gosu 0.7.10.3 → 0.7.11

data/windows/zlib, libpng/zlib/contrib/masmx64/inffas8664.c

@@ -0,0 +1,186 @@
+/* inffas8664.c is a hand tuned assembler version of inffast.c - fast decoding
+ * version for AMD64 on Windows using Microsoft C compiler
+ *
+ * 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.
+ *
+ * 2005 - Adaptation to Microsoft C Compiler for AMD64 by Gilles Vollant
+ *
+ * inffas8664.c call function inffas8664fnc in inffasx64.asm
+ *  inffasx64.asm is automatically convert from AMD64 portion of inffas86.c
+ *
+ * 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,
+ * 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/
+ *
+ */
+
+#include <stdio.h>
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+/* Mark Adler's comments from inffast.c: */
+
+/*
+   Decode literal, length, and distance codes and write out the resulting
+   literal and match bytes until either not enough input or output is
+   available, an end-of-block is encountered, or a data error is encountered.
+   When large enough input and output buffers are supplied to inflate(), for
+   example, a 16K input buffer and a 64K output buffer, more than 95% of the
+   inflate execution time is spent in this routine.
+
+   Entry assumptions:
+
+        state->mode == LEN
+        strm->avail_in >= 6
+        strm->avail_out >= 258
+        start >= strm->avail_out
+        state->bits < 8
+
+   On return, state->mode is one of:
+
+        LEN -- ran out of enough output space or enough available input
+        TYPE -- reached end of block code, inflate() to interpret next block
+        BAD -- error in block data
+
+   Notes:
+
+    - The maximum input bits used by a length/distance pair is 15 bits for the
+      length code, 5 bits for the length extra, 15 bits for the distance code,
+      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
+      Therefore if strm->avail_in >= 6, then there is enough input to avoid
+      checking for available input while decoding.
+
+    - The maximum bytes that a single length/distance pair can output is 258
+      bytes, which is the maximum length that can be coded.  inflate_fast()
+      requires strm->avail_out >= 258 for each loop to avoid checking for
+      output space.
+ */
+
+
+
+    typedef struct inffast_ar {
+/* 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 */ size_t /*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*/
+    } type_ar;
+#ifdef ASMINF
+
+void inflate_fast(strm, start)
+z_streamp strm;
+unsigned start;         /* inflate()'s starting value for strm->avail_out */
+{
+    struct inflate_state FAR *state;
+    type_ar ar;
+    void inffas8664fnc(struct inffast_ar * par);
+
+    
+
+#if (defined( __GNUC__ ) && defined( __amd64__ ) && ! defined( __i386 )) || (defined(_MSC_VER) && defined(_M_AMD64))
+#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 - PAD_AVAIL_IN);
+    ar.out = strm->next_out;
+    ar.beg = ar.out - (start - strm->avail_out);
+    ar.end = ar.out + (strm->avail_out - PAD_AVAIL_OUT);
+    ar.wsize = state->wsize;
+    ar.write = state->write;
+    ar.window = state->window;
+    ar.hold = state->hold;
+    ar.bits = state->bits;
+    ar.lcode = state->lencode;
+    ar.dcode = state->distcode;
+    ar.lmask = (1U << state->lenbits) - 1;
+    ar.dmask = (1U << state->distbits) - 1;
+
+    /* decode literals and length/distances until end-of-block or not enough
+       input data or output space */
+
+    /* align in on 1/2 hold size boundary */
+    while (((size_t)(void *)ar.in & (sizeof(ar.hold) / 2 - 1)) != 0) {
+        ar.hold += (unsigned long)*ar.in++ << ar.bits;
+        ar.bits += 8;
+    }
+
+    inffas8664fnc(&ar);
+
+    if (ar.status > 1) {
+        if (ar.status == 2)
+            strm->msg = "invalid literal/length code";
+        else if (ar.status == 3)
+            strm->msg = "invalid distance code";
+        else
+            strm->msg = "invalid distance too far back";
+        state->mode = BAD;
+    }
+    else if ( ar.status == 1 ) {
+        state->mode = TYPE;
+    }
+
+    /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
+    ar.len = ar.bits >> 3;
+    ar.in -= ar.len;
+    ar.bits -= ar.len << 3;
+    ar.hold &= (1U << ar.bits) - 1;
+
+    /* update state and return */
+    strm->next_in = ar.in;
+    strm->next_out = ar.out;
+    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 = (unsigned long)ar.hold;
+    state->bits = ar.bits;
+    return;
+}
+
+#endif

data/windows/zlib, libpng/zlib/contrib/masmx64/inffasx64.asm

@@ -0,0 +1,392 @@
+; inffasx64.asm is a hand tuned assembler version of inffast.c - fast decoding
+; version for AMD64 on Windows using Microsoft C compiler
+;
+; inffasx64.asm is automatically convert from AMD64 portion of inffas86.c
+; inffasx64.asm is called by inffas8664.c, which contain more info.
