Attachment #75916 for bug #113255

View | Details | Raw Unified | Return to bug 113255
Collapse All | Expand All





#ifdef EMU_QTX_API
#include "wrapper.h"
int report_func(void *stack_base, int stack_size, reg386_t *reg, u_int32_t *flags);
int report_func_ret(void *stack_base, int stack_size, reg386_t *reg, u_int32_t *flags);
#endif

//#undef TRACE

	    printf ("wine/module: QuickTime.qts patched!!! old entry=%p\n",ptr[0]);

#ifdef EMU_QTX_API
	    report_entry = report_func;
	    report_ret   = report_func_ret;
	    wrapper_target=ptr[0];
	    ptr[0]=wrapper;
#endif

static u_int32_t ret_array[4096];
static int ret_i=0;

int report_func(void *stack_base, int stack_size, reg386_t *reg, u_int32_t *flags)
{
#ifdef DEBUG_QTX_API
  int i;

    return 0;
}

int report_func_ret(void *stack_base, int stack_size, reg386_t *reg, u_int32_t *flags)
{
#ifdef DEBUG_QTX_API
  int i;

//      || !strcmp(function,"_CallComponent")
      ){
	fprintf(stderr,"theQuickTimeDispatcher caught -> %p\n",retproc);
      report_entry = report_func;
      report_ret   = report_func_ret;
      wrapper_target=(void *)retproc;
      retproc=(void *)wrapper;
    }

Lines 33-35 exp_EH_prolog: Link Here

(-)xine-lib-1.1.1-old/src/libw32dll/wine/stubs.s (+3 lines)
33	leal 12(%esp), %ebp	33	leal 12(%esp), %ebp
34	pushl %eax	34	pushl %eax
35	ret	35	ret
		36
		37	.section .note.GNU-stack,"",@progbits
		38




.section .data
.globl caller_return
caller_return:
	.long 0
.globl report_entry
report_entry:
	.long null_call
.globl report_ret
report_ret:	
	.long null_call
.global wrapper_target
wrapper_target:	
	.long null_call

#undef __i686 /* gcc define gets in our way */
	.section .gnu.linkonce.t.__i686.get_pc_thunk.bx,"ax",@progbits
.globl __i686.get_pc_thunk.bx
	.hidden	__i686.get_pc_thunk.bx
	.type	__i686.get_pc_thunk.bx,@function
__i686.get_pc_thunk.bx:
	movl (%esp), %ebx
	ret

.section .text
.globl null_call
	.type null_call, @function

	.type wrapper, @function
	.balign 16,0x90
wrapper:
	pushl $0
	pusha			# store registers (EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI) 
	pushf			# store flags
	
	push %ebp		# set up a stack frame 
	movl %esp, %ebp

	call __i686.get_pc_thunk.bx
	addl $_GLOBAL_OFFSET_TABLE_, %ebx

	leal 4(%ebp), %eax	# push flags addr
	push %eax
	leal 8(%ebp), %eax	# push registers addr
	push %eax
	
	leal 44(%ebp), %edx	
	movl (%ebp), %eax
	subl %edx, %eax
	push %eax
	push %edx
	
	call report_func@PLT	# report entry
	
	test %eax, %eax
	jnz .Ldone

	movl 44(%ebp), %eax	# switch return addresses
	movl %eax, caller_return@GOTOFF(%ebx)
	leal .Lwrapper_return@GOTOFF(%ebx), %eax
	movl %eax, 40(%ebp)

	movl wrapper_target@GOTOFF(%ebx), %eax
	mov %eax, 40(%ebp)	# wrapper_target should return at .Lwrapper_return

	leave			# restore %esp, %ebp
	popf			# restore flags
	popa			# restore registers

	ret			# fake 'return' to wrapper_target actually
	pushl $.Lwrapper_return 
	
	jmp *wrapper_target	# wrapper_target should return at .Lwrapper_return

	.balign 16, 0x90
.Lwrapper_return:	
	pushl $0
	pusha			# more for reference sake here
	pushf
	
	push %ebp		# set up a stack frame
	movl %esp, %ebp

	call __i686.get_pc_thunk.bx
	addl $_GLOBAL_OFFSET_TABLE_, %ebx
	movl caller_return@GOTOFF(%ebx), %eax
	movl %eax, 40(%ebp)	# restore the original return address

	leal 4(%ebp), %eax	# push flags addr
	push %eax
	leal 8(%ebp), %eax	# push registers addr

	push %eax
	push %edx

	call report_func_ret@PLT# report the return information (same args)
.Ldone:	

	leave
	popf
	popa
	ret

.section .note.GNU-stack,"",@progbits


Lines 11-20 typedef struct { Link Here

(-)xine-lib-1.1.1-old/src/libw32dll/wine/wrapper.h (-4 lines)
11	u_int32_t edi, esi, ebp, esp, ebx, edx, ecx, eax;	11	u_int32_t edi, esi, ebp, esp, ebx, edx, ecx, eax;
12	} reg386_t;	12	} reg386_t;
13		13
14	typedef int (wrapper_func_t)(void stack_base, int stack_size, reg386_t reg, u_int32_t flags);
15
16	extern wrapper_func_t report_entry, report_ret;
17
18	extern void (*wrapper_target)(void);	14	extern void (*wrapper_target)(void);
19		15
20	extern int wrapper(void);	16	extern int wrapper(void);




        * See above for a description of the algorithm.
	*/
            ".align 8 \n\t"
            "movq %4, %%mm6			\n\t"

            "movq %0, %%mm1			\n\t"     // T1
            "movq %1, %%mm0			\n\t"     // M1

            "movq %3, %%mm2			\n\t"     // M0
            : /* no output */
            : "m" (*T1), "m" (*M1), 
              "m" (*B1), "m" (*M0), "m" (Mask) );
          

          asm volatile(

#endif

            /* if |M1-M0| > Threshold we want dword worth of twos */
            "pcmpgtb %3, %%mm4			\n\t"
            "pand    %4, %%mm4			\n\t" /* get rid of sign bit */
            "pcmpgtd %5, %%mm4			\n\t" /* do we want to bob */
            "pandn   %6, %%mm4			\n\t"

            "movq    %1, %%mm2			\n\t" /* mm2 = T0 */


            "pand    %%mm6, %%mm5			\n\t"

            /* if |T1-T0| > Threshold we want dword worth of ones */
            "pcmpgtb %3, %%mm5			\n\t"
            "pand    %%mm6, %%mm5		\n\t" /* get rid of sign bit */

            "pcmpgtd %5, %%mm5			\n\t" 
            "pandn   %5, %%mm5			\n\t"
            "paddd   %%mm5, %%mm4			\n\t"

            "movq    %2, %%mm2			\n\t"     /* B0 */

            "pand    %%mm6, %%mm5			\n\t"

            /* if |B1-B0| > Threshold we want dword worth of ones */
            "pcmpgtb %3, %%mm5		\n\t"
            "pand    %%mm6, %%mm5	\n\t"     /* get rid of any sign bit */
            "pcmpgtd %5, %%mm5			\n\t"
            "pandn   %5, %%mm5			\n\t"
            "paddd   %%mm5, %%mm4			\n\t"

            "pcmpgtd %6, %%mm4			\n\t"

/* debugging feature
 * output the value of mm4 at this point which is pink where we will weave

#endif

          : "=m" (*Dest2)
          : "m" (*T0), "m" (*B0), "m" (qwGreedyTwoFrameThreshold), "m" (Mask), "m" (DwordOne), "m" (DwordTwo) );

          /* Advance to the next set of pixels. */
          T1 += 8;





    int Line;
    long LoopCtr;
    long oldbx;
    unsigned int Pitch = stride*2;
    int FieldHeight = height / 2;


    unsigned char* L3;					// ptr to Line3

    unsigned char* L2P;					// ptr to prev Line2
    unsigned char* temp;
    unsigned char* Dest = output;

    int64_t LastAvg=0;			//interp value from left qword

#define asmLastAvg      "%0"
#define asmL1           "%1"
#define asmL3           "%2"
#define asmtemp         "%3"
#define asmL2           "%4"
#define asmDest         "%5"
#define asmLoopCtr      "%6"
#define asmoldbx        "%7"
#endif

        // For ease of reading, the comments below assume that we're operating on an odd
        // field (i.e., that InfoIsOdd is true).  Assume the obvious for even lines..
        temp = L2P;
        __asm__ __volatile__
            (
             // save ebx (-fPIC)
	     MOVX" %%"XBX", "asmoldbx"\n\t"

             MOVX"  "asmL1",          %%"XAX"\n\t"
             LEAX"  8(%%"XAX"),     %%"XDX"\n\t"    // next qword needed by DJR
             MOVX"  "asmL3",          %%"XCX"\n\t"
             SUBX"  %%"XAX",        %%"XCX"\n\t"    // carry L3 addr as an offset
             MOVX"  "asmL2P",         %%"XDX"\n\t"
             MOVX"  "asmL2",          %%"XSI"\n\t"
             MOVX"  "asmDest",        %%"XDI"\n\t"    // DL1 if Odd or DL2 if Even



             "movq  (%%"XSI"),      %%mm0\n\t"      // L2 - the newest weave pixel value
             "movq  (%%"XAX"),      %%mm1\n\t"      // L1 - the top pixel
             PUSHX" %%"XDX              "\n\t"
             MOVX"  "asmtemp",    %%"XDX"\n\t"
             "movq  (%%"XDX"),      %%mm2\n\t"      // L2P - the prev weave pixel
             POPX" %%"XDX               "\n\t"
             "movq  (%%"XAX", %%"XCX"), %%mm3\n\t"  // L3, next odd row
             "movq  %%mm1,          %%mm6\n\t"      // L1 - get simple single pixel interp
             //	pavgb   mm6, mm3                    // use macro below
             V_PAVGB ("%%mm6", "%%mm3", "%%mm4", "%8")

             // DJR - Diagonal Jaggie Reduction
             // In the event that we are going to use an average (Bob) pixel we do not want a jagged

             "psllq $16,            %%mm7\n\t"      // left justify 3 pixels
             "por   %%mm7,          %%mm4\n\t"      // and combine

             "movq  (%%"XDX"),      %%mm5\n\t"      // next horiz qword from L1
             //			pavgb   mm5, qword ptr[ebx+ecx] // next horiz qword from L3, use macro below
             V_PAVGB ("%%mm5", "(%%"XDX",%%"XCX")", "%%mm7", "%8")
             "psllq $48,            %%mm5\n\t"      // left just 1 pixel
             "movq  %%mm6,          %%mm7\n\t"      // another copy of simple bob pixel
             "psrlq $16,            %%mm7\n\t"      // right just 3 pixels
             "por   %%mm7,          %%mm5\n\t"      // combine
             //			pavgb	mm4, mm5			// avg of forward and prev by 1 pixel, use macro
             V_PAVGB ("%%mm4", "%%mm5", "%%mm5", "%8")   // mm5 gets modified if MMX
             //			pavgb	mm6, mm4			// avg of center and surround interp vals, use macro
             V_PAVGB ("%%mm6", "%%mm4", "%%mm7", "%8")

             // Don't do any more averaging than needed for mmx. It hurts performance and causes rounding errors.
#ifndef IS_MMX
             //          pavgb	mm4, mm6			// 1/4 center, 3/4 adjacent
             V_PAVGB ("%%mm4", "%%mm6", "%%mm7", "%8")
             //    		pavgb	mm6, mm4			// 3/8 center, 5/8 adjacent
             V_PAVGB ("%%mm6", "%%mm4", "%%mm7", "%8")
#endif

             // get abs value of possible L2 comb

             // pminub	mm5, mm3                    // now = Min(L1,L3), use macro
             V_PMINUB ("%%mm5", "%%mm3", "%%mm7")
             // allow the value to be above the high or below the low by amt of MaxComb
             "psubusb %9,           %%mm5\n\t"      // lower min by diff
             "paddusb %9,           %%mm2\n\t"      // increase max by diff
             // pmaxub	mm4, mm5                    // now = Max(best,Min(L1,L3) use macro
             V_PMAXUB ("%%mm4", "%%mm5")
             // pminub	mm4, mm2                    // now = Min( Max(best, Min(L1,L3), L2 )=L2 clipped
             V_PMINUB ("%%mm4", "%%mm2", "%%mm7")

             // Blend weave pixel with bob pixel, depending on motion val in mm0
             "psubusb %10,          %%mm0\n\t"// test Threshold, clear chroma change >>>??
             "pmullw  %11,          %%mm0\n\t"    // mul by user factor, keep low 16 bits
             "movq    %12,          %%mm7\n\t"
#ifdef IS_SSE
             "pminsw  %%mm7,        %%mm0\n\t"      // max = 256
#else
             "paddusw %13,          %%mm0\n\t"      // add, may sat at fff..
             "psubusw %13,          %%mm0\n\t"      // now = Min(L1,256)
#endif
             "psubusw %%mm0,        %%mm7\n\t"      // so the 2 sum to 256, weighted avg
             "movq    %%mm4,        %%mm2\n\t"      // save weave chroma info before trashing
             "pand    %14,          %%mm4\n\t"      // keep only luma from calc'd value
             "pmullw  %%mm7,        %%mm4\n\t"      // use more weave for less motion
             "pand    %14,          %%mm6\n\t"      // keep only luma from calc'd value
             "pmullw  %%mm0,        %%mm6\n\t"      // use more bob for large motion
             "paddusw %%mm6,        %%mm4\n\t"      // combine
             "psrlw   $8,           %%mm4\n\t"      // div by 256 to get weighted avg

             // chroma comes from weave pixel
             "pand    %15,          %%mm2\n\t"      // keep chroma
             "por     %%mm4,        %%mm2\n\t"      // and combine

             V_MOVNTQ ("(%%"XDI")", "%%mm2")        // move in our clipped best, use macro

             // bump ptrs and loop
             LEAX"    8(%%"XAX"),   %%"XAX"\n\t"
             LEAX"    8(%%"XBX"),   %%"XBX"\n\t"
             LEAX"    8(%%"XDX"),   %%"XDX"\n\t"
             ADDX"    $8,         "asmtemp"\n\t"
             LEAX"    8(%%"XDI"),   %%"XDI"\n\t"
             LEAX"    8(%%"XSI"),   %%"XSI"\n\t"
             DECX"    "asmLoopCtr"\n\t"
             "jg      1b\n\t"                       // loop if not to last line
                                                    // note P-III default assumes backward branches taken
             "jl      1f\n\t"                       // done
             MOVX"    %%"XAX",      %%"XDX"\n\t"  // sharpness lookahead 1 byte only, be wrong on 1
             "jmp     1b\n\t"

             "1:\n\t"
	     MOVX" "asmoldbx", %%"XBX"\n\t"

