xvid.c

mpeg4 video codec mpeg4 video codec

		interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm;
		interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm;
		interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm;

		/* Quantization */
		quant_mpeg_intra = quant_mpeg_intra_xmm;
		quant_mpeg_inter = quant_mpeg_inter_xmm;

		dequant_h263_intra = dequant_h263_intra_xmm;
		dequant_h263_inter = dequant_h263_inter_xmm;

		/* Buffer transfer */
		transfer_8to16sub2 = transfer_8to16sub2_xmm;
		transfer_8to16sub2ro = transfer_8to16sub2ro_xmm;

		/* Colorspace transformation */
		yv12_to_yv12  = yv12_to_yv12_xmm;
		yuyv_to_yv12  = yuyv_to_yv12_xmm;
		uyvy_to_yv12  = uyvy_to_yv12_xmm;

		/* ME functions */
		sad16 = sad16_xmm;
		sad8  = sad8_xmm;
		sad16bi = sad16bi_xmm;
		sad8bi  = sad8bi_xmm;
		dev16 = dev16_xmm;
		sad16v	 = sad16v_xmm;
	}

	if ((cpu_flags & XVID_CPU_3DNOW)) {

		/* Interpolation */
		interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
		interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
		interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;

		interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dn;
		interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dn;
		interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dn;
	}

	if ((cpu_flags & XVID_CPU_3DNOWEXT)) {

		/* Buffer transfer */
		transfer_8to16copy =  transfer_8to16copy_3dne;
		transfer_16to8copy = transfer_16to8copy_3dne;
		transfer_8to16sub =  transfer_8to16sub_3dne;
		transfer_8to16subro =  transfer_8to16subro_3dne;
		transfer_16to8add = transfer_16to8add_3dne;
		transfer8x8_copy = transfer8x8_copy_3dne;
		transfer8x4_copy = transfer8x4_copy_3dne;

		if ((cpu_flags & XVID_CPU_MMXEXT)) {
			/* Inverse DCT */
			idct =  idct_3dne;

			/* Buffer transfer */
			transfer_8to16sub2 =  transfer_8to16sub2_3dne;

			/* Interpolation */
			interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne;
			interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne;
			interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;

			interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dne;
			interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dne;
			interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dne;

			/* Quantization */
			quant_h263_intra = quant_h263_intra_3dne;		/* cmov only */
			quant_h263_inter = quant_h263_inter_3dne;
			dequant_mpeg_intra = dequant_mpeg_intra_3dne;	/* cmov only */
			dequant_mpeg_inter = dequant_mpeg_inter_3dne;
			dequant_h263_intra = dequant_h263_intra_3dne;
			dequant_h263_inter = dequant_h263_inter_3dne;

			/* ME functions */
			calc_cbp = calc_cbp_3dne;

			sad16 = sad16_3dne;
			sad8 = sad8_3dne;
			sad16bi = sad16bi_3dne;
			sad8bi = sad8bi_3dne;
			dev16 = dev16_3dne;
		}
	}

	if ((cpu_flags & XVID_CPU_SSE2)) {

		calc_cbp = calc_cbp_sse2;

		/* Quantization */
		quant_h263_intra   = quant_h263_intra_sse2;
		quant_h263_inter   = quant_h263_inter_sse2;
		dequant_h263_intra = dequant_h263_intra_sse2;
		dequant_h263_inter = dequant_h263_inter_sse2;

		/* SAD operators */
		sad16    = sad16_sse2;
		dev16    = dev16_sse2;

		/* DCT operators */
		fdct = fdct_sse2_skal;
    /* idct = idct_sse2_skal; */   /* Is now IEEE1180 and Walken compliant. Disabled until fully tested. */

		/* postprocessing */
		image_brightness = image_brightness_sse2;
	}
#endif /* ARCH_IS_IA32 */

#if defined(ARCH_IS_IA64)
	if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
	  idct_ia64_init();
	  fdct = fdct_ia64;
	  idct = idct_ia64;   /*not yet working, crashes */
	  interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
	  interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
	  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
	  sad16 = sad16_ia64;
	  sad16bi = sad16bi_ia64;
	  sad8 = sad8_ia64;
	  dev16 = dev16_ia64;
/*	  Halfpel8_Refine = Halfpel8_Refine_ia64; */
	  quant_h263_intra = quant_h263_intra_ia64;
	  quant_h263_inter = quant_h263_inter_ia64;
	  dequant_h263_intra = dequant_h263_intra_ia64;
	  dequant_h263_inter = dequant_h263_inter_ia64;
	  transfer_8to16copy = transfer_8to16copy_ia64;
	  transfer_16to8copy = transfer_16to8copy_ia64;
	  transfer_8to16sub = transfer_8to16sub_ia64;
	  transfer_8to16sub2 = transfer_8to16sub2_ia64;
	  transfer_16to8add = transfer_16to8add_ia64;
	  transfer8x8_copy = transfer8x8_copy_ia64;
	}
#endif

