ffmpeg/libavutil/tx_priv.h

314 lines
14 KiB
C
Raw Normal View History

/*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVUTIL_TX_PRIV_H
#define AVUTIL_TX_PRIV_H
#include "tx.h"
#include "thread.h"
#include "mem_internal.h"
#include "attributes.h"
#ifdef TX_FLOAT
#define TX_TAB(x) x ## _float
#define TX_NAME(x) x ## _float_c
#define TX_NAME_STR(x) NULL_IF_CONFIG_SMALL(x "_float_c")
#define TX_TYPE(x) AV_TX_FLOAT_ ## x
#define TX_FN_NAME(fn, suffix) ff_tx_ ## fn ## _float_ ## suffix
#define TX_FN_NAME_STR(fn, suffix) NULL_IF_CONFIG_SMALL(#fn "_float_" #suffix)
#define MULT(x, m) ((x) * (m))
#define SCALE_TYPE float
typedef float TXSample;
typedef AVComplexFloat TXComplex;
#elif defined(TX_DOUBLE)
#define TX_TAB(x) x ## _double
#define TX_NAME(x) x ## _double_c
#define TX_NAME_STR(x) NULL_IF_CONFIG_SMALL(x "_double_c")
#define TX_TYPE(x) AV_TX_DOUBLE_ ## x
#define TX_FN_NAME(fn, suffix) ff_tx_ ## fn ## _double_ ## suffix
#define TX_FN_NAME_STR(fn, suffix) NULL_IF_CONFIG_SMALL(#fn "_double_" #suffix)
#define MULT(x, m) ((x) * (m))
#define SCALE_TYPE double
typedef double TXSample;
typedef AVComplexDouble TXComplex;
#elif defined(TX_INT32)
#define TX_TAB(x) x ## _int32
#define TX_NAME(x) x ## _int32_c
#define TX_NAME_STR(x) NULL_IF_CONFIG_SMALL(x "_int32_c")
#define TX_TYPE(x) AV_TX_INT32_ ## x
#define TX_FN_NAME(fn, suffix) ff_tx_ ## fn ## _int32_ ## suffix
#define TX_FN_NAME_STR(fn, suffix) NULL_IF_CONFIG_SMALL(#fn "_int32_" #suffix)
#define MULT(x, m) (((((int64_t)(x)) * (int64_t)(m)) + 0x40000000) >> 31)
#define SCALE_TYPE float
typedef int32_t TXSample;
typedef AVComplexInt32 TXComplex;
#else
typedef void TXComplex;
#endif
#define TX_DECL_FN(fn, suffix) \
void TX_FN_NAME(fn, suffix)(AVTXContext *s, void *o, void *i, ptrdiff_t st);
#define TX_DEF(fn, tx_type, len_min, len_max, f1, f2, \
p, init_fn, suffix, cf, cd_flags, cf2) \
&(const FFTXCodelet){ \
.name = TX_FN_NAME_STR(fn, suffix), \
.function = TX_FN_NAME(fn, suffix), \
.type = TX_TYPE(tx_type), \
.flags = FF_TX_ALIGNED | FF_TX_OUT_OF_PLACE | cd_flags, \
.factors = { f1, f2 }, \
.min_len = len_min, \
.max_len = len_max, \
.init = init_fn, \
.cpu_flags = cf2 | AV_CPU_FLAG_ ## cf, \
.prio = p, \
}
#if defined(TX_FLOAT) || defined(TX_DOUBLE)
#define CMUL(dre, dim, are, aim, bre, bim) \
do { \
(dre) = (are) * (bre) - (aim) * (bim); \
(dim) = (are) * (bim) + (aim) * (bre); \
} while (0)
#define SMUL(dre, dim, are, aim, bre, bim) \
do { \
(dre) = (are) * (bre) - (aim) * (bim); \
(dim) = (are) * (bim) - (aim) * (bre); \
} while (0)
#define UNSCALE(x) (x)
#define RESCALE(x) (x)
#define FOLD(a, b) ((a) + (b))
#elif defined(TX_INT32)
/* Properly rounds the result */
#define CMUL(dre, dim, are, aim, bre, bim) \
do { \
int64_t accu; \
(accu) = (int64_t)(bre) * (are); \
(accu) -= (int64_t)(bim) * (aim); \
(dre) = (int)(((accu) + 0x40000000) >> 31); \
(accu) = (int64_t)(bim) * (are); \
(accu) += (int64_t)(bre) * (aim); \
(dim) = (int)(((accu) + 0x40000000) >> 31); \
} while (0)
#define SMUL(dre, dim, are, aim, bre, bim) \
do { \
int64_t accu; \