             : /* no outputs */

             : "m"(LastAvg),
               "m"(L1),
               "m"(L3),
               "m"(temp),
               "m"(L2),
               "m"(Dest),
               "m"(LoopCtr),
               "m"(temp),
               "m"(ShiftMask),
               "m"(MaxComb),
               "m"(MotionThreshold),
               "m"(MotionSense),
               "m"(QW256),
               "m"(QW256B),
               "m"(YMask),
               "m"(UVMask)

             : XAX, XCX, XDX, XSI, XDI,
#ifdef ARCH_X86

Lines 7-15 Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/SearchLoop0A.inc (-2 / +2 lines)
7	// up by a little, and adjust later	7	// up by a little, and adjust later
8		8
9	#ifdef IS_SSE2	9	#ifdef IS_SSE2
10	"paddusb "MANGLE(ONES)", %%xmm7\n\t" // bias toward no motion	10	"paddusb "_ONES", %%xmm7\n\t" // bias toward no motion
11	#else	11	#else
12	"paddusb "MANGLE(ONES)", %%mm7\n\t" // bias toward no motion	12	"paddusb "_ONES", %%mm7\n\t" // bias toward no motion
13	#endif	13	#endif
14		14
15	MERGE4PIXavg("(%%"XDI", %%"XCX")", "(%%"XSI", %%"XCX")") // center, in old and new	15	MERGE4PIXavg("(%%"XDI", %%"XCX")", "(%%"XSI", %%"XCX")") // center, in old and new




            // Use the best weave if diffs less than 10 as that
            // means the image is still or moving cleanly
            // if there is motion we will clip which will catch anything
            "psubusb "_FOURS", %%mm7\n\t"          // sets bits to zero if weave diff < 4
            "pxor    %%mm0, %%mm0\n\t"
            "pcmpeqb %%mm0, %%mm7\n\t"            // all ff where weave better, else 00
            "pcmpeqb %%mm7, %%mm0\n\t"            // all ff where bob better, else 00

#else
            // Use the better of bob or weave
            //      pminub  mm4, TENS           // the most we care about
            V_PMINUB ("%%mm4", _TENS, "%%mm0")   // the most we care about
            
            "psubusb %%mm4, %%mm7\n\t"            // foregive that much from weave est?
            "psubusb "_FOURS", %%mm7\n\t"       // bias it a bit toward weave
            "pxor    %%mm0, %%mm0\n\t"
            "pcmpeqb %%mm0, %%mm7\n\t"            // all ff where weave better, else 00
            "pcmpeqb %%mm7, %%mm0\n\t"            // all ff where bob better, else 00

            
            
                //      pminub  mm0, Max_Vals       // but clip to catch the stray error
//                V_PMINUB ("%%mm0", _Max_Vals, "%%mm1") // but clip to catch the stray error
                //      pmaxub  mm0, Min_Vals
//                V_PMAXUB ("%%mm0", _Min_Vals)
                
#endif


                
#ifdef USE_VERTICAL_FILTER
            "movq    %%mm0, %%mm1\n\t"
            //      pavgb   mm0, qword ptr["XDX"]
            V_PAVGB ("%%mm0", "(%%"XDX")", "%%mm2", _ShiftMask)
            //      movntq  qword ptr["XAX"+"olddx"], mm0
            ADDX"   "_olddx", %%"XAX"\n\t"
            V_MOVNTQ ("(%%"XAX")", "%%mm0")
            //      pavgb   mm1, qword ptr["XDX"+"XCX"]
            V_PAVGB ("%%mm1", "(%%"XDX", %%"XCX")", "%%mm2", _ShiftMask)
            "addq   "_dst_pitchw", %%"XDX
            //      movntq  qword ptr["XAX"+"olddx"], mm1
            V_MOVNTQ ("(%%"XAX")", "%%mm1")
#else
                
            //      movntq  qword ptr["XAX"+"olddx"], mm0
            ADDX"   "_olddx", %%"XAX"\n\t"
            V_MOVNTQ ("(%%"XAX")", "%%mm0")
#endif
                
           ADDX"    $8, "_olddx"\n\t"       // bump offset pointer
           MOVX"    "_olddx", %%"XAX"\n\t"
           CMPX"    "_Last8", %%"XAX"\n\t"       // done with line?
           "jb      1b\n\t"                    // y
#endif

           MOVX" "_oldbx", %%"XBX"\n\t"

        : /* no outputs */

        : "m"(pBob),

          "m"(pSrc),
          "m"(pSrcP),
          "m"(pBobP),
          "m"(olddx),
          "m"(UVMask),
          "m"(ShiftMask),
          "m"(FOURS),
          "m"(TENS),
          "m"(Max_Vals),
          "m"(Min_Vals),
          "m"(YMask),
          "m"(Max_Mov),
          "m"(ONES),
          "m"(DiffThres)

        : XAX, XCX, XDX, XSI, XDI,
#ifdef ARCH_X86




#define _pSrc       "%5"
#define _pSrcP      "%6"
#define _pBobP      "%7"
#define _olddx      "%8"
#define _UVMask     "%9"
#define _ShiftMask  "%10"
#define _FOURS      "%11"
#define _TENS       "%12"
#define _Max_Vals   "%13"
#define _Min_Vals   "%14"
#define _YMask      "%15"
#define _Max_Mov    "%16"
#define _ONES       "%17"
#define _DiffThres  "%18"
#endif

	for (y=1; y < FldHeight-1; y++)

             // Loop general reg usage
             //
             // XAX - pBobP, then pDest 
             // XDX - pBob
             // XCX - src_pitch2
             // _olddx - current offset
             // XDI - prev weave pixels, 1 line up
             // XSI - next weave pixels, 1 line up

             // Save "XBX" (-fPIC)
	     MOVX" %%"XBX", "_oldbx"\n\t"
             
#ifdef IS_SSE2
             
             // sse2 code deleted for now

#else
             // simple bob first 8 bytes
             MOVX"	"_pBob",        %%"XDX"\n\t"
             MOVX"	"_src_pitch2",  %%"XCX"\n\t"

#ifdef USE_VERTICAL_FILTER
             "movq	    (%%"XDX"),        %%mm0\n\t"
             "movq	    (%%"XDX", %%"XCX"), %%mm1\n\t" //, qword ptr["XDX"+"XCX"]
             "movq	    %%mm0,          %%mm2\n\t"
             V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)		// halfway between
             V_PAVGB ("%%mm0", "%%mm2", "%%mm3", _ShiftMask)		// 1/4 way
             V_PAVGB ("%%mm1", "%%mm2", "%%mm3", _ShiftMask)		// 3/4 way
             MOVX"		"_pDest",       %%"XDI"\n\t"
             MOVX"		"_dst_pitchw",  %%"XAX"\n\t"
             V_MOVNTQ	("(%%"XDI")", "%%mm0")
             V_MOVNTQ	("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1

             // simple bob last 8 bytes
             MOVX"		"_Last8", %%"XSI"\n\t"
             ADDX"		%%"XDX", %%"XSI"\n\t"  // ["XDX"+"olddx"]
             "movq	    (%%"XSI"), %%mm0\n\t"
             "movq	    (%%"XSI", %%"XCX"), %%mm1\n\t"    // qword ptr["XSI"+"XCX"]
             "movq	    %%mm0, %%mm2\n\t"
             V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)		// halfway between
             V_PAVGB ("%%mm0", "%%mm2", "%%mm3", _ShiftMask)		// 1/4 way
             V_PAVGB ("%%mm1", "%%mm2", "%%mm3", _ShiftMask)		// 3/4 way
             ADDX"		"_olddx", %%"XDI"\n\t"						// last 8 bytes of dest
             V_MOVNTQ	("%%"XDI"", "%%mm0")
             V_MOVNTQ	("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1)

#else
             "movq	(%%"XDX"), %%mm0\n\t"
             //		pavgb	mm0, qword ptr["XDX"+"XCX"]
             V_PAVGB ("%%mm0", "(%%"XDX", %%"XCX")", "%%mm2", _ShiftMask) // qword ptr["XDX"+"XCX"], mm2, ShiftMask)
             MOVX"		"_pDest", %%"XDI"\n\t"
             V_MOVNTQ	("(%%"XDI")", "%%mm0")

             // simple bob last 8 bytes
             MOVX"		"_Last8", %%"XSI"\n\t"
             ADDX"		%%"XDX", %%"XSI"\n\t" //"XSI", ["XDX"+"olddx"]
             "movq	    (%%"XSI"), %%mm0\n\t"
             //		pavgb	mm0, qword ptr["XSI"+"XCX"]
             V_PAVGB	("%%mm0", "(%%"XSI", %%"XCX")", "%%mm2", _ShiftMask) // qword ptr["XSI"+"XCX"], mm2, ShiftMask)
             ADDX"		"_olddx", %%"XDI"\n\t"
             V_MOVNTQ	("(%%"XDI")", "%%mm0") // qword ptr["XDI"+"olddx"], mm0)
#endif
             // now loop and get the middle qwords
             MOVX"		"_pSrc", %%"XSI"\n\t"
             MOVX"		"_pSrcP", %%"XDI"\n\t"
             MOVX"		$8, "_olddx"\n\t"				// curr offset longo all lines

             "1:\n\t"	
             MOVX"		"_pBobP", %%"XAX"\n\t"
             ADDX"		$8, %%"XDI"\n\t"
             ADDX"		$8, %%"XSI"\n\t"
             ADDX"		$8, %%"XDX"\n\t"
             ADDX"		"_olddx", %%"XAX"\n\t"

#ifdef USE_STRANGE_BOB
#include "StrangeBob.inc"




        "pxor %%mm6, %%mm6\n\t"
        "pxor %%mm7, %%mm7\n\t"

		"movq    -2(%%"XDX"), %%mm0\n\t"		// value a from top left		
		"movq    -4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value m from bottom right			
        
		"movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"
		"por		%%mm1, %%mm3\n\t"					// abs(a,m)

		"psubusb "_DiffThres", %%mm3\n\t"		// nonzero where abs(a,m) > Thres else 0
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where abs(a,m) < Thres, else 00	
		"pcmpeqb	%%mm3, %%mm4\n\t"			// here ff where abs(a,m) > Thres, else 00


		"movq    -4(%%"XDX"), %%mm0\n\t"		// value j
		"movq    4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value n
		"movq	%%mm0, %%mm2\n\t"					
		"pavgb	%%mm1, %%mm2\n\t"					// avg(j,n)
        "movq	%%mm0, %%mm3\n\t"

		"por		%%mm1, %%mm0\n\t"					// abs(j,n)

        "movq    %%mm0, %%mm1\n\t"
		"psubusb "_DiffThres", %%mm1\n\t"		// nonzero where abs(j,n) > Thres else 0
		"pxor	%%mm3, %%mm3\n\t"
		"pcmpeqb %%mm3, %%mm1\n\t"			// now ff where abs(j,n) < Thres, else 00	


        "por     %%mm0, %%mm7\n\t"
        
        // k & m
		"movq    2(%%"XDX"), %%mm0\n\t"		// value c from top left		
		"movq    4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value n from bottom right			

		"movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"
		"por		%%mm1, %%mm3\n\t"					// abs(c,n)

		"psubusb "_DiffThres", %%mm3\n\t"		// nonzero where abs(c,n) > Thres else 0
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where abs(c,n) < Thres, else 00	
		"pcmpeqb	%%mm3, %%mm4\n\t"			// here ff where abs(c,n) > Thres, else 00


		"movq    4(%%"XDX"), %%mm0\n\t"		// value k
		"movq    -4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value m
		"movq	%%mm0, %%mm2\n\t"					
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(k,m)
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm0\n\t"
		"psubusb %%mm3, %%mm1\n\t"
		"por		%%mm1, %%mm0\n\t"					// abs(k,m)

        "movq    %%mm0, %%mm1\n\t"
		"psubusb "_DiffThres", %%mm1\n\t"		// nonzero where abs(k,m) > Thres else 0
		"pxor	%%mm3, %%mm3\n\t"
		"pcmpeqb %%mm3, %%mm1\n\t"			// now ff where abs(k,m) < Thres, else 00	




        // c & d
		"movq    (%%"XDX"), %%mm0\n\t"		// value b from top left		
		"movq    2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value f from bottom right			

		"movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"
		"por		%%mm1, %%mm3\n\t"					// abs(b,f)

		"psubusb "_DiffThres", %%mm3\n\t"		// nonzero where abs(b,f) > Thres else 0
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where abs(b,f) < Thres, else 00	
		"pcmpeqb	%%mm3, %%mm4\n\t"			// here ff where abs(b,f) > Thres, else 00

		"movq    2(%%"XDX"), %%mm0\n\t"		// value c
		"movq    -2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value d
		"movq	%%mm0, %%mm2\n\t"					
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(c,d)
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm0\n\t"
		"psubusb %%mm3, %%mm1\n\t"
		"por		%%mm1, %%mm0\n\t"					// abs(c,d)

        "movq    %%mm0, %%mm1\n\t"
		"psubusb "_DiffThres", %%mm1\n\t"		// nonzero where abs(c,d) > Thres else 0
		"pxor	%%mm3, %%mm3\n\t"
        "pcmpeqb %%mm3, %%mm1\n\t"			// now ff where abs(c,d) < Thres, else 00	


        "por     %%mm0, %%mm7\n\t"

        // a & f
		"movq    (%%"XDX"), %%mm0\n\t"		// value b from top left		
		"movq    -2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value d from bottom right			

		"movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"
		"por		%%mm1, %%mm3\n\t"					// abs(b,d)

		"psubusb "_DiffThres", %%mm3\n\t"	// nonzero where abs(b,d) > Thres else 0
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where abs(b,d) < Thres, else 00	
		"pcmpeqb	%%mm3, %%mm4\n\t"			// here ff where abs(b,d) > Thres, else 00

		"movq    -2(%%"XDX"), %%mm0\n\t"		// value a
		"movq    2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value f
		"movq	%%mm0, %%mm2\n\t"					
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(a,f)
        "movq	%%mm0, %%mm3\n\t"
        "psubusb	%%mm1, %%mm0\n\t"
		"psubusb %%mm3, %%mm1\n\t"
		"por		%%mm1, %%mm0\n\t"					// abs(a,f)

        "movq    %%mm0, %%mm1\n\t"
		"psubusb "_DiffThres", %%mm1\n\t"		// nonzero where abs(a,f) > Thres else 0
		"pxor	%%mm3, %%mm3\n\t"
		"pcmpeqb %%mm3, %%mm1\n\t"			// now ff where abs(a,f) < Thres, else 00	


        "por     %%mm2, %%mm6\n\t"
        "por     %%mm0, %%mm7\n\t"
           
 		"pand	"_YMask", %%mm5\n\t"		// mask out chroma from here
 		"pand	"_YMask", %%mm6\n\t"			// mask out chroma from here
 		"pand	"_YMask", %%mm7\n\t"			// mask out chroma from here

