DEADSOFTWARE

render: fix archvile fire animation
[d2df-sdl.git] / src / lib / vampimg / JpegLib / imjidctflt.pas
1 unit imjidctflt;
3 {$N+}
4 { This file contains a floating-point implementation of the
5 inverse DCT (Discrete Cosine Transform). In the IJG code, this routine
6 must also perform dequantization of the input coefficients.
8 This implementation should be more accurate than either of the integer
9 IDCT implementations. However, it may not give the same results on all
10 machines because of differences in roundoff behavior. Speed will depend
11 on the hardware's floating point capacity.
13 A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
14 on each row (or vice versa, but it's more convenient to emit a row at
15 a time). Direct algorithms are also available, but they are much more
16 complex and seem not to be any faster when reduced to code.
18 This implementation is based on Arai, Agui, and Nakajima's algorithm for
19 scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in
20 Japanese, but the algorithm is described in the Pennebaker & Mitchell
21 JPEG textbook (see REFERENCES section in file README). The following code
22 is based directly on figure 4-8 in P&M.
23 While an 8-point DCT cannot be done in less than 11 multiplies, it is
24 possible to arrange the computation so that many of the multiplies are
25 simple scalings of the final outputs. These multiplies can then be
26 folded into the multiplications or divisions by the JPEG quantization
27 table entries. The AA&N method leaves only 5 multiplies and 29 adds
28 to be done in the DCT itself.
29 The primary disadvantage of this method is that with a fixed-point
30 implementation, accuracy is lost due to imprecise representation of the
31 scaled quantization values. However, that problem does not arise if
32 we use floating point arithmetic. }
34 { Original: jidctflt.c ; Copyright (C) 1994-1996, Thomas G. Lane. }
36 interface
38 {$I imjconfig.inc}
40 uses
41 imjmorecfg,
42 imjinclude,
43 imjpeglib,
44 imjdct; { Private declarations for DCT subsystem }
46 { Perform dequantization and inverse DCT on one block of coefficients. }
48 {GLOBAL}
49 procedure jpeg_idct_float (cinfo : j_decompress_ptr;
50 compptr : jpeg_component_info_ptr;
51 coef_block : JCOEFPTR;
52 output_buf : JSAMPARRAY;
53 output_col : JDIMENSION);
55 implementation
57 { This module is specialized to the case DCTSIZE = 8. }
59 {$ifndef DCTSIZE_IS_8}
60 Sorry, this code only copes with 8x8 DCTs. { deliberate syntax err }
61 {$endif}
64 { Dequantize a coefficient by multiplying it by the multiplier-table
65 entry; produce a float result. }
67 function DEQUANTIZE(coef : int; quantval : FAST_FLOAT) : FAST_FLOAT;
68 begin
69 Dequantize := ( (coef) * quantval);
70 end;
72 { Descale and correctly round an INT32 value that's scaled by N bits.
73 We assume RIGHT_SHIFT rounds towards minus infinity, so adding
74 the fudge factor is correct for either sign of X. }
76 function DESCALE(x : INT32; n : int) : INT32;
77 var
78 shift_temp : INT32;
79 begin
80 {$ifdef RIGHT_SHIFT_IS_UNSIGNED}
81 shift_temp := x + (INT32(1) shl (n-1));
82 if shift_temp < 0 then
83 Descale := (shift_temp shr n) or ((not INT32(0)) shl (32-n))
84 else
85 Descale := (shift_temp shr n);
86 {$else}
87 Descale := (x + (INT32(1) shl (n-1)) shr n;
88 {$endif}
89 end;
92 { Perform dequantization and inverse DCT on one block of coefficients. }
94 {GLOBAL}
95 procedure jpeg_idct_float (cinfo : j_decompress_ptr;
96 compptr : jpeg_component_info_ptr;
97 coef_block : JCOEFPTR;
98 output_buf : JSAMPARRAY;
99 output_col : JDIMENSION);
100 type
101 PWorkspace = ^TWorkspace;
102 TWorkspace = array[0..DCTSIZE2-1] of FAST_FLOAT;
103 var
104 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7 : FAST_FLOAT;
105 tmp10, tmp11, tmp12, tmp13 : FAST_FLOAT;
106 z5, z10, z11, z12, z13 : FAST_FLOAT;
107 inptr : JCOEFPTR;
108 quantptr : FLOAT_MULT_TYPE_FIELD_PTR;
109 wsptr : PWorkSpace;
110 outptr : JSAMPROW;
111 range_limit : JSAMPROW;
112 ctr : int;
113 workspace : TWorkspace; { buffers data between passes }
114 {SHIFT_TEMPS}
115 var
116 dcval : FAST_FLOAT;
117 begin
118 { Each IDCT routine is responsible for range-limiting its results and
119 converting them to unsigned form (0..MAXJSAMPLE). The raw outputs could
120 be quite far out of range if the input data is corrupt, so a bulletproof
121 range-limiting step is required. We use a mask-and-table-lookup method
122 to do the combined operations quickly. See the comments with
123 prepare_range_limit_table (in jdmaster.c) for more info. }
125 range_limit := JSAMPROW(@(cinfo^.sample_range_limit^[CENTERJSAMPLE]));
127 { Pass 1: process columns from input, store into work array. }
129 inptr := coef_block;
130 quantptr := FLOAT_MULT_TYPE_FIELD_PTR (compptr^.dct_table);
131 wsptr := @workspace;