(accu) = (int64_t)(bre) * (are); \
(accu) -= (int64_t)(bim) * (aim); \
(dre) = (int)(((accu) + 0x40000000) >> 31); \
(accu) = (int64_t)(bim) * (are); \
(accu) -= (int64_t)(bre) * (aim); \
(dim) = (int)(((accu) + 0x40000000) >> 31); \
} while (0)
#define UNSCALE(x) ((double)(x)/2147483648.0)
#define RESCALE(x) (av_clip64(lrintf((x) * 2147483648.0), INT32_MIN, INT32_MAX))
#define FOLD(x, y) ((int32_t)((x) + (unsigned)(y) + 32) >> 6)
#endif /* TX_INT32 */
#define BF(x, y, a, b) \
do { \
x = (a) - (b); \
y = (a) + (b); \
} while (0)
#define CMUL3(c, a, b) CMUL((c).re, (c).im, (a).re, (a).im, (b).re, (b).im)
/* Codelet flags, used to pick codelets. Must be a superset of enum AVTXFlags,
* but if it runs out of bits, it can be made separate. */
#define FF_TX_OUT_OF_PLACE (1ULL << 63) /* Can be OR'd with AV_TX_INPLACE */
#define FF_TX_ALIGNED (1ULL << 62) /* Cannot be OR'd with AV_TX_UNALIGNED */
#define FF_TX_PRESHUFFLE (1ULL << 61) /* Codelet expects permuted coeffs */
#define FF_TX_INVERSE_ONLY (1ULL << 60) /* For non-orthogonal inverse-only transforms */
#define FF_TX_FORWARD_ONLY (1ULL << 59) /* For non-orthogonal forward-only transforms */
typedef enum FFTXCodeletPriority {
FF_TX_PRIO_BASE = 0, /* Baseline priority */
/* For SIMD, set base prio to the register size in bits and increment in
* steps of 64 depending on faster/slower features, like FMA. */
FF_TX_PRIO_MIN = -131072, /* For naive implementations */
FF_TX_PRIO_MAX = 32768, /* For custom implementations/ASICs */
} FFTXCodeletPriority;
/* Codelet options */
typedef struct FFTXCodeletOptions {
int invert_lookup; /* If codelet is flagged as FF_TX_CODELET_PRESHUFFLE,
invert the lookup direction for the map generated */
} FFTXCodeletOptions;
/* Maximum amount of subtransform functions, subtransforms and factors. Arbitrary. */
#define TX_MAX_SUB 4
typedef struct FFTXCodelet {
const char *name; /* Codelet name, for debugging */
av_tx_fn function; /* Codelet function, != NULL */
enum AVTXType type; /* Type of codelet transform */
#define TX_TYPE_ANY INT32_MAX /* Special type to allow all types */
uint64_t flags; /* A combination of AVTXFlags and codelet
* flags that describe its properties. */
int factors[TX_MAX_SUB]; /* Length factors */
#define TX_FACTOR_ANY -1 /* When used alone, signals that the codelet
* supports all factors. Otherwise, if other
* factors are present, it signals that whatever
* remains will be supported, as long as the
* other factors are a component of the length */
int min_len; /* Minimum length of transform, must be >= 1 */
int max_len; /* Maximum length of transform */
#define TX_LEN_UNLIMITED -1 /* Special length value to permit all lengths */
int (*init)(AVTXContext *s, /* Optional callback for current context initialization. */
const struct FFTXCodelet *cd,
uint64_t flags,
FFTXCodeletOptions *opts,
int len, int inv,
const void *scale);
int (*uninit)(AVTXContext *s); /* Optional callback for uninitialization. */
int cpu_flags; /* CPU flags. If any negative flags like
* SLOW are present, will avoid picking.