		// b,e
		"movq    (%%"XDX"), %%mm0\n\t"		// value b from top 		
		"movq    (%%"XDX", %%"XCX"), %%mm1\n\t"	// value e from bottom 
		"movq	%%mm0, %%mm2\n\t"					
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(b,e)
        "movq	%%mm0, %%mm3\n\t"
        "psubusb	%%mm1, %%mm0\n\t"
		"psubusb %%mm3, %%mm1\n\t"
		"por		%%mm1, %%mm0\n\t"					// abs(b,e)

        "movq    %%mm0, %%mm1\n\t"
		"psubusb "_DiffThres", %%mm1\n\t"		// nonzero where abs(b,e) > Thres else 0
		"pxor	%%mm3, %%mm3\n\t"
		"pcmpeqb %%mm3, %%mm1\n\t"		// now ff where abs(b,e) < Thres, else 00	


        "por     %%mm0, %%mm7\n\t"

		// bob in any leftovers
		"movq    (%%"XDX"), %%mm0\n\t"		// value b from top 		
		"movq    (%%"XDX", %%"XCX"), %%mm1\n\t"	// value e from bottom 


// We will also calc here the max/min values to later limit comb

		"por		%%mm2, %%mm3\n\t"			// abs diff
//		pmaxub  %%mm3, %%mm4			// top or bottom pixel moved most
		V_PMAXUB ("%%mm3", "%%mm4")			// top or bottom pixel moved most
		"psubusb "_DiffThres", %%mm3\n\t"		// moved more than allowed? or goes to 0?
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where low motion, else high motion
		

		V_PMAXUB ("%%mm6", "%%mm2")

        "psubusb %%mm3, %%mm2\n\t"			// maybe decrease it to 0000.. if no surround motion
//		"movq	%%mm2, "_Min_Vals"\n\t"

		"movq	%%mm0, %%mm2\n\t"
		V_PMAXUB ("%%mm2", "%%mm1")
//		pminub	%%mm6, %%mm2			// clip our current results so far to be below this
		V_PMINUB ("%%mm6", "%%mm2", "%%mm4")
        "paddusb %%mm3, %%mm2\n\t"			// maybe increase it to ffffff if no surround motion
//		"movq	%%mm2, "_Max_Vals"\n\t"
#endif
			
		"movq	%%mm0, %%mm2\n\t"						
//		pavgb	%%mm2, %%mm1					// avg(b,e)
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(b,e)
				
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"

Lines 33-40 static const int64_t __attribute__((__us Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/TomsMoCompAll.inc (-3 / +3 lines)
33	static const int64_t __attribute__((__used__)) FOURS = 0x0404040404040404ull;	33	static const int64_t __attribute__((__used__)) FOURS = 0x0404040404040404ull;
34	static const int64_t __attribute__((__used__)) ONES = 0x0101010101010101ull;	34	static const int64_t __attribute__((__used__)) ONES = 0x0101010101010101ull;
35	static const int64_t __attribute__((__used__)) ShiftMask = 0xfefffefffefffeffull;	35	static const int64_t __attribute__((__used__)) ShiftMask = 0xfefffefffefffeffull;
36	//static int64_t Min_Vals = 0x0000000000000000ull;	36	static int64_t Min_Vals = 0x0000000000000000ull;
37	//static int64_t Max_Vals = 0x0000000000000000ull;	37	static int64_t Max_Vals = 0x0000000000000000ull;
38	#endif	38	#endif
39		39
40	#ifndef TopFirst	40	#ifndef TopFirst
Lines 69-75 static void FUNCT_NAME(uint8_t *output, Link Here
69	int rowsize;	69	int rowsize;
70	int FldHeight;	70	int FldHeight;
71	int stride = (width*2);	71	int stride = (width*2);
72	long oldbx;	72	long olddx;
73		73
74		74
75	src_pitch = stride*2;	75	src_pitch = stride*2;




		// selected for the	smallest of abs(a,f), abs(c,d), or abs(b,e), etc.

		// a,f
		"movq    -2(%%"XDX"), %%mm0\n\t"		// value a from top left		
		"movq    2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value f from bottom right			
		"movq	%%mm0, %%mm6\n\t"					
//		pavgb	%%mm6, %%mm1					// avg(a,f), also best so far
		V_PAVGB ("%%mm6", "%%mm1", "%%mm7", _ShiftMask)	// avg(a,f), also best so far
        "movq	%%mm0, %%mm7\n\t"
		"psubusb	 %%mm1, %%mm7\n\t"
		"psubusb %%mm0, %%mm1\n\t"
		"por		%%mm1, %%mm7\n\t"					// abs diff, also best so far

		// c,d
		"movq    2(%%"XDX"), %%mm0\n\t"		// value a from top left		
		"movq    -2(%%"XDX", %%"XCX"), %%mm1\n\t"	// value f from bottom right			
		"movq	%%mm0, %%mm2\n\t"						
//		pavgb	%%mm2, %%mm1					// avg(c,d)
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(c,d)
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"


		"por		%%mm2, %%mm6\n\t"			// and merge new & old vals keeping best
		"por		%%mm1, %%mm7\n\t"
		"por		"_UVMask", %%mm7\n\t"			// but we know chroma is worthless so far
		"pand	"_YMask", %%mm5\n\t"			// mask out chroma from here also

		// j,n
		"movq    -4(%%"XDX"), %%mm0\n\t"		// value j from top left		
		"movq    4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value n from bottom right			
		"movq	%%mm0, %%mm2\n\t"						
//		pavgb	%%mm2, %%mm1					// avg(j,n)
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(j,n)
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"
		"psubusb %%mm0, %%mm1\n\t"

		"por		%%mm1, %%mm7\n\t"			// "

		// k, m
		"movq    4(%%"XDX"), %%mm0\n\t"		// value k from top right		
		"movq    -4(%%"XDX", %%"XCX"), %%mm1\n\t"	// value n from bottom left			
		"movq	%%mm0, %%mm4\n\t"						
//		pavgb	%%mm4, %%mm1					// avg(k,m)
		V_PAVGB ("%%mm4", "%%mm1", "%%mm3", _ShiftMask)	// avg(k,m)

        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"

		"por		%%mm1, %%mm7\n\t"			// "

		// b,e
		"movq    (%%"XDX"), %%mm0\n\t"		// value b from top 		
		"movq    (%%"XDX", %%"XCX"), %%mm1\n\t"	// value e from bottom 

// We will also calc here the max/min values to later limit comb
// so the max excursion will not exceed the Max_Comb constant

		"por		%%mm2, %%mm3\n\t"			// abs diff
//		pmaxub  %%mm3, %%mm4			// top or bottom pixel moved most
		V_PMAXUB ("%%mm3", "%%mm4")			// top or bottom pixel moved most
		"psubusb "_Max_Mov", %%mm3\n\t"		// moved more than allowed? or goes to 0?
		"pxor	%%mm4, %%mm4\n\t"
		"pcmpeqb %%mm4, %%mm3\n\t"			// now ff where low motion, else high motion
		

		V_PMAXUB ("%%mm6", "%%mm2")

		"psubusb %%mm3, %%mm2\n\t"			// maybe decrease it to 0000.. if no surround motion
//		"movq	%%mm2, "_Min_Vals"\n\t"

		"movq	%%mm0, %%mm2\n\t"
		V_PMAXUB ("%%mm2", "%%mm1")
//		pminub	%%mm6, %%mm2			// clip our current results so far to be below this
		V_PMINUB ("%%mm6", "%%mm2", "%%mm4")
        "paddusb %%mm3, %%mm2\n\t"			// maybe increase it to ffffff if no surround motion
//		"movq	%%mm2, "_Max_Vals"\n\t"
#endif
        
		"movq	%%mm0, %%mm2\n\t"						
//		pavgb	%%mm2, %%mm1					// avg(b,e)
		V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask)	// avg(b,e)
				
        "movq	%%mm0, %%mm3\n\t"
		"psubusb	%%mm1, %%mm3\n\t"




    "por     %%xmm0,      %%xmm5\n\t"      /* and merge new & old vals */                \
    "por     %%xmm2,      %%xmm7\n\t"

#define RESET_CHROMA "por "_UVMask", %%xmm7\n\t"

#else // ifdef IS_SSE2


    "psubusb %%mm1,      %%mm2\n\t"                                                     \
    "psubusb %%mm0,      %%mm3\n\t"                                                     \
    "por     %%mm3,      %%mm2\n\t"                                                     \
    V_PAVGB ("%%mm0", "%%mm1", "%%mm3", _ShiftMask) /* avg of 2 pixels */               \
    "movq    %%mm2,      %%mm3\n\t"       /* another copy of our our weights */         \
    "pxor    %%mm1,      %%mm1\n\t"                                                     \
    "psubusb %%mm7,      %%mm3\n\t"       /* nonzero where old weights lower, else 0 */ \

    "movq    "PADDR2A",   %%mm1\n\t"      /* our pixel2 value */                        \
    "movq    "PADDR1B",   %%mm2\n\t"      /* our 4 pixels */                            \
    "movq    "PADDR2B",   %%mm3\n\t"      /* our pixel2 value */                        \
    V_PAVGB("%%mm0", "%%mm2", "%%mm2", _ShiftMask)                                      \
    V_PAVGB("%%mm1", "%%mm3", "%%mm3", _ShiftMask)                                      \
    "movq    %%mm0,       %%mm2\n\t"      /* another copy of our pixel1 value */        \
    "movq    %%mm1,       %%mm3\n\t"      /* another copy of our pixel1 value */        \
    "psubusb %%mm1,       %%mm2\n\t"                                                    \
    "psubusb %%mm0,       %%mm3\n\t"                                                    \
    "por     %%mm3,       %%mm2\n\t"                                                    \
    V_PAVGB("%%mm0", "%%mm1", "%%mm3", _ShiftMask)   /* avg of 2 pixels */              \
    "movq    %%mm2,       %%mm3\n\t"      /* another copy of our our weights */         \
    "pxor    %%mm1,       %%mm1\n\t"                                                    \
    "psubusb %%mm7,       %%mm3\n\t"      /* nonzero where old weights lower, else 0 */ \

    "por     %%mm0,       %%mm5\n\t"      /* and merge new & old vals */                \
    "por     %%mm2,       %%mm7\n\t"

#define RESET_CHROMA "por "_UVMask", %%mm7\n\t"

#endif


Return to bug 113255

Lines 38-45 Link Here

(-)xine-lib-1.1.1-old/src/libw32dll/wine/module.c (-8 / +4 lines)
38		38
39	#ifdef EMU_QTX_API	39	#ifdef EMU_QTX_API
40	#include "wrapper.h"	40	#include "wrapper.h"
41	static int report_func(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags);	41	int report_func(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags);
42	static int report_func_ret(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags);	42	int report_func_ret(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags);
43	#endif	43	#endif
44		44
45	//#undef TRACE	45	//#undef TRACE
Lines 533-540 HMODULE WINAPI LoadLibraryExA(LPCSTR lib Link Here
533	printf ("wine/module: QuickTime.qts patched!!! old entry=%p\n",ptr[0]);	533	printf ("wine/module: QuickTime.qts patched!!! old entry=%p\n",ptr[0]);
534		534
535	#ifdef EMU_QTX_API	535	#ifdef EMU_QTX_API
536	report_entry = report_func;
537	report_ret = report_func_ret;
538	wrapper_target=ptr[0];	536	wrapper_target=ptr[0];
539	ptr[0]=wrapper;	537	ptr[0]=wrapper;
540	#endif	538	#endif
Lines 697-703 static int dump_component(char* name,int Link Here
697	static u_int32_t ret_array[4096];	695	static u_int32_t ret_array[4096];
698	static int ret_i=0;	696	static int ret_i=0;
699		697
700	static int report_func(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags)	698	int report_func(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags)
701	{	699	{
702	#ifdef DEBUG_QTX_API	700	#ifdef DEBUG_QTX_API
703	int i;	701	int i;
Lines 896-902 static int report_func(void *stack_base, Link Here
896	return 0;	894	return 0;
897	}	895	}
898		896
899	static int report_func_ret(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags)	897	int report_func_ret(void stack_base, int stack_size, reg386_t reg, u_int32_t *flags)
900	{	898	{
901	#ifdef DEBUG_QTX_API	899	#ifdef DEBUG_QTX_API
902	int i;	900	int i;
Lines 1011-1018 FARPROC MODULE_GetProcAddress( Link Here
1011	// \|\| !strcmp(function,"_CallComponent")	1009	// \|\| !strcmp(function,"_CallComponent")
1012	){	1010	){
1013	fprintf(stderr,"theQuickTimeDispatcher caught -> %p\n",retproc);	1011	fprintf(stderr,"theQuickTimeDispatcher caught -> %p\n",retproc);
1014	report_entry = report_func;
1015	report_ret = report_func_ret;
1016	wrapper_target=(void *)retproc;	1012	wrapper_target=(void *)retproc;
1017	retproc=(void *)wrapper;	1013	retproc=(void *)wrapper;
1018	}	1014	}

