4,mm0
movd eax,mm4
movq mm4,mm3
mov eax, [ebx+eax*4]
L_dolen_mmx:
movzx ecx,ah
movd mm1,ecx
sub ebp,ecx
test al,al
jnz L_test_for_length_base_mmx
shr eax,16
stosb
L_while_test_mmx:
cmp [esp+16],edi
jbe L_break_loop
cmp [esp+20],esi
ja L_do_loop_mmx
jmp L_break_loop
L_test_for_length_base_mmx:
mov edx,eax
shr edx,16
test al,16
jz L_test_for_second_level_length_mmx
and eax,15
jz L_decode_distance_mmx
psrlq mm0,mm1
movd mm1,eax
movd ecx,mm0
sub ebp,eax
and ecx, [inflate_fast_mask+eax*4]
add edx,ecx
L_decode_distance_mmx:
psrlq mm0,mm1
cmp ebp,32
ja L_get_dist_code_mmx
movd mm6,ebp
movd mm7,[esi]
add esi,4
psllq mm7,mm6
add ebp,32
por mm0,mm7
L_get_dist_code_mmx:
mov ebx, [esp+12]
pand mm5,mm0
movd eax,mm5
movq mm5,mm2
mov eax, [ebx+eax*4]
L_dodist_mmx:
movzx ecx,ah
mov ebx,eax
shr ebx,16
sub ebp,ecx
movd mm1,ecx
test al,16
jz L_test_for_second_level_dist_mmx
and eax,15
jz L_check_dist_one_mmx
L_add_bits_to_dist_mmx:
psrlq mm0,mm1
movd mm1,eax
movd ecx,mm0
sub ebp,eax
and ecx, [inflate_fast_mask+eax*4]
add ebx,ecx
L_check_window_mmx:
mov [esp+44],esi
mov eax,edi
sub eax, [esp+40]
cmp eax,ebx
jb L_clip_window_mmx
mov ecx,edx
mov esi,edi
sub esi,ebx
sub ecx,3
mov al, [esi]
mov [edi],al
mov al, [esi+1]
mov dl, [esi+2]
add esi,3
mov [edi+1],al
mov [edi+2],dl
add edi,3
rep movsb
mov esi, [esp+44]
mov ebx, [esp+8]
jmp L_while_test_mmx
ALIGN 4
L_check_dist_one_mmx:
cmp ebx,1
jne L_check_window_mmx
cmp [esp+40],edi
je L_check_window_mmx
dec edi
mov ecx,edx
mov al, [edi]
sub ecx,3
mov [edi+1],al
mov [edi+2],al
mov [edi+3],al
add edi,4
rep stosb
mov ebx, [esp+8]
jmp L_while_test_mmx
ALIGN 4
L_test_for_second_level_length_mmx:
test al,64
jnz L_test_for_end_of_block
and eax,15
psrlq mm0,mm1
movd ecx,mm0
and ecx, [inflate_fast_mask+eax*4]
add ecx,edx
mov eax, [ebx+ecx*4]
jmp L_dolen_mmx
ALIGN 4
L_test_for_second_level_dist_mmx:
test al,64
jnz L_invalid_distance_code
and eax,15
psrlq mm0,mm1
movd ecx,mm0
and ecx, [inflate_fast_mask+eax*4]
mov eax, [esp+12]
add ecx,ebx
mov eax, [eax+ecx*4]
jmp L_dodist_mmx
ALIGN 4
L_clip_window_mmx:
mov ecx,eax
mov eax, [esp+52]
neg ecx
mov esi, [esp+56]
cmp eax,ebx
jb L_invalid_distance_too_far
add ecx,ebx
cmp dword ptr [esp+48],0
jne L_wrap_around_window_mmx
sub eax,ecx
add esi,eax
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
L_wrap_around_window_mmx:
mov eax, [esp+48]
cmp ecx,eax
jbe L_contiguous_in_window_mmx
add esi, [esp+52]
add esi,eax
sub esi,ecx
sub ecx,eax
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi, [esp+56]
mov ecx, [esp+48]
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
L_contiguous_in_window_mmx:
add esi,eax
sub esi,ecx
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
L_do_copy1_mmx:
mov ecx,edx
rep movsb
mov esi, [esp+44]
mov ebx, [esp+8]
jmp L_while_test_mmx
; 1174 "inffast.S"
L_invalid_distance_code:
mov ecx, invalid_distance_code_msg
mov edx,26
jmp L_update_stream_state
L_test_for_end_of_block:
test al,32
jz L_invalid_literal_length_code
mov ecx,0
mov edx,11
jmp L_update_stream_state
L_invalid_literal_length_code:
mov ecx, invalid_literal_length_code_msg
mov edx,26
jmp L_update_stream_state
L_invalid_distance_too_far:
mov esi, [esp+44]
mov ecx, invalid_distance_too_far_msg
mov edx,26
jmp L_update_stream_state
L_update_stream_state:
mov eax, [esp+88]
test ecx,ecx
jz L_skip_msg
mov [eax+24],ecx
L_skip_msg:
mov eax, [eax+28]
mov [eax+mode_state],edx
jmp L_break_loop
ALIGN 4
L_break_loop:
; 1243 "inffast.S"
cmp dword ptr [inflate_fast_use_mmx],2
jne L_update_next_in
mov ebx,ebp
L_update_next_in:
; 1266 "inffast.S"
mov eax, [esp+88]
mov ecx,ebx
mov edx, [eax+28]
shr ecx,3
sub esi,ecx
shl ecx,3
sub ebx,ecx
mov [eax+12],edi
mov [edx+bits_state],ebx
mov ecx,ebx
lea ebx, [esp+28]
cmp [esp+20],ebx
jne L_buf_not_used
sub esi,ebx
mov ebx, [eax+0]
mov [esp+20],ebx
add esi,ebx
mov ebx, [eax+4]
sub ebx,11
add [esp+20],ebx
L_buf_not_used:
mov [eax+0],esi
mov ebx,1
shl ebx,cl
dec ebx
cmp dword ptr [inflate_fast_use_mmx],2
jne L_update_hold
psrlq mm0,mm1
movd ebp,mm0
emms
L_update_hold:
and ebp,ebx
mov [edx+hold_state],ebp
mov ebx, [esp+20]
cmp ebx,esi
jbe L_last_is_smaller
sub ebx,esi
add ebx,11
mov [eax+4],ebx
jmp L_fixup_out
L_last_is_smaller:
sub esi,ebx
neg esi
add esi,11
mov [eax+4],esi
L_fixup_out:
mov ebx, [esp+16]
cmp ebx,edi
jbe L_end_is_smaller
sub ebx,edi
add ebx,257
mov [eax+16],ebx
jmp L_done
L_end_is_smaller:
sub edi,ebx
neg edi
add edi,257
mov [eax+16],edi
L_done:
add esp,64
popfd
pop ebx
pop ebp
pop esi
pop edi
ret
_TEXT ends
end
;/* inffas32.asm is a hand tuned assembler version of inffast.c -- fast decoding
; *
; * inffas32.asm is derivated from inffas86.c, with translation of assembly code
; *
; * Copyright (C) 1995-2003 Mark Adler
; * For conditions of distribution and use, see copyright notice in zlib.h
; *
; * Copyright (C) 2003 Chris Anderson <christop@charm.net>
; * Please use the copyright conditions above.
