IJG R6b with x86SIMD V1.02
Independent JPEG Group's JPEG software release 6b with x86 SIMD extension for IJG JPEG library version 1.02
This commit is contained in:
MIYASAKA Masaru
DRC
216
jdct.h
216
jdct.h
@@ -5,6 +5,13 @@
|
||||
* This file is part of the Independent JPEG Group's software.
|
||||
* For conditions of distribution and use, see the accompanying README file.
|
||||
*
|
||||
* ---------------------------------------------------------------------
|
||||
* x86 SIMD extension for IJG JPEG library
|
||||
* Copyright (C) 1999-2006, MIYASAKA Masaru.
|
||||
* This file has been modified for SIMD extension.
|
||||
* Last Modified : January 5, 2006
|
||||
* ---------------------------------------------------------------------
|
||||
*
|
||||
* This include file contains common declarations for the forward and
|
||||
* inverse DCT modules. These declarations are private to the DCT managers
|
||||
* (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
|
||||
@@ -13,6 +20,13 @@
|
||||
*/
|
||||
|
||||
|
||||
/* SIMD Ext: configuration check */
|
||||
|
||||
#if BITS_IN_JSAMPLE != 8
|
||||
#error "Sorry, this SIMD code only copes with 8-bit sample values."
|
||||
#endif
|
||||
|
||||
|
||||
/*
|
||||
* A forward DCT routine is given a pointer to a work area of type DCTELEM[];
|
||||
* the DCT is to be performed in-place in that buffer. Type DCTELEM is int
|
||||
@@ -26,14 +40,25 @@
|
||||
* Quantization of the output coefficients is done by jcdctmgr.c.
|
||||
*/
|
||||
|
||||
#if BITS_IN_JSAMPLE == 8
|
||||
typedef int DCTELEM; /* 16 or 32 bits is fine */
|
||||
#else
|
||||
typedef INT32 DCTELEM; /* must have 32 bits */
|
||||
#endif
|
||||
/* SIMD Ext: To maximize parallelism, Type DCTELEM is changed to short
|
||||
* (originally, int).
|
||||
*/
|
||||
typedef short DCTELEM; /* SIMD Ext: must be short */
|
||||
|
||||
typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data));
|
||||
typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
|
||||
typedef JMETHOD(void, convsamp_int_method_ptr,
|
||||
(JSAMPARRAY sample_data, JDIMENSION start_col,
|
||||
DCTELEM * workspace));
|
||||
typedef JMETHOD(void, convsamp_float_method_ptr,
|
||||
(JSAMPARRAY sample_data, JDIMENSION start_col,
|
||||
FAST_FLOAT *workspace));
|
||||
typedef JMETHOD(void, quantize_int_method_ptr,
|
||||
(JCOEFPTR coef_block, DCTELEM * divisors,
|
||||
DCTELEM * workspace));
|
||||
typedef JMETHOD(void, quantize_float_method_ptr,
|
||||
(JCOEFPTR coef_block, FAST_FLOAT * divisors,
|
||||
FAST_FLOAT * workspace));
|
||||
|
||||
|
||||
/*
|
||||
@@ -49,19 +74,22 @@ typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
|
||||
|
||||
/* typedef inverse_DCT_method_ptr is declared in jpegint.h */
|
||||
|
||||
/* SIMD Ext: To maximize parallelism, Type MULTIPLIER is changed to short.
|
||||
* Macro definitions of MULTIPLIER and FAST_FLOAT in jmorecfg.h are ignored.
|
||||
*/
|
||||
#undef MULTIPLIER
|
||||
#define MULTIPLIER short /* SIMD Ext: must be short */
|
||||
#undef FAST_FLOAT
|
||||
#define FAST_FLOAT float /* SIMD Ext: must be float */
|
||||
|
||||
/*
|
||||
* Each IDCT routine has its own ideas about the best dct_table element type.