* 0x0 to signal it's a C codelet */
#define FF_TX_CPU_FLAGS_ALL 0x0 /* Special CPU flag for C */
int prio; /* < 0 = least, 0 = no pref, > 0 = prefer */
} FFTXCodelet;
struct AVTXContext {
/* Fields the root transform and subtransforms use or may use.
* NOTE: This section is used by assembly, do not reorder or change */
int len; /* Length of the transform */
int inv; /* If transform is inverse */
int *map; /* Lookup table(s) */
TXComplex *exp; /* Any non-pre-baked multiplication factors needed */
TXComplex *tmp; /* Temporary buffer, if needed */
AVTXContext *sub; /* Subtransform context(s), if needed */
av_tx_fn fn[TX_MAX_SUB]; /* Function(s) for the subtransforms */
int nb_sub; /* Number of subtransforms.
* The reason all of these are set here
* rather than in each separate context
* is to eliminate extra pointer
* dereferences. */
/* Fields mainly useul/applicable for the root transform or initialization.
* Fields below are not used by assembly code. */
const FFTXCodelet *cd[TX_MAX_SUB]; /* Subtransform codelets */
const FFTXCodelet *cd_self; /* Codelet for the current context */
enum AVTXType type; /* Type of transform */
uint64_t flags; /* A combination of AVTXFlags and
* codelet flags used when creating */
float scale_f;
double scale_d;
void *opaque; /* Free to use by implementations */
};
/* Create a subtransform in the current context with the given parameters.
* The flags parameter from FFTXCodelet.init() should be preserved as much
* as that's possible.
* MUST be called during the sub() callback of each codelet. */
int ff_tx_init_subtx(AVTXContext *s, enum AVTXType type,
uint64_t flags, FFTXCodeletOptions *opts,
int len, int inv, const void *scale);
/*
* Generates the PFA permutation table into AVTXContext->pfatab. The end table
* is appended to the start table.
*/
int ff_tx_gen_compound_mapping(AVTXContext *s, int n, int m);
/*
* Generates a standard-ish (slightly modified) Split-Radix revtab into
* AVTXContext->map. Invert lookup changes how the mapping needs to be applied.
* If it's set to 0, it has to be applied like out[map[i]] = in[i], otherwise
* if it's set to 1, has to be applied as out[i] = in[map[i]]
*/
int ff_tx_gen_ptwo_revtab(AVTXContext *s, int invert_lookup);
/*
* Generates an index into AVTXContext->inplace_idx that if followed in the
* specific order, allows the revtab to be done in-place. The sub-transform
* and its map should already be initialized.
*/
int ff_tx_gen_ptwo_inplace_revtab_idx(AVTXContext *s);
/*
* This generates a parity-based revtab of length len and direction inv.
*
* Parity means even and odd complex numbers will be split, e.g. the even
* coefficients will come first, after which the odd coefficients will be
* placed. For example, a 4-point transform's coefficients after reordering:
* z[0].re, z[0].im, z[2].re, z[2].im, z[1].re, z[1].im, z[3].re, z[3].im
*
* The basis argument is the length of the largest non-composite transform
* supported, and also implies that the basis/2 transform is supported as well,
* as the split-radix algorithm requires it to be.
*
* The dual_stride argument indicates that both the basis, as well as the
* basis/2 transforms support doing two transforms at once, and the coefficients
* will be interleaved between each pair in a split-radix like so (stride == 2):
* tx1[0], tx1[2], tx2[0], tx2[2], tx1[1], tx1[3], tx2[1], tx2[3]
* A non-zero number switches this on, with the value indicating the stride
* (how many values of 1 transform to put first before switching to the other).
* Must be a power of two or 0. Must be less than the basis.
* Value will be clipped to the transform size, so for a basis of 16 and a
* dual_stride of 8, dual 8-point transforms will be laid out as if dual_stride
* was set to 4.
* Usually you'll set this to half the complex numbers that fit in a single
* register or 0. This allows to reuse SSE functions as dual-transform
* functions in AVX mode.
*
* If length is smaller than basis/2 this function will not do anything.
