dct-test: Move arch-specific bits into arch-specific subdirectories

Diego Biurrun committed 12 years ago

Commit 81b9bf31 parent a9f3f5f4 files · +185 −74Browse files

Unified · Split

4 files

libavcodec/arm/dct-test.c +40 −0
libavcodec/dct-test.c +31 −74
libavcodec/ppc/dct-test.c +32 −0
libavcodec/x86/dct-test.c +82 −0

@@ -0,0 +1,40 @@
	1	/*
	2	* This file is part of Libav.
	3	*
	4	* Libav is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU Lesser General Public
	6	* License as published by the Free Software Foundation; either
	7	* version 2.1 of the License, or (at your option) any later version.
	8	*
	9	* Libav is distributed in the hope that it will be useful,
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	12	* Lesser General Public License for more details.
	13	*
	14	* You should have received a copy of the GNU Lesser General Public
	15	* License along with Libav; if not, write to the Free Software
	16	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
	17	*/
	18
	19	#include "config.h"
	20
	21	#include "idct.h"
	22
	23	static const struct algo fdct_tab_arch[] = {
	24	{ 0 }
	25	};
	26
	27	static const struct algo idct_tab_arch[] = {
	28	{ "SIMPLE-ARM", ff_simple_idct_arm, FF_IDCT_PERM_NONE },
	29	{ "INT-ARM", ff_j_rev_dct_arm, FF_IDCT_PERM_LIBMPEG2 },
	30	#if HAVE_ARMV5TE
	31	{ "SIMPLE-ARMV5TE", ff_simple_idct_armv5te, FF_IDCT_PERM_NONE, AV_CPU_FLAG_ARMV5TE },
	32	#endif
	33	#if HAVE_ARMV6
	34	{ "SIMPLE-ARMV6", ff_simple_idct_armv6, FF_IDCT_PERM_LIBMPEG2, AV_CPU_FLAG_ARMV6 },
	35	#endif
	36	#if HAVE_NEON
	37	{ "SIMPLE-NEON", ff_simple_idct_neon, FF_IDCT_PERM_PARTTRANS, AV_CPU_FLAG_NEON },
	38	#endif
	39	{ 0 }
	40	};

