1 /**************************************************************************
3 * Copyright © 2010 Luca Barbieri
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
24 **************************************************************************/
27 #include <translate/translate.h>
28 #include <util/u_memory.h>
29 #include <util/u_format.h>
30 #include <util/u_cpu_detect.h>
31 #include <rtasm/rtasm_cpu.h>
33 int main(int argc
, char** argv
)
35 struct translate
*(*create_fn
)(const struct translate_key
*key
) = 0;
37 struct translate_key key
;
38 unsigned output_format
;
39 unsigned input_format
;
40 unsigned char* buffer
[5];
45 const float error
= 0.03125;
53 else if (!strcmp(argv
[1], "generic"))
54 create_fn
= translate_generic_create
;
55 else if (!strcmp(argv
[1], "x86"))
56 create_fn
= translate_sse2_create
;
57 else if (!strcmp(argv
[1], "nosse"))
59 util_cpu_caps
.has_sse
= 0;
60 util_cpu_caps
.has_sse2
= 0;
61 util_cpu_caps
.has_sse3
= 0;
62 util_cpu_caps
.has_sse4_1
= 0;
63 create_fn
= translate_sse2_create
;
65 else if (!strcmp(argv
[1], "sse"))
67 if(!util_cpu_caps
.has_sse
|| !rtasm_cpu_has_sse())
69 printf("Error: CPU doesn't support SSE (test with qemu)\n");
72 util_cpu_caps
.has_sse2
= 0;
73 util_cpu_caps
.has_sse3
= 0;
74 util_cpu_caps
.has_sse4_1
= 0;
75 create_fn
= translate_sse2_create
;
77 else if (!strcmp(argv
[1], "sse2"))
79 if(!util_cpu_caps
.has_sse2
|| !rtasm_cpu_has_sse())
81 printf("Error: CPU doesn't support SSE2 (test with qemu)\n");
84 util_cpu_caps
.has_sse3
= 0;
85 util_cpu_caps
.has_sse4_1
= 0;
86 create_fn
= translate_sse2_create
;
88 else if (!strcmp(argv
[1], "sse3"))
90 if(!util_cpu_caps
.has_sse3
|| !rtasm_cpu_has_sse())
92 printf("Error: CPU doesn't support SSE3 (test with qemu)\n");
95 util_cpu_caps
.has_sse4_1
= 0;
96 create_fn
= translate_sse2_create
;
98 else if (!strcmp(argv
[1], "sse4.1"))
100 if(!util_cpu_caps
.has_sse4_1
|| !rtasm_cpu_has_sse())
102 printf("Error: CPU doesn't support SSE4.1 (test with qemu)\n");
105 create_fn
= translate_sse2_create
;
110 printf("Usage: ./translate_test [generic|x86|nosse|sse|sse2|sse3|sse4.1]\n");
114 for (i
= 0; i
< Elements(buffer
); ++i
)
115 buffer
[i
] = align_malloc(4096, 4096);
118 key
.element
[0].input_buffer
= 0;
119 key
.element
[0].input_offset
= 0;
120 key
.element
[0].output_offset
= 0;
121 key
.element
[0].type
= TRANSLATE_ELEMENT_NORMAL
;
122 key
.element
[0].instance_divisor
= 0;
125 for (i
= 0; i
< 4096; ++i
)
126 buffer
[0][i
] = rand() & 0x7f; /* avoid negative values that work badly when converted to unsigned format*/
128 for (output_format
= 1; output_format
< PIPE_FORMAT_COUNT
; ++output_format
)
130 const struct util_format_description
* output_format_desc
= util_format_description(output_format
);
131 unsigned output_format_size
;
132 if (!output_format_desc
133 || !output_format_desc
->fetch_rgba_float
134 || !output_format_desc
->pack_rgba_float
135 || output_format_desc
->colorspace
!= UTIL_FORMAT_COLORSPACE_RGB
136 || output_format_desc
->layout
!= UTIL_FORMAT_LAYOUT_PLAIN
137 || !translate_is_output_format_supported(output_format
))
140 output_format_size
= util_format_get_stride(output_format
, 1);
142 for (input_format
= 1; input_format
< PIPE_FORMAT_COUNT
; ++input_format
)
144 const struct util_format_description
* input_format_desc
= util_format_description(input_format
);
145 unsigned input_format_size
;
146 struct translate
* translate
[2];
148 unsigned used_generic
= 0;
150 if (!input_format_desc
151 || !input_format_desc
->fetch_rgba_float
152 || !input_format_desc
->pack_rgba_float
153 || input_format_desc
->colorspace
!= UTIL_FORMAT_COLORSPACE_RGB
154 || input_format_desc
->layout
!= UTIL_FORMAT_LAYOUT_PLAIN
155 || !translate_is_output_format_supported(input_format
))
158 input_format_size
= util_format_get_stride(input_format
, 1);
160 key
.element
[0].input_format
= input_format
;
161 key
.element
[0].output_format
= output_format
;
162 key
.output_stride
= output_format_size
;
163 translate
[0] = create_fn(&key
);
167 key
.element
[0].input_format
= output_format
;
168 key
.element
[0].output_format
= input_format
;
169 key
.output_stride
= input_format_size
;
170 translate
[1] = create_fn(&key
);
174 translate
[1] = translate_generic_create(&key
);
179 translate
[0]->set_buffer(translate
[0], 0, buffer
[0], input_format_size
, ~0);
180 translate
[0]->run(translate
[0], 0, count
, 0, buffer
[1]);
181 translate
[1]->set_buffer(translate
[1], 0, buffer
[1], output_format_size
, ~0);
182 translate
[1]->run(translate
[1], 0, count
, 0, buffer
[2]);
183 translate
[0]->set_buffer(translate
[0], 0, buffer
[2], input_format_size
, ~0);
184 translate
[0]->run(translate
[0], 0, count
, 0, buffer
[3]);
185 translate
[1]->set_buffer(translate
[1], 0, buffer
[3], output_format_size
, ~0);
186 translate
[1]->run(translate
[1], 0, count
, 0, buffer
[4]);
188 for (i
= 0; i
< count
; ++i
)
192 input_format_desc
->fetch_rgba_float(a
, buffer
[2] + i
* input_format_size
, 0, 0);
193 input_format_desc
->fetch_rgba_float(b
, buffer
[4] + i
* input_format_size
, 0, 0);
195 for (j
= 0; j
< count
; ++j
)
197 float d
= a
[j
] - b
[j
];
198 if (d
> error
|| d
< -error
)
206 printf("%s%s: %s -> %s -> %s -> %s -> %s\n",
207 fail
? "FAIL" : "PASS",
208 used_generic
? "[GENERIC]" : "",
209 input_format_desc
->name
, output_format_desc
->name
, input_format_desc
->name
, output_format_desc
->name
, input_format_desc
->name
);
213 for (i
= 0; i
< Elements(buffer
); ++i
)
215 unsigned format_size
= (i
& 1) ? output_format_size
: input_format_size
;
216 printf("%c ", (i
== 2 || i
== 4) ? '*' : ' ');
217 for (j
= 0; j
< count
; ++j
)
219 for (k
= 0; k
< format_size
; ++k
)
221 printf("%02x", buffer
[i
][j
* format_size
+ k
]);
234 translate
[1]->release(translate
[1]);
235 translate
[0]->release(translate
[0]);
239 printf("%u/%u tests passed for translate_%s\n", passed
, total
, argv
[1]);
240 return passed
!= total
;