132 for ctr := pred(DCTSIZE) downto 0 do
133 begin
134 { Due to quantization, we will usually find that many of the input
135 coefficients are zero, especially the AC terms. We can exploit this
136 by short-circuiting the IDCT calculation for any column in which all
137 the AC terms are zero. In that case each output is equal to the
138 DC coefficient (with scale factor as needed).
139 With typical images and quantization tables, half or more of the
140 column DCT calculations can be simplified this way. }
142 if (inptr^[DCTSIZE*1]=0) and (inptr^[DCTSIZE*2]=0) and
143 (inptr^[DCTSIZE*3]=0) and (inptr^[DCTSIZE*4]=0) and
144 (inptr^[DCTSIZE*5]=0) and (inptr^[DCTSIZE*6]=0) and
145 (inptr^[DCTSIZE*7]=0) then
146 begin
147 { AC terms all zero }
148 FAST_FLOAT(dcval) := DEQUANTIZE(inptr^[DCTSIZE*0], quantptr^[DCTSIZE*0]);
150 wsptr^[DCTSIZE*0] := dcval;
151 wsptr^[DCTSIZE*1] := dcval;
152 wsptr^[DCTSIZE*2] := dcval;
153 wsptr^[DCTSIZE*3] := dcval;
154 wsptr^[DCTSIZE*4] := dcval;
155 wsptr^[DCTSIZE*5] := dcval;
156 wsptr^[DCTSIZE*6] := dcval;
157 wsptr^[DCTSIZE*7] := dcval;
159 Inc(JCOEF_PTR(inptr)); { advance pointers to next column }
160 Inc(FLOAT_MULT_TYPE_PTR(quantptr));
161 Inc(FAST_FLOAT_PTR(wsptr));
162 continue;
163 end;
165 { Even part }
167 tmp0 := DEQUANTIZE(inptr^[DCTSIZE*0], quantptr^[DCTSIZE*0]);
168 tmp1 := DEQUANTIZE(inptr^[DCTSIZE*2], quantptr^[DCTSIZE*2]);
169 tmp2 := DEQUANTIZE(inptr^[DCTSIZE*4], quantptr^[DCTSIZE*4]);
170 tmp3 := DEQUANTIZE(inptr^[DCTSIZE*6], quantptr^[DCTSIZE*6]);
172 tmp10 := tmp0 + tmp2; { phase 3 }
173 tmp11 := tmp0 - tmp2;
175 tmp13 := tmp1 + tmp3; { phases 5-3 }
176 tmp12 := (tmp1 - tmp3) * ({FAST_FLOAT}(1.414213562)) - tmp13; { 2*c4 }
178 tmp0 := tmp10 + tmp13; { phase 2 }
179 tmp3 := tmp10 - tmp13;
180 tmp1 := tmp11 + tmp12;
181 tmp2 := tmp11 - tmp12;
183 { Odd part }
185 tmp4 := DEQUANTIZE(inptr^[DCTSIZE*1], quantptr^[DCTSIZE*1]);
186 tmp5 := DEQUANTIZE(inptr^[DCTSIZE*3], quantptr^[DCTSIZE*3]);
187 tmp6 := DEQUANTIZE(inptr^[DCTSIZE*5], quantptr^[DCTSIZE*5]);
188 tmp7 := DEQUANTIZE(inptr^[DCTSIZE*7], quantptr^[DCTSIZE*7]);
190 z13 := tmp6 + tmp5; { phase 6 }
191 z10 := tmp6 - tmp5;
192 z11 := tmp4 + tmp7;
193 z12 := tmp4 - tmp7;
195 tmp7 := z11 + z13; { phase 5 }
196 tmp11 := (z11 - z13) * ({FAST_FLOAT}(1.414213562)); { 2*c4 }
198 z5 := (z10 + z12) * ({FAST_FLOAT}(1.847759065)); { 2*c2 }
199 tmp10 := ({FAST_FLOAT}(1.082392200)) * z12 - z5; { 2*(c2-c6) }
200 tmp12 := ({FAST_FLOAT}(-2.613125930)) * z10 + z5; { -2*(c2+c6) }
202 tmp6 := tmp12 - tmp7; { phase 2 }
203 tmp5 := tmp11 - tmp6;
204 tmp4 := tmp10 + tmp5;
206 wsptr^[DCTSIZE*0] := tmp0 + tmp7;
207 wsptr^[DCTSIZE*7] := tmp0 - tmp7;
208 wsptr^[DCTSIZE*1] := tmp1 + tmp6;