libavcodec/dct-test.c+31−74

@@ -45,11 +45,6 @@ · 44 unmodified lines
45	#include "aandcttab.h"	45	#include "aandcttab.h"
46	#include "faandct.h"	46	#include "faandct.h"
47	#include "faanidct.h"	47	#include "faanidct.h"
48	#include "arm/idct.h"
49	#include "ppc/fdct.h"
50	#include "x86/fdct.h"
51	#include "x86/idct_xvid.h"
52	#include "x86/simple_idct.h"
53	#include "dctref.h"	48	#include "dctref.h"
54		49
55	struct algo {	50	struct algo {
@@ -60,81 +55,36 @@ struct algo { · 4 unmodified lines
60	int nonspec;	55	int nonspec;
61	};	56	};
62		57
63	static const struct algo fdct_tab[] = {	58	static const struct algo fdct_tab[4] = {
64	{ "REF-DBL", ff_ref_fdct, FF_IDCT_PERM_NONE },	59	{ "REF-DBL", ff_ref_fdct, FF_IDCT_PERM_NONE },
65	{ "FAAN", ff_faandct, FF_IDCT_PERM_NONE },	60	{ "FAAN", ff_faandct, FF_IDCT_PERM_NONE },
66	{ "IJG-AAN-INT", ff_fdct_ifast, FF_IDCT_PERM_NONE },	61	{ "IJG-AAN-INT", ff_fdct_ifast, FF_IDCT_PERM_NONE },
67	{ "IJG-LLM-INT", ff_jpeg_fdct_islow_8, FF_IDCT_PERM_NONE },	62	{ "IJG-LLM-INT", ff_jpeg_fdct_islow_8, FF_IDCT_PERM_NONE },
68
69	#if HAVE_MMX_INLINE
70	{ "MMX", ff_fdct_mmx, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMX },
71	#endif
72	#if HAVE_MMXEXT_INLINE
73	{ "MMXEXT", ff_fdct_mmxext, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMXEXT },
74	#endif
75	#if HAVE_SSE2_INLINE
76	{ "SSE2", ff_fdct_sse2, FF_IDCT_PERM_NONE, AV_CPU_FLAG_SSE2 },
77	#endif
78
79	#if HAVE_ALTIVEC
80	{ "altivecfdct", ff_fdct_altivec, FF_IDCT_PERM_NONE, AV_CPU_FLAG_ALTIVEC },
81	#endif
82
83	{ 0 }
84	};	63	};
85		64
86	static const struct algo idct_tab[] = {	65	static const struct algo idct_tab[4] = {
87	{ "FAANI", ff_faanidct, FF_IDCT_PERM_NONE },	66	{ "FAANI", ff_faanidct, FF_IDCT_PERM_NONE },
88	{ "REF-DBL", ff_ref_idct, FF_IDCT_PERM_NONE },	67	{ "REF-DBL", ff_ref_idct, FF_IDCT_PERM_NONE },
89	{ "INT", ff_j_rev_dct, FF_IDCT_PERM_LIBMPEG2 },	68	{ "INT", ff_j_rev_dct, FF_IDCT_PERM_LIBMPEG2 },
90	{ "SIMPLE-C", ff_simple_idct_8, FF_IDCT_PERM_NONE },	69	{ "SIMPLE-C", ff_simple_idct_8, FF_IDCT_PERM_NONE },
91		70	};
92	#if HAVE_MMX_INLINE
93	{ "SIMPLE-MMX", ff_simple_idct_mmx, FF_IDCT_PERM_SIMPLE, AV_CPU_FLAG_MMX },
94	{ "XVID-MMX", ff_idct_xvid_mmx, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMX, 1 },
95	#endif
96	#if HAVE_MMXEXT_INLINE
97	{ "XVID-MMXEXT", ff_idct_xvid_mmxext, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMXEXT, 1 },
98	#endif
99	#if HAVE_SSE2_INLINE
100	{ "XVID-SSE2", ff_idct_xvid_sse2, FF_IDCT_PERM_SSE2, AV_CPU_FLAG_SSE2, 1 },
101	#endif
102		71
103	#if ARCH_ARM	72	#if ARCH_ARM
104	{ "SIMPLE-ARM", ff_simple_idct_arm, FF_IDCT_PERM_NONE },	73	#include "arm/dct-test.c"
105	{ "INT-ARM", ff_j_rev_dct_arm, FF_IDCT_PERM_LIBMPEG2 },	74	#elif ARCH_PPC
106	#endif	75	#include "ppc/dct-test.c"
107	#if HAVE_ARMV5TE	76	#elif ARCH_X86
108	{ "SIMPLE-ARMV5TE", ff_simple_idct_armv5te, FF_IDCT_PERM_NONE, AV_CPU_FLAG_ARMV5TE },	77	#include "x86/dct-test.c"
109	#endif	78	#else
110	#if HAVE_ARMV6	79	static const struct algo fdct_tab_arch[] = { 0 };
111	{ "SIMPLE-ARMV6", ff_simple_idct_armv6, FF_IDCT_PERM_LIBMPEG2, AV_CPU_FLAG_ARMV6 },	80	static const struct algo idct_tab_arch[] = { 0 };
112	#endif
113	#if HAVE_NEON && ARCH_ARM
114	{ "SIMPLE-NEON", ff_simple_idct_neon, FF_IDCT_PERM_PARTTRANS, AV_CPU_FLAG_NEON },