+
+
+; to compile this file, I use option
+;   ml64.exe /Flinffasx64 /c /Zi inffasx64.asm
+;   with Microsoft Macro Assembler (x64) for AMD64
+;
+;   ml64.exe is given with Visual Studio 2005, Windows 2003 server DDK
+;
+;   (you can get Windows 2003 server DDK with ml64 and cl.exe for AMD64 from 
+;      http://www.microsoft.com/whdc/devtools/ddk/default.mspx for low price)
+;
+
+.code
+inffas8664fnc PROC
+
+; see http://weblogs.asp.net/oldnewthing/archive/2004/01/14/58579.aspx and
+; http://msdn.microsoft.com/library/en-us/kmarch/hh/kmarch/64bitAMD_8e951dd2-ee77-4728-8702-55ce4b5dd24a.xml.asp
+;
+; All registers must be preserved across the call, except for
+;   rax, rcx, rdx, r8, r-9, r10, and r11, which are scratch.
+
+
+	mov [rsp-8],rsi
+	mov [rsp-16],rdi
+	mov [rsp-24],r12
+	mov [rsp-32],r13
+	mov [rsp-40],r14
+	mov [rsp-48],r15
+	mov [rsp-56],rbx
+
+	mov rax,rcx
+
+	mov	[rax+8], rbp       ; /* save regs rbp and rsp */
+	mov	[rax], rsp
+
+	mov	rsp, rax          ; /* make rsp point to &ar */
+
+	mov	rsi, [rsp+16]      ; /* rsi  = in */
+	mov	rdi, [rsp+32]      ; /* rdi  = out */
+	mov	r9, [rsp+24]       ; /* r9   = last */
+	mov	r10, [rsp+48]      ; /* r10  = end */
+	mov	rbp, [rsp+64]      ; /* rbp  = lcode */
+	mov	r11, [rsp+72]      ; /* r11  = dcode */
+	mov	rdx, [rsp+80]      ; /* rdx  = hold */
+	mov	ebx, [rsp+88]      ; /* ebx  = bits */
+	mov	r12d, [rsp+100]    ; /* r12d = lmask */
+	mov	r13d, [rsp+104]    ; /* r13d = dmask */
+                                          ; /* r14d = len */
+                                          ; /* r15d = dist */
+
+
+	cld
+	cmp	r10, rdi
+	je	L_one_time           ; /* if only one decode left */
+	cmp	r9, rsi
+
+    jne L_do_loop
+
+
+L_one_time:
+	mov	r8, r12           ; /* r8 = lmask */
+	cmp	bl, 32
+	ja	L_get_length_code_one_time
+
+	lodsd                         ; /* eax = *(uint *)in++ */
+	mov	cl, bl            ; /* cl = bits, needs it for shifting */
+	add	bl, 32             ; /* bits += 32 */
+	shl	rax, cl
+	or	rdx, rax          ; /* hold |= *((uint *)in)++ << bits */
+	jmp	L_get_length_code_one_time
+
+ALIGN 4
+L_while_test:
+	cmp	r10, rdi
+	jbe	L_break_loop
+	cmp	r9, rsi
+	jbe	L_break_loop
+
+L_do_loop:
+	mov	r8, r12           ; /* r8 = lmask */
+	cmp	bl, 32
+	ja	L_get_length_code    ; /* if (32 < bits) */
+
+	lodsd                         ; /* eax = *(uint *)in++ */
+	mov	cl, bl            ; /* cl = bits, needs it for shifting */
+	add	bl, 32             ; /* bits += 32 */
+	shl	rax, cl