Lines 183-189 static void DeinterlaceGreedy2Frame_MMX( Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/greedy2frame_template.c (-14 / +14 lines)
183	* See above for a description of the algorithm.	183	* See above for a description of the algorithm.
184	*/	184	*/
185	".align 8 \n\t"	185	".align 8 \n\t"
186	"movq "MANGLE(Mask)", %%mm6 \n\t"	186	"movq %4, %%mm6 \n\t"
187		187
188	"movq %0, %%mm1 \n\t" // T1	188	"movq %0, %%mm1 \n\t" // T1
189	"movq %1, %%mm0 \n\t" // M1	189	"movq %1, %%mm0 \n\t" // M1
Lines 191-197 static void DeinterlaceGreedy2Frame_MMX( Link Here
191	"movq %3, %%mm2 \n\t" // M0	191	"movq %3, %%mm2 \n\t" // M0
192	: /* no output */	192	: /* no output */
193	: "m" (T1), "m" (M1),	193	: "m" (T1), "m" (M1),
194	"m" (B1), "m" (M0) );	194	"m" (B1), "m" (M0), "m" (Mask) );
195		195
196		196
197	asm volatile(	197	asm volatile(
Lines 248-257 static void DeinterlaceGreedy2Frame_MMX( Link Here
248	#endif	248	#endif
249		249
250	/* if \|M1-M0\| > Threshold we want dword worth of twos */	250	/* if \|M1-M0\| > Threshold we want dword worth of twos */
251	"pcmpgtb "MANGLE(qwGreedyTwoFrameThreshold)", %%mm4 \n\t"	251	"pcmpgtb %3, %%mm4 \n\t"
252	"pand "MANGLE(Mask)", %%mm4 \n\t" /* get rid of sign bit */	252	"pand %4, %%mm4 \n\t" /* get rid of sign bit */
253	"pcmpgtd "MANGLE(DwordOne)", %%mm4 \n\t" /* do we want to bob */	253	"pcmpgtd %5, %%mm4 \n\t" /* do we want to bob */
254	"pandn "MANGLE(DwordTwo)", %%mm4 \n\t"	254	"pandn %6, %%mm4 \n\t"
255		255
256	"movq %1, %%mm2 \n\t" /* mm2 = T0 */	256	"movq %1, %%mm2 \n\t" /* mm2 = T0 */
257		257
Lines 264-274 static void DeinterlaceGreedy2Frame_MMX( Link Here
264	"pand %%mm6, %%mm5 \n\t"	264	"pand %%mm6, %%mm5 \n\t"
265		265
266	/* if \|T1-T0\| > Threshold we want dword worth of ones */	266	/* if \|T1-T0\| > Threshold we want dword worth of ones */
267	"pcmpgtb "MANGLE(qwGreedyTwoFrameThreshold)", %%mm5 \n\t"	267	"pcmpgtb %3, %%mm5 \n\t"
268	"pand %%mm6, %%mm5 \n\t" /* get rid of sign bit */	268	"pand %%mm6, %%mm5 \n\t" /* get rid of sign bit */
269		269
270	"pcmpgtd "MANGLE(DwordOne)", %%mm5 \n\t"	270	"pcmpgtd %5, %%mm5 \n\t"
271	"pandn "MANGLE(DwordOne)", %%mm5 \n\t"	271	"pandn %5, %%mm5 \n\t"
272	"paddd %%mm5, %%mm4 \n\t"	272	"paddd %%mm5, %%mm4 \n\t"
273		273
274	"movq %2, %%mm2 \n\t" /* B0 */	274	"movq %2, %%mm2 \n\t" /* B0 */
Lines 282-294 static void DeinterlaceGreedy2Frame_MMX( Link Here
282	"pand %%mm6, %%mm5 \n\t"	282	"pand %%mm6, %%mm5 \n\t"
283		283
284	/* if \|B1-B0\| > Threshold we want dword worth of ones */	284	/* if \|B1-B0\| > Threshold we want dword worth of ones */
285	"pcmpgtb "MANGLE(qwGreedyTwoFrameThreshold)", %%mm5 \n\t"	285	"pcmpgtb %3, %%mm5 \n\t"
286	"pand %%mm6, %%mm5 \n\t" /* get rid of any sign bit */	286	"pand %%mm6, %%mm5 \n\t" /* get rid of any sign bit */
287	"pcmpgtd "MANGLE(DwordOne)", %%mm5 \n\t"	287	"pcmpgtd %5, %%mm5 \n\t"
288	"pandn "MANGLE(DwordOne)", %%mm5 \n\t"	288	"pandn %5, %%mm5 \n\t"
289	"paddd %%mm5, %%mm4 \n\t"	289	"paddd %%mm5, %%mm4 \n\t"
290		290
291	"pcmpgtd "MANGLE(DwordTwo)", %%mm4 \n\t"	291	"pcmpgtd %6, %%mm4 \n\t"
292		292
293	/* debugging feature	293	/* debugging feature
294	* output the value of mm4 at this point which is pink where we will weave	294	* output the value of mm4 at this point which is pink where we will weave
Lines 314-320 static void DeinterlaceGreedy2Frame_MMX( Link Here
314	#endif	314	#endif
315		315
316	: "=m" (*Dest2)	316	: "=m" (*Dest2)
317	: "m" (T0), "m" (B0) );	317	: "m" (T0), "m" (B0), "m" (qwGreedyTwoFrameThreshold), "m" (Mask), "m" (DwordOne), "m" (DwordTwo) );
318		318
319	/* Advance to the next set of pixels. */	319	/* Advance to the next set of pixels. */
320	T1 += 8;	320	T1 += 8;

Lines 43-49 static void FUNCT_NAME(uint8_t *output, Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/greedyh.asm (-30 / +36 lines)
43		43
44	int Line;	44	int Line;
45	long LoopCtr;	45	long LoopCtr;
46	long oldbx;
47	unsigned int Pitch = stride*2;	46	unsigned int Pitch = stride*2;
48	int FieldHeight = height / 2;	47	int FieldHeight = height / 2;
49		48
Lines 52-57 static void FUNCT_NAME(uint8_t *output, Link Here
52	unsigned char* L3; // ptr to Line3	51	unsigned char* L3; // ptr to Line3
53		52
54	unsigned char* L2P; // ptr to prev Line2	53	unsigned char* L2P; // ptr to prev Line2
		54	unsigned char* temp;
55	unsigned char* Dest = output;	55	unsigned char* Dest = output;
56		56
57	int64_t LastAvg=0; //interp value from left qword	57	int64_t LastAvg=0; //interp value from left qword
Lines 121-145 static void FUNCT_NAME(uint8_t *output, Link Here
121	#define asmLastAvg "%0"	121	#define asmLastAvg "%0"
122	#define asmL1 "%1"	122	#define asmL1 "%1"
123	#define asmL3 "%2"	123	#define asmL3 "%2"
124	#define asmL2P "%3"	124	#define asmtemp "%3"
125	#define asmL2 "%4"	125	#define asmL2 "%4"
126	#define asmDest "%5"	126	#define asmDest "%5"
127	#define asmLoopCtr "%6"	127	#define asmLoopCtr "%6"
128	#define asmoldbx "%7"
129	#endif	128	#endif
130		129
131	// For ease of reading, the comments below assume that we're operating on an odd	130	// For ease of reading, the comments below assume that we're operating on an odd
132	// field (i.e., that InfoIsOdd is true). Assume the obvious for even lines..	131	// field (i.e., that InfoIsOdd is true). Assume the obvious for even lines..
		132	temp = L2P;
133	__asm__ __volatile__	133	__asm__ __volatile__
134	(	134	(
135	// save ebx (-fPIC)
136	MOVX" %%"XBX", "asmoldbx"\n\t"
137
138	MOVX" "asmL1", %%"XAX"\n\t"	135	MOVX" "asmL1", %%"XAX"\n\t"
139	LEAX" 8(%%"XAX"), %%"XBX"\n\t" // next qword needed by DJR	136	LEAX" 8(%%"XAX"), %%"XDX"\n\t" // next qword needed by DJR
140	MOVX" "asmL3", %%"XCX"\n\t"	137	MOVX" "asmL3", %%"XCX"\n\t"
141	SUBX" %%"XAX", %%"XCX"\n\t" // carry L3 addr as an offset	138	SUBX" %%"XAX", %%"XCX"\n\t" // carry L3 addr as an offset
142	MOVX" "asmL2P", %%"XDX"\n\t"
143	MOVX" "asmL2", %%"XSI"\n\t"	139	MOVX" "asmL2", %%"XSI"\n\t"
144	MOVX" "asmDest", %%"XDI"\n\t" // DL1 if Odd or DL2 if Even	140	MOVX" "asmDest", %%"XDI"\n\t" // DL1 if Odd or DL2 if Even
145		141
Lines 148-158 static void FUNCT_NAME(uint8_t *output, Link Here
148		144
149	"movq (%%"XSI"), %%mm0\n\t" // L2 - the newest weave pixel value	145	"movq (%%"XSI"), %%mm0\n\t" // L2 - the newest weave pixel value
150	"movq (%%"XAX"), %%mm1\n\t" // L1 - the top pixel	146	"movq (%%"XAX"), %%mm1\n\t" // L1 - the top pixel
		147	PUSHX" %%"XDX "\n\t"
		148	MOVX" "asmtemp", %%"XDX"\n\t"
151	"movq (%%"XDX"), %%mm2\n\t" // L2P - the prev weave pixel	149	"movq (%%"XDX"), %%mm2\n\t" // L2P - the prev weave pixel
		150	POPX" %%"XDX "\n\t"
152	"movq (%%"XAX", %%"XCX"), %%mm3\n\t" // L3, next odd row	151	"movq (%%"XAX", %%"XCX"), %%mm3\n\t" // L3, next odd row
153	"movq %%mm1, %%mm6\n\t" // L1 - get simple single pixel interp	152	"movq %%mm1, %%mm6\n\t" // L1 - get simple single pixel interp
154	// pavgb mm6, mm3 // use macro below	153	// pavgb mm6, mm3 // use macro below
155	V_PAVGB ("%%mm6", "%%mm3", "%%mm4", MANGLE(ShiftMask))	154	V_PAVGB ("%%mm6", "%%mm3", "%%mm4", "%8")
156		155
157	// DJR - Diagonal Jaggie Reduction	156	// DJR - Diagonal Jaggie Reduction
158	// In the event that we are going to use an average (Bob) pixel we do not want a jagged	157	// In the event that we are going to use an average (Bob) pixel we do not want a jagged
Lines 166-189 static void FUNCT_NAME(uint8_t *output, Link Here
166	"psllq $16, %%mm7\n\t" // left justify 3 pixels	165	"psllq $16, %%mm7\n\t" // left justify 3 pixels
167	"por %%mm7, %%mm4\n\t" // and combine	166	"por %%mm7, %%mm4\n\t" // and combine
168		167
169	"movq (%%"XBX"), %%mm5\n\t" // next horiz qword from L1	168	"movq (%%"XDX"), %%mm5\n\t" // next horiz qword from L1
170	// pavgb mm5, qword ptr[ebx+ecx] // next horiz qword from L3, use macro below	169	// pavgb mm5, qword ptr[ebx+ecx] // next horiz qword from L3, use macro below
171	V_PAVGB ("%%mm5", "(%%"XBX",%%"XCX")", "%%mm7", MANGLE(ShiftMask))	170	V_PAVGB ("%%mm5", "(%%"XDX",%%"XCX")", "%%mm7", "%8")
172	"psllq $48, %%mm5\n\t" // left just 1 pixel	171	"psllq $48, %%mm5\n\t" // left just 1 pixel
173	"movq %%mm6, %%mm7\n\t" // another copy of simple bob pixel	172	"movq %%mm6, %%mm7\n\t" // another copy of simple bob pixel
174	"psrlq $16, %%mm7\n\t" // right just 3 pixels	173	"psrlq $16, %%mm7\n\t" // right just 3 pixels
175	"por %%mm7, %%mm5\n\t" // combine	174	"por %%mm7, %%mm5\n\t" // combine
176	// pavgb mm4, mm5 // avg of forward and prev by 1 pixel, use macro	175	// pavgb mm4, mm5 // avg of forward and prev by 1 pixel, use macro
177	V_PAVGB ("%%mm4", "%%mm5", "%%mm5", MANGLE(ShiftMask)) // mm5 gets modified if MMX	176	V_PAVGB ("%%mm4", "%%mm5", "%%mm5", "%8") // mm5 gets modified if MMX
178	// pavgb mm6, mm4 // avg of center and surround interp vals, use macro	177	// pavgb mm6, mm4 // avg of center and surround interp vals, use macro
179	V_PAVGB ("%%mm6", "%%mm4", "%%mm7", MANGLE(ShiftMask))	178	V_PAVGB ("%%mm6", "%%mm4", "%%mm7", "%8")
180		179
181	// Don't do any more averaging than needed for mmx. It hurts performance and causes rounding errors.	180	// Don't do any more averaging than needed for mmx. It hurts performance and causes rounding errors.
182	#ifndef IS_MMX	181	#ifndef IS_MMX
183	// pavgb mm4, mm6 // 1/4 center, 3/4 adjacent	182	// pavgb mm4, mm6 // 1/4 center, 3/4 adjacent
184	V_PAVGB ("%%mm4", "%%mm6", "%%mm7", MANGLE(ShiftMask))	183	V_PAVGB ("%%mm4", "%%mm6", "%%mm7", "%8")
185	// pavgb mm6, mm4 // 3/8 center, 5/8 adjacent	184	// pavgb mm6, mm4 // 3/8 center, 5/8 adjacent
186	V_PAVGB ("%%mm6", "%%mm4", "%%mm7", MANGLE(ShiftMask))	185	V_PAVGB ("%%mm6", "%%mm4", "%%mm7", "%8")
187	#endif	186	#endif
188		187
189	// get abs value of possible L2 comb	188	// get abs value of possible L2 comb
Lines 236-299 static void FUNCT_NAME(uint8_t *output, Link Here
236	// pminub mm5, mm3 // now = Min(L1,L3), use macro	235	// pminub mm5, mm3 // now = Min(L1,L3), use macro
237	V_PMINUB ("%%mm5", "%%mm3", "%%mm7")	236	V_PMINUB ("%%mm5", "%%mm3", "%%mm7")
238	// allow the value to be above the high or below the low by amt of MaxComb	237	// allow the value to be above the high or below the low by amt of MaxComb
239	"psubusb "MANGLE(MaxComb)", %%mm5\n\t" // lower min by diff	238	"psubusb %9, %%mm5\n\t" // lower min by diff
240	"paddusb "MANGLE(MaxComb)", %%mm2\n\t" // increase max by diff	239	"paddusb %9, %%mm2\n\t" // increase max by diff
241	// pmaxub mm4, mm5 // now = Max(best,Min(L1,L3) use macro	240	// pmaxub mm4, mm5 // now = Max(best,Min(L1,L3) use macro
242	V_PMAXUB ("%%mm4", "%%mm5")	241	V_PMAXUB ("%%mm4", "%%mm5")
243	// pminub mm4, mm2 // now = Min( Max(best, Min(L1,L3), L2 )=L2 clipped	242	// pminub mm4, mm2 // now = Min( Max(best, Min(L1,L3), L2 )=L2 clipped
244	V_PMINUB ("%%mm4", "%%mm2", "%%mm7")	243	V_PMINUB ("%%mm4", "%%mm2", "%%mm7")
245		244
246	// Blend weave pixel with bob pixel, depending on motion val in mm0	245	// Blend weave pixel with bob pixel, depending on motion val in mm0
247	"psubusb "MANGLE(MotionThreshold)", %%mm0\n\t"// test Threshold, clear chroma change >>>??	246	"psubusb %10, %%mm0\n\t"// test Threshold, clear chroma change >>>??
248	"pmullw "MANGLE(MotionSense)", %%mm0\n\t" // mul by user factor, keep low 16 bits	247	"pmullw %11, %%mm0\n\t" // mul by user factor, keep low 16 bits
249	"movq "MANGLE(QW256)", %%mm7\n\t"	248	"movq %12, %%mm7\n\t"
250	#ifdef IS_SSE	249	#ifdef IS_SSE
251	"pminsw %%mm7, %%mm0\n\t" // max = 256	250	"pminsw %%mm7, %%mm0\n\t" // max = 256
252	#else	251	#else
253	"paddusw "MANGLE(QW256B)", %%mm0\n\t" // add, may sat at fff..	252	"paddusw %13, %%mm0\n\t" // add, may sat at fff..
254	"psubusw "MANGLE(QW256B)", %%mm0\n\t" // now = Min(L1,256)	253	"psubusw %13, %%mm0\n\t" // now = Min(L1,256)
255	#endif	254	#endif
256	"psubusw %%mm0, %%mm7\n\t" // so the 2 sum to 256, weighted avg	255	"psubusw %%mm0, %%mm7\n\t" // so the 2 sum to 256, weighted avg
257	"movq %%mm4, %%mm2\n\t" // save weave chroma info before trashing	256	"movq %%mm4, %%mm2\n\t" // save weave chroma info before trashing
258	"pand "MANGLE(YMask)", %%mm4\n\t" // keep only luma from calc'd value	257	"pand %14, %%mm4\n\t" // keep only luma from calc'd value
259	"pmullw %%mm7, %%mm4\n\t" // use more weave for less motion	258	"pmullw %%mm7, %%mm4\n\t" // use more weave for less motion
260	"pand "MANGLE(YMask)", %%mm6\n\t" // keep only luma from calc'd value	259	"pand %14, %%mm6\n\t" // keep only luma from calc'd value
261	"pmullw %%mm0, %%mm6\n\t" // use more bob for large motion	260	"pmullw %%mm0, %%mm6\n\t" // use more bob for large motion
262	"paddusw %%mm6, %%mm4\n\t" // combine	261	"paddusw %%mm6, %%mm4\n\t" // combine
263	"psrlw $8, %%mm4\n\t" // div by 256 to get weighted avg	262	"psrlw $8, %%mm4\n\t" // div by 256 to get weighted avg
264		263
265	// chroma comes from weave pixel	264	// chroma comes from weave pixel
266	"pand "MANGLE(UVMask)", %%mm2\n\t" // keep chroma	265	"pand %15, %%mm2\n\t" // keep chroma
267	"por %%mm4, %%mm2\n\t" // and combine	266	"por %%mm4, %%mm2\n\t" // and combine
268		267
269	V_MOVNTQ ("(%%"XDI")", "%%mm2") // move in our clipped best, use macro	268	V_MOVNTQ ("(%%"XDI")", "%%mm2") // move in our clipped best, use macro
270		269
271	// bump ptrs and loop	270	// bump ptrs and loop
272	LEAX" 8(%%"XAX"), %%"XAX"\n\t"	271	LEAX" 8(%%"XAX"), %%"XAX"\n\t"
273	LEAX" 8(%%"XBX"), %%"XBX"\n\t"
274	LEAX" 8(%%"XDX"), %%"XDX"\n\t"	272	LEAX" 8(%%"XDX"), %%"XDX"\n\t"
		273	ADDX" $8, "asmtemp"\n\t"
275	LEAX" 8(%%"XDI"), %%"XDI"\n\t"	274	LEAX" 8(%%"XDI"), %%"XDI"\n\t"
276	LEAX" 8(%%"XSI"), %%"XSI"\n\t"	275	LEAX" 8(%%"XSI"), %%"XSI"\n\t"
277	DECX" "asmLoopCtr"\n\t"	276	DECX" "asmLoopCtr"\n\t"
278	"jg 1b\n\t" // loop if not to last line	277	"jg 1b\n\t" // loop if not to last line
279	// note P-III default assumes backward branches taken	278	// note P-III default assumes backward branches taken
280	"jl 1f\n\t" // done	279	"jl 1f\n\t" // done
281	MOVX" %%"XAX", %%"XBX"\n\t" // sharpness lookahead 1 byte only, be wrong on 1	280	MOVX" %%"XAX", %%"XDX"\n\t" // sharpness lookahead 1 byte only, be wrong on 1
282	"jmp 1b\n\t"	281	"jmp 1b\n\t"
283		282
284	"1:\n\t"	283	"1:\n\t"
285	MOVX" "asmoldbx", %%"XBX"\n\t"
286		284
287	: /* no outputs */	285	: /* no outputs */
288		286
289	: "m"(LastAvg),	287	: "m"(LastAvg),
290	"m"(L1),	288	"m"(L1),
291	"m"(L3),	289	"m"(L3),
292	"m"(L2P),	290	"m"(temp),
293	"m"(L2),	291	"m"(L2),
294	"m"(Dest),	292	"m"(Dest),
295	"m"(LoopCtr),	293	"m"(LoopCtr),
296	"m"(oldbx)	294	"m"(temp),
		295	"m"(ShiftMask),
		296	"m"(MaxComb),
		297	"m"(MotionThreshold),
		298	"m"(MotionSense),
		299	"m"(QW256),
		300	"m"(QW256B),
		301	"m"(YMask),
		302	"m"(UVMask)
297		303
298	: XAX, XCX, XDX, XSI, XDI,	304	: XAX, XCX, XDX, XSI, XDI,
299	#ifdef ARCH_X86	305	#ifdef ARCH_X86