209 wsptr^[DCTSIZE*6] := tmp1 - tmp6;
210 wsptr^[DCTSIZE*2] := tmp2 + tmp5;
211 wsptr^[DCTSIZE*5] := tmp2 - tmp5;
212 wsptr^[DCTSIZE*4] := tmp3 + tmp4;
213 wsptr^[DCTSIZE*3] := tmp3 - tmp4;
215 Inc(JCOEF_PTR(inptr)); { advance pointers to next column }
216 Inc(FLOAT_MULT_TYPE_PTR(quantptr));
217 Inc(FAST_FLOAT_PTR(wsptr));
218 end;
220 { Pass 2: process rows from work array, store into output array. }
221 { Note that we must descale the results by a factor of 8 = 2**3. }
223 wsptr := @workspace;
224 for ctr := 0 to pred(DCTSIZE) do
225 begin
226 outptr := JSAMPROW(@(output_buf^[ctr]^[output_col]));
227 { Rows of zeroes can be exploited in the same way as we did with columns.
228 However, the column calculation has created many nonzero AC terms, so
229 the simplification applies less often (typically 5% to 10% of the time).
230 And testing floats for zero is relatively expensive, so we don't bother. }
232 { Even part }
234 tmp10 := wsptr^[0] + wsptr^[4];
235 tmp11 := wsptr^[0] - wsptr^[4];
237 tmp13 := wsptr^[2] + wsptr^[6];
238 tmp12 := (wsptr^[2] - wsptr^[6]) * ({FAST_FLOAT}(1.414213562)) - tmp13;
240 tmp0 := tmp10 + tmp13;
241 tmp3 := tmp10 - tmp13;
242 tmp1 := tmp11 + tmp12;
243 tmp2 := tmp11 - tmp12;
245 { Odd part }
247 z13 := wsptr^[5] + wsptr^[3];
248 z10 := wsptr^[5] - wsptr^[3];
249 z11 := wsptr^[1] + wsptr^[7];
250 z12 := wsptr^[1] - wsptr^[7];
252 tmp7 := z11 + z13;
253 tmp11 := (z11 - z13) * ({FAST_FLOAT}(1.414213562));
255 z5 := (z10 + z12) * ({FAST_FLOAT}(1.847759065)); { 2*c2 }
256 tmp10 := ({FAST_FLOAT}(1.082392200)) * z12 - z5; { 2*(c2-c6) }
257 tmp12 := ({FAST_FLOAT}(-2.613125930)) * z10 + z5; { -2*(c2+c6) }
259 tmp6 := tmp12 - tmp7;
260 tmp5 := tmp11 - tmp6;
261 tmp4 := tmp10 + tmp5;
263 { Final output stage: scale down by a factor of 8 and range-limit }
265 outptr^[0] := range_limit^[ int(DESCALE( INT32(Round((tmp0 + tmp7))), 3))
266 and RANGE_MASK];
267 outptr^[7] := range_limit^[ int(DESCALE( INT32(Round((tmp0 - tmp7))), 3))
268 and RANGE_MASK];
269 outptr^[1] := range_limit^[ int(DESCALE( INT32(Round((tmp1 + tmp6))), 3))
270 and RANGE_MASK];
271 outptr^[6] := range_limit^[ int(DESCALE( INT32(Round((tmp1 - tmp6))), 3))
272 and RANGE_MASK];
273 outptr^[2] := range_limit^[ int(DESCALE( INT32(Round((tmp2 + tmp5))), 3))
274 and RANGE_MASK];
275 outptr^[5] := range_limit^[ int(DESCALE( INT32(Round((tmp2 - tmp5))), 3))
276 and RANGE_MASK];
277 outptr^[4] := range_limit^[ int(DESCALE( INT32(Round((tmp3 + tmp4))), 3))
278 and RANGE_MASK];
279 outptr^[3] := range_limit^[ int(DESCALE( INT32(Round((tmp3 - tmp4))), 3))
280 and RANGE_MASK];
282 Inc(FAST_FLOAT_PTR(wsptr), DCTSIZE); { advance pointer to next row }
283 end;
284 end;
286 end.