+	or	rdx, rax          ; /* hold |= *((uint *)in)++ << bits */
+
+L_get_length_code:
+	and	r8, rdx            ; /* r8 &= hold */
+	mov	eax, [rbp+r8*4]  ; /* eax = lcode[hold & lmask] */
+
+	mov	cl, ah            ; /* cl = this.bits */
+	sub	bl, ah            ; /* bits -= this.bits */
+	shr	rdx, cl           ; /* hold >>= this.bits */
+
+	test	al, al
+	jnz	L_test_for_length_base ; /* if (op != 0) 45.7% */
+
+	mov	r8, r12            ; /* r8 = lmask */
+	shr	eax, 16            ; /* output this.val char */
+	stosb
+
+L_get_length_code_one_time:
+	and	r8, rdx            ; /* r8 &= hold */
+	mov	eax, [rbp+r8*4] ; /* eax = lcode[hold & lmask] */
+
+L_dolen:
+	mov	cl, ah            ; /* cl = this.bits */
+	sub	bl, ah            ; /* bits -= this.bits */
+	shr	rdx, cl           ; /* hold >>= this.bits */
+
+	test	al, al
+	jnz	L_test_for_length_base ; /* if (op != 0) 45.7% */
+
+	shr	eax, 16            ; /* output this.val char */
+	stosb
+	jmp	L_while_test
+
+ALIGN 4
+L_test_for_length_base:
+	mov	r14d, eax         ; /* len = this */
+	shr	r14d, 16           ; /* len = this.val */
+	mov	cl, al
+
+	test	al, 16
+	jz	L_test_for_second_level_length ; /* if ((op & 16) == 0) 8% */
+	and	cl, 15             ; /* op &= 15 */
+	jz	L_decode_distance    ; /* if (!op) */
+
+L_add_bits_to_len:
+	sub	bl, cl
+	xor	eax, eax
+	inc	eax
+	shl	eax, cl
+	dec	eax
+	and	eax, edx          ; /* eax &= hold */
+	shr	rdx, cl
+	add	r14d, eax         ; /* len += hold & mask[op] */
+
+L_decode_distance:
+	mov	r8, r13           ; /* r8 = dmask */
+	cmp	bl, 32
+	ja	L_get_distance_code  ; /* if (32 < bits) */
+
+	lodsd                         ; /* eax = *(uint *)in++ */
+	mov	cl, bl            ; /* cl = bits, needs it for shifting */
+	add	bl, 32             ; /* bits += 32 */
+	shl	rax, cl
+	or	rdx, rax          ; /* hold |= *((uint *)in)++ << bits */
+
+L_get_distance_code:
+	and	r8, rdx           ; /* r8 &= hold */
+	mov	eax, [r11+r8*4] ; /* eax = dcode[hold & dmask] */
+
+L_dodist:
+	mov	r15d, eax         ; /* dist = this */
+	shr	r15d, 16           ; /* dist = this.val */
+	mov	cl, ah
+	sub	bl, ah            ; /* bits -= this.bits */
+	shr	rdx, cl           ; /* hold >>= this.bits */
+	mov	cl, al            ; /* cl = this.op */
+
+	test	al, 16             ; /* if ((op & 16) == 0) */
+	jz	L_test_for_second_level_dist
+	and	cl, 15             ; /* op &= 15 */
+	jz	L_check_dist_one
+
+L_add_bits_to_dist:
+	sub	bl, cl
+	xor	eax, eax
+	inc	eax
+	shl	eax, cl
+	dec	eax                 ; /* (1 << op) - 1 */
+	and	eax, edx          ; /* eax &= hold */
