1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
|
#
# MMX surface clear routines for HERMES
# Copyright (c) 1998 Christian Nentwich (c.nentwich@cs.ucl.ac.uk)
# This source code is licensed under the GNU LGPL
#
# Please refer to the file COPYING.LIB contained in the distribution for
# licensing conditions
#
.globl _ClearMMX_32
.globl _ClearMMX_24
.globl _ClearMMX_16
.globl _ClearMMX_8
.text
##
## --------------------------------------------------------------------------
## HermesClearInterface (ebp+..)
## 0: char8 *dest
## 4: int32 value
## 8: unsigned int width (already checked to be >0!)
## 12: unsigned int height (already checked to be >0!)
## 16: int add
_ClearMMX_32:
pushl %ebp
movl %esp,%ebp
movl 8(%ebp),%ebp
movl 4(%ebp),%eax # pixel value
movd 4(%ebp),%mm0
movl 12(%ebp),%edx # height
movq %mm0,%mm1
psllq $32,%mm0
movl (%ebp),%edi # destination
por %mm1,%mm0
_ClearMMX_32.L_y:
movl 8(%ebp),%ecx
movl %ecx,%ebx
shrl %ecx
jz _ClearMMX_32.L_last
_ClearMMX_32.L_x:
movq %mm0,(%edi)
addl $8,%edi
decl %ecx
jnz _ClearMMX_32.L_x
_ClearMMX_32.L_last:
testl $1,%ebx
jz _ClearMMX_32.L_endline
movl %eax,(%edi)
addl $4,%edi
_ClearMMX_32.L_endline:
addl 16(%ebp),%edi
decl %edx
jnz _ClearMMX_32.L_y
emms
popl %ebp
ret
_ClearMMX_24:
ret
_ClearMMX_16:
pushl %ebp
movl %esp,%ebp
movl 8(%ebp),%ebp
movl 4(%ebp),%eax # pixel value
movl 4(%ebp),%ebx
movl 12(%ebp),%edx # height
movl (%ebp),%edi # destination
shll $16,%eax # Duplicate pixel value
andl $0x0ffff,%ebx
orl %ebx,%eax
movd %eax,%mm0
movd %eax,%mm1
psllq $32,%mm0
por %mm1,%mm0
_ClearMMX_16.L_y:
movl 8(%ebp),%ecx
testl $3,%edi # Check if destination is aligned mod 4
jz _ClearMMX_16.L_aligned
movw %ax,(%edi) # otherwise write one pixel
addl $2,%edi
decl %ecx
jz _ClearMMX_16.L_endline
_ClearMMX_16.L_aligned:
movl %ecx,%ebx
shrl $2,%ecx
jz _ClearMMX_16.L_last
_ClearMMX_16.L_x:
movq %mm0,(%edi)
addl $8,%edi
decl %ecx
jnz _ClearMMX_16.L_x
_ClearMMX_16.L_last:
andl $3,%ebx
jz _ClearMMX_16.L_endline
movw %ax,(%edi) # Write trailing pixels
addl $2,%edi
decl %ebx
jz _ClearMMX_16.L_endline
movw %ax,(%edi)
addl $2,%edi
decl %ebx
jz _ClearMMX_16.L_endline
movw %ax,(%edi)
addl $2,%edi
decl %ebx
jnz _ClearMMX_16.L_endline
_ClearMMX_16.L_endline:
addl 16(%ebp),%edi
decl %edx
jnz _ClearMMX_16.L_y
emms
popl %ebp
ret
## Clear8_x86 isnt optimised fully yet as it seems to be a tiny bit slower
## than the C routine
_ClearMMX_8:
pushl %ebp
movl %esp,%ebp
movl 8(%ebp),%ebp
movl 4(%ebp),%eax # pixel value
movl 4(%ebp),%ebx
movl 12(%ebp),%edx # height
andl $0x0ff,%ebx
shll $8,%eax # Put the byte pixel value in all four bytes
movl (%ebp),%edi # destination
movb %bl,%al
movb %bl,%bh
shll $16,%eax
movb %bh,%ah
movb %bl,%al
movd %eax,%mm0
movd %eax,%mm1
psllq $32,%mm0
por %mm1,%mm0
_ClearMMX_8.L_y:
movl 8(%ebp),%ecx
testl $3,%edi # Align mod 4
jz _ClearMMX_8.L_aligned
movl %edi,%ebx
andl $3,%ebx
movb %al,(%edi) # Unrolled (copy & paste), align and jump
incl %edi # if finished, faster than a loop...
decl %ecx
jz _ClearMMX_8.L_endline
decl %ebx
jz _ClearMMX_8.L_aligned
movb %al,(%edi) # Second pixel
incl %edi
decl %ecx
jz _ClearMMX_8.L_endline
decl %ebx
jz _ClearMMX_8.L_aligned
movb %al,(%edi) # Third pixel
incl %edi
decl %ecx
jz _ClearMMX_8.L_endline
decl %ebx
jz _ClearMMX_8.L_aligned
_ClearMMX_8.L_aligned:
movl %ecx,%ebx # Store ecx for later
shrl $3,%ecx # We write 8 pixels at once
jz _ClearMMX_8.L_last
_ClearMMX_8.L_x:
movq %mm0,(%edi)
addl $8,%edi
decl %ecx
jnz _ClearMMX_8.L_x
_ClearMMX_8.L_last:
movl %ebx,%ecx # Clean up trailing pixels
andl $7,%ecx # Could be up to 7 left
jz _ClearMMX_8.L_endline
testb $0b100,%cl # If theres less than four jump
jz _ClearMMX_8.L_lessthanfour
movl %eax,(%edi) # Otherwise write a dword
addl $4,%edi
subl $4,%ecx
_ClearMMX_8.L_lessthanfour:
rep
stosb # Clean up the very rest
_ClearMMX_8.L_endline:
addl 16(%ebp),%edi
decl %edx
jnz _ClearMMX_8.L_y
emms
popl %ebp
ret
|