x86: MC AVX2

meson.build

@@ -189,20 +189,6 @@ libdav1d_tmpl_sources = files(
     'src/recon.c'
 )
 
-# Build a helper library for each bitdepth
-bitdepth_objs = []
-foreach bitdepth : dav1d_bitdepths
-    bitdepth_lib = static_library(
-        'dav1d_bitdepth_@0@'.format(bitdepth),
-        libdav1d_tmpl_sources, config_h_target,
-        include_directories: dav1d_inc_dirs,
-        c_args: ['-DBITDEPTH=@0@'.format(bitdepth)] + stackalign_flag,
-        install: false,
-        build_by_default: false,
-    )
-    bitdepth_objs += bitdepth_lib.extract_all_objects()
-endforeach
-
 entrypoints_src = files(
     'src/lib.c',
     'src/thread_task.c'
@@ -241,8 +227,12 @@ if is_asm_enabled
     libdav1d_sources += files(
         'src/x86/cpu.c',
     )
+    libdav1d_tmpl_sources += files(
+        'src/x86/mc_init.c',
+    )
     libdav1d_sources_asm = files(
         'src/x86/cpuid.asm',
+        'src/x86/mc.asm',
     )
 
     nasm = find_program('nasm')
@@ -281,6 +271,20 @@ if host_machine.system() == 'windows'
     libdav1d_sources += files('src/win32/thread.c')
 endif
 
+# Build a helper library for each bitdepth
+bitdepth_objs = []
+foreach bitdepth : dav1d_bitdepths
+    bitdepth_lib = static_library(
+        'dav1d_bitdepth_@0@'.format(bitdepth),
+        libdav1d_tmpl_sources, config_h_target,
+        include_directories: dav1d_inc_dirs,
+        c_args: ['-DBITDEPTH=@0@'.format(bitdepth)] + stackalign_flag,
+        install: false,
+        build_by_default: false,
+    )
+    bitdepth_objs += bitdepth_lib.extract_all_objects()
+endforeach
+
 libdav1d = library('dav1d',
     libdav1d_sources, rev_target, nasm_objs,
     version: '0.0.1',

src/mc.c

@@ -530,4 +530,8 @@ void bitfn(dav1d_mc_dsp_init)(Dav1dMCDSPContext *const c) {
     c->w_mask[2] = w_mask_420_c;
     c->warp8x8  = warp_affine_8x8_c;
     c->warp8x8t = warp_affine_8x8t_c;
+
+#if HAVE_ASM && ARCH_X86
+    bitfn(dav1d_mc_dsp_init_x86)(c);
+#endif
 }

src/mc.h

@@ -101,4 +101,7 @@ typedef struct Dav1dMCDSPContext {
 void dav1d_mc_dsp_init_8bpc(Dav1dMCDSPContext *c);
 void dav1d_mc_dsp_init_10bpc(Dav1dMCDSPContext *c);
 
+void dav1d_mc_dsp_init_x86_8bpc(Dav1dMCDSPContext *c);
+void dav1d_mc_dsp_init_x86_10bpc(Dav1dMCDSPContext *c);
+
 #endif /* __DAV1D_SRC_MC_H__ */

src/x86/mc_init.c

+/*
+ * Copyright © 2018, VideoLAN and dav1d authors
+ * Copyright © 2018, Two Orioles, LLC
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice, this
+ *    list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
+ * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "config.h"
+
+#include "src/mc.h"
+#include "src/x86/cpu.h"
+
+decl_mc_fn(dav1d_put_8tap_regular_avx2);
+decl_mc_fn(dav1d_put_8tap_regular_smooth_avx2);
+decl_mc_fn(dav1d_put_8tap_regular_sharp_avx2);
+decl_mc_fn(dav1d_put_8tap_smooth_avx2);
+decl_mc_fn(dav1d_put_8tap_smooth_regular_avx2);
+decl_mc_fn(dav1d_put_8tap_smooth_sharp_avx2);
+decl_mc_fn(dav1d_put_8tap_sharp_avx2);
+decl_mc_fn(dav1d_put_8tap_sharp_regular_avx2);
+decl_mc_fn(dav1d_put_8tap_sharp_smooth_avx2);
+decl_mc_fn(dav1d_put_bilin_avx2);
+
+decl_mct_fn(dav1d_prep_8tap_regular_avx2);
+decl_mct_fn(dav1d_prep_8tap_regular_smooth_avx2);
+decl_mct_fn(dav1d_prep_8tap_regular_sharp_avx2);
+decl_mct_fn(dav1d_prep_8tap_smooth_avx2);
+decl_mct_fn(dav1d_prep_8tap_smooth_regular_avx2);
+decl_mct_fn(dav1d_prep_8tap_smooth_sharp_avx2);
+decl_mct_fn(dav1d_prep_8tap_sharp_avx2);
+decl_mct_fn(dav1d_prep_8tap_sharp_regular_avx2);
+decl_mct_fn(dav1d_prep_8tap_sharp_smooth_avx2);
+decl_mct_fn(dav1d_prep_bilin_avx2);
+
+decl_avg_fn(dav1d_avg_avx2);
+decl_w_avg_fn(dav1d_w_avg_avx2);
+decl_mask_fn(dav1d_mask_avx2);
+decl_w_mask_fn(dav1d_w_mask_420_avx2);
+
+void bitfn(dav1d_mc_dsp_init_x86)(Dav1dMCDSPContext *const c) {
+#define init_mc_fn(type, name, suffix) \
+    c->mc[type] = dav1d_put_##name##_##suffix
+#define init_mct_fn(type, name, suffix) \
+    c->mct[type] = dav1d_prep_##name##_##suffix
+    const enum CpuFlags flags = dav1d_get_cpu_flags_x86();
+
+    if (!(flags & DAV1D_X86_CPU_FLAG_AVX2)) return;
+
+#if BITDEPTH == 8 && ARCH_X86_64 && !defined(_WIN32) // FIXME: Windows
+    init_mc_fn (FILTER_2D_8TAP_REGULAR,        8tap_regular,        avx2);
+    init_mc_fn (FILTER_2D_8TAP_REGULAR_SMOOTH, 8tap_regular_smooth, avx2);
+    init_mc_fn (FILTER_2D_8TAP_REGULAR_SHARP,  8tap_regular_sharp,  avx2);
+    init_mc_fn (FILTER_2D_8TAP_SMOOTH_REGULAR, 8tap_smooth_regular, avx2);
+    init_mc_fn (FILTER_2D_8TAP_SMOOTH,         8tap_smooth,         avx2);
+    init_mc_fn (FILTER_2D_8TAP_SMOOTH_SHARP,   8tap_smooth_sharp,   avx2);
+    init_mc_fn (FILTER_2D_8TAP_SHARP_REGULAR,  8tap_sharp_regular,  avx2);
+    init_mc_fn (FILTER_2D_8TAP_SHARP_SMOOTH,   8tap_sharp_smooth,   avx2);
+    init_mc_fn (FILTER_2D_8TAP_SHARP,          8tap_sharp,          avx2);
+    init_mc_fn (FILTER_2D_BILINEAR,            bilin,               avx2);
+
+    init_mct_fn(FILTER_2D_8TAP_REGULAR,        8tap_regular,        avx2);
+    init_mct_fn(FILTER_2D_8TAP_REGULAR_SMOOTH, 8tap_regular_smooth, avx2);
+    init_mct_fn(FILTER_