; *
; * Mar-13-2003 -- Most of this is derived from inffast.S which is derived from
; * the gcc -S output of zlib-1.2.0/inffast.c. Zlib-1.2.0 is in beta release at
; * the moment. I have successfully compiled and tested this code with gcc2.96,
; * gcc3.2, icc5.0, msvc6.0. It is very close to the speed of inffast.S
; * compiled with gcc -DNO_MMX, but inffast.S is still faster on the P3 with MMX
; * enabled. I will attempt to merge the MMX code into this version. Newer
; * versions of this and inffast.S can be found at
; * http://www.eetbeetee.com/zlib/ and http://www.charm.net/~christop/zlib/
; *
; * 2005 : modification by Gilles Vollant
; */
; For Visual C++ 4.x and higher and ML 6.x and higher
; ml.exe is in directory \MASM611C of Win95 DDK
; ml.exe is also distributed in http://www.masm32.com/masmdl.htm
; and in VC++2003 toolkit at http://msdn.microsoft.com/visualc/vctoolkit2003/
;
;
; compile with command line option
; ml /coff /Zi /c /Flinffas32.lst inffas32.asm
; if you define NO_GZIP (see inflate.h), compile with
; ml /coff /Zi /c /Flinffas32.lst /DNO_GUNZIP inffas32.asm
; zlib122sup is 0 fort zlib 1.2.2.1 and lower
; zlib122sup is 8 fort zlib 1.2.2.2 and more (with addition of dmax and head
; in inflate_state in inflate.h)
zlib1222sup equ 8
IFDEF GUNZIP
INFLATE_MODE_TYPE equ 11
INFLATE_MODE_BAD equ 26
ELSE
IFNDEF NO_GUNZIP
INFLATE_MODE_TYPE equ 11
INFLATE_MODE_BAD equ 26
ELSE
INFLATE_MODE_TYPE equ 3
INFLATE_MODE_BAD equ 17
ENDIF
ENDIF
; 75 "inffast.S"
;FILE "inffast.S"
;;;GLOBAL _inflate_fast
;;;SECTION .text
.586p
.mmx
name inflate_fast_x86
.MODEL FLAT
_DATA segment
inflate_fast_use_mmx:
dd 1
_TEXT segment
PUBLIC _inflate_fast
ALIGN 4
_inflate_fast:
jmp inflate_fast_entry
ALIGN 4
db 'Fast decoding Code from Chris Anderson'
db 0
ALIGN 4
invalid_literal_length_code_msg:
db 'invalid literal/length code'
db 0
ALIGN 4
invalid_distance_code_msg:
db 'invalid distance code'
db 0
ALIGN 4
invalid_distance_too_far_msg:
db 'invalid distance too far back'
db 0
ALIGN 4
inflate_fast_mask:
dd 0
dd 1
dd 3
dd 7
dd 15
dd 31
dd 63
dd 127
dd 255
dd 511
dd 1023
dd 2047
dd 4095
dd 8191
dd 16383
dd 32767
dd 65535
dd 131071
dd 262143
dd 524287
dd 1048575
dd 2097151
dd 4194303
dd 8388607
dd 16777215
dd 33554431
dd 67108863
dd 134217727
dd 268435455
dd 536870911
dd 1073741823
dd 2147483647
dd 4294967295
mode_state equ 0 ;/* state->mode */
wsize_state equ (32+zlib1222sup) ;/* state->wsize */
write_state equ (36+4+zlib1222sup) ;/* state->write */
window_state equ (40+4+zlib1222sup) ;/* state->window */
hold_state equ (44+4+zlib1222sup) ;/* state->hold */
bits_state equ (48+4+zlib1222sup) ;/* state->bits */
lencode_state equ (64+4+zlib1222sup) ;/* state->lencode */
distcode_state equ (68+4+zlib1222sup) ;/* state->distcode */
lenbits_state equ (72+4+zlib1222sup) ;/* state->lenbits */
distbits_state equ (76+4+zlib1222sup) ;/* state->distbits */
;;SECTION .text
; 205 "inffast.S"
;GLOBAL inflate_fast_use_mmx
;SECTION .data
; GLOBAL inflate_fast_use_mmx:object
;.size inflate_fast_use_mmx, 4
; 226 "inffast.S"
;SECTION .text
ALIGN 4
inflate_fast_entry:
push edi
push esi
push ebp
push ebx
pushfd
sub esp,64
cld
mov esi, [esp+88]
mov edi, [esi+28]
mov edx, [esi+4]
mov eax, [esi+0]
add edx,eax
sub edx,11
mov [esp+44],eax
mov [esp+20],edx
mov ebp, [esp+92]
mov ecx, [esi+16]
mov ebx, [esi+12]
sub ebp,ecx
neg ebp
add ebp,ebx
sub ecx,257
add ecx,ebx
mov [esp+60],ebx
mov [esp+40],ebp
mov [esp+16],ecx
; 285 "inffast.S"
mov eax, [edi+lencode_state]
mov ecx, [edi+distcode_state]
mov [esp+8],eax
mov [esp+12],ecx
mov eax,1
mov ecx, [edi+lenbits_state]
shl eax,cl
dec eax
mov [esp+0],eax
mov eax,1
mov ecx, [edi+distbits_state]
shl eax,cl
dec eax
mov [esp+4],eax
mov eax, [edi+wsize_state]
mov ecx, [edi+write_state]
mov edx, [edi+window_state]
mov [esp+52],eax
mov [esp+48],ecx
mov [esp+56],edx
mov ebp, [edi+hold_state]
mov ebx, [edi+bits_state]
; 321 "inffast.S"
mov esi, [esp+44]
mov ecx, [esp+20]
cmp ecx,esi
ja L_align_long
add ecx,11
sub ecx,esi
mov eax,12
sub eax,ecx
lea edi, [esp+28]
rep movsb
mov ecx,eax
xor eax,eax
rep stosb
lea esi, [esp+28]
mov [esp+20],esi
jmp L_is_aligned
L_align_long:
test esi,3
jz L_is_aligned
xor eax,eax
mov al, [esi]
inc esi
mov ecx,ebx
add ebx,8
shl eax,cl
or ebp,eax
jmp L_align_long
L_is_aligned:
mov edi, [esp+60]
; 366 "inffast.S"
L_check_mmx:
cmp dword ptr [inflate_fast_use_mmx],2
je L_init_mmx
ja L_do_loop
push eax
push ebx
push ecx
push edx
pushfd
mov eax, [esp]
xor dword ptr [esp],0200000h
popfd
pushfd
pop edx
xor edx,eax
jz L_dont_use_mmx
xor eax,eax
cpuid
cmp ebx,0756e6547h
jne L_dont_use_mmx
cmp ecx,06c65746eh
jne L_dont_use_mmx
cmp edx,049656e69h
jne L_dont_use_mmx
mov eax,1
cpuid
shr eax,8
and eax,15
cmp eax,6
jne L_dont_use_mmx
test edx,0800000h
jnz L_use_mmx
jmp L_dont_use_mmx
L_use_mmx:
mov dword ptr [inflate_fast_use_mmx],2
jmp L_check_mmx_pop
L_dont_use_mmx:
mov dword ptr [inflate_fast_use_mmx],3
L_check_mmx_pop:
pop edx
pop ecx
pop ebx
pop eax
jmp L_check_mmx
; 426 "inffast.S"
ALIGN 4
L_do_loop:
; 437 "inffast.S"
cmp bl,15
ja L_get_length_code
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
L_get_length_code:
mov edx, [esp+0]
mov ecx, [esp+8]
and edx,ebp
mov eax, [ecx+edx*4]
L_dolen:
mov cl,ah
sub bl,ah
shr ebp,cl
test al,al
jnz L_test_for_length_base
shr eax,16
stosb
L_while_test:
cmp [esp+16],edi
jbe L_break_loop
cmp [esp+20],esi
ja L_do_loop
jmp L_break_loop
L_test_for_length_base:
; 502 "inffast.S"
mov edx,eax
shr edx,16
mov cl,al
test al,16
jz L_test_for_second_level_length
and cl,15
jz L_save_len
cmp bl,cl
jae L_add_bits_to_len
mov ch,cl
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
mov cl,ch
L_add_bits_to_len:
mov eax,1
shl eax,cl
dec eax
sub bl,cl
and eax,ebp
shr ebp,cl
add edx,eax
L_save_len:
mov [esp+24],edx
L_decode_distance:
; 549 "inffast.S"
cmp bl,15
ja L_get_distance_code
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
L_get_distance_code:
mov edx, [esp+4]
mov ecx, [esp+12]
and edx,ebp
mov eax, [ecx+edx*4]
L_dodist:
mov edx,eax
shr edx,16
mov cl,ah
sub bl,ah
shr ebp,cl
; 584 "inffast.S"
mov cl,al
test al,16
jz L_test_for_second_level_dist
and cl,15
jz L_check_dist_one
cmp bl,cl
jae L_add_bits_to_dist
mov ch,cl
xor eax,eax
lodsw
mov cl,bl
add bl,16
shl eax,cl
or ebp,eax
mov cl,ch
L_add_bits_to_dist:
mov eax,1
shl eax,cl
dec eax
sub bl,cl
and eax,ebp
shr ebp,cl
add edx,eax
jmp L_check_window
L_check_window:
; 625 "inffast.S"
mov [esp+44],esi
mov eax,edi
sub eax, [esp+40]
cmp eax,edx
jb L_clip_window
mov ecx, [esp+24]
mov esi,edi
sub esi,edx
sub ecx,3
mov al, [esi]
mov [edi],al
mov al, [esi+1]
mov dl, [esi+2]
add esi,3
mov [edi+1],al
mov [edi+2],dl
add edi,3
rep movsb
mov esi, [esp+44]
jmp L_while_test
ALIGN 4
L_check_dist_one:
cmp edx,1
jne L_check_window
cmp [esp+40],edi
je L_check_window
dec edi
mov ecx, [esp+24]
mov al, [edi]
sub ecx,3
mov [edi+1],al
mov [edi+2],al
mov [edi+3],al
add edi,4
rep stosb
jmp L_while_test
ALIGN 4
L_test_for_second_level_length:
test al,64
jnz L_test_for_end_of_block
mov eax,1
shl eax,cl
dec eax
and eax,ebp
add eax,edx
mov edx, [esp+8]
mov eax, [edx+eax*4]
jmp L_dolen
ALIGN 4
L_test_for_second_level_dist:
test al,64
jnz L_invalid_distance_code
mov eax,1
shl eax,cl
dec eax
and eax,ebp
add eax,edx
mov edx, [esp+12]
mov eax, [edx+eax*4]
jmp L_dodist
ALIGN 4
L_clip_window:
; 721 "inffast.S"
mov ecx,eax
mov eax, [esp+52]
neg ecx
mov esi, [esp+56]
cmp eax,edx
jb L_invalid_distance_too_far
add ecx,edx
cmp dword ptr [esp+48],0
jne L_wrap_around_window
sub eax,ecx
add esi,eax
; 749 "inffast.S"
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
L_wrap_around_window:
; 793 "inffast.S"
mov eax, [esp+48]
cmp ecx,eax
jbe L_contiguous_in_window
add esi, [esp+52]
add esi,eax
sub esi,ecx
sub ecx,eax
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi, [esp+56]
mov ecx, [esp+48]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
jmp L_do_copy1
L_contiguous_in_window:
; 836 "inffast.S"
add esi,eax
sub esi,ecx
mov eax, [esp+24]
cmp eax,ecx
jbe L_do_copy1
sub eax,ecx
rep movsb
mov esi,edi
sub esi,edx
L_do_copy1:
; 862 "inffast.S"
mov ecx,eax
rep movsb
mov esi, [esp+44]
jmp L_while_test
; 878 "inffast.S"
ALIGN 4
L_init_mmx:
emms
movd mm0,ebp
mov ebp,ebx
; 896 "inffast.S"