|
||||
*/
|
||||
|
||||
typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
|
||||
#if BITS_IN_JSAMPLE == 8
|
||||
typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
|
||||
typedef MULTIPLIER ISLOW_MULT_TYPE; /* SIMD Ext: must be short */
|
||||
typedef MULTIPLIER IFAST_MULT_TYPE; /* SIMD Ext: must be short */
|
||||
#define IFAST_SCALE_BITS 2 /* fractional bits in scale factors */
|
||||
#else
|
||||
typedef INT32 IFAST_MULT_TYPE; /* need 32 bits for scaled quantizers */
|
||||
#define IFAST_SCALE_BITS 13 /* fractional bits in scale factors */
|
||||
#endif
|
||||
typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
|
||||
typedef FAST_FLOAT FLOAT_MULT_TYPE; /* SIMD Ext: must be float */
|
||||
|
||||
|
||||
/*
|
||||
@@ -81,15 +109,64 @@ typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
|
||||
/* Short forms of external names for systems with brain-damaged linkers. */
|
||||
|
||||
#ifdef NEED_SHORT_EXTERNAL_NAMES
|
||||
#define jpeg_fdct_islow jFDislow
|
||||
#define jpeg_fdct_ifast jFDifast
|
||||
#define jpeg_fdct_float jFDfloat
|
||||
#define jpeg_idct_islow jRDislow
|
||||
#define jpeg_idct_ifast jRDifast
|
||||
#define jpeg_idct_float jRDfloat
|
||||
#define jpeg_idct_4x4 jRD4x4
|
||||
#define jpeg_idct_2x2 jRD2x2
|
||||
#define jpeg_idct_1x1 jRD1x1
|
||||
#define jpeg_fdct_islow jFDislow /* jfdctint.asm */
|
||||
#define jpeg_fdct_ifast jFDifast /* jfdctfst.asm */
|
||||
#define jpeg_fdct_float jFDfloat /* jfdctflt.asm */
|
||||
#define jpeg_fdct_islow_mmx jFDMislow /* jfmmxint.asm */
|
||||
#define jpeg_fdct_ifast_mmx jFDMifast /* jfmmxfst.asm */
|
||||
#define jpeg_fdct_float_3dnow jFD3float /* jf3dnflt.asm */
|
||||
#define jpeg_fdct_islow_sse2 jFDSislow /* jfss2int.asm */
|
||||
#define jpeg_fdct_ifast_sse2 jFDSifast /* jfss2fst.asm */
|
||||
#define jpeg_fdct_float_sse jFDSfloat /* jfsseflt.asm */
|
||||
#define jpeg_convsamp_int jCnvInt /* jcqntint.asm */
|
||||
#define jpeg_quantize_int jQntInt /* jcqntint.asm */
|
||||
#define jpeg_quantize_idiv jQntIDiv /* jcqntint.asm */
|
||||
#define jpeg_convsamp_float jCnvFloat /* jcqntflt.asm */
|
||||
#define jpeg_quantize_float jQntFloat /* jcqntflt.asm */
|
||||
#define jpeg_convsamp_int_mmx jCnvMmx /* jcqntmmx.asm */
|
||||
#define jpeg_quantize_int_mmx jQntMmx /* jcqntmmx.asm */
|
||||
#define jpeg_convsamp_flt_3dnow jCnv3dnow /* jcqnt3dn.asm */
|
||||
#define jpeg_quantize_flt_3dnow jQnt3dnow /* jcqnt3dn.asm */
|
||||
#define jpeg_convsamp_int_sse2 jCnvISse2 /* jcqnts2i.asm */
|
||||
#define jpeg_quantize_int_sse2 jQntISse2 /* jcqnts2i.asm */
|
||||
#define jpeg_convsamp_flt_sse jCnvSse /* jcqntsse.asm */
|
||||
#define jpeg_quantize_flt_sse jQntSse /* jcqntsse.asm */
|
||||
#define jpeg_convsamp_flt_sse2 jCnvFSse2 /* jcqnts2f.asm */
|
||||
#define jpeg_quantize_flt_sse2 jQntFSse2 /* jcqnts2f.asm */
|
||||
#define jpeg_idct_islow jRDislow /* jidctint.asm */
|
||||
#define jpeg_idct_ifast jRDifast /* jidctfst.asm */
|
||||
#define jpeg_idct_float jRDfloat /* jidctflt.asm */
|
||||
#define jpeg_idct_4x4 jRD4x4 /* jidctred.asm */
|
||||
#define jpeg_idct_2x2 jRD2x2 /* jidctred.asm */
|
||||
#define jpeg_idct_1x1 jRD1x1 /* jidctred.asm */
|
||||
#define jpeg_idct_islow_mmx jRDMislow /* jimmxint.asm */
|
||||
#define jpeg_idct_ifast_mmx jRDMifast /* jimmxfst.asm */
|
||||
#define jpeg_idct_float_3dnow jRD3float /* ji3dnflt.asm */
|
||||
#define jpeg_idct_4x4_mmx jRDM4x4 /* jimmxred.asm */
|
||||
#define jpeg_idct_2x2_mmx jRDM2x2 /* jimmxred.asm */
|
||||
#define jpeg_idct_islow_sse2 jRDSislow /* jiss2int.asm */
|
||||
#define jpeg_idct_ifast_sse2 jRDSifast /* jiss2fst.asm */
|
||||
#define jpeg_idct_float_sse jRDSfloat /* jisseflt.asm */
|
||||
#define jpeg_idct_float_sse2 jRD2float /* jiss2flt.asm */
|
||||
#define jpeg_idct_4x4_sse2 jRDS4x4 /* jiss2red.asm */
|
||||
#define jpeg_idct_2x2_sse2 jRDS2x2 /* jiss2red.asm */
|
||||
#define jconst_fdct_float jFCfloat /* jfdctflt.asm */
|
||||
#define jconst_fdct_islow_mmx jFCMislow /* jfmmxint.asm */
|
||||
#define jconst_fdct_ifast_mmx jFCMifast /* jfmmxfst.asm */
|
||||
#define jconst_fdct_float_3dnow jFC3float /* jf3dnflt.asm */
|
||||
#define jconst_fdct_islow_sse2 jFCSislow /* jfss2int.asm */
|
||||
#define jconst_fdct_ifast_sse2 jFCSifast /* jfss2fst.asm */
|
||||
#define jconst_fdct_float_sse jFCSfloat /* jfsseflt.asm */
|
||||
#define jconst_idct_float jRCfloat /* jidctflt.asm */
|
||||
#define jconst_idct_islow_mmx jRCMislow /* jimmxint.asm */
|
||||
#define jconst_idct_ifast_mmx jRCMifast /* jimmxfst.asm */
|
||||
#define jconst_idct_float_3dnow jRC3float /* ji3dnflt.asm */