115	#endif	81	#endif
116		82
117	{ 0 }
118	};
119
120	#define AANSCALE_BITS 12	83	#define AANSCALE_BITS 12
121		84
122	#define NB_ITS 20000	85	#define NB_ITS 20000
123	#define NB_ITS_SPEED 50000	86	#define NB_ITS_SPEED 50000
124		87
125	static short idct_simple_mmx_perm[64] = {
126	0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
127	0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
128	0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
129	0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
130	0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
131	0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
132	0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
133	0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
134	};
135
136	static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
137
138	DECLARE_ALIGNED(16, static int16_t, block)[64];	88	DECLARE_ALIGNED(16, static int16_t, block)[64];
139	DECLARE_ALIGNED(8, static int16_t, block1)[64];	89	DECLARE_ALIGNED(8, static int16_t, block1)[64];
140		90
@@ -171,19 +121,16 @@ static void permute(int16_t dst[64], const int16_t src[64], · 30 unmodified lines
171	{	121	{
172	int i;	122	int i;
173		123
		124	#if ARCH_X86
		125	if (permute_x86(dst, src, perm_type))
		126	return;
		127	#endif
		128
174	switch (perm_type) {	129	switch (perm_type) {
175	case FF_IDCT_PERM_LIBMPEG2:	130	case FF_IDCT_PERM_LIBMPEG2:
176	for (i = 0; i < 64; i++)	131	for (i = 0; i < 64; i++)
177	dst[(i & 0x38) \| ((i & 6) >> 1) \| ((i & 1) << 2)] = src[i];	132	dst[(i & 0x38) \| ((i & 6) >> 1) \| ((i & 1) << 2)] = src[i];
178	break;	133	break;
179	case FF_IDCT_PERM_SIMPLE:
180	for (i = 0; i < 64; i++)
181	dst[idct_simple_mmx_perm[i]] = src[i];
182	break;
183	case FF_IDCT_PERM_SSE2:
184	for (i = 0; i < 64; i++)
185	dst[(i & 0x38) \| idct_sse2_row_perm[i & 7]] = src[i];
186	break;
187	case FF_IDCT_PERM_PARTTRANS:	134	case FF_IDCT_PERM_PARTTRANS:
188	for (i = 0; i < 64; i++)	135	for (i = 0; i < 64; i++)
189	dst[(i & 0x24) \| ((i & 3) << 3) \| ((i >> 3) & 3)] = src[i];	136	dst[(i & 0x24) \| ((i & 3) << 3) \| ((i >> 3) & 3)] = src[i];
@@ -489,11 +436,21 @@ int main(int argc, char **argv) · 299 unmodified lines
489	idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);	436	idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);
490	} else {	437	} else {
491	const int cpu_flags = av_get_cpu_flags();	438	const int cpu_flags = av_get_cpu_flags();
492	const struct algo *algos = test_idct ? idct_tab : fdct_tab;	439	if (test_idct) {
493	for (i = 0; algos[i].name; i++)	440	for (i = 0; i < FF_ARRAY_ELEMS(idct_tab); i++)
494	if (!(~cpu_flags & algos[i].cpu_flag)) {	441	err \|= dct_error(&idct_tab[i], test, test_idct, speed);
495	err \|= dct_error(&algos[i], test, test_idct, speed);	442
496	}	443	for (i = 0; idct_tab_arch[i].name; i++)
		444	if (!(~cpu_flags & idct_tab_arch[i].cpu_flag))
		445	err \|= dct_error(&idct_tab_arch[i], test, test_idct, speed);
		446	} else {
		447	for (i = 0; i < FF_ARRAY_ELEMS(fdct_tab); i++)
		448	err \|= dct_error(&fdct_tab[i], test, test_idct, speed);
		449
		450	for (i = 0; fdct_tab_arch[i].name; i++)
		451	if (!(~cpu_flags & fdct_tab_arch[i].cpu_flag))
		452	err \|= dct_error(&fdct_tab_arch[i], test, test_idct, speed);
		453	}
497	}	454	}
498		455
499	if (err)	456	if (err)