movd mm4,[esp+0]
movq mm3,mm4
movd mm5,[esp+4]
movq mm2,mm5
pxor mm1,mm1
mov ebx, [esp+8]
jmp L_do_loop_mmx
ALIGN 4
L_do_loop_mmx:
psrlq mm0,mm1
cmp ebp,32
ja L_get_length_code_mmx
movd mm6,ebp
movd mm7,[esi]
add esi,4
psllq mm7,mm6
add ebp,32
por mm0,mm7
L_get_length_code_mmx:
pand mm4,mm0
movd eax,mm4
movq mm4,mm3
mov eax, [ebx+eax*4]
L_dolen_mmx:
movzx ecx,ah
movd mm1,ecx
sub ebp,ecx
test al,al
jnz L_test_for_length_base_mmx
shr eax,16
stosb
L_while_test_mmx:
cmp [esp+16],edi
jbe L_break_loop
cmp [esp+20],esi
ja L_do_loop_mmx
jmp L_break_loop
L_test_for_length_base_mmx:
mov edx,eax
shr edx,16
test al,16
jz L_test_for_second_level_length_mmx
and eax,15
jz L_decode_distance_mmx
psrlq mm0,mm1
movd mm1,eax
movd ecx,mm0
sub ebp,eax
and ecx, [inflate_fast_mask+eax*4]
add edx,ecx
L_decode_distance_mmx:
psrlq mm0,mm1
cmp ebp,32
ja L_get_dist_code_mmx
movd mm6,ebp
movd mm7,[esi]
add esi,4
psllq mm7,mm6
add ebp,32
por mm0,mm7
L_get_dist_code_mmx:
mov ebx, [esp+12]
pand mm5,mm0
movd eax,mm5
movq mm5,mm2
mov eax, [ebx+eax*4]
L_dodist_mmx:
movzx ecx,ah
mov ebx,eax
shr ebx,16
sub ebp,ecx
movd mm1,ecx
test al,16
jz L_test_for_second_level_dist_mmx
and eax,15
jz L_check_dist_one_mmx
L_add_bits_to_dist_mmx:
psrlq mm0,mm1
movd mm1,eax
movd ecx,mm0
sub ebp,eax
and ecx, [inflate_fast_mask+eax*4]
add ebx,ecx
L_check_window_mmx:
mov [esp+44],esi
mov eax,edi
sub eax, [esp+40]
cmp eax,ebx
jb L_clip_window_mmx
mov ecx,edx
mov esi,edi
sub esi,ebx
sub ecx,3
mov al, [esi]
mov [edi],al
mov al, [esi+1]
mov dl, [esi+2]
add esi,3
mov [edi+1],al
mov [edi+2],dl
add edi,3
rep movsb
mov esi, [esp+44]
mov ebx, [esp+8]
jmp L_while_test_mmx
ALIGN 4
L_check_dist_one_mmx:
cmp ebx,1
jne L_check_window_mmx
cmp [esp+40],edi
je L_check_window_mmx
dec edi
mov ecx,edx
mov al, [edi]
sub ecx,3
mov [edi+1],al
mov [edi+2],al
mov [edi+3],al
add edi,4
rep stosb
mov ebx, [esp+8]
jmp L_while_test_mmx
ALIGN 4
L_test_for_second_level_length_mmx:
test al,64
jnz L_test_for_end_of_block
and eax,15
psrlq mm0,mm1
movd ecx,mm0
and ecx, [inflate_fast_mask+eax*4]
add ecx,edx
mov eax, [ebx+ecx*4]
jmp L_dolen_mmx
ALIGN 4
L_test_for_second_level_dist_mmx:
test al,64
jnz L_invalid_distance_code
and eax,15
psrlq mm0,mm1
movd ecx,mm0
and ecx, [inflate_fast_mask+eax*4]
mov eax, [esp+12]
add ecx,ebx
mov eax, [eax+ecx*4]
jmp L_dodist_mmx
ALIGN 4
L_clip_window_mmx:
mov ecx,eax
mov eax, [esp+52]
neg ecx
mov esi, [esp+56]
cmp eax,ebx
jb L_invalid_distance_too_far
add ecx,ebx
cmp dword ptr [esp+48],0
jne L_wrap_around_window_mmx
sub eax,ecx
add esi,eax
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
L_wrap_around_window_mmx:
mov eax, [esp+48]
cmp ecx,eax
jbe L_contiguous_in_window_mmx
add esi, [esp+52]
add esi,eax
sub esi,ecx
sub ecx,eax
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi, [esp+56]
mov ecx, [esp+48]
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
jmp L_do_copy1_mmx
L_contiguous_in_window_mmx:
add esi,eax
sub esi,ecx
cmp edx,ecx
jbe L_do_copy1_mmx
sub edx,ecx
rep movsb
mov esi,edi
sub esi,ebx
L_do_copy1_mmx:
mov ecx,edx
rep movsb
mov esi, [esp+44]
mov ebx, [esp+8]
jmp L_while_test_mmx
; 1174 "inffast.S"
L_invalid_distance_code:
mov ecx, invalid_distance_code_msg
mov edx,INFLATE_MODE_BAD
jmp L_update_stream_state
L_test_for_end_of_block:
test al,32
jz L_invalid_literal_length_code
mov ecx,0
mov edx,INFLATE_MODE_TYPE
jmp L_update_stream_state
L_invalid_literal_length_code:
mov ecx, invalid_literal_length_code_msg
mov edx,INFLATE_MODE_BAD
jmp L_update_stream_state
L_invalid_distance_too_far:
mov esi, [esp+44]
mov ecx, invalid_distance_too_far_msg
mov edx,INFLATE_MODE_BAD
jmp L_update_stream_state
L_update_stream_state:
mov eax, [esp+88]
test ecx,ecx
jz L_skip_msg
mov [eax+24],ecx
L_skip_msg:
mov eax, [eax+28]
mov [eax+mode_state],edx
jmp L_break_loop
ALIGN 4
L_break_loop:
; 1243 "inffast.S"
cmp dword ptr [inflate_fast_use_mmx],2
jne L_update_next_in
mov ebx,ebp
L_update_next_in:
; 1266 "inffast.S"
mov eax, [esp+88]
mov ecx,ebx
mov edx, [eax+28]
shr ecx,3
sub esi,ecx
shl ecx,3
sub ebx,ecx
mov [eax+12],edi
mov [edx+bits_state],ebx
mov ecx,ebx
lea ebx, [esp+28]
cmp [esp+20],ebx
jne L_buf_not_used
sub esi,ebx
mov ebx, [eax+0]
mov [esp+20],ebx
add esi,ebx
mov ebx, [eax+4]
sub ebx,11
add [esp+20],ebx
L_buf_not_used:
mov [eax+0],esi
mov ebx,1
shl ebx,cl
dec ebx
cmp dword ptr [inflate_fast_use_mmx],2
jne L_update_hold
psrlq mm0,mm1
movd ebp,mm0
emms
L_update_hold:
and ebp,ebx
mov [edx+hold_state],ebp
mov ebx, [esp+20]
cmp ebx,esi
jbe L_last_is_smaller
sub ebx,esi
add ebx,11
mov [eax+4],ebx
jmp L_fixup_out
L_last_is_smaller:
sub esi,ebx
neg esi
add esi,11
mov [eax+4],esi
L_fixup_out:
mov ebx, [esp+16]
cmp ebx,edi
jbe L_end_is_smaller
sub ebx,edi
add ebx,257
mov [eax+16],ebx
jmp L_done
L_end_is_smaller:
sub edi,ebx
neg edi
add edi,257
mov [eax+16],edi
L_done:
add esp,64
popfd
pop ebx
pop ebp
pop esi
pop edi
ret
_TEXT ends
end
zlib/contrib/masmx86/mkasm.bat
cl /I..\.. /O2 /c gvmat32c.c
cl /DASMV /I..\.. /O2 /c gvmat32c.c
ml /coff /Zi /c /Flgvmat32.lst gvmat32.asm
ml /coff /Zi /c /Flinffas32.lst inffas32.asm
zlib/contrib/minizip/crypt.h
/* crypt.h -- base code for crypt/uncrypt ZIPfile
Version 1.00, September 10th, 2003
Version 1.01e, February 12th, 2005
Copyright (C) 1998-2003 Gilles Vollant
Copyright (C) 1998-2005 Gilles Vollant
This code is a modified version of crypting code in Infozip distribution
......
zlib/contrib/minizip/ioapi.c
/* ioapi.c -- IO base function header for compress/uncompress .zip
files using zlib + zip or unzip API
Version 1.00, September 10th, 2003
Version 1.01e, February 12th, 2005
Copyright (C) 1998-2003 Gilles Vollant
Copyright (C) 1998-2005 Gilles Vollant
*/
#include <stdio.h>
......@@ -94,7 +94,7 @@ uLong ZCALLBACK fread_file_func (opaque, stream, buf, size)
uLong size;
{
uLong ret;
ret = fread(buf, 1, (size_t)size, (FILE *)stream);
ret = (uLong)fread(buf, 1, (size_t)size, (FILE *)stream);
return ret;
}
......@@ -106,7 +106,7 @@ uLong ZCALLBACK fwrite_file_func (opaque, stream, buf, size)
uLong size;
{
uLong ret;
ret = fwrite(buf, 1, (size_t)size, (FILE *)stream);
ret = (uLong)fwrite(buf, 1, (size_t)size, (FILE *)stream);
return ret;
}
......
zlib/contrib/minizip/ioapi.h
/* ioapi.h -- IO base function header for compress/uncompress .zip
files using zlib + zip or unzip API
Version 1.00, September 10th, 2003
Version 1.01e, February 12th, 2005
Copyright (C) 1998-2003 Gilles Vollant
Copyright (C) 1998-2005 Gilles Vollant
*/
#ifndef _ZLIBIOAPI_H
......
zlib/contrib/minizip/iowin32.c
......@@ -2,9 +2,9 @@
files using zlib + zip or unzip API
This IO API version uses the Win32 API (for Microsoft Windows)
Version 1.00, September 10th, 2003
Version 1.01e, February 12th, 2005
Copyright (C) 1998-2003 Gilles Vollant
Copyright (C) 1998-2005 Gilles Vollant
*/
#include <stdlib.h>
......
zlib/contrib/minizip/iowin32.h
......@@ -2,9 +2,9 @@
files using zlib + zip or unzip API
This IO API version uses the Win32 API (for Microsoft Windows)
Version 1.00, September 10th, 2003
Version 1.01e, February 12th, 2005
Copyright (C) 1998-2003 Gilles Vollant
Copyright (C) 1998-2005 Gilles Vollant
*/
#include <windows.h>
......
zlib/contrib/pascal/zlibpas.pas
......@@ -10,7 +10,7 @@ unit zlibpas;
interface
const
ZLIB_VERSION = '1.2.1';
ZLIB_VERSION = '1.2.3';
type
alloc_func = function(opaque: Pointer; items, size: Integer): Pointer;
......
zlib/contrib/puff/puff.c
/*
* puff.c
* Copyright (C) 2002, 2003 Mark Adler
* Copyright (C) 2002-2004 Mark Adler
* For conditions of distribution and use, see copyright notice in puff.h
* version 1.7, 3 Mar 2003
* version 1.8, 9 Jan 2004
*
* puff.c is a simple inflate written to be an unambiguous way to specify the
* deflate format. It is not written for speed but rather simplicity. As a
......@@ -60,6 +60,7 @@
* 1.6 7 Aug 2002 - Minor format changes
* 1.7 3 Mar 2003 - Added test code for distribution
* - Added zlib-like license
* 1.8 9 Jan 2004 - Added some comments on no distance codes case
*/
#include <setjmp.h> /* for setjmp(), longjmp(), and jmp_buf */
......@@ -577,6 +578,9 @@ local int fixed(struct state *s)
* block is fewer bits), but it is allowed by the format. So incomplete
* literal/length codes of one symbol should also be permitted.