Lines 66-72 long dst_pitchw = dst_pitch; // local s Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/SearchLoopTop.inc (-28 / +36 lines)
66	#define _pSrc "%5"	66	#define _pSrc "%5"
67	#define _pSrcP "%6"	67	#define _pSrcP "%6"
68	#define _pBobP "%7"	68	#define _pBobP "%7"
69	#define _oldbx "%8"	69	#define _olddx "%8"
		70	#define _UVMask "%9"
		71	#define _ShiftMask "%10"
		72	#define _FOURS "%11"
		73	#define _TENS "%12"
		74	#define _Max_Vals "%13"
		75	#define _Min_Vals "%14"
		76	#define _YMask "%15"
		77	#define _Max_Mov "%16"
		78	#define _ONES "%17"
		79	#define _DiffThres "%18"
70	#endif	80	#endif
71		81
72	for (y=1; y < FldHeight-1; y++)	82	for (y=1; y < FldHeight-1; y++)
Lines 77-151 long dst_pitchw = dst_pitch; // local s Link Here
77	// Loop general reg usage	87	// Loop general reg usage
78	//	88	//
79	// XAX - pBobP, then pDest	89	// XAX - pBobP, then pDest
80	// XBX - pBob	90	// XDX - pBob
81	// XCX - src_pitch2	91	// XCX - src_pitch2
82	// XDX - current offset	92	// _olddx - current offset
83	// XDI - prev weave pixels, 1 line up	93	// XDI - prev weave pixels, 1 line up
84	// XSI - next weave pixels, 1 line up	94	// XSI - next weave pixels, 1 line up
85		95
86	// Save "XBX" (-fPIC)
87	MOVX" %%"XBX", "_oldbx"\n\t"
88
89	#ifdef IS_SSE2	96	#ifdef IS_SSE2
90		97
91	// sse2 code deleted for now	98	// sse2 code deleted for now
92		99
93	#else	100	#else
94	// simple bob first 8 bytes	101	// simple bob first 8 bytes
95	MOVX" "_pBob", %%"XBX"\n\t"	102	MOVX" "_pBob", %%"XDX"\n\t"
96	MOVX" "_src_pitch2", %%"XCX"\n\t"	103	MOVX" "_src_pitch2", %%"XCX"\n\t"
97		104
98	#ifdef USE_VERTICAL_FILTER	105	#ifdef USE_VERTICAL_FILTER
99	"movq (%%"XBX"), %%mm0\n\t"	106	"movq (%%"XDX"), %%mm0\n\t"
100	"movq (%%"XBX", %%"XCX"), %%mm1\n\t" //, qword ptr["XBX"+"XCX"]	107	"movq (%%"XDX", %%"XCX"), %%mm1\n\t" //, qword ptr["XDX"+"XCX"]
101	"movq %%mm0, %%mm2\n\t"	108	"movq %%mm0, %%mm2\n\t"
102	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // halfway between	109	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // halfway between
103	V_PAVGB ("%%mm0", "%%mm2", "%%mm3", MANGLE(ShiftMask)) // 1/4 way	110	V_PAVGB ("%%mm0", "%%mm2", "%%mm3", _ShiftMask) // 1/4 way
104	V_PAVGB ("%%mm1", "%%mm2", "%%mm3", MANGLE(ShiftMask)) // 3/4 way	111	V_PAVGB ("%%mm1", "%%mm2", "%%mm3", _ShiftMask) // 3/4 way
105	MOVX" "_pDest", %%"XDI"\n\t"	112	MOVX" "_pDest", %%"XDI"\n\t"
106	MOVX" "_dst_pitchw", %%"XAX"\n\t"	113	MOVX" "_dst_pitchw", %%"XAX"\n\t"
107	V_MOVNTQ ("(%%"XDI")", "%%mm0")	114	V_MOVNTQ ("(%%"XDI")", "%%mm0")
108	V_MOVNTQ ("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1	115	V_MOVNTQ ("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1
109		116
110	// simple bob last 8 bytes	117	// simple bob last 8 bytes
111	MOVX" "_Last8", %%"XDX"\n\t"	118	MOVX" "_Last8", %%"XSI"\n\t"
112	LEAX" (%%"XBX", %%"XDX"), %%"XSI"\n\t" // ["XBX"+"XDX"]	119	ADDX" %%"XDX", %%"XSI"\n\t" // ["XDX"+"olddx"]
113	"movq (%%"XSI"), %%mm0\n\t"	120	"movq (%%"XSI"), %%mm0\n\t"
114	"movq (%%"XSI", %%"XCX"), %%mm1\n\t" // qword ptr["XSI"+"XCX"]	121	"movq (%%"XSI", %%"XCX"), %%mm1\n\t" // qword ptr["XSI"+"XCX"]
115	"movq %%mm0, %%mm2\n\t"	122	"movq %%mm0, %%mm2\n\t"
116	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // halfway between	123	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // halfway between
117	V_PAVGB ("%%mm0", "%%mm2", "%%mm3", MANGLE(ShiftMask)) // 1/4 way	124	V_PAVGB ("%%mm0", "%%mm2", "%%mm3", _ShiftMask) // 1/4 way
118	V_PAVGB ("%%mm1", "%%mm2", "%%mm3", MANGLE(ShiftMask)) // 3/4 way	125	V_PAVGB ("%%mm1", "%%mm2", "%%mm3", _ShiftMask) // 3/4 way
119	ADDX" %%"XDX", %%"XDI"\n\t" // last 8 bytes of dest	126	ADDX" "_olddx", %%"XDI"\n\t" // last 8 bytes of dest
120	V_MOVNTQ ("%%"XDI"", "%%mm0")	127	V_MOVNTQ ("%%"XDI"", "%%mm0")
121	V_MOVNTQ ("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1)	128	V_MOVNTQ ("(%%"XDI", %%"XAX")", "%%mm1") // qword ptr["XDI"+"XAX"], mm1)
122		129
123	#else	130	#else
124	"movq (%%"XBX"), %%mm0\n\t"	131	"movq (%%"XDX"), %%mm0\n\t"
125	// pavgb mm0, qword ptr["XBX"+"XCX"]	132	// pavgb mm0, qword ptr["XDX"+"XCX"]
126	V_PAVGB ("%%mm0", "(%%"XBX", %%"XCX")", "%%mm2", MANGLE(ShiftMask)) // qword ptr["XBX"+"XCX"], mm2, ShiftMask)	133	V_PAVGB ("%%mm0", "(%%"XDX", %%"XCX")", "%%mm2", _ShiftMask) // qword ptr["XDX"+"XCX"], mm2, ShiftMask)
127	MOVX" "_pDest", %%"XDI"\n\t"	134	MOVX" "_pDest", %%"XDI"\n\t"
128	V_MOVNTQ ("(%%"XDI")", "%%mm0")	135	V_MOVNTQ ("(%%"XDI")", "%%mm0")
129		136
130	// simple bob last 8 bytes	137	// simple bob last 8 bytes
131	MOVX" "_Last8", %%"XDX"\n\t"	138	MOVX" "_Last8", %%"XSI"\n\t"
132	LEAX" (%%"XBX", %%"XDX"), %%"XSI"\n\t" //"XSI", ["XBX"+"XDX"]	139	ADDX" %%"XDX", %%"XSI"\n\t" //"XSI", ["XDX"+"olddx"]
133	"movq (%%"XSI"), %%mm0\n\t"	140	"movq (%%"XSI"), %%mm0\n\t"
134	// pavgb mm0, qword ptr["XSI"+"XCX"]	141	// pavgb mm0, qword ptr["XSI"+"XCX"]
135	V_PAVGB ("%%mm0", "(%%"XSI", %%"XCX")", "%%mm2", MANGLE(ShiftMask)) // qword ptr["XSI"+"XCX"], mm2, ShiftMask)	142	V_PAVGB ("%%mm0", "(%%"XSI", %%"XCX")", "%%mm2", _ShiftMask) // qword ptr["XSI"+"XCX"], mm2, ShiftMask)
136	V_MOVNTQ ("(%%"XDI", %%"XDX")", "%%mm0") // qword ptr["XDI"+"XDX"], mm0)	143	ADDX" "_olddx", %%"XDI"\n\t"
		144	V_MOVNTQ ("(%%"XDI")", "%%mm0") // qword ptr["XDI"+"olddx"], mm0)
137	#endif	145	#endif
138	// now loop and get the middle qwords	146	// now loop and get the middle qwords
139	MOVX" "_pSrc", %%"XSI"\n\t"	147	MOVX" "_pSrc", %%"XSI"\n\t"
140	MOVX" "_pSrcP", %%"XDI"\n\t"	148	MOVX" "_pSrcP", %%"XDI"\n\t"
141	MOVX" $8, %%"XDX"\n\t" // curr offset longo all lines	149	MOVX" $8, "_olddx"\n\t" // curr offset longo all lines
142		150
143	"1:\n\t"	151	"1:\n\t"
144	MOVX" "_pBobP", %%"XAX"\n\t"	152	MOVX" "_pBobP", %%"XAX"\n\t"
145	ADDX" $8, %%"XDI"\n\t"	153	ADDX" $8, %%"XDI"\n\t"
146	ADDX" $8, %%"XSI"\n\t"	154	ADDX" $8, %%"XSI"\n\t"
147	ADDX" $8, %%"XBX"\n\t"	155	ADDX" $8, %%"XDX"\n\t"
148	ADDX" %%"XDX", %%"XAX"\n\t"	156	ADDX" "_olddx", %%"XAX"\n\t"
149		157
150	#ifdef USE_STRANGE_BOB	158	#ifdef USE_STRANGE_BOB
151	#include "StrangeBob.inc"	159	#include "StrangeBob.inc"