libavcodec/ppc/dct-test.c+32−0

@@ -0,0 +1,32 @@
	1	/*
	2	* This file is part of Libav.
	3	*
	4	* Libav is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU Lesser General Public
	6	* License as published by the Free Software Foundation; either
	7	* version 2.1 of the License, or (at your option) any later version.
	8	*
	9	* Libav is distributed in the hope that it will be useful,
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	12	* Lesser General Public License for more details.
	13	*
	14	* You should have received a copy of the GNU Lesser General Public
	15	* License along with Libav; if not, write to the Free Software
	16	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
	17	*/
	18
	19	#include "config.h"
	20
	21	#include "fdct.h"
	22
	23	static const struct algo fdct_tab_arch[] = {
	24	#if HAVE_ALTIVEC
	25	{ "altivecfdct", ff_fdct_altivec, FF_IDCT_PERM_NONE, AV_CPU_FLAG_ALTIVEC },
	26	#endif
	27	{ 0 }
	28	};
	29
	30	static const struct algo idct_tab_arch[] = {
	31	{ 0 }
	32	};

libavcodec/x86/dct-test.c+82−0

@@ -0,0 +1,82 @@
	1	/*
	2	* This file is part of Libav.
	3	*
	4	* Libav is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU Lesser General Public
	6	* License as published by the Free Software Foundation; either
	7	* version 2.1 of the License, or (at your option) any later version.
	8	*
	9	* Libav is distributed in the hope that it will be useful,
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	12	* Lesser General Public License for more details.
	13	*
	14	* You should have received a copy of the GNU Lesser General Public
	15	* License along with Libav; if not, write to the Free Software
	16	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
	17	*/
	18
	19	#include "config.h"
	20
	21	#include "fdct.h"
	22	#include "idct_xvid.h"
	23	#include "simple_idct.h"
	24
	25	static const struct algo fdct_tab_arch[] = {
	26	#if HAVE_MMX_INLINE
	27	{ "MMX", ff_fdct_mmx, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMX },
	28	#endif
	29	#if HAVE_MMXEXT_INLINE
	30	{ "MMXEXT", ff_fdct_mmxext, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMXEXT },
	31	#endif
	32	#if HAVE_SSE2_INLINE
	33	{ "SSE2", ff_fdct_sse2, FF_IDCT_PERM_NONE, AV_CPU_FLAG_SSE2 },
	34	#endif
	35	{ 0 }
	36	};
	37
	38	static const struct algo idct_tab_arch[] = {
	39	#if HAVE_MMX_INLINE
	40	{ "SIMPLE-MMX", ff_simple_idct_mmx, FF_IDCT_PERM_SIMPLE, AV_CPU_FLAG_MMX },
	41	{ "XVID-MMX", ff_idct_xvid_mmx, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMX, 1 },
	42	#endif
	43	#if HAVE_MMXEXT_INLINE
	44	{ "XVID-MMXEXT", ff_idct_xvid_mmxext, FF_IDCT_PERM_NONE, AV_CPU_FLAG_MMXEXT, 1 },
	45	#endif
	46	#if HAVE_SSE2_INLINE
	47	{ "XVID-SSE2", ff_idct_xvid_sse2, FF_IDCT_PERM_SSE2, AV_CPU_FLAG_SSE2, 1 },
	48	#endif
	49	{ 0 }
	50	};
	51
	52	static short idct_simple_mmx_perm[64] = {
	53	0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
	54	0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
	55	0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
	56	0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
	57	0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
	58	0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
	59	0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
	60	0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
	61	};
	62
	63	static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
	64
	65	static int permute_x86(int16_t dst[64], const int16_t src[64],
	66	enum idct_permutation_type perm_type)
	67	{
	68	int i;
	69
	70	switch (perm_type) {
	71	case FF_IDCT_PERM_SIMPLE:
	72	for (i = 0; i < 64; i++)
	73	dst[idct_simple_mmx_perm[i]] = src[i];
	74	return 1;
	75	case FF_IDCT_PERM_SSE2:
	76	for (i = 0; i < 64; i++)
	77	dst[(i & 0x38) \| idct_sse2_row_perm[i & 7]] = src[i];
	78	return 1;
	79	}
	80
	81	return 0;
	82	}

dct-test: Move arch-specific bits into arch-specific subdirectories

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code