+	shr	rdx, cl
+	add	r15d, eax         ; /* dist += hold & ((1 << op) - 1) */
+
+L_check_window:
+	mov	r8, rsi           ; /* save in so from can use it's reg */
+	mov	rax, rdi
+	sub	rax, [rsp+40]      ; /* nbytes = out - beg */
+
+	cmp	eax, r15d
+	jb	L_clip_window        ; /* if (dist > nbytes) 4.2% */
+
+	mov	ecx, r14d         ; /* ecx = len */
+	mov	rsi, rdi
+	sub	rsi, r15          ; /* from = out - dist */
+
+	sar	ecx, 1
+	jnc	L_copy_two           ; /* if len % 2 == 0 */
+
+	rep     movsw
+	mov	al, [rsi]
+	mov	[rdi], al
+	inc	rdi
+
+	mov	rsi, r8           ; /* move in back to %rsi, toss from */
+	jmp	L_while_test
+
+L_copy_two:
+	rep     movsw
+	mov	rsi, r8           ; /* move in back to %rsi, toss from */
+	jmp	L_while_test
+
+ALIGN 4
+L_check_dist_one:
+	cmp	r15d, 1            ; /* if dist 1, is a memset */
+	jne	L_check_window
+	cmp	[rsp+40], rdi      ; /* if out == beg, outside window */
+	je	L_check_window
+
+	mov	ecx, r14d         ; /* ecx = len */
+	mov	al, [rdi-1]
+	mov	ah, al
+
+	sar	ecx, 1
+	jnc	L_set_two
+	mov	[rdi], al
+	inc	rdi
+
+L_set_two:
+	rep     stosw
+	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) */
+
+	xor	eax, eax
+	inc	eax
+	shl	eax, cl
+	dec	eax
+	and	eax, edx         ; /* eax &= hold */
+	add	eax, r14d        ; /* eax += len */
+	mov	eax, [rbp+rax*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) */
+
+	xor	eax, eax
+	inc	eax
+	shl	eax, cl
+	dec	eax
+	and	eax, edx         ; /* eax &= hold */
+	add	eax, r15d        ; /* eax += dist */
+	mov	eax, [r11+rax*4] ; /* eax = dcode[val+(hold&mask[op])]*/
+	jmp	L_dodist
+
+ALIGN 4
+L_clip_window:
+	mov	ecx, eax         ; /* ecx = nbytes */
+	mov	eax, [rsp+92]     ; /* eax = wsize, prepare for dist cmp */
+	neg	ecx                ; /* nbytes = -nbytes */
+
+	cmp	eax, r15d
+	jb	L_invalid_distance_too_far ; /* if (dist > wsize) */
+
+	add	ecx, r15d         ; /* nbytes = dist - nbytes */
+	cmp	dword ptr [rsp+96], 0
+	jne	L_wrap_around_window ; /* if (write != 0) */
+
+	mov	rsi, [rsp+56]     ; /* from  = window */
+	sub	eax, ecx         ; /* eax  -= nbytes */
+	add	rsi, rax         ; /* from += wsize - nbytes */
+
+	mov	eax, r14d        ; /* eax = len */
+	cmp	r14d, ecx
+	jbe	L_do_copy           ; /* if (nbytes >= len) */
+
+	sub	eax, ecx         ; /* eax -= nbytes */
+	rep     movsb
+	mov	rsi, rdi
+	sub	rsi, r15         ; /* from = &out[ -dist ] */
+	jmp	L_do_copy
+
+ALIGN 4
+L_wrap_around_window:
+	mov	eax, [rsp+96]     ; /* eax = write */
+	cmp	ecx, eax