Lines 31-52 Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/StrangeBob.inc (-40 / +40 lines)
31	"pxor %%mm6, %%mm6\n\t"	31	"pxor %%mm6, %%mm6\n\t"
32	"pxor %%mm7, %%mm7\n\t"	32	"pxor %%mm7, %%mm7\n\t"
33		33
34	"movq -2(%%"XBX"), %%mm0\n\t" // value a from top left	34	"movq -2(%%"XDX"), %%mm0\n\t" // value a from top left
35	"movq -4(%%"XBX", %%"XCX"), %%mm1\n\t" // value m from bottom right	35	"movq -4(%%"XDX", %%"XCX"), %%mm1\n\t" // value m from bottom right
36		36
37	"movq %%mm0, %%mm3\n\t"	37	"movq %%mm0, %%mm3\n\t"
38	"psubusb %%mm1, %%mm3\n\t"	38	"psubusb %%mm1, %%mm3\n\t"
39	"psubusb %%mm0, %%mm1\n\t"	39	"psubusb %%mm0, %%mm1\n\t"
40	"por %%mm1, %%mm3\n\t" // abs(a,m)	40	"por %%mm1, %%mm3\n\t" // abs(a,m)
41		41
42	"psubusb "MANGLE(DiffThres)", %%mm3\n\t" // nonzero where abs(a,m) > Thres else 0	42	"psubusb "_DiffThres", %%mm3\n\t" // nonzero where abs(a,m) > Thres else 0
43	"pxor %%mm4, %%mm4\n\t"	43	"pxor %%mm4, %%mm4\n\t"
44	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(a,m) < Thres, else 00	44	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(a,m) < Thres, else 00
45	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(a,m) > Thres, else 00	45	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(a,m) > Thres, else 00
46		46
47		47
48	"movq -4(%%"XBX"), %%mm0\n\t" // value j	48	"movq -4(%%"XDX"), %%mm0\n\t" // value j
49	"movq 4(%%"XBX", %%"XCX"), %%mm1\n\t" // value n	49	"movq 4(%%"XDX", %%"XCX"), %%mm1\n\t" // value n
50	"movq %%mm0, %%mm2\n\t"	50	"movq %%mm0, %%mm2\n\t"
51	"pavgb %%mm1, %%mm2\n\t" // avg(j,n)	51	"pavgb %%mm1, %%mm2\n\t" // avg(j,n)
52	"movq %%mm0, %%mm3\n\t"	52	"movq %%mm0, %%mm3\n\t"
Lines 55-61 Link Here
55	"por %%mm1, %%mm0\n\t" // abs(j,n)	55	"por %%mm1, %%mm0\n\t" // abs(j,n)
56		56
57	"movq %%mm0, %%mm1\n\t"	57	"movq %%mm0, %%mm1\n\t"
58	"psubusb "MANGLE(DiffThres)", %%mm1\n\t" // nonzero where abs(j,n) > Thres else 0	58	"psubusb "_DiffThres", %%mm1\n\t" // nonzero where abs(j,n) > Thres else 0
59	"pxor %%mm3, %%mm3\n\t"	59	"pxor %%mm3, %%mm3\n\t"
60	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(j,n) < Thres, else 00	60	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(j,n) < Thres, else 00
61		61
Lines 75-105 Link Here
75	"por %%mm0, %%mm7\n\t"	75	"por %%mm0, %%mm7\n\t"
76		76
77	// k & m	77	// k & m
78	"movq 2(%%"XBX"), %%mm0\n\t" // value c from top left	78	"movq 2(%%"XDX"), %%mm0\n\t" // value c from top left
79	"movq 4(%%"XBX", %%"XCX"), %%mm1\n\t" // value n from bottom right	79	"movq 4(%%"XDX", %%"XCX"), %%mm1\n\t" // value n from bottom right
80		80
81	"movq %%mm0, %%mm3\n\t"	81	"movq %%mm0, %%mm3\n\t"
82	"psubusb %%mm1, %%mm3\n\t"	82	"psubusb %%mm1, %%mm3\n\t"
83	"psubusb %%mm0, %%mm1\n\t"	83	"psubusb %%mm0, %%mm1\n\t"
84	"por %%mm1, %%mm3\n\t" // abs(c,n)	84	"por %%mm1, %%mm3\n\t" // abs(c,n)
85		85
86	"psubusb "MANGLE(DiffThres)", %%mm3\n\t" // nonzero where abs(c,n) > Thres else 0	86	"psubusb "_DiffThres", %%mm3\n\t" // nonzero where abs(c,n) > Thres else 0
87	"pxor %%mm4, %%mm4\n\t"	87	"pxor %%mm4, %%mm4\n\t"
88	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(c,n) < Thres, else 00	88	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(c,n) < Thres, else 00
89	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(c,n) > Thres, else 00	89	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(c,n) > Thres, else 00
90		90
91		91
92	"movq 4(%%"XBX"), %%mm0\n\t" // value k	92	"movq 4(%%"XDX"), %%mm0\n\t" // value k
93	"movq -4(%%"XBX", %%"XCX"), %%mm1\n\t" // value m	93	"movq -4(%%"XDX", %%"XCX"), %%mm1\n\t" // value m
94	"movq %%mm0, %%mm2\n\t"	94	"movq %%mm0, %%mm2\n\t"
95	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(k,m)	95	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(k,m)
96	"movq %%mm0, %%mm3\n\t"	96	"movq %%mm0, %%mm3\n\t"
97	"psubusb %%mm1, %%mm0\n\t"	97	"psubusb %%mm1, %%mm0\n\t"
98	"psubusb %%mm3, %%mm1\n\t"	98	"psubusb %%mm3, %%mm1\n\t"
99	"por %%mm1, %%mm0\n\t" // abs(k,m)	99	"por %%mm1, %%mm0\n\t" // abs(k,m)
100		100
101	"movq %%mm0, %%mm1\n\t"	101	"movq %%mm0, %%mm1\n\t"
102	"psubusb "MANGLE(DiffThres)", %%mm1\n\t" // nonzero where abs(k,m) > Thres else 0	102	"psubusb "_DiffThres", %%mm1\n\t" // nonzero where abs(k,m) > Thres else 0
103	"pxor %%mm3, %%mm3\n\t"	103	"pxor %%mm3, %%mm3\n\t"
104	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(k,m) < Thres, else 00	104	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(k,m) < Thres, else 00
105		105
Lines 120-149 Link Here
120		120
121		121
122	// c & d	122	// c & d
123	"movq (%%"XBX"), %%mm0\n\t" // value b from top left	123	"movq (%%"XDX"), %%mm0\n\t" // value b from top left
124	"movq 2(%%"XBX", %%"XCX"), %%mm1\n\t" // value f from bottom right	124	"movq 2(%%"XDX", %%"XCX"), %%mm1\n\t" // value f from bottom right
125		125
126	"movq %%mm0, %%mm3\n\t"	126	"movq %%mm0, %%mm3\n\t"
127	"psubusb %%mm1, %%mm3\n\t"	127	"psubusb %%mm1, %%mm3\n\t"
128	"psubusb %%mm0, %%mm1\n\t"	128	"psubusb %%mm0, %%mm1\n\t"
129	"por %%mm1, %%mm3\n\t" // abs(b,f)	129	"por %%mm1, %%mm3\n\t" // abs(b,f)
130		130
131	"psubusb "MANGLE(DiffThres)", %%mm3\n\t" // nonzero where abs(b,f) > Thres else 0	131	"psubusb "_DiffThres", %%mm3\n\t" // nonzero where abs(b,f) > Thres else 0
132	"pxor %%mm4, %%mm4\n\t"	132	"pxor %%mm4, %%mm4\n\t"
133	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(b,f) < Thres, else 00	133	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(b,f) < Thres, else 00
134	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(b,f) > Thres, else 00	134	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(b,f) > Thres, else 00
135		135
136	"movq 2(%%"XBX"), %%mm0\n\t" // value c	136	"movq 2(%%"XDX"), %%mm0\n\t" // value c
137	"movq -2(%%"XBX", %%"XCX"), %%mm1\n\t" // value d	137	"movq -2(%%"XDX", %%"XCX"), %%mm1\n\t" // value d
138	"movq %%mm0, %%mm2\n\t"	138	"movq %%mm0, %%mm2\n\t"
139	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(c,d)	139	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(c,d)
140	"movq %%mm0, %%mm3\n\t"	140	"movq %%mm0, %%mm3\n\t"
141	"psubusb %%mm1, %%mm0\n\t"	141	"psubusb %%mm1, %%mm0\n\t"
142	"psubusb %%mm3, %%mm1\n\t"	142	"psubusb %%mm3, %%mm1\n\t"
143	"por %%mm1, %%mm0\n\t" // abs(c,d)	143	"por %%mm1, %%mm0\n\t" // abs(c,d)
144		144
145	"movq %%mm0, %%mm1\n\t"	145	"movq %%mm0, %%mm1\n\t"
146	"psubusb "MANGLE(DiffThres)", %%mm1\n\t" // nonzero where abs(c,d) > Thres else 0	146	"psubusb "_DiffThres", %%mm1\n\t" // nonzero where abs(c,d) > Thres else 0
147	"pxor %%mm3, %%mm3\n\t"	147	"pxor %%mm3, %%mm3\n\t"
148	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(c,d) < Thres, else 00	148	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(c,d) < Thres, else 00
149		149
Lines 163-192 Link Here
163	"por %%mm0, %%mm7\n\t"	163	"por %%mm0, %%mm7\n\t"
164		164
165	// a & f	165	// a & f
166	"movq (%%"XBX"), %%mm0\n\t" // value b from top left	166	"movq (%%"XDX"), %%mm0\n\t" // value b from top left
167	"movq -2(%%"XBX", %%"XCX"), %%mm1\n\t" // value d from bottom right	167	"movq -2(%%"XDX", %%"XCX"), %%mm1\n\t" // value d from bottom right
168		168
169	"movq %%mm0, %%mm3\n\t"	169	"movq %%mm0, %%mm3\n\t"
170	"psubusb %%mm1, %%mm3\n\t"	170	"psubusb %%mm1, %%mm3\n\t"
171	"psubusb %%mm0, %%mm1\n\t"	171	"psubusb %%mm0, %%mm1\n\t"
172	"por %%mm1, %%mm3\n\t" // abs(b,d)	172	"por %%mm1, %%mm3\n\t" // abs(b,d)
173		173
174	"psubusb "MANGLE(DiffThres)", %%mm3\n\t" // nonzero where abs(b,d) > Thres else 0	174	"psubusb "_DiffThres", %%mm3\n\t" // nonzero where abs(b,d) > Thres else 0
175	"pxor %%mm4, %%mm4\n\t"	175	"pxor %%mm4, %%mm4\n\t"
176	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(b,d) < Thres, else 00	176	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where abs(b,d) < Thres, else 00
177	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(b,d) > Thres, else 00	177	"pcmpeqb %%mm3, %%mm4\n\t" // here ff where abs(b,d) > Thres, else 00
178		178
179	"movq -2(%%"XBX"), %%mm0\n\t" // value a	179	"movq -2(%%"XDX"), %%mm0\n\t" // value a
180	"movq 2(%%"XBX", %%"XCX"), %%mm1\n\t" // value f	180	"movq 2(%%"XDX", %%"XCX"), %%mm1\n\t" // value f
181	"movq %%mm0, %%mm2\n\t"	181	"movq %%mm0, %%mm2\n\t"
182	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(a,f)	182	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(a,f)
183	"movq %%mm0, %%mm3\n\t"	183	"movq %%mm0, %%mm3\n\t"
184	"psubusb %%mm1, %%mm0\n\t"	184	"psubusb %%mm1, %%mm0\n\t"
185	"psubusb %%mm3, %%mm1\n\t"	185	"psubusb %%mm3, %%mm1\n\t"
186	"por %%mm1, %%mm0\n\t" // abs(a,f)	186	"por %%mm1, %%mm0\n\t" // abs(a,f)
187		187
188	"movq %%mm0, %%mm1\n\t"	188	"movq %%mm0, %%mm1\n\t"
189	"psubusb "MANGLE(DiffThres)", %%mm1\n\t" // nonzero where abs(a,f) > Thres else 0	189	"psubusb "_DiffThres", %%mm1\n\t" // nonzero where abs(a,f) > Thres else 0
190	"pxor %%mm3, %%mm3\n\t"	190	"pxor %%mm3, %%mm3\n\t"
191	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(a,f) < Thres, else 00	191	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(a,f) < Thres, else 00
192		192
Lines 205-226 Link Here
205	"por %%mm2, %%mm6\n\t"	205	"por %%mm2, %%mm6\n\t"
206	"por %%mm0, %%mm7\n\t"	206	"por %%mm0, %%mm7\n\t"
207		207
208	"pand "MANGLE(YMask)", %%mm5\n\t" // mask out chroma from here	208	"pand "_YMask", %%mm5\n\t" // mask out chroma from here
209	"pand "MANGLE(YMask)", %%mm6\n\t" // mask out chroma from here	209	"pand "_YMask", %%mm6\n\t" // mask out chroma from here
210	"pand "MANGLE(YMask)", %%mm7\n\t" // mask out chroma from here	210	"pand "_YMask", %%mm7\n\t" // mask out chroma from here
211		211
212	// b,e	212	// b,e
213	"movq (%%"XBX"), %%mm0\n\t" // value b from top	213	"movq (%%"XDX"), %%mm0\n\t" // value b from top
214	"movq (%%"XBX", %%"XCX"), %%mm1\n\t" // value e from bottom	214	"movq (%%"XDX", %%"XCX"), %%mm1\n\t" // value e from bottom
215	"movq %%mm0, %%mm2\n\t"	215	"movq %%mm0, %%mm2\n\t"
216	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(b,e)	216	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(b,e)
217	"movq %%mm0, %%mm3\n\t"	217	"movq %%mm0, %%mm3\n\t"
218	"psubusb %%mm1, %%mm0\n\t"	218	"psubusb %%mm1, %%mm0\n\t"
219	"psubusb %%mm3, %%mm1\n\t"	219	"psubusb %%mm3, %%mm1\n\t"
220	"por %%mm1, %%mm0\n\t" // abs(b,e)	220	"por %%mm1, %%mm0\n\t" // abs(b,e)
221		221
222	"movq %%mm0, %%mm1\n\t"	222	"movq %%mm0, %%mm1\n\t"
223	"psubusb "MANGLE(DiffThres)", %%mm1\n\t" // nonzero where abs(b,e) > Thres else 0	223	"psubusb "_DiffThres", %%mm1\n\t" // nonzero where abs(b,e) > Thres else 0
224	"pxor %%mm3, %%mm3\n\t"	224	"pxor %%mm3, %%mm3\n\t"
225	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(b,e) < Thres, else 00	225	"pcmpeqb %%mm3, %%mm1\n\t" // now ff where abs(b,e) < Thres, else 00
226		226
Lines 238-245 Link Here
238	"por %%mm0, %%mm7\n\t"	238	"por %%mm0, %%mm7\n\t"
239		239
240	// bob in any leftovers	240	// bob in any leftovers
241	"movq (%%"XBX"), %%mm0\n\t" // value b from top	241	"movq (%%"XDX"), %%mm0\n\t" // value b from top
242	"movq (%%"XBX", %%"XCX"), %%mm1\n\t" // value e from bottom	242	"movq (%%"XDX", %%"XCX"), %%mm1\n\t" // value e from bottom
243		243
244		244
245	// We will also calc here the max/min values to later limit comb	245	// We will also calc here the max/min values to later limit comb
Lines 271-277 Link Here
271	"por %%mm2, %%mm3\n\t" // abs diff	271	"por %%mm2, %%mm3\n\t" // abs diff
272	// pmaxub %%mm3, %%mm4 // top or bottom pixel moved most	272	// pmaxub %%mm3, %%mm4 // top or bottom pixel moved most
273	V_PMAXUB ("%%mm3", "%%mm4") // top or bottom pixel moved most	273	V_PMAXUB ("%%mm3", "%%mm4") // top or bottom pixel moved most
274	"psubusb "MANGLE(DiffThres)", %%mm3\n\t" // moved more than allowed? or goes to 0?	274	"psubusb "_DiffThres", %%mm3\n\t" // moved more than allowed? or goes to 0?
275	"pxor %%mm4, %%mm4\n\t"	275	"pxor %%mm4, %%mm4\n\t"
276	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where low motion, else high motion	276	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where low motion, else high motion
277		277
Lines 283-301 Link Here
283	V_PMAXUB ("%%mm6", "%%mm2")	283	V_PMAXUB ("%%mm6", "%%mm2")
284		284
285	"psubusb %%mm3, %%mm2\n\t" // maybe decrease it to 0000.. if no surround motion	285	"psubusb %%mm3, %%mm2\n\t" // maybe decrease it to 0000.. if no surround motion
286	// "movq %%mm2, "MANGLE(Min_Vals)"\n\t"	286	// "movq %%mm2, "_Min_Vals"\n\t"
287		287
288	"movq %%mm0, %%mm2\n\t"	288	"movq %%mm0, %%mm2\n\t"
289	V_PMAXUB ("%%mm2", "%%mm1")	289	V_PMAXUB ("%%mm2", "%%mm1")
290	// pminub %%mm6, %%mm2 // clip our current results so far to be below this	290	// pminub %%mm6, %%mm2 // clip our current results so far to be below this
291	V_PMINUB ("%%mm6", "%%mm2", "%%mm4")	291	V_PMINUB ("%%mm6", "%%mm2", "%%mm4")
292	"paddusb %%mm3, %%mm2\n\t" // maybe increase it to ffffff if no surround motion	292	"paddusb %%mm3, %%mm2\n\t" // maybe increase it to ffffff if no surround motion
293	// "movq %%mm2, "MANGLE(Max_Vals)"\n\t"	293	// "movq %%mm2, "_Max_Vals"\n\t"
294	#endif	294	#endif
295		295
296	"movq %%mm0, %%mm2\n\t"	296	"movq %%mm0, %%mm2\n\t"
297	// pavgb %%mm2, %%mm1 // avg(b,e)	297	// pavgb %%mm2, %%mm1 // avg(b,e)
298	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(b,e)	298	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(b,e)
299		299
300	"movq %%mm0, %%mm3\n\t"	300	"movq %%mm0, %%mm3\n\t"
301	"psubusb %%mm1, %%mm3\n\t"	301	"psubusb %%mm1, %%mm3\n\t"