*
* - If there are only literal codes and no lengths, then there are no distance
* codes. This is represented by one distance code with zero bits.
*
* - The list of up to 286 length/literal lengths and up to 30 distance lengths
* are themselves compressed using Huffman codes and run-length encoding. In
* the list of code lengths, a 0 symbol means no code, a 1..15 symbol means
......
zlib/contrib/testzlib/testzlib.c
#include <stdio.h>
#include <stdlib.h>
#include <windows.h>
#include "zlib.h"
int ReadFileMemory(const char* filename,long* plFileSize,void** pFilePtr)
{
FILE* stream;
void* ptr;
int retVal=1;
stream=fopen(filename, "rb");
if (stream==NULL)
return 0;
fseek(stream,0,SEEK_END);
*plFileSize=ftell(stream);
fseek(stream,0,SEEK_SET);
ptr=malloc((*plFileSize)+1);
if (ptr==NULL)
retVal=0;
else
{
if (fread(ptr, 1, *plFileSize,stream) != (*plFileSize))
retVal=0;
}
fclose(stream);
*pFilePtr=ptr;
return retVal;
}
int main(int argc, char *argv[])
{
int BlockSizeCompress=0x8000;
int BlockSizeUncompress=0x8000;
int cprLevel=Z_DEFAULT_COMPRESSION ;
long lFileSize;
unsigned char* FilePtr;
long lBufferSizeCpr;
long lBufferSizeUncpr;
long lCompressedSize=0;
unsigned char* CprPtr;
unsigned char* UncprPtr;
long lSizeCpr,lSizeUncpr;
DWORD dwGetTick;
if (argc<=1)
{
printf("run TestZlib <File> [BlockSizeCompress] [BlockSizeUncompress] [compres. level]\n");
return 0;
}
if (ReadFileMemory(argv[1],&lFileSize,&FilePtr)==0)
{
printf("error reading %s\n",argv[1]);
return 1;
}
else printf("file %s read, %u bytes\n",argv[1],lFileSize);
if (argc>=3)
BlockSizeCompress=atol(argv[2]);
if (argc>=4)
BlockSizeUncompress=atol(argv[3]);
if (argc>=5)
cprLevel=(int)atol(argv[4]);
lBufferSizeCpr = lFileSize + (lFileSize/0x10) + 0x200;
lBufferSizeUncpr = lBufferSizeCpr;
CprPtr=(unsigned char*)malloc(lBufferSizeCpr + BlockSizeCompress);
UncprPtr=(unsigned char*)malloc(lBufferSizeUncpr + BlockSizeUncompress);
dwGetTick=GetTickCount();
{
z_stream zcpr;
int ret=Z_OK;
long lOrigToDo = lFileSize;
long lOrigDone = 0;
int step=0;
memset(&zcpr,0,sizeof(z_stream));
deflateInit(&zcpr,cprLevel);
zcpr.next_in = FilePtr;
zcpr.next_out = CprPtr;
do
{
long all_read_before = zcpr.total_in;
zcpr.avail_in = min(lOrigToDo,BlockSizeCompress);
zcpr.avail_out = BlockSizeCompress;
ret=deflate(&zcpr,(zcpr.avail_in==lOrigToDo) ? Z_FINISH : Z_SYNC_FLUSH);
lOrigDone += (zcpr.total_in-all_read_before);
lOrigToDo -= (zcpr.total_in-all_read_before);
step++;
} while (ret==Z_OK);
lSizeCpr=zcpr.total_out;
deflateEnd(&zcpr);
dwGetTick=GetTickCount()-dwGetTick;
printf("total compress size = %u, in %u step\n",lSizeCpr,step);
printf("time = %u msec = %f sec\n\n",dwGetTick,dwGetTick/(double)1000.);
}
dwGetTick=GetTickCount();
{
z_stream zcpr;
int ret=Z_OK;
long lOrigToDo = lSizeCpr;
long lOrigDone = 0;
int step=0;
memset(&zcpr,0,sizeof(z_stream));
inflateInit(&zcpr);
zcpr.next_in = CprPtr;
zcpr.next_out = UncprPtr;
do
{
long all_read_before = zcpr.total_in;
zcpr.avail_in = min(lOrigToDo,BlockSizeUncompress);
zcpr.avail_out = BlockSizeUncompress;
ret=inflate(&zcpr,Z_SYNC_FLUSH);
lOrigDone += (zcpr.total_in-all_read_before);
lOrigToDo -= (zcpr.total_in-all_read_before);
step++;
} while (ret==Z_OK);
lSizeUncpr=zcpr.total_out;
inflateEnd(&zcpr);
dwGetTick=GetTickCount()-dwGetTick;
printf("total uncompress size = %u, in %u step\n",lSizeUncpr,step);
printf("time = %u msec = %f sec\n\n",dwGetTick,dwGetTick/(double)1000.);
}
if (lSizeUncpr==lFileSize)
{
if (memcmp(FilePtr,UncprPtr,lFileSize)==0)
printf("compare ok\n");
}
return 0;
}
#include <stdio.h>
#include <stdlib.h>
#include <windows.h>
#include "zlib.h"
void MyDoMinus64(LARGE_INTEGER *R,LARGE_INTEGER A,LARGE_INTEGER B)
{
R->HighPart = A.HighPart - B.HighPart;
if (A.LowPart >= B.LowPart)
R->LowPart = A.LowPart - B.LowPart;
else
{
R->LowPart = A.LowPart - B.LowPart;
R->HighPart --;
}
}
#ifdef _M_X64
// see http://msdn2.microsoft.com/library/twchhe95(en-us,vs.80).aspx for __rdtsc
unsigned __int64 __rdtsc(void);
void BeginCountRdtsc(LARGE_INTEGER * pbeginTime64)
{
// printf("rdtsc = %I64x\n",__rdtsc());
pbeginTime64->QuadPart=__rdtsc();
}
LARGE_INTEGER GetResRdtsc(LARGE_INTEGER beginTime64,BOOL fComputeTimeQueryPerf)
{
LARGE_INTEGER LIres;
unsigned _int64 res=__rdtsc()-((unsigned _int64)(beginTime64.QuadPart));
LIres.QuadPart=res;
// printf("rdtsc = %I64x\n",__rdtsc());
return LIres;
}
#else
#ifdef _M_IX86
void myGetRDTSC32(LARGE_INTEGER * pbeginTime64)
{
DWORD dwEdx,dwEax;
_asm
{
rdtsc
mov dwEax,eax
mov dwEdx,edx
}
pbeginTime64->LowPart=dwEax;
pbeginTime64->HighPart=dwEdx;
}
void BeginCountRdtsc(LARGE_INTEGER * pbeginTime64)
{
myGetRDTSC32(pbeginTime64);
}
LARGE_INTEGER GetResRdtsc(LARGE_INTEGER beginTime64,BOOL fComputeTimeQueryPerf)
{
LARGE_INTEGER LIres,endTime64;
myGetRDTSC32(&endTime64);
LIres.LowPart=LIres.HighPart=0;
MyDoMinus64(&LIres,endTime64,beginTime64);
return LIres;
}
#else
void myGetRDTSC32(LARGE_INTEGER * pbeginTime64)
{
}
void BeginCountRdtsc(LARGE_INTEGER * pbeginTime64)
{
}
LARGE_INTEGER GetResRdtsc(LARGE_INTEGER beginTime64,BOOL fComputeTimeQueryPerf)
{
LARGE_INTEGER lr;
lr.QuadPart=0;
return lr;
}
#endif
#endif
void BeginCountPerfCounter(LARGE_INTEGER * pbeginTime64,BOOL fComputeTimeQueryPerf)
{
if ((!fComputeTimeQueryPerf) || (!QueryPerformanceCounter(pbeginTime64)))
{
pbeginTime64->LowPart = GetTickCount();
pbeginTime64->HighPart = 0;
}
}
DWORD GetMsecSincePerfCounter(LARGE_INTEGER beginTime64,BOOL fComputeTimeQueryPerf)
{
LARGE_INTEGER endTime64,ticksPerSecond,ticks;
DWORDLONG ticksShifted,tickSecShifted;
DWORD dwLog=16+0;
DWORD dwRet;
if ((!fComputeTimeQueryPerf) || (!QueryPerformanceCounter(&endTime64)))
dwRet = (GetTickCount() - beginTime64.LowPart)*1;
else
{
MyDoMinus64(&ticks,endTime64,beginTime64);
QueryPerformanceFrequency(&ticksPerSecond);
{
ticksShifted = Int64ShrlMod32(*(DWORDLONG*)&ticks,dwLog);
tickSecShifted = Int64ShrlMod32(*(DWORDLONG*)&ticksPerSecond,dwLog);
}
dwRet = (DWORD)((((DWORD)ticksShifted)*1000)/(DWORD)(tickSecShifted));
dwRet *=1;
}
return dwRet;
}
int ReadFileMemory(const char* filename,long* plFileSize,void** pFilePtr)
{
FILE* stream;
void* ptr;
int retVal=1;
stream=fopen(filename, "rb");
if (stream==NULL)
return 0;
fseek(stream,0,SEEK_END);
*plFileSize=ftell(stream);
fseek(stream,0,SEEK_SET);
ptr=malloc((*plFileSize)+1);
if (ptr==NULL)
retVal=0;
else
{
if (fread(ptr, 1, *plFileSize,stream) != (*plFileSize))
retVal=0;
}
fclose(stream);
*pFilePtr=ptr;
return retVal;
}
int main(int argc, char *argv[])
{
int BlockSizeCompress=0x8000;
int BlockSizeUncompress=0x8000;
int cprLevel=Z_DEFAULT_COMPRESSION ;
long lFileSize;
unsigned char* FilePtr;
long lBufferSizeCpr;
long lBufferSizeUncpr;
long lCompressedSize=0;
unsigned char* CprPtr;
unsigned char* UncprPtr;
long lSizeCpr,lSizeUncpr;
DWORD dwGetTick,dwMsecQP;
LARGE_INTEGER li_qp,li_rdtsc,dwResRdtsc;
if (argc<=1)
{
printf("run TestZlib <File> [BlockSizeCompress] [BlockSizeUncompress] [compres. level]\n");
return 0;
}
if (ReadFileMemory(argv[1],&lFileSize,&FilePtr)==0)
{
