x86/mlpdec: add ff_mlp_rematrix_channel_{sse4,avx2}

2x to 2.5x faster than the C version.

Reviewed-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: James Almer <jamrial@gmail.com>

1 files changed, 206 insertions, 0 deletions

diff --git a/libavcodec/x86/mlpdsp_init.c b/libavcodec/x86/mlpdsp_init.c
new file mode 100644
index 0000000000..dc0bc585c7
--- /dev/null
+++ b/libavcodec/x86/mlpdsp_init.c
@@ -0,0 +1,206 @@
+/*
+ * MLP DSP functions x86-optimized
+ * Copyright (c) 2009 Ramiro Polla
+ *
+ * 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
+ */
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/x86/asm.h"
+#include "libavutil/x86/cpu.h"
+#include "libavcodec/mlpdsp.h"
+#include "libavcodec/mlp.h"
+
+#define REMATRIX_CHANNEL_FUNC(opt) \
+void ff_mlp_rematrix_channel_##opt(int32_t *samples, \
+                                   const int32_t *coeffs, \
+                                   const uint8_t *bypassed_lsbs, \
+                                   const int8_t *noise_buffer, \
+                                   int index, \
+                                   unsigned int dest_ch, \
+                                   uint16_t blockpos, \
+                                   unsigned int maxchan, \
+                                   int matrix_noise_shift, \
+                                   int access_unit_size_pow2, \
+                                   int32_t mask);
+
+REMATRIX_CHANNEL_FUNC(sse4)
+REMATRIX_CHANNEL_FUNC(avx2_bmi2)
+
+#if HAVE_7REGS && HAVE_INLINE_ASM && HAVE_INLINE_ASM_NONLOCAL_LABELS
+
+extern char ff_mlp_firorder_8;
+extern char ff_mlp_firorder_7;
+extern char ff_mlp_firorder_6;
+extern char ff_mlp_firorder_5;
+extern char ff_mlp_firorder_4;
+extern char ff_mlp_firorder_3;
+extern char ff_mlp_firorder_2;
+extern char ff_mlp_firorder_1;
+extern char ff_mlp_firorder_0;
+
+extern char ff_mlp_iirorder_4;
+extern char ff_mlp_iirorder_3;
+extern char ff_mlp_iirorder_2;
+extern char ff_mlp_iirorder_1;
+extern char ff_mlp_iirorder_0;
+
+static const void * const firtable[9] = { &ff_mlp_firorder_0, &ff_mlp_firorder_1,
+                                   &ff_mlp_firorder_2, &ff_mlp_firorder_3,
+                                   &ff_mlp_firorder_4, &ff_mlp_firorder_5,
+                                   &ff_mlp_firorder_6, &ff_mlp_firorder_7,
+                                   &ff_mlp_firorder_8 };
+static const void * const iirtable[5] = { &ff_mlp_iirorder_0, &ff_mlp_iirorder_1,
+                                   &ff_mlp_iirorder_2, &ff_mlp_iirorder_3,
+                                   &ff_mlp_iirorder_4 };
+
+#if ARCH_X86_64
+
+#define MLPMUL(label, offset, offs, offc)   \
+    LABEL_MANGLE(label)":             \n\t" \
+    "movslq "offset"+"offs"(%0), %%rax\n\t" \
+    "movslq "offset"+"offc"(%1), %%rdx\n\t" \
+    "imul                 %%rdx, %%rax\n\t" \
+    "add                  %%rax, %%rsi\n\t"
+
+#define FIRMULREG(label, offset, firc)\
+    LABEL_MANGLE(label)":       \n\t" \
+    "movslq "#offset"(%0), %%rax\n\t" \
+    "imul        %"#firc", %%rax\n\t" \
+    "add            %%rax, %%rsi\n\t"
+
+#define CLEAR_ACCUM                   \
+    "xor            %%rsi, %%rsi\n\t"
+
+#define SHIFT_ACCUM                   \
+    "shr     %%cl,         %%rsi\n\t"
+
+#define ACCUM    "%%rdx"
+#define RESULT   "%%rsi"
+#define RESULT32 "%%esi"
+
+#else /* if ARCH_X86_32 */
+
+#define MLPMUL(label, offset, offs, offc)  \
+    LABEL_MANGLE(label)":            \n\t" \
+    "mov   "offset"+"offs"(%0), %%eax\n\t" \
+    "imull "offset"+"offc"(%1)       \n\t" \
+    "add                %%eax , %%esi\n\t" \
+    "adc                %%edx , %%ecx\n\t"
+
+#define FIRMULREG(label, offset, firc)  \
+    MLPMUL(label, #offset, "0", "0")
+