Lines 1-17 Link Here

(-)xine-lib-1.1.1-old/src/libw32dll/wine/wrapper.S (-17 / +35 lines)
1	.section .data	1	.section .data
2	.globl caller_return
3	caller_return:	2	caller_return:
4	.long 0	3	.long 0
5	.globl report_entry
6	report_entry:
7	.long null_call
8	.globl report_ret
9	report_ret:
10	.long null_call
11	.global wrapper_target	4	.global wrapper_target
12	wrapper_target:	5	wrapper_target:
13	.long null_call	6	.long null_call
14		7
		8	#undef __i686 /* gcc define gets in our way */
		9	.section .gnu.linkonce.t.__i686.get_pc_thunk.bx,"ax",@progbits
		10	.globl __i686.get_pc_thunk.bx
		11	.hidden __i686.get_pc_thunk.bx
		12	.type __i686.get_pc_thunk.bx,@function
		13	__i686.get_pc_thunk.bx:
		14	movl (%esp), %ebx
		15	ret
		16
15	.section .text	17	.section .text
16	.globl null_call	18	.globl null_call
17	.type null_call, @function	19	.type null_call, @function
Lines 22-67 null_call: Link Here
22	.type wrapper, @function	24	.type wrapper, @function
23	.balign 16,0x90	25	.balign 16,0x90
24	wrapper:	26	wrapper:
		27	pushl $0
25	pusha # store registers (EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI)	28	pusha # store registers (EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI)
26	pushf # store flags	29	pushf # store flags
27		30
28	push %ebp # set up a stack frame	31	push %ebp # set up a stack frame
29	movl %esp, %ebp	32	movl %esp, %ebp
30		33
		34	call __i686.get_pc_thunk.bx
		35	addl $_GLOBAL_OFFSET_TABLE_, %ebx
		36
31	leal 4(%ebp), %eax # push flags addr	37	leal 4(%ebp), %eax # push flags addr
32	push %eax	38	push %eax
33	leal 8(%ebp), %eax # push registers addr	39	leal 8(%ebp), %eax # push registers addr
34	push %eax	40	push %eax
35		41
36	leal 40(%ebp), %edx	42	leal 44(%ebp), %edx
37	movl (%ebp), %eax	43	movl (%ebp), %eax
38	subl %edx, %eax	44	subl %edx, %eax
39	push %eax	45	push %eax
40	push %edx	46	push %edx
41		47
42	call *report_entry # report entry	48	call report_func@PLT # report entry
43		49
44	test %eax, %eax	50	test %eax, %eax
45	jnz .Ldone	51	jnz .Ldone
46		52
		53	movl 44(%ebp), %eax # switch return addresses
		54	movl %eax, caller_return@GOTOFF(%ebx)
		55	leal .Lwrapper_return@GOTOFF(%ebx), %eax
		56	movl %eax, 40(%ebp)
		57
		58	movl wrapper_target@GOTOFF(%ebx), %eax
		59	mov %eax, 40(%ebp) # wrapper_target should return at .Lwrapper_return
		60
47	leave # restore %esp, %ebp	61	leave # restore %esp, %ebp
48	popf # restore flags	62	popf # restore flags
49	popa # restore registers	63	popa # restore registers
50		64
51	popl caller_return # switch return addresses	65	ret # fake 'return' to wrapper_target actually
52	pushl $.Lwrapper_return
53
54	jmp *wrapper_target # wrapper_target should return at .Lwrapper_return
55		66
56	.balign 16, 0x90	67	.balign 16, 0x90
57	.Lwrapper_return:	68	.Lwrapper_return:
58	pushl caller_return # restore the original return address	69	pushl $0
59	pusha # more for reference sake here	70	pusha # more for reference sake here
60	pushf	71	pushf
61		72
62	push %ebp # set up a stack frame	73	push %ebp # set up a stack frame
63	movl %esp, %ebp	74	movl %esp, %ebp
64		75
		76	call __i686.get_pc_thunk.bx
		77	addl $_GLOBAL_OFFSET_TABLE_, %ebx
		78	movl caller_return@GOTOFF(%ebx), %eax
		79	movl %eax, 40(%ebp) # restore the original return address
		80
65	leal 4(%ebp), %eax # push flags addr	81	leal 4(%ebp), %eax # push flags addr
66	push %eax	82	push %eax
67	leal 8(%ebp), %eax # push registers addr	83	leal 8(%ebp), %eax # push registers addr
Lines 73-83 wrapper: Link Here
73	push %eax	89	push %eax
74	push %edx	90	push %edx
75		91
76	call *report_ret # report the return information (same args)	92	call report_func_ret@PLT# report the return information (same args)
77	.Ldone:	93	.Ldone:
78		94
79	leave	95	leave
80	popf	96	popf
81	popa	97	popa
82	ret	98	ret
83		99
		100	.section .note.GNU-stack,"",@progbits
		101

Lines 18-24 Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/SearchLoopBottom.inc (-21 / +32 lines)
18	// Use the best weave if diffs less than 10 as that	18	// Use the best weave if diffs less than 10 as that
19	// means the image is still or moving cleanly	19	// means the image is still or moving cleanly
20	// if there is motion we will clip which will catch anything	20	// if there is motion we will clip which will catch anything
21	"psubusb "MANGLE(FOURS)", %%mm7\n\t" // sets bits to zero if weave diff < 4	21	"psubusb "_FOURS", %%mm7\n\t" // sets bits to zero if weave diff < 4
22	"pxor %%mm0, %%mm0\n\t"	22	"pxor %%mm0, %%mm0\n\t"
23	"pcmpeqb %%mm0, %%mm7\n\t" // all ff where weave better, else 00	23	"pcmpeqb %%mm0, %%mm7\n\t" // all ff where weave better, else 00
24	"pcmpeqb %%mm7, %%mm0\n\t" // all ff where bob better, else 00	24	"pcmpeqb %%mm7, %%mm0\n\t" // all ff where bob better, else 00
Lines 28-37 Link Here
28	#else	28	#else
29	// Use the better of bob or weave	29	// Use the better of bob or weave
30	// pminub mm4, TENS // the most we care about	30	// pminub mm4, TENS // the most we care about
31	V_PMINUB ("%%mm4", MANGLE(TENS), "%%mm0") // the most we care about	31	V_PMINUB ("%%mm4", _TENS, "%%mm0") // the most we care about
32		32
33	"psubusb %%mm4, %%mm7\n\t" // foregive that much from weave est?	33	"psubusb %%mm4, %%mm7\n\t" // foregive that much from weave est?
34	"psubusb "MANGLE(FOURS)", %%mm7\n\t" // bias it a bit toward weave	34	"psubusb "_FOURS", %%mm7\n\t" // bias it a bit toward weave
35	"pxor %%mm0, %%mm0\n\t"	35	"pxor %%mm0, %%mm0\n\t"
36	"pcmpeqb %%mm0, %%mm7\n\t" // all ff where weave better, else 00	36	"pcmpeqb %%mm0, %%mm7\n\t" // all ff where weave better, else 00
37	"pcmpeqb %%mm7, %%mm0\n\t" // all ff where bob better, else 00	37	"pcmpeqb %%mm7, %%mm0\n\t" // all ff where bob better, else 00
Lines 42-50 Link Here
42		42
43		43
44	// pminub mm0, Max_Vals // but clip to catch the stray error	44	// pminub mm0, Max_Vals // but clip to catch the stray error
45	// V_PMINUB ("%%mm0", MANGLE(Max_Vals), "%%mm1") // but clip to catch the stray error	45	// V_PMINUB ("%%mm0", _Max_Vals, "%%mm1") // but clip to catch the stray error
46	// pmaxub mm0, Min_Vals	46	// pmaxub mm0, Min_Vals
47	// V_PMAXUB ("%%mm0", MANGLE(Min_Vals))	47	// V_PMAXUB ("%%mm0", _Min_Vals)
48		48
49	#endif	49	#endif
50		50
Lines 53-80 Link Here
53		53
54	#ifdef USE_VERTICAL_FILTER	54	#ifdef USE_VERTICAL_FILTER
55	"movq %%mm0, %%mm1\n\t"	55	"movq %%mm0, %%mm1\n\t"
56	// pavgb mm0, qword ptr["XBX"]	56	// pavgb mm0, qword ptr["XDX"]
57	V_PAVGB ("%%mm0", "(%%"XBX")", "%%mm2", MANGLE(ShiftMask))	57	V_PAVGB ("%%mm0", "(%%"XDX")", "%%mm2", _ShiftMask)
58	// movntq qword ptr["XAX"+"XDX"], mm0	58	// movntq qword ptr["XAX"+"olddx"], mm0
59	V_MOVNTQ ("(%"XAX", %%"XDX")", "%%mm0")	59	ADDX" "_olddx", %%"XAX"\n\t"
60	// pavgb mm1, qword ptr["XBX"+"XCX"]	60	V_MOVNTQ ("(%%"XAX")", "%%mm0")
61	V_PAVGB ("%%mm1", "(%%"XBX", %%"XCX")", "%%mm2", MANGLE(ShiftMask))	61	// pavgb mm1, qword ptr["XDX"+"XCX"]
62	"addq "_dst_pitchw", %%"XBX	62	V_PAVGB ("%%mm1", "(%%"XDX", %%"XCX")", "%%mm2", _ShiftMask)
63	// movntq qword ptr["XAX"+"XDX"], mm1	63	"addq "_dst_pitchw", %%"XDX
64	V_MOVNTQ ("(%%"XAX", %%"XDX")", "%%mm1")	64	// movntq qword ptr["XAX"+"olddx"], mm1
		65	V_MOVNTQ ("(%%"XAX")", "%%mm1")
65	#else	66	#else
66		67
67	// movntq qword ptr["XAX"+"XDX"], mm0	68	// movntq qword ptr["XAX"+"olddx"], mm0
68	V_MOVNTQ ("(%%"XAX", %%"XDX")", "%%mm0")	69	ADDX" "_olddx", %%"XAX"\n\t"
		70	V_MOVNTQ ("(%%"XAX")", "%%mm0")
69	#endif	71	#endif
70		72
71	LEAX" 8(%%"XDX"), %%"XDX"\n\t" // bump offset pointer	73	ADDX" $8, "_olddx"\n\t" // bump offset pointer
72	CMPX" "_Last8", %%"XDX"\n\t" // done with line?	74	MOVX" "_olddx", %%"XAX"\n\t"
		75	CMPX" "_Last8", %%"XAX"\n\t" // done with line?
73	"jb 1b\n\t" // y	76	"jb 1b\n\t" // y
74	#endif	77	#endif
75		78
76	MOVX" "_oldbx", %%"XBX"\n\t"
77
78	: /* no outputs */	79	: /* no outputs */
79		80
80	: "m"(pBob),	81	: "m"(pBob),
Lines 85-91 Link Here
85	"m"(pSrc),	86	"m"(pSrc),
86	"m"(pSrcP),	87	"m"(pSrcP),
87	"m"(pBobP),	88	"m"(pBobP),
88	"m"(oldbx)	89	"m"(olddx),
		90	"m"(UVMask),
		91	"m"(ShiftMask),
		92	"m"(FOURS),
		93	"m"(TENS),
		94	"m"(Max_Vals),
		95	"m"(Min_Vals),
		96	"m"(YMask),
		97	"m"(Max_Mov),
		98	"m"(ONES),
		99	"m"(DiffThres)
89		100
90	: XAX, XCX, XDX, XSI, XDI,	101	: XAX, XCX, XDX, XSI, XDI,
91	#ifdef ARCH_X86	102	#ifdef ARCH_X86