printf("error reading %s\n",argv[1]);
return 1;
}
else printf("file %s read, %u bytes\n",argv[1],lFileSize);
if (argc>=3)
BlockSizeCompress=atol(argv[2]);
if (argc>=4)
BlockSizeUncompress=atol(argv[3]);
if (argc>=5)
cprLevel=(int)atol(argv[4]);
lBufferSizeCpr = lFileSize + (lFileSize/0x10) + 0x200;
lBufferSizeUncpr = lBufferSizeCpr;
CprPtr=(unsigned char*)malloc(lBufferSizeCpr + BlockSizeCompress);
BeginCountPerfCounter(&li_qp,TRUE);
dwGetTick=GetTickCount();
BeginCountRdtsc(&li_rdtsc);
{
z_stream zcpr;
int ret=Z_OK;
long lOrigToDo = lFileSize;
long lOrigDone = 0;
int step=0;
memset(&zcpr,0,sizeof(z_stream));
deflateInit(&zcpr,cprLevel);
zcpr.next_in = FilePtr;
zcpr.next_out = CprPtr;
do
{
long all_read_before = zcpr.total_in;
zcpr.avail_in = min(lOrigToDo,BlockSizeCompress);
zcpr.avail_out = BlockSizeCompress;
ret=deflate(&zcpr,(zcpr.avail_in==lOrigToDo) ? Z_FINISH : Z_SYNC_FLUSH);
lOrigDone += (zcpr.total_in-all_read_before);
lOrigToDo -= (zcpr.total_in-all_read_before);
step++;
} while (ret==Z_OK);
lSizeCpr=zcpr.total_out;
deflateEnd(&zcpr);
dwGetTick=GetTickCount()-dwGetTick;
dwMsecQP=GetMsecSincePerfCounter(li_qp,TRUE);
dwResRdtsc=GetResRdtsc(li_rdtsc,TRUE);
printf("total compress size = %u, in %u step\n",lSizeCpr,step);
printf("time = %u msec = %f sec\n",dwGetTick,dwGetTick/(double)1000.);
printf("defcpr time QP = %u msec = %f sec\n",dwMsecQP,dwMsecQP/(double)1000.);
printf("defcpr result rdtsc = %I64x\n\n",dwResRdtsc.QuadPart);
}
CprPtr=(unsigned char*)realloc(CprPtr,lSizeCpr);
UncprPtr=(unsigned char*)malloc(lBufferSizeUncpr + BlockSizeUncompress);
BeginCountPerfCounter(&li_qp,TRUE);
dwGetTick=GetTickCount();
BeginCountRdtsc(&li_rdtsc);
{
z_stream zcpr;
int ret=Z_OK;
long lOrigToDo = lSizeCpr;
long lOrigDone = 0;
int step=0;
memset(&zcpr,0,sizeof(z_stream));
inflateInit(&zcpr);
zcpr.next_in = CprPtr;
zcpr.next_out = UncprPtr;
do
{
long all_read_before = zcpr.total_in;
zcpr.avail_in = min(lOrigToDo,BlockSizeUncompress);
zcpr.avail_out = BlockSizeUncompress;
ret=inflate(&zcpr,Z_SYNC_FLUSH);
lOrigDone += (zcpr.total_in-all_read_before);
lOrigToDo -= (zcpr.total_in-all_read_before);
step++;
} while (ret==Z_OK);
lSizeUncpr=zcpr.total_out;
inflateEnd(&zcpr);
dwGetTick=GetTickCount()-dwGetTick;
dwMsecQP=GetMsecSincePerfCounter(li_qp,TRUE);
dwResRdtsc=GetResRdtsc(li_rdtsc,TRUE);
printf("total uncompress size = %u, in %u step\n",lSizeUncpr,step);
printf("time = %u msec = %f sec\n",dwGetTick,dwGetTick/(double)1000.);
printf("uncpr time QP = %u msec = %f sec\n",dwMsecQP,dwMsecQP/(double)1000.);
printf("uncpr result rdtsc = %I64x\n\n",dwResRdtsc.QuadPart);
}
if (lSizeUncpr==lFileSize)
{
if (memcmp(FilePtr,UncprPtr,lFileSize)==0)
printf("compare ok\n");
}
return 0;
}
zlib/contrib/vstudio/readme.txt
Building instructions for the DLL versions of Zlib 1.21
=======================================================
Building instructions for the DLL versions of Zlib 1.2.3
========================================================
This directory contains projects that build zlib and minizip using
Microsoft Visual C++ 7.0/7.1.
Microsoft Visual C++ 7.0/7.1, and Visual C++ .
You don't need to build these projects yourself. You can download the
binaries from:
......@@ -11,18 +11,36 @@ binaries from:
More information can be found at this site.
Build instructions
------------------
- Unzip zlib*.zip and copy the files from contrib\vstudio\vc7,
from contrib\vstudio\masmx86 and from contrib\minizip into the same
directory.
Build instructions for Visual Studio 7.x (32 bits)
--------------------------------------------------
- Uncompress current zlib, including all contrib/* files
- Download the crtdll library from
http://www.winimage.com/zLibDll/crtdll.zip
Unzip crtdll.zip to extract crtdll.lib.
- If you are using x86, use the Release target.
- Open zlibvc.sln with Microsoft Visual C++ 7.0 or 7.1
Unzip crtdll.zip to extract crtdll.lib on contrib\vstudio\vc7.
- Open contrib\vstudio\vc7\zlibvc.sln with Microsoft Visual C++ 7.x
(Visual Studio .Net 2002 or 2003).
Build instructions for Visual Studio 2005 (32 bits or 64 bits)
--------------------------------------------------------------
- Uncompress current zlib, including all contrib/* files
- For 32 bits only: download the crtdll library from
http://www.winimage.com/zLibDll/crtdll.zip
Unzip crtdll.zip to extract crtdll.lib on contrib\vstudio\vc8.
- Open contrib\vstudio\vc8\zlibvc.sln with Microsoft Visual C++ 8.0
Build instructions for Visual Studio 2005 64 bits, PSDK compiler
----------------------------------------------------------------
at the time of writing this text file, Visual Studio 2005 (and
Microsoft Visual C++ 8.0) is on the beta 2 stage.
Using you can get the free 64 bits compiler from Platform SDK,
which is NOT a beta, and compile using the Visual studio 2005 IDE
see http://www.winimage.com/misc/sdk64onvs2005/ for instruction
- Uncompress current zlib, including all contrib/* files
- start Visual Studio 2005 from a platform SDK command prompt, using
the /useenv switch
- Open contrib\vstudio\vc8\zlibvc.sln with Microsoft Visual C++ 8.0
Important
---------
......
zlib/contrib/vstudio/vc7/miniunz.vcproj
......@@ -19,6 +19,7 @@
<Tool
Name="VCCLCompilerTool"
Optimization="0"
AdditionalIncludeDirectories="..\..\..;..\..\minizip"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI;_DEBUG;_CONSOLE"
MinimalRebuild="TRUE"
BasicRuntimeChecks="3"
......@@ -63,6 +64,7 @@
Optimization="2"
InlineFunctionExpansion="1"
OmitFramePointers="TRUE"
AdditionalIncludeDirectories="..\..\..;..\..\minizip"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI;NDEBUG;_CONSOLE"
StringPooling="TRUE"
RuntimeLibrary="4"
......@@ -104,7 +106,7 @@
Name="Source Files"
Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm">
<File
RelativePath="miniunz.c">
RelativePath="..\..\minizip\miniunz.c">
</File>
</Filter>
<Filter
......@@ -116,7 +118,7 @@
Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe">
</Filter>
<File
RelativePath="zlibwapi.lib">
RelativePath="ReleaseDll\zlibwapi.lib">
</File>
</Files>
<Globals>
......
zlib/contrib/vstudio/vc7/minizip.vcproj
......@@ -19,6 +19,7 @@
<Tool
Name="VCCLCompilerTool"
Optimization="0"
AdditionalIncludeDirectories="..\..\..;..\..\minizip"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI;_DEBUG;_CONSOLE"
MinimalRebuild="TRUE"
BasicRuntimeChecks="3"
......@@ -63,6 +64,7 @@
Optimization="2"
InlineFunctionExpansion="1"
OmitFramePointers="TRUE"
AdditionalIncludeDirectories="..\..\..;..\..\minizip"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI;NDEBUG;_CONSOLE"
StringPooling="TRUE"
RuntimeLibrary="4"
......@@ -104,7 +106,7 @@
Name="Source Files"
Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm">
<File
RelativePath="minizip.c">
RelativePath="..\..\minizip\minizip.c">
</File>
</Filter>
<Filter
......@@ -116,7 +118,7 @@
Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe">
</Filter>
<File
RelativePath="zlibwapi.lib">
RelativePath="ReleaseDll\zlibwapi.lib">
</File>
</Files>
<Globals>
......