#if defined(ARCH_IS_PPC)
	if ((cpu_flags & XVID_CPU_ALTIVEC)) {
          /* sad operators */
		  sad16 = sad16_altivec_c;
		  sad16bi = sad16bi_altivec_c;
		  sad8 = sad8_altivec_c;
		  dev16 = dev16_altivec_c;
          
          sse8_16bit = sse8_16bit_altivec_c;
          
          /* mem transfer */
          transfer_8to16copy = transfer_8to16copy_altivec_c;
          transfer_16to8copy = transfer_16to8copy_altivec_c;
          transfer_8to16sub = transfer_8to16sub_altivec_c;
          transfer_8to16subro = transfer_8to16subro_altivec_c;
          transfer_8to16sub2 = transfer_8to16sub2_altivec_c;
          transfer_16to8add = transfer_16to8add_altivec_c;
          transfer8x8_copy = transfer8x8_copy_altivec_c;
           
          /* Inverse DCT */
          idct = idct_altivec_c;
          
          /* Interpolation */
          interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c;
          interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c;
          interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c;
		  
          interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c;
          interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c;
		  
		  interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c;
		  interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c;
		  interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c;
		  interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c;
          
          /* Colorspace conversion */
          bgra_to_yv12 = bgra_to_yv12_altivec_c;
          abgr_to_yv12 = abgr_to_yv12_altivec_c;
          rgba_to_yv12 = rgba_to_yv12_altivec_c;
          argb_to_yv12 = argb_to_yv12_altivec_c;
          
          yuyv_to_yv12 = yuyv_to_yv12_altivec_c;
          uyvy_to_yv12 = uyvy_to_yv12_altivec_c;
          
          yv12_to_yuyv = yv12_to_yuyv_altivec_c;
          yv12_to_uyvy = yv12_to_uyvy_altivec_c;
          
          /* Quantization */
          quant_h263_intra = quant_h263_intra_altivec_c;
          quant_h263_inter = quant_h263_inter_altivec_c;
          dequant_h263_intra = dequant_h263_intra_altivec_c;
          dequant_h263_inter = dequant_h263_inter_altivec_c;

		  dequant_mpeg_intra = dequant_mpeg_intra_altivec_c;
		  dequant_mpeg_inter = dequant_mpeg_inter_altivec_c;
		  
		  /* Qpel stuff */
		  xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C;
		  xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C;
        }
#endif

#if defined(ARCH_IS_X86_64)
	/* For now, only XVID_CPU_ASM is looked for, so user can still
	 * disable asm usage the usual way. When Intel EMT64 cpus will
	 * be out, maybe we'll have to check more precisely what cpu
	 * features there really are. */
	if (cpu_flags & XVID_CPU_ASM) {
		/* SIMD state flusher */
		emms = emms_3dn;

		/* DCT operators */
		fdct = fdct_skal_x86_64;
		idct = idct_x86_64;

		/* SAD operators */
		sad16      = sad16_x86_64;
		sad8       = sad8_x86_64;
		sad16bi    = sad16bi_x86_64;
		sad8bi     = sad8bi_x86_64;
		dev16      = dev16_x86_64;
		sad16v	   = sad16v_x86_64;
		sse8_16bit = sse8_16bit_x86_64;
		sse8_8bit  = sse8_8bit_x86_64;

		/* Interpolation operators */
		interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_x86_64;
		interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_x86_64;
		interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_x86_64;

		interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_x86_64;
		interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_x86_64;
		interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_x86_64;
		interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_x86_64;

		interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_x86_64;
		interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_x86_64;

		interpolate8x8_avg2 = interpolate8x8_avg2_x86_64;
		interpolate8x8_avg4 = interpolate8x8_avg4_x86_64;

		/* Quantization related functions */
		quant_h263_intra   = quant_h263_intra_x86_64;
		quant_h263_inter   = quant_h263_inter_x86_64;
		dequant_h263_intra = dequant_h263_intra_x86_64;
		dequant_h263_inter = dequant_h263_inter_x86_64;
		quant_mpeg_intra   = quant_mpeg_intra_x86_64;
		quant_mpeg_inter   = quant_mpeg_inter_x86_64;
		dequant_mpeg_intra   = dequant_mpeg_intra_x86_64;
		dequant_mpeg_inter   = dequant_mpeg_inter_x86_64;