|
||||
#define jconst_idct_red_mmx jRCMred /* jimmxred.asm */
|
||||
#define jconst_idct_islow_sse2 jRCSislow /* jiss2int.asm */
|
||||
#define jconst_idct_ifast_sse2 jRCSifast /* jiss2fst.asm */
|
||||
#define jconst_idct_float_sse jRCSfloat /* jisseflt.asm */
|
||||
#define jconst_idct_float_sse2 jRC2float /* jiss2flt.asm */
|
||||
#define jconst_idct_red_sse2 jRCSred /* jiss2red.asm */
|
||||
#endif /* NEED_SHORT_EXTERNAL_NAMES */
|
||||
|
||||
/* Extern declarations for the forward and inverse DCT routines. */
|
||||
@@ -98,6 +175,47 @@ EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
|
||||
|
||||
EXTERN(void) jpeg_fdct_islow_mmx JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_ifast_mmx JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_float_3dnow JPP((FAST_FLOAT * data));
|
||||
|
||||
EXTERN(void) jpeg_fdct_islow_sse2 JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_ifast_sse2 JPP((DCTELEM * data));
|
||||
EXTERN(void) jpeg_fdct_float_sse JPP((FAST_FLOAT * data));
|
||||
|
||||
EXTERN(void) jpeg_convsamp_int
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_quantize_int
|
||||
JPP((JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_quantize_idiv
|
||||
JPP((JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_convsamp_float
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace));
|
||||
EXTERN(void) jpeg_quantize_float
|
||||
JPP((JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace));
|
||||
|
||||
EXTERN(void) jpeg_convsamp_int_mmx
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_quantize_int_mmx
|
||||
JPP((JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_convsamp_flt_3dnow
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace));
|
||||
EXTERN(void) jpeg_quantize_flt_3dnow
|
||||
JPP((JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace));
|
||||
|
||||
EXTERN(void) jpeg_convsamp_int_sse2
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_quantize_int_sse2
|
||||
JPP((JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace));
|
||||
EXTERN(void) jpeg_convsamp_flt_sse
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace));
|
||||
EXTERN(void) jpeg_quantize_flt_sse
|
||||
JPP((JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace));
|
||||
EXTERN(void) jpeg_convsamp_flt_sse2
|
||||
JPP((JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace));
|
||||
EXTERN(void) jpeg_quantize_flt_sse2
|
||||
JPP((JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace));
|
||||
|
||||
EXTERN(void) jpeg_idct_islow
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
@@ -117,6 +235,60 @@ EXTERN(void) jpeg_idct_1x1
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
|
||||
EXTERN(void) jpeg_idct_islow_mmx
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_ifast_mmx
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_4x4_mmx
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_2x2_mmx
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
|
||||
EXTERN(void) jpeg_idct_float_3dnow
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_float_sse
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_float_sse2
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
|
||||
EXTERN(void) jpeg_idct_islow_sse2
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_ifast_sse2
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_4x4_sse2
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
EXTERN(void) jpeg_idct_2x2_sse2
|
||||
JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
|
||||
JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
|
||||
|
||||
extern const int jconst_fdct_float[];
|
||||
extern const int jconst_fdct_islow_mmx[];
|
||||
extern const int jconst_fdct_ifast_mmx[];
|
||||
extern const int jconst_fdct_float_3dnow[];
|
||||
extern const int jconst_fdct_islow_sse2[];
|
||||
extern const int jconst_fdct_ifast_sse2[];
|
||||
extern const int jconst_fdct_float_sse[];
|
||||
extern const int jconst_idct_float[];
|
||||
extern const int jconst_idct_islow_mmx[];
|
||||
extern const int jconst_idct_ifast_mmx[];
|
||||
extern const int jconst_idct_float_3dnow[];
|
||||
extern const int jconst_idct_red_mmx[];
|
||||
extern const int jconst_idct_islow_sse2[];
|
||||
extern const int jconst_idct_ifast_sse2[];
|
||||
extern const int jconst_idct_float_sse[];
|
||||
extern const int jconst_idct_float_sse2[];
|
||||
extern const int jconst_idct_red_sse2[];
|
||||
|
||||
|
||||
/*
|
||||
* Macros for handling fixed-point arithmetic; these are used by many
|
||||
|
||||
Reference in New Issue
Block a user