zlib/contrib/vstudio/vc7/zlib.rc
......@@ -2,8 +2,8 @@
#define IDR_VERSION1 1
IDR_VERSION1 VERSIONINFO MOVEABLE IMPURE LOADONCALL DISCARDABLE
FILEVERSION 1,2,1,0
PRODUCTVERSION 1,2,1,0
FILEVERSION 1,2,3,0
PRODUCTVERSION 1,2,3,0
FILEFLAGSMASK VS_FFI_FILEFLAGSMASK
FILEFLAGS 0
FILEOS VOS_DOS_WINDOWS32
......@@ -17,7 +17,7 @@ BEGIN
BEGIN
VALUE "FileDescription", "zlib data compression library\0"
VALUE "FileVersion", "1.2.1.0\0"
VALUE "FileVersion", "1.2.3.0\0"
VALUE "InternalName", "zlib\0"
VALUE "OriginalFilename", "zlib.dll\0"
VALUE "ProductName", "ZLib.DLL\0"
......
zlib/contrib/vstudio/vc7/zlibstat.vcproj
......@@ -20,6 +20,7 @@
<Tool
Name="VCCLCompilerTool"
Optimization="0"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI"
ExceptionHandling="FALSE"
RuntimeLibrary="5"
......@@ -61,6 +62,7 @@
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -102,6 +104,7 @@
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI;ASMV;ASMINF"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -117,7 +120,7 @@
Name="VCCustomBuildTool"/>
<Tool
Name="VCLibrarianTool"
AdditionalOptions="gvmat32.obj inffas32.obj /NODEFAULTLIB "
AdditionalOptions="..\..\masmx86\gvmat32.obj ..\..\masmx86\inffas32.obj /NODEFAULTLIB "
OutputFile=".\zlibstat\zlibstat.lib"
SuppressStartupBanner="TRUE"/>
<Tool
......@@ -144,6 +147,7 @@
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32;ZLIB_WINAPI"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -182,49 +186,49 @@
Name="Source Files"
Filter="">
<File
RelativePath=".\adler32.c">
RelativePath="..\..\..\adler32.c">
</File>
<File
RelativePath=".\compress.c">
RelativePath="..\..\..\compress.c">
</File>
<File
RelativePath=".\crc32.c">
RelativePath="..\..\..\crc32.c">
</File>
<File
RelativePath=".\deflate.c">
RelativePath="..\..\..\deflate.c">
</File>
<File
RelativePath=".\gvmat32c.c">
RelativePath="..\..\masmx86\gvmat32c.c">
</File>
<File
RelativePath=".\gzio.c">
RelativePath="..\..\..\gzio.c">
</File>
<File
RelativePath=".\infback.c">
RelativePath="..\..\..\infback.c">
</File>
<File
RelativePath=".\inffast.c">
RelativePath="..\..\..\inffast.c">
</File>
<File
RelativePath=".\inflate.c">
RelativePath="..\..\..\inflate.c">
</File>
<File
RelativePath=".\inftrees.c">
RelativePath="..\..\..\inftrees.c">
</File>
<File
RelativePath=".\ioapi.c">
RelativePath="..\..\minizip\ioapi.c">
</File>
<File
RelativePath=".\trees.c">
RelativePath="..\..\..\trees.c">
</File>
<File
RelativePath=".\uncompr.c">
RelativePath="..\..\..\uncompr.c">
</File>
<File
RelativePath=".\unzip.c">
RelativePath="..\..\minizip\unzip.c">
</File>
<File
RelativePath=".\zip.c">
RelativePath="..\..\minizip\zip.c">
</File>
<File
RelativePath=".\zlib.rc">
......@@ -233,7 +237,7 @@
RelativePath=".\zlibvc.def">
</File>
<File
RelativePath=".\zutil.c">
RelativePath="..\..\..\zutil.c">
</File>
</Filter>
</Files>
......
zlib/contrib/vstudio/vc7/zlibvc.sln
......@@ -7,6 +7,8 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "minizip", "minizip.vcproj",
EndProject
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "miniunz", "miniunz.vcproj", "{C52F9E7B-498A-42BE-8DB4-85A15694382A}"
EndProject
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "testZlibDll", "testzlib.vcproj", "{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}"
EndProject
Global
GlobalSection(SolutionConfiguration) = preSolution
ConfigName.0 = Debug
......@@ -58,6 +60,16 @@ Global
{C52F9E7B-498A-42BE-8DB4-85A15694382A}.ReleaseWithoutAsm.Build.0 = Release|Win32
{C52F9E7B-498A-42BE-8DB4-85A15694382A}.ReleaseWithoutCrtdll.ActiveCfg = Release|Win32
{C52F9E7B-498A-42BE-8DB4-85A15694382A}.ReleaseWithoutCrtdll.Build.0 = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.Debug.ActiveCfg = Debug|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.Debug.Build.0 = Debug|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.Release.ActiveCfg = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.Release.Build.0 = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseAxp.ActiveCfg = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseAxp.Build.0 = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseWithoutAsm.ActiveCfg = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseWithoutAsm.Build.0 = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseWithoutCrtdll.ActiveCfg = Release|Win32
{AA6666AA-E09F-4135-9C0C-4FE50C3C654C}.ReleaseWithoutCrtdll.Build.0 = Release|Win32
EndGlobalSection
GlobalSection(ExtensibilityGlobals) = postSolution
EndGlobalSection
......
zlib/contrib/vstudio/vc7/zlibvc.vcproj
......@@ -20,6 +20,7 @@
<Tool
Name="VCCLCompilerTool"
Optimization="0"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32,ZLIB_WINAPI,ASMV,ASMINF"
ExceptionHandling="FALSE"
RuntimeLibrary="1"
......@@ -35,7 +36,7 @@
<Tool
Name="VCLinkerTool"
AdditionalOptions="/MACHINE:I386"
AdditionalDependencies="gvmat32.obj inffas32.obj"
AdditionalDependencies="..\..\masmx86\gvmat32.obj ..\..\masmx86\inffas32.obj"
OutputFile=".\DebugDll\zlibwapi.dll"
LinkIncremental="2"
SuppressStartupBanner="TRUE"
......@@ -72,10 +73,12 @@
IntermediateDirectory=".\zlibDllWithoutAsm"
ConfigurationType="2"
UseOfMFC="0"
ATLMinimizesCRunTimeLibraryUsage="FALSE">
ATLMinimizesCRunTimeLibraryUsage="FALSE"
WholeProgramOptimization="TRUE">
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32,ZLIB_WINAPI"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -134,10 +137,12 @@
IntermediateDirectory=".\zlibDllWithoutCrtDll"
ConfigurationType="2"
UseOfMFC="0"
ATLMinimizesCRunTimeLibraryUsage="FALSE">
ATLMinimizesCRunTimeLibraryUsage="FALSE"
WholeProgramOptimization="TRUE">
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32,ZLIB_WINAPI,ASMV,ASMINF"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -156,7 +161,7 @@
<Tool
Name="VCLinkerTool"
AdditionalOptions="/MACHINE:I386"
AdditionalDependencies="gvmat32.obj inffas32.obj "
AdditionalDependencies="..\..\masmx86\gvmat32.obj ..\..\masmx86\inffas32.obj "
OutputFile=".\zlibDllWithoutCrtDll\zlibwapi.dll"
LinkIncremental="1"
SuppressStartupBanner="TRUE"
......@@ -196,10 +201,12 @@
IntermediateDirectory=".\zlibvc__"
ConfigurationType="2"
UseOfMFC="0"
ATLMinimizesCRunTimeLibraryUsage="FALSE">
ATLMinimizesCRunTimeLibraryUsage="FALSE"
WholeProgramOptimization="TRUE">
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32,ZLIB_WINAPI"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -256,10 +263,12 @@
IntermediateDirectory=".\ReleaseDll"
ConfigurationType="2"
UseOfMFC="0"
ATLMinimizesCRunTimeLibraryUsage="FALSE">
ATLMinimizesCRunTimeLibraryUsage="FALSE"
WholeProgramOptimization="TRUE">
<Tool
Name="VCCLCompilerTool"
InlineFunctionExpansion="1"
AdditionalIncludeDirectories="..\..\..;..\..\masmx86"
PreprocessorDefinitions="WIN32,ZLIB_WINAPI,ASMV,ASMINF"
StringPooling="TRUE"
ExceptionHandling="FALSE"
......@@ -278,7 +287,7 @@
<Tool
Name="VCLinkerTool"
AdditionalOptions="/MACHINE:I386"
AdditionalDependencies="gvmat32.obj inffas32.obj crtdll.lib"
AdditionalDependencies="..\..\masmx86\gvmat32.obj ..\..\masmx86\inffas32.obj crtdll.lib"
OutputFile=".\ReleaseDll\zlibwapi.dll"
LinkIncremental="1"
SuppressStartupBanner="TRUE"
......@@ -318,19 +327,19 @@
Name="Source Files"
Filter="cpp;c;cxx;rc;def;r;odl;hpj;bat;for;f90">
<File
RelativePath=".\adler32.c">
RelativePath="..\..\..\adler32.c">
</File>
<File
RelativePath=".\compress.c">
RelativePath="..\..\..\compress.c">
</File>
<File
RelativePath=".\crc32.c">
RelativePath="..\..\..\crc32.c">
</File>
<File
RelativePath=".\deflate.c">
RelativePath="..\..\..\deflate.c">
</File>
<File
RelativePath=".\gvmat32c.c">
RelativePath="..\..\masmx86\gvmat32c.c">
<FileConfiguration
Name="ReleaseWithoutAsm|Win32"
ExcludedFromBuild="TRUE">
......@@ -339,34 +348,34 @@
</FileConfiguration>
</File>
<File
RelativePath=".\gzio.c">
RelativePath="..\..\..\gzio.c">
</File>
<File
RelativePath=".\infback.c">
RelativePath="..\..\..\infback.c">
</File>
<File
RelativePath=".\inffast.c">
RelativePath="..\..\..\inffast.c">
</File>
<File
RelativePath=".\inflate.c">
RelativePath="..\..\..\inflate.c">
</File>
<File
RelativePath=".\inftrees.c">
RelativePath="..\..\..\inftrees.c">
</File>
<File
RelativePath=".\ioapi.c">
RelativePath="..\..\minizip\ioapi.c">
</File>
<File
RelativePath=".\iowin32.c">
RelativePath="..\..\minizip\iowin32.c">
</File>
<File
RelativePath=".\trees.c">
RelativePath="..\..\..\trees.c">
</File>
<File
RelativePath=".\uncompr.c">
RelativePath="..\..\..\uncompr.c">
</File>
<File
RelativePath=".\unzip.c">
RelativePath="..\..\minizip\unzip.c">
<FileConfiguration
Name="Release|Win32">
<Tool
......@@ -376,7 +385,7 @@
</FileConfiguration>
</File>
<File
RelativePath=".\zip.c">
RelativePath="..\..\minizip\zip.c">
<FileConfiguration
Name="Release|Win32">
<Tool
......@@ -392,38 +401,38 @@
RelativePath=".\zlibvc.def">
</File>
<File
RelativePath=".\zutil.c">
RelativePath="..\..\..\zutil.c">
</File>
</Filter>
<Filter
Name="Header Files"
Filter="h;hpp;hxx;hm;inl;fi;fd">
<File
RelativePath=".\deflate.h">
RelativePath="..\..\..\deflate.h">
</File>
<File
RelativePath=".\infblock.h">
RelativePath="..\..\..\infblock.h">
</File>
<File
RelativePath=".\infcodes.h">
RelativePath="..\..\..\infcodes.h">
</File>
<File
RelativePath=".\inffast.h">
RelativePath="..\..\..\inffast.h">
</File>
<File
RelativePath=".\inftrees.h">
RelativePath="..\..\..\inftrees.h">
</File>
<File
RelativePath=".\infutil.h">
RelativePath="..\..\..\infutil.h">
</File>
<File
RelativePath=".\zconf.h">
RelativePath="..\..\..\zconf.h">
</File>
<File
RelativePath=".\zlib.h">
RelativePath="..\..\..\zlib.h">
</File>
<File
RelativePath=".\zutil.h">
RelativePath="..\..\..\zutil.h">
</File>
</Filter>
<Filter
......