*/
int ff_tx_gen_split_radix_parity_revtab(AVTXContext *s, int invert_lookup,
int basis, int dual_stride);
/* Typed init function to initialize shared tables. Will initialize all tables
* for all factors of a length. */
void ff_tx_init_tabs_float (int len);
void ff_tx_init_tabs_double(int len);
void ff_tx_init_tabs_int32 (int len);
/* Typed init function to initialize an MDCT exptab in a context. */
int ff_tx_mdct_gen_exp_float (AVTXContext *s);
int ff_tx_mdct_gen_exp_double(AVTXContext *s);
int ff_tx_mdct_gen_exp_int32 (AVTXContext *s);
/* Lists of codelets */
extern const FFTXCodelet * const ff_tx_codelet_list_float_c [];
extern const FFTXCodelet * const ff_tx_codelet_list_float_x86 [];
extern const FFTXCodelet * const ff_tx_codelet_list_double_c [];
extern const FFTXCodelet * const ff_tx_codelet_list_int32_c [];
lavu/x86: add FFT assembly This commit adds a pure x86 assembly SIMD version of the FFT in libavutil/tx. The design of this pure assembly FFT is pretty unconventional. On the lowest level, instead of splitting the complex numbers into real and imaginary parts, we keep complex numbers together but split them in terms of parity. This saves a number of shuffles in each transform, but more importantly, it splits each transform into two independent paths, which we process using separate registers in parallel. This allows us to keep all units saturated and lets us use all available registers to avoid dependencies. Moreover, it allows us to double the granularity of our per-load permutation, skipping many expensive lookups and allowing us to use just 4 loads per register, rather than 8, or in case FMA3 (and by extension, AVX2), use the vgatherdpd instruction, which is at least as fast as 4 separate loads on old hardware, and quite a bit faster on modern CPUs). Higher up, we go for a bottom-up construction of large transforms, foregoing the traditional per-transform call-return recursion chains. Instead, we always start at the bottom-most basis transform (in this case, a 32-point transform), and continue constructing larger and larger transforms until we return to the top-most transform. This way, we only touch the stack 3 times per a complete target transform: once for the 1/2 length transform and two times for the 1/4 length transform. The combination algorithm we use is a standard Split-Radix algorithm, as used in our C code. Although a version with less operations exists (Steven G. Johnson and Matteo Frigo's "A modified split-radix FFT with fewer arithmetic operations", IEEE Trans. Signal Process. 55 (1), 111–119 (2007), which is the one FFTW uses), it only has 2% less operations and requires at least 4x the binary code (due to it needing 4 different paths to do a single transform). That version also has other issues which prevent it from being implemented with SIMD code as efficiently, which makes it lose the marginal gains it offered, and cannot be performed bottom-up, requiring many recursive call-return chains, whose overhead adds up. We go through a lot of effort to minimize load/stores by keeping as much in registers in between construcring transforms. This saves us around 32 cycles, on paper, but in reality a lot more due to load/store aliasing (a load from a memory location cannot be issued while there's a store pending, and there are only so many (2 for Zen 3) load/store units in a CPU). Also, we interleave coefficients during the last stage to save on a store+load per register. Each of the smallest, basis transforms (4, 8 and 16-point in our case) has been extremely optimized. Our 8-point transform is barely 20 instructions in total, beating our old implementation 8-point transform by 1 instruction. Our 2x8-point transform is 23 instructions, beating our old implementation by 6 instruction and needing 50% less cycles. Our 16-point transform's combination code takes slightly more instructions than our old implementation, but makes up for it by requiring a lot less arithmetic operations. Overall, the transform was optimized for the timings of Zen 3, which at the time of writing has the most IPC from all documented CPUs. Shuffles were preferred over arithmetic operations due to their 1/0.5 latency/throughput. On average, this code is 30% faster than our old libavcodec implementation. It's able to trade blows with the previously-untouchable FFTW on small transforms, and due to its tiny size and better prediction, outdoes FFTW on larger transforms by 11% on the largest currently supported size.
2021-04-10 01:54:22 +00:00
#endif /* AVUTIL_TX_PRIV_H */