+	jbe	L_contiguous_in_window ; /* if (write >= nbytes) */
+
+	mov	esi, [rsp+92]     ; /* from  = wsize */
+	add	rsi, [rsp+56]     ; /* from += window */
+	add	rsi, rax         ; /* from += write */
+	sub	rsi, rcx         ; /* from -= nbytes */
+	sub	ecx, eax         ; /* nbytes -= write */
+
+	mov	eax, r14d        ; /* eax = len */
+	cmp	eax, ecx
+	jbe	L_do_copy           ; /* if (nbytes >= len) */
+
+	sub	eax, ecx         ; /* len -= nbytes */
+	rep     movsb
+	mov	rsi, [rsp+56]     ; /* from = window */
+	mov	ecx, [rsp+96]     ; /* nbytes = write */
+	cmp	eax, ecx
+	jbe	L_do_copy           ; /* if (nbytes >= len) */
+
+	sub	eax, ecx         ; /* len -= nbytes */
+	rep     movsb
+	mov	rsi, rdi
+	sub	rsi, r15         ; /* from = out - dist */
+	jmp	L_do_copy
+
+ALIGN 4
+L_contiguous_in_window:
+	mov	rsi, [rsp+56]     ; /* rsi = window */
+	add	rsi, rax
+	sub	rsi, rcx         ; /* from += write - nbytes */
+
+	mov	eax, r14d        ; /* eax = len */
+	cmp	eax, ecx
+	jbe	L_do_copy           ; /* if (nbytes >= len) */
+
+	sub	eax, ecx         ; /* len -= nbytes */
+	rep     movsb
+	mov	rsi, rdi
+	sub	rsi, r15         ; /* from = out - dist */
+	jmp	L_do_copy           ; /* if (nbytes >= len) */
+
+ALIGN 4
+L_do_copy:
+	mov	ecx, eax         ; /* ecx = len */
+	rep     movsb
+
+	mov	rsi, r8          ; /* move in back to %esi, toss from */
+	jmp	L_while_test
+
+L_test_for_end_of_block:
+	test	al, 32
+	jz	L_invalid_literal_length_code
+	mov	dword ptr [rsp+116], 1
+	jmp	L_break_loop_with_status
+
+L_invalid_literal_length_code:
+	mov	dword ptr [rsp+116], 2
+	jmp	L_break_loop_with_status
+
+L_invalid_distance_code:
+	mov	dword ptr [rsp+116], 3
+	jmp	L_break_loop_with_status
+
+L_invalid_distance_too_far:
+	mov	dword ptr [rsp+116], 4
+	jmp	L_break_loop_with_status
+
+L_break_loop:
+	mov	dword ptr [rsp+116], 0
+
+L_break_loop_with_status:
+; /* put in, out, bits, and hold back into ar and pop esp */
+	mov	[rsp+16], rsi     ; /* in */
+	mov	[rsp+32], rdi     ; /* out */
+	mov	[rsp+88], ebx     ; /* bits */
+	mov	[rsp+80], rdx     ; /* hold */
+
+	mov	rax, [rsp]       ; /* restore rbp and rsp */
+	mov	rbp, [rsp+8]
+	mov	rsp, rax
+
+
+
+	mov rsi,[rsp-8]
+	mov rdi,[rsp-16]
+	mov r12,[rsp-24]
+	mov r13,[rsp-32]
+	mov r14,[rsp-40]
+	mov r15,[rsp-48]
+	mov rbx,[rsp-56]
+	
+    ret 0
+;          :
+;          : "m" (ar)
+;          : "memory", "%rax", "%rbx", "%rcx", "%rdx", "%rsi", "%rdi",
+;            "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15"
+;    );
+
+inffas8664fnc 	ENDP
+;_TEXT	ENDS
+END