+#define CLEAR_ACCUM                  \
+    "xor           %%esi, %%esi\n\t" \
+    "xor           %%ecx, %%ecx\n\t"
+
+#define SHIFT_ACCUM                  \
+    "mov           %%ecx, %%edx\n\t" \
+    "mov           %%esi, %%eax\n\t" \
+    "movzbl        %7   , %%ecx\n\t" \
+    "shrd    %%cl, %%edx, %%eax\n\t" \
+
+#define ACCUM    "%%edx"
+#define RESULT   "%%eax"
+#define RESULT32 "%%eax"
+
+#endif /* !ARCH_X86_64 */
+
+#define BINC  AV_STRINGIFY(4* MAX_CHANNELS)
+#define IOFFS AV_STRINGIFY(4*(MAX_FIR_ORDER + MAX_BLOCKSIZE))
+#define IOFFC AV_STRINGIFY(4* MAX_FIR_ORDER)
+
+#define FIRMUL(label, offset) MLPMUL(label, #offset,   "0",   "0")
+#define IIRMUL(label, offset) MLPMUL(label, #offset, IOFFS, IOFFC)
+
+static void mlp_filter_channel_x86(int32_t *state, const int32_t *coeff,
+                                   int firorder, int iirorder,
+                                   unsigned int filter_shift, int32_t mask,
+                                   int blocksize, int32_t *sample_buffer)
+{
+    const void *firjump = firtable[firorder];
+    const void *iirjump = iirtable[iirorder];
+
+    blocksize = -blocksize;
+
+    __asm__ volatile(
+        "1:                           \n\t"
+        CLEAR_ACCUM
+        "jmp  *%5                     \n\t"
+        FIRMUL   (ff_mlp_firorder_8, 0x1c   )
+        FIRMUL   (ff_mlp_firorder_7, 0x18   )
+        FIRMUL   (ff_mlp_firorder_6, 0x14   )
+        FIRMUL   (ff_mlp_firorder_5, 0x10   )
+        FIRMUL   (ff_mlp_firorder_4, 0x0c   )
+        FIRMULREG(ff_mlp_firorder_3, 0x08,10)
+        FIRMULREG(ff_mlp_firorder_2, 0x04, 9)
+        FIRMULREG(ff_mlp_firorder_1, 0x00, 8)
+        LABEL_MANGLE(ff_mlp_firorder_0)":\n\t"
+        "jmp  *%6                     \n\t"
+        IIRMUL   (ff_mlp_iirorder_4, 0x0c   )
+        IIRMUL   (ff_mlp_iirorder_3, 0x08   )
+        IIRMUL   (ff_mlp_iirorder_2, 0x04   )
+        IIRMUL   (ff_mlp_iirorder_1, 0x00   )
+        LABEL_MANGLE(ff_mlp_iirorder_0)":\n\t"
+        SHIFT_ACCUM
+        "mov  "RESULT"  ,"ACCUM"      \n\t"
+        "add  (%2)      ,"RESULT"     \n\t"
+        "and   %4       ,"RESULT"     \n\t"
+        "sub   $4       ,  %0         \n\t"
+        "mov  "RESULT32", (%0)        \n\t"
+        "mov  "RESULT32", (%2)        \n\t"
+        "add $"BINC"    ,  %2         \n\t"
+        "sub  "ACCUM"   ,"RESULT"     \n\t"
+        "mov  "RESULT32","IOFFS"(%0)  \n\t"
+        "incl              %3         \n\t"
+        "js 1b                        \n\t"
+        : /* 0*/"+r"(state),
+          /* 1*/"+r"(coeff),
+          /* 2*/"+r"(sample_buffer),
+#if ARCH_X86_64
+          /* 3*/"+r"(blocksize)
+        : /* 4*/"r"((x86_reg)mask), /* 5*/"r"(firjump),
+          /* 6*/"r"(iirjump)      , /* 7*/"c"(filter_shift)
+        , /* 8*/"r"((int64_t)coeff[0])
+        , /* 9*/"r"((int64_t)coeff[1])
+        , /*10*/"r"((int64_t)coeff[2])
+        : "rax", "rdx", "rsi"
+#else /* ARCH_X86_32 */
+          /* 3*/"+m"(blocksize)
+        : /* 4*/"m"(         mask), /* 5*/"m"(firjump),
+          /* 6*/"m"(iirjump)      , /* 7*/"m"(filter_shift)
+        : "eax", "edx", "esi", "ecx"
+#endif /* !ARCH_X86_64 */
+    );
+}
+
+#endif /* HAVE_7REGS && HAVE_INLINE_ASM */
+
+av_cold void ff_mlpdsp_init_x86(MLPDSPContext *c)
+{
+    int cpu_flags = av_get_cpu_flags();
+#if HAVE_7REGS && HAVE_INLINE_ASM && HAVE_INLINE_ASM_NONLOCAL_LABELS
+    if (INLINE_MMX(cpu_flags))
+        c->mlp_filter_channel = mlp_filter_channel_x86;
+#endif
+    if (ARCH_X86_64 && EXTERNAL_SSE4(cpu_flags))
+        c->mlp_rematrix_channel = ff_mlp_rematrix_channel_sse4;
+    if (ARCH_X86_64 && EXTERNAL_AVX2(cpu_flags) && cpu_flags & AV_CPU_FLAG_BMI2)
+        c->mlp_rematrix_channel = ff_mlp_rematrix_channel_avx2_bmi2;
+}