zlib/old/Makefile.riscos
......@@ -3,12 +3,12 @@
# test works out-of-the-box, installs `somewhere' on demand
# Toolflags:
CCflags = -c -depend !Depend -IC: -g -throwback -DRISCOS -fah
CCflags = -c -depend !Depend -IC: -g -throwback -DRISCOS -fah
C++flags = -c -depend !Depend -IC: -throwback
Linkflags = -aif -c++ -o $@
Linkflags = -aif -c++ -o $@
ObjAsmflags = -throwback -NoCache -depend !Depend
CMHGflags =
LibFileflags = -c -l -o $@
CMHGflags =
LibFileflags = -c -l -o $@
Squeezeflags = -o $@
# change the line below to where _you_ want the library installed.
......@@ -17,10 +17,10 @@ libdest = lib:zlib
# Final targets:
@.lib: @.o.adler32 @.o.compress @.o.crc32 @.o.deflate @.o.gzio \
@.o.infblock @.o.infcodes @.o.inffast @.o.inflate @.o.inftrees @.o.infutil @.o.trees \
@.o.uncompr @.o.zutil
@.o.uncompr @.o.zutil
LibFile $(LibFileflags) @.o.adler32 @.o.compress @.o.crc32 @.o.deflate \
@.o.gzio @.o.infblock @.o.infcodes @.o.inffast @.o.inflate @.o.inftrees @.o.infutil \
@.o.trees @.o.uncompr @.o.zutil
@.o.trees @.o.uncompr @.o.zutil
test: @.minigzip @.example @.lib
@copy @.lib @.libc A~C~DF~L~N~P~Q~RS~TV
@echo running tests: hang on.
......@@ -41,9 +41,9 @@ test: @.minigzip @.example @.lib
@/@.example @.fred @.fred
@echo that will have given lots of hello!'s.
@.minigzip: @.o.minigzip @.lib C:o.Stubs
Link $(Linkflags) @.o.minigzip @.lib C:o.Stubs
@.example: @.o.example @.lib C:o.Stubs
@.minigzip: @.o.minigzip @.lib C:o.Stubs
Link $(Linkflags) @.o.minigzip @.lib C:o.Stubs
@.example: @.o.example @.lib C:o.Stubs
Link $(Linkflags) @.o.example @.lib C:o.Stubs
install: @.lib
......
zlib/old/README
This directory contains files that have not been updated for zlib 1.2.1
This directory contains files that have not been updated for zlib 1.2.x
(Volunteers are encouraged to help clean this up. Thanks.)
zlib/qnx/package.qpg
......@@ -25,10 +25,10 @@
<QPG:Files>
<QPG:Add file="../zconf.h" install="/opt/include/" user="root:sys" permission="644"/>
<QPG:Add file="../zlib.h" install="/opt/include/" user="root:sys" permission="644"/>
<QPG:Add file="../libz.so.1.2.1" install="/opt/lib/" user="root:bin" permission="644"/>
<QPG:Add file="libz.so" install="/opt/lib/" component="dev" filetype="symlink" linkto="libz.so.1.2.1"/>
<QPG:Add file="libz.so.1" install="/opt/lib/" filetype="symlink" linkto="libz.so.1.2.1"/>
<QPG:Add file="../libz.so.1.2.1" install="/opt/lib/" component="slib"/>
<QPG:Add file="../libz.so.1.2.3" install="/opt/lib/" user="root:bin" permission="644"/>
<QPG:Add file="libz.so" install="/opt/lib/" component="dev" filetype="symlink" linkto="libz.so.1.2.3"/>
<QPG:Add file="libz.so.1" install="/opt/lib/" filetype="symlink" linkto="libz.so.1.2.3"/>
<QPG:Add file="../libz.so.1.2.3" install="/opt/lib/" component="slib"/>
</QPG:Files>
<QPG:PackageFilter>
......@@ -63,7 +63,7 @@
</QPM:ProductDescription>
<QPM:ReleaseDescription>
<QPM:ReleaseVersion>1.2.1</QPM:ReleaseVersion>
<QPM:ReleaseVersion>1.2.3</QPM:ReleaseVersion>
<QPM:ReleaseUrgency>Medium</QPM:ReleaseUrgency>
<QPM:ReleaseStability>Stable</QPM:ReleaseStability>
<QPM:ReleaseNoteMinor></QPM:ReleaseNoteMinor>
......@@ -105,7 +105,7 @@
</QPM:Script>
</QPM:ProductInstallationProcedure>
</QPM:PackageManifest>
<QPM:Launch>
</QPM:Launch>
</QPG:PackageFilter>
......@@ -119,7 +119,7 @@
</QPM:OrderDependency>
</QPM:ProductInstallationProcedure>
</QPM:PackageManifest>
<QPM:Launch>
</QPM:Launch>
</QPG:PackageFilter>
......@@ -133,7 +133,7 @@
</QPM:OrderDependency>
</QPM:ProductInstallationProcedure>
</QPM:PackageManifest>
<QPM:Launch>
</QPM:Launch>
</QPG:PackageFilter>
......
zlib/win32/DLL_FAQ.txt
......@@ -12,7 +12,7 @@ in the zlib distribution, or at the following location:
1. What is ZLIB1.DLL, and how can I get it?
- ZLIB1.DLL is the official build of zlib as a DLL.
(Please remark the symbol '1' in the name.)
(Please remark the character '1' in the name.)
Pointers to a precompiled ZLIB1.DLL can be found in the zlib
web site at:
......@@ -37,20 +37,19 @@ in the zlib distribution, or at the following location:
and build settings. If you do build the DLL yourself, please
make sure that it complies with all the above requirements,
and it runs with the precompiled test programs, bundled with
the original ZLIB1.DLL distribution and available at the zlib
web site.
the original ZLIB1.DLL distribution.
If, for any reason, you need to build an incompatible DLL,
please use a different name.
please use a different file name.
2. Why did you change the name of the DLL to ZLIB1.DLL?
What happened to the old ZLIB.DLL?
- The old ZLIB.DLL, built from zlib-1.1.x and earlier, required
compilation settings that were incompatible to those used by a
static build. The DLL settings were supposed to be enabled by
defining the macro ZLIB_DLL, before including "zlib.h".
- The old ZLIB.DLL, built from zlib-1.1.4 or earlier, required
compilation settings that were incompatible to those used by
a static build. The DLL settings were supposed to be enabled
by defining the macro ZLIB_DLL, before including "zlib.h".
Incorrect handling of this macro was silently accepted at
build time, resulting in two major problems:
......@@ -65,8 +64,8 @@ in the zlib distribution, or at the following location:
functions. Failure to do so resulted in creating binaries
that were unable to run with the official ZLIB.DLL build.
The only possible solution that we could foresee was to make a
binary-incompatible change in the DLL interfacing, in order to
The only possible solution that we could foresee was to make
a binary-incompatible change in the DLL interface, in order to
remove the dependency on the ZLIB_DLL macro, and to release
the new DLL under a different name.
......@@ -85,17 +84,13 @@ in the zlib distribution, or at the following location:
- In principle, you can do it by assigning calling convention
keywords to the macros ZEXPORT and ZEXPORTVA. In practice,
it depends on what you mean by "an old ZLIB.DLL", because
the old DLL exists in several mutually-incompatible versions.
If you have a compiled application that works with a certain
ZLIB.DLL without any known security issues, there is hardly
a need to rebuild the DLL from new sources only to link it to
the old app binary. But if you really want to do it, you have
to find out first what kind of calling convention uses your
particular ZLIB.DLL build, and to use the same one in the new
build. If you don't know what this is all about, you might be
better off if you would just forget it.
it depends on what you mean by "an old ZLIB.DLL", because the
old DLL exists in several mutually-incompatible versions.
You have to find out first what kind of calling convention is
being used in your particular ZLIB.DLL build, and to use the
same one in the new build. If you don't know what this is all
about, you might be better off if you would just leave the old
DLL intact.
4. Can I compile my application using the new zlib interface, and
......@@ -170,19 +165,19 @@ in the zlib distribution, or at the following location:
the K&R-style function prototypes, where the argument types
are not specified; but that is another story for another day.
The fact that remains is that CDECL is the default convention.
Even if an explicit convention (such as STDCALL or FASTCALL)
is hard-coded into the function prototypes inside C headers,
problems may appear. One problem, for example, deals with the
necessity to expose the convention in users' callbacks.