Lines 14-35 Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/WierdBob.inc (-20 / +20 lines)
14	// selected for the smallest of abs(a,f), abs(c,d), or abs(b,e), etc.	14	// selected for the smallest of abs(a,f), abs(c,d), or abs(b,e), etc.
15		15
16	// a,f	16	// a,f
17	"movq -2(%%"XBX"), %%mm0\n\t" // value a from top left	17	"movq -2(%%"XDX"), %%mm0\n\t" // value a from top left
18	"movq 2(%%"XBX", %%"XCX"), %%mm1\n\t" // value f from bottom right	18	"movq 2(%%"XDX", %%"XCX"), %%mm1\n\t" // value f from bottom right
19	"movq %%mm0, %%mm6\n\t"	19	"movq %%mm0, %%mm6\n\t"
20	// pavgb %%mm6, %%mm1 // avg(a,f), also best so far	20	// pavgb %%mm6, %%mm1 // avg(a,f), also best so far
21	V_PAVGB ("%%mm6", "%%mm1", "%%mm7", MANGLE(ShiftMask)) // avg(a,f), also best so far	21	V_PAVGB ("%%mm6", "%%mm1", "%%mm7", _ShiftMask) // avg(a,f), also best so far
22	"movq %%mm0, %%mm7\n\t"	22	"movq %%mm0, %%mm7\n\t"
23	"psubusb %%mm1, %%mm7\n\t"	23	"psubusb %%mm1, %%mm7\n\t"
24	"psubusb %%mm0, %%mm1\n\t"	24	"psubusb %%mm0, %%mm1\n\t"
25	"por %%mm1, %%mm7\n\t" // abs diff, also best so far	25	"por %%mm1, %%mm7\n\t" // abs diff, also best so far
26		26
27	// c,d	27	// c,d
28	"movq 2(%%"XBX"), %%mm0\n\t" // value a from top left	28	"movq 2(%%"XDX"), %%mm0\n\t" // value a from top left
29	"movq -2(%%"XBX", %%"XCX"), %%mm1\n\t" // value f from bottom right	29	"movq -2(%%"XDX", %%"XCX"), %%mm1\n\t" // value f from bottom right
30	"movq %%mm0, %%mm2\n\t"	30	"movq %%mm0, %%mm2\n\t"
31	// pavgb %%mm2, %%mm1 // avg(c,d)	31	// pavgb %%mm2, %%mm1 // avg(c,d)
32	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(c,d)	32	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(c,d)
33	"movq %%mm0, %%mm3\n\t"	33	"movq %%mm0, %%mm3\n\t"
34	"psubusb %%mm1, %%mm3\n\t"	34	"psubusb %%mm1, %%mm3\n\t"
35	"psubusb %%mm0, %%mm1\n\t"	35	"psubusb %%mm0, %%mm1\n\t"
Lines 49-63 Link Here
49		49
50	"por %%mm2, %%mm6\n\t" // and merge new & old vals keeping best	50	"por %%mm2, %%mm6\n\t" // and merge new & old vals keeping best
51	"por %%mm1, %%mm7\n\t"	51	"por %%mm1, %%mm7\n\t"
52	"por "MANGLE(UVMask)", %%mm7\n\t" // but we know chroma is worthless so far	52	"por "_UVMask", %%mm7\n\t" // but we know chroma is worthless so far
53	"pand "MANGLE(YMask)", %%mm5\n\t" // mask out chroma from here also	53	"pand "_YMask", %%mm5\n\t" // mask out chroma from here also
54		54
55	// j,n	55	// j,n
56	"movq -4(%%"XBX"), %%mm0\n\t" // value j from top left	56	"movq -4(%%"XDX"), %%mm0\n\t" // value j from top left
57	"movq 4(%%"XBX", %%"XCX"), %%mm1\n\t" // value n from bottom right	57	"movq 4(%%"XDX", %%"XCX"), %%mm1\n\t" // value n from bottom right
58	"movq %%mm0, %%mm2\n\t"	58	"movq %%mm0, %%mm2\n\t"
59	// pavgb %%mm2, %%mm1 // avg(j,n)	59	// pavgb %%mm2, %%mm1 // avg(j,n)
60	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(j,n)	60	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(j,n)
61	"movq %%mm0, %%mm3\n\t"	61	"movq %%mm0, %%mm3\n\t"
62	"psubusb %%mm1, %%mm3\n\t"	62	"psubusb %%mm1, %%mm3\n\t"
63	"psubusb %%mm0, %%mm1\n\t"	63	"psubusb %%mm0, %%mm1\n\t"
Lines 79-89 Link Here
79	"por %%mm1, %%mm7\n\t" // "	79	"por %%mm1, %%mm7\n\t" // "
80		80
81	// k, m	81	// k, m
82	"movq 4(%%"XBX"), %%mm0\n\t" // value k from top right	82	"movq 4(%%"XDX"), %%mm0\n\t" // value k from top right
83	"movq -4(%%"XBX", %%"XCX"), %%mm1\n\t" // value n from bottom left	83	"movq -4(%%"XDX", %%"XCX"), %%mm1\n\t" // value n from bottom left
84	"movq %%mm0, %%mm4\n\t"	84	"movq %%mm0, %%mm4\n\t"
85	// pavgb %%mm4, %%mm1 // avg(k,m)	85	// pavgb %%mm4, %%mm1 // avg(k,m)
86	V_PAVGB ("%%mm4", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(k,m)	86	V_PAVGB ("%%mm4", "%%mm1", "%%mm3", _ShiftMask) // avg(k,m)
87		87
88	"movq %%mm0, %%mm3\n\t"	88	"movq %%mm0, %%mm3\n\t"
89	"psubusb %%mm1, %%mm3\n\t"	89	"psubusb %%mm1, %%mm3\n\t"
Lines 108-115 Link Here
108	"por %%mm1, %%mm7\n\t" // "	108	"por %%mm1, %%mm7\n\t" // "
109		109
110	// b,e	110	// b,e
111	"movq (%%"XBX"), %%mm0\n\t" // value b from top	111	"movq (%%"XDX"), %%mm0\n\t" // value b from top
112	"movq (%%"XBX", %%"XCX"), %%mm1\n\t" // value e from bottom	112	"movq (%%"XDX", %%"XCX"), %%mm1\n\t" // value e from bottom
113		113
114	// We will also calc here the max/min values to later limit comb	114	// We will also calc here the max/min values to later limit comb
115	// so the max excursion will not exceed the Max_Comb constant	115	// so the max excursion will not exceed the Max_Comb constant
Lines 140-146 Link Here
140	"por %%mm2, %%mm3\n\t" // abs diff	140	"por %%mm2, %%mm3\n\t" // abs diff
141	// pmaxub %%mm3, %%mm4 // top or bottom pixel moved most	141	// pmaxub %%mm3, %%mm4 // top or bottom pixel moved most
142	V_PMAXUB ("%%mm3", "%%mm4") // top or bottom pixel moved most	142	V_PMAXUB ("%%mm3", "%%mm4") // top or bottom pixel moved most
143	"psubusb "MANGLE(Max_Mov)", %%mm3\n\t" // moved more than allowed? or goes to 0?	143	"psubusb "_Max_Mov", %%mm3\n\t" // moved more than allowed? or goes to 0?
144	"pxor %%mm4, %%mm4\n\t"	144	"pxor %%mm4, %%mm4\n\t"
145	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where low motion, else high motion	145	"pcmpeqb %%mm4, %%mm3\n\t" // now ff where low motion, else high motion
146		146
Lines 152-170 Link Here
152	V_PMAXUB ("%%mm6", "%%mm2")	152	V_PMAXUB ("%%mm6", "%%mm2")
153		153
154	"psubusb %%mm3, %%mm2\n\t" // maybe decrease it to 0000.. if no surround motion	154	"psubusb %%mm3, %%mm2\n\t" // maybe decrease it to 0000.. if no surround motion
155	// "movq %%mm2, "MANGLE(Min_Vals)"\n\t"	155	// "movq %%mm2, "_Min_Vals"\n\t"
156		156
157	"movq %%mm0, %%mm2\n\t"	157	"movq %%mm0, %%mm2\n\t"
158	V_PMAXUB ("%%mm2", "%%mm1")	158	V_PMAXUB ("%%mm2", "%%mm1")
159	// pminub %%mm6, %%mm2 // clip our current results so far to be below this	159	// pminub %%mm6, %%mm2 // clip our current results so far to be below this
160	V_PMINUB ("%%mm6", "%%mm2", "%%mm4")	160	V_PMINUB ("%%mm6", "%%mm2", "%%mm4")
161	"paddusb %%mm3, %%mm2\n\t" // maybe increase it to ffffff if no surround motion	161	"paddusb %%mm3, %%mm2\n\t" // maybe increase it to ffffff if no surround motion
162	// "movq %%mm2, "MANGLE(Max_Vals)"\n\t"	162	// "movq %%mm2, "_Max_Vals"\n\t"
163	#endif	163	#endif
164		164
165	"movq %%mm0, %%mm2\n\t"	165	"movq %%mm0, %%mm2\n\t"
166	// pavgb %%mm2, %%mm1 // avg(b,e)	166	// pavgb %%mm2, %%mm1 // avg(b,e)
167	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", MANGLE(ShiftMask)) // avg(b,e)	167	V_PAVGB ("%%mm2", "%%mm1", "%%mm3", _ShiftMask) // avg(b,e)
168		168
169	"movq %%mm0, %%mm3\n\t"	169	"movq %%mm0, %%mm3\n\t"
170	"psubusb %%mm1, %%mm3\n\t"	170	"psubusb %%mm1, %%mm3\n\t"

Lines 114-120 Link Here

(-)xine-lib-1.1.1-old/src/post/deinterlace/plugins/tomsmocomp/tomsmocompmacros.h (-6 / +6 lines)
114	"por %%xmm0, %%xmm5\n\t" /* and merge new & old vals */ \	114	"por %%xmm0, %%xmm5\n\t" /* and merge new & old vals */ \
115	"por %%xmm2, %%xmm7\n\t"	115	"por %%xmm2, %%xmm7\n\t"
116		116
117	#define RESET_CHROMA "por "MANGLE(UVMask)", %%xmm7\n\t"	117	#define RESET_CHROMA "por "_UVMask", %%xmm7\n\t"
118		118
119	#else // ifdef IS_SSE2	119	#else // ifdef IS_SSE2
120		120
Lines 126-132 Link Here
126	"psubusb %%mm1, %%mm2\n\t" \	126	"psubusb %%mm1, %%mm2\n\t" \
127	"psubusb %%mm0, %%mm3\n\t" \	127	"psubusb %%mm0, %%mm3\n\t" \
128	"por %%mm3, %%mm2\n\t" \	128	"por %%mm3, %%mm2\n\t" \
129	V_PAVGB ("%%mm0", "%%mm1", "%%mm3", MANGLE(ShiftMask)) /* avg of 2 pixels */ \	129	V_PAVGB ("%%mm0", "%%mm1", "%%mm3", _ShiftMask) /* avg of 2 pixels */ \
130	"movq %%mm2, %%mm3\n\t" /* another copy of our our weights */ \	130	"movq %%mm2, %%mm3\n\t" /* another copy of our our weights */ \
131	"pxor %%mm1, %%mm1\n\t" \	131	"pxor %%mm1, %%mm1\n\t" \
132	"psubusb %%mm7, %%mm3\n\t" /* nonzero where old weights lower, else 0 */ \	132	"psubusb %%mm7, %%mm3\n\t" /* nonzero where old weights lower, else 0 */ \
Lines 144-157 Link Here
144	"movq "PADDR2A", %%mm1\n\t" /* our pixel2 value */ \	144	"movq "PADDR2A", %%mm1\n\t" /* our pixel2 value */ \
145	"movq "PADDR1B", %%mm2\n\t" /* our 4 pixels */ \	145	"movq "PADDR1B", %%mm2\n\t" /* our 4 pixels */ \
146	"movq "PADDR2B", %%mm3\n\t" /* our pixel2 value */ \	146	"movq "PADDR2B", %%mm3\n\t" /* our pixel2 value */ \
147	V_PAVGB("%%mm0", "%%mm2", "%%mm2", MANGLE(ShiftMask)) \	147	V_PAVGB("%%mm0", "%%mm2", "%%mm2", _ShiftMask) \
148	V_PAVGB("%%mm1", "%%mm3", "%%mm3", MANGLE(ShiftMask)) \	148	V_PAVGB("%%mm1", "%%mm3", "%%mm3", _ShiftMask) \
149	"movq %%mm0, %%mm2\n\t" /* another copy of our pixel1 value */ \	149	"movq %%mm0, %%mm2\n\t" /* another copy of our pixel1 value */ \
150	"movq %%mm1, %%mm3\n\t" /* another copy of our pixel1 value */ \	150	"movq %%mm1, %%mm3\n\t" /* another copy of our pixel1 value */ \
151	"psubusb %%mm1, %%mm2\n\t" \	151	"psubusb %%mm1, %%mm2\n\t" \
152	"psubusb %%mm0, %%mm3\n\t" \	152	"psubusb %%mm0, %%mm3\n\t" \
153	"por %%mm3, %%mm2\n\t" \	153	"por %%mm3, %%mm2\n\t" \
154	V_PAVGB("%%mm0", "%%mm1", "%%mm3", MANGLE(ShiftMask)) /* avg of 2 pixels */ \	154	V_PAVGB("%%mm0", "%%mm1", "%%mm3", _ShiftMask) /* avg of 2 pixels */ \
155	"movq %%mm2, %%mm3\n\t" /* another copy of our our weights */ \	155	"movq %%mm2, %%mm3\n\t" /* another copy of our our weights */ \
156	"pxor %%mm1, %%mm1\n\t" \	156	"pxor %%mm1, %%mm1\n\t" \
157	"psubusb %%mm7, %%mm3\n\t" /* nonzero where old weights lower, else 0 */ \	157	"psubusb %%mm7, %%mm3\n\t" /* nonzero where old weights lower, else 0 */ \
Lines 164-170 Link Here
164	"por %%mm0, %%mm5\n\t" /* and merge new & old vals */ \	164	"por %%mm0, %%mm5\n\t" /* and merge new & old vals */ \
165	"por %%mm2, %%mm7\n\t"	165	"por %%mm2, %%mm7\n\t"
166		166
167	#define RESET_CHROMA "por "MANGLE(UVMask)", %%mm7\n\t"	167	#define RESET_CHROMA "por "_UVMask", %%mm7\n\t"
168		168
169	#endif	169	#endif
170		170