		/* Block related functions */
		transfer_8to16copy  = transfer_8to16copy_x86_64;
		transfer_16to8copy  = transfer_16to8copy_x86_64;
		transfer_8to16sub   = transfer_8to16sub_x86_64;
		transfer_8to16subro = transfer_8to16subro_x86_64;
		transfer_8to16sub2  = transfer_8to16sub2_x86_64;
		transfer_8to16sub2ro= transfer_8to16sub2ro_x86_64;
		transfer_16to8add   = transfer_16to8add_x86_64;
		transfer8x8_copy    = transfer8x8_copy_x86_64;

		/* Qpel stuff */
		xvid_QP_Funcs = &xvid_QP_Funcs_x86_64;
		xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_x86_64;

		/* Interlacing Functions */
		MBFieldTest = MBFieldTest_x86_64;
	}
#endif

#if defined(_DEBUG)
    xvid_debug = init->debug;
#endif

    return(0);
}


static int
xvid_gbl_info(xvid_gbl_info_t * info)
{
	if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */
		return XVID_ERR_VERSION;

	info->actual_version = XVID_VERSION;
	info->build = "xvid-1.1.0";
	info->cpu_flags = detect_cpu_flags();

#if defined(_SMP) && defined(WIN32)
	info->num_threads = pthread_num_processors_np();;
#else
	info->num_threads = 0;
#endif

	return 0;
}


static int
xvid_gbl_convert(xvid_gbl_convert_t* convert)
{
	int width;
	int height;
	int width2;
	int height2;
	IMAGE img;

	if (XVID_VERSION_MAJOR(convert->version) != 1)   /* v1.x.x */
	      return XVID_ERR_VERSION;

#if 0
	const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);
#endif
	width = convert->width;
	height = convert->height;
	width2 = convert->width/2;
	height2 = convert->height/2;

	switch (convert->input.csp & ~XVID_CSP_VFLIP)
	{
		case XVID_CSP_YV12 :
			img.y = convert->input.plane[0];
			img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height;
			img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2;
			image_output(&img, width, height, width,
						(uint8_t**)convert->output.plane, convert->output.stride,
						convert->output.csp, convert->interlacing);
			break;

		default :
			return XVID_ERR_FORMAT;
	}


	emms();
	return 0;
}

/*****************************************************************************
 * XviD Global Entry point
 *
 * Well this function initialize all internal function pointers according
 * to the CPU features forced by the library client or autodetected (depending
 * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
 * image colorspace transformation tables.
 *
 ****************************************************************************/


int
xvid_global(void *handle,
		  int opt,
		  void *param1,
		  void *param2)
{
	switch(opt)
	{
		case XVID_GBL_INIT :
			return xvid_gbl_init((xvid_gbl_init_t*)param1);

        case XVID_GBL_INFO :
            return xvid_gbl_info((xvid_gbl_info_t*)param1);

		case XVID_GBL_CONVERT :
			return xvid_gbl_convert((xvid_gbl_convert_t*)param1);

		default :
			return XVID_ERR_FAIL;
	}
}

/*****************************************************************************
 * XviD Native decoder entry point
 *
 * This function is just a wrapper to all the option cases.
 *
 * Returned values : XVID_ERR_FAIL when opt is invalid
 *                   else returns the wrapped function result
 *
 ****************************************************************************/

int
xvid_decore(void *handle,
			int opt,
			void *param1,
			void *param2)
{
	switch (opt) {
	case XVID_DEC_CREATE:
		return decoder_create((xvid_dec_create_t *) param1);

	case XVID_DEC_DESTROY:
		return decoder_destroy((DECODER *) handle);

	case XVID_DEC_DECODE:
		return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2);

	default:
		return XVID_ERR_FAIL;
	}
}


/*****************************************************************************
 * XviD Native encoder entry point
 *
 * This function is just a wrapper to all the option cases.
 *
 * Returned values : XVID_ERR_FAIL when opt is invalid
 *                   else returns the wrapped function result
 *
 ****************************************************************************/

int
xvid_encore(void *handle,
			int opt,
			void *param1,
			void *param2)
{
	switch (opt) {
	case XVID_ENC_ENCODE:

		return enc_encode((Encoder *) handle,
							  (xvid_enc_frame_t *) param1,
							  (xvid_enc_stats_t *) param2);

	case XVID_ENC_CREATE:
		return enc_create((xvid_enc_create_t *) param1);

	case XVID_ENC_DESTROY:
		return enc_destroy((Encoder *) handle);

	default:
		return XVID_ERR_FAIL;
	}
}