The remaining fact is that CDECL is the default convention.
Even if an explicit convention is hard-coded into the function
prototypes inside C headers, problems may appear. The
necessity to expose the convention in users' callbacks is one
of these problems.
The calling convention issues are also important when using
zlib in other programming languages. Some of them, like Ada
(GNAT) and Fortran (GNU G77), have C bindings implemented
initially on Unix, and relying on the C calling convention.
On the other hand, the pre- .NET versions of Microsoft Visual
Basic require STDCALL, while Borland Delphi prefers (although
it does not require) FASTCALL.
Basic require STDCALL, while Borland Delphi prefers, although
it does not require, FASTCALL.
In fairness to all possible uses of zlib outside the C
programming language, we choose the default "C" convention.
......@@ -208,7 +203,14 @@ in the zlib distribution, or at the following location:
zlib distribution.
8. If my application uses ZLIB1.DLL, should I link it to
8. I need to use zlib in my Microsoft .NET project. What can I
do?
- Henrik Ravn has contributed a .NET wrapper around zlib. Look
into contrib/dotzlib/, inside the zlib distribution.
9. If my application uses ZLIB1.DLL, should I link it to
MSVCRT.DLL? Why?
- It is not required, but it is recommended to link your
......@@ -223,8 +225,8 @@ in the zlib distribution, or at the following location:
depend on it should also be linked to MSVCRT.DLL.
9. Why are you saying that ZLIB1.DLL and my application must be
linked to the same C run-time (CRT) library? I linked my
10. Why are you saying that ZLIB1.DLL and my application should
be linked to the same C run-time (CRT) library? I linked my
application and my DLLs to different C libraries (e.g. my
application to a static library, and my DLLs to MSVCRT.DLL),
and everything works fine.
......@@ -255,11 +257,11 @@ in the zlib distribution, or at the following location:
and DLLs are avoiding the corruption of each of the CRTs'
internal states, maybe by careful design, or maybe by fortune.
Also note that linking ZLIB1.DLL to non-Microsoft CRTs (such
as those provided by Borland) raises similar problems.
Also note that linking ZLIB1.DLL to non-Microsoft CRTs, such
as those provided by Borland, raises similar problems.
10. Why are you linking ZLIB1.DLL to MSVCRT.DLL?
11. Why are you linking ZLIB1.DLL to MSVCRT.DLL?
- MSVCRT.DLL exists on every Windows 95 with a new service pack
installed, or with Microsoft Internet Explorer 4 or later, and
......@@ -269,21 +271,14 @@ in the zlib distribution, or at the following location:
software provider for free.
The fact that MSVCRT.DLL does not exist on a virgin Windows 95
is not so problematic. The number of Windows 95 installations
is rapidly decreasing, Microsoft stopped supporting it a long
time ago, and many recent applications from various vendors,
including Microsoft, do not even run on it. Furthermore, no
serious user should run Windows 95 without a proper update
installed.
There is also the fact that the mainstream C compilers for
Windows are Microsoft Visual C++ 6.0, and gcc/MinGW. Both
are producing executables that link to MSVCRT.DLL by default,
without offering other dynamic CRTs as alternatives easy to
select by users.
is not so problematic. Windows 95 is scarcely found nowadays,
Microsoft ended its support a long time ago, and many recent
applications from various vendors, including Microsoft, do not
even run on it. Furthermore, no serious user should run
Windows 95 without a proper update installed.
11. Why are you not linking ZLIB1.DLL to
12. Why are you not linking ZLIB1.DLL to
<<my favorite C run-time library>> ?
- We considered and abandoned the following alternatives:
......@@ -294,27 +289,60 @@ in the zlib distribution, or at the following location:
to a static C library, you may as well consider linking zlib
in statically, too.
* Linking ZLIB1.DLL to CRTDLL.DLL looks very appealing,
because CRTDLL.DLL is present on every Win32 installation.
Unfortunately, it has a series of problems: it raises
difficulties when using it with C++ code, it does not work
with 64-bit file offsets, (and so on...), and Microsoft
discontinued its support a long time ago.
* Linking ZLIB1.DLL to MSVCR70.DLL, supplied with the
Microsoft .NET platform and Visual C++ 7.0 or newer, is not
a good option. Although it is available for free download
and distribution, its presence is scarce on today's Win32
installations. If it will ever become more popular than
MSVCRT.DLL and will be pre-installed on the future Win32
systems, we will probably think again about it.
* Linking ZLIB1.DLL to NTDLL.DLL is not possible.
NTDLL.DLL exports only a part of the C library, and only on
Windows NT systems.
12. I need to link my own DLL build to a CRT different than
* Linking ZLIB1.DLL to CRTDLL.DLL looks appealing, because
CRTDLL.DLL is present on every Win32 installation.
Unfortunately, it has a series of problems: it does not
work properly with Microsoft's C++ libraries, it does not
provide support for 64-bit file offsets, (and so on...),
and Microsoft discontinued its support a long time ago.
* Linking ZLIB1.DLL to MSVCR70.DLL or MSVCR71.DLL, supplied
with the Microsoft .NET platform, and Visual C++ 7.0/7.1,
raises problems related to the status of ZLIB1.DLL as a
system component. According to the Microsoft Knowledge Base
article KB326922 "INFO: Redistribution of the Shared C
Runtime Component in Visual C++ .NET", MSVCR70.DLL and
MSVCR71.DLL are not supposed to function as system DLLs,
because they may clash with MSVCRT.DLL. Instead, the
application's installer is supposed to put these DLLs
(if needed) in the application's private directory.
If ZLIB1.DLL depends on a non-system runtime, it cannot
function as a redistributable system component.
* Linking ZLIB1.DLL to non-Microsoft runtimes, such as
Borland's, or Cygwin's, raises problems related to the
reliable presence of these runtimes on Win32 systems.
It's easier to let the DLL build of zlib up to the people
who distribute these runtimes, and who may proceed as
explained in the answer to Question 14.
13. If ZLIB1.DLL cannot be linked to MSVCR70.DLL or MSVCR71.DLL,
how can I build/use ZLIB1.DLL in Microsoft Visual C++ 7.0
(Visual Studio .NET) or newer?
- Due to the problems explained in the Microsoft Knowledge Base
article KB326922 (see the previous answer), the C runtime that
comes with the VC7 environment is no longer considered a
system component. That is, it should not be assumed that this
runtime exists, or may be installed in a system directory.
Since ZLIB1.DLL is supposed to be a system component, it may
not depend on a non-system component.
In order to link ZLIB1.DLL and your application to MSVCRT.DLL
in VC7, you need the library of Visual C++ 6.0 or older. If
you don't have this library at hand, it's probably best not to
use ZLIB1.DLL.
We are hoping that, in the future, Microsoft will provide a
way to build applications linked to a proper system runtime,
from the Visual C++ environment. Until then, you have a
couple of alternatives, such as linking zlib in statically.
If your application requires dynamic linking, you may proceed
as explained in the answer to Question 14.
14. I need to link my own DLL build to a CRT different than
MSVCRT.DLL. What can I do?
- Feel free to rebuild the DLL from the zlib sources, and link
......@@ -330,7 +358,7 @@ in the zlib distribution, or at the following location:
CYGWIN1.DLL, and it is distributed under the name CYGZ.DLL.
13. May I include additional pieces of code that I find useful,
15. May I include additional pieces of code that I find useful,
link them in ZLIB1.DLL, and export them?
- No. A legitimate build of ZLIB1.DLL must not include code
......@@ -338,14 +366,12 @@ in the zlib distribution, or at the following location:
But you can make your own private DLL build, under a different
file name, as suggested in the previous answer.
For example, in Borland Delphi and C++ Builder, zlib is a part
of the standard VCL library. If an application links to VCL
dynamically, the name of the distributable binary (VCLxx.DLL)
does not posess any danger of clashing with a legitimate but
incompatible ZLIB1.DLL.
For example, zlib is a part of the VCL library, distributed
with Borland Delphi and C++ Builder. The DLL build of VCL
is a redistributable file, named VCLxx.DLL.
14. May I remove some functionality out of ZLIB1.DLL, by enabling
16. May I remove some functionality out of ZLIB1.DLL, by enabling
macros like NO_GZCOMPRESS or NO_GZIP at compile time?
- No. A legitimate build of ZLIB1.DLL must provide the complete
......@@ -354,7 +380,7 @@ in the zlib distribution, or at the following location:
different file name, as suggested in the previous answer.
15. I made my own ZLIB1.DLL build. Can I test it for compliance?
17. I made my own ZLIB1.DLL build. Can I test it for compliance?
- We prefer that you download the official DLL from the zlib
web site. If you need something peculiar from this DLL, you
......
zlib/win32/zlib1.rc
......@@ -5,8 +5,8 @@ VS_VERSION_INFO VERSIONINFO
#else
VS_VERSION_INFO VERSIONINFO MOVEABLE IMPURE LOADONCALL DISCARDABLE
#endif
FILEVERSION 1,2,1,0
PRODUCTVERSION 1,2,1,0
FILEVERSION 1,2,2,0
PRODUCTVERSION 1,2,2,0
FILEFLAGSMASK VS_FFI_FILEFLAGSMASK
#ifdef _DEBUG
FILEFLAGS 1
......@@ -23,12 +23,12 @@ BEGIN
//language ID = U.S. English, char set = Windows, Multilingual
BEGIN
VALUE "FileDescription", "zlib data compression library\0"
VALUE "FileVersion", "1.2.1\0"
VALUE "FileVersion", "1.2.3\0"
VALUE "InternalName", "zlib1.dll\0"
VALUE "LegalCopyright", "(C) 1995-2003 Jean-loup Gailly & Mark Adler\0"
VALUE "LegalCopyright", "(C) 1995-2004 Jean-loup Gailly & Mark Adler\0"
VALUE "OriginalFilename", "zlib1.dll\0"
VALUE "ProductName", "zlib\0"
VALUE "ProductVersion", "1.2.1\0"
VALUE "ProductVersion", "1.2.3\0"
VALUE "Comments","DLL support by Alessandro Iacopetti & Gilles Vollant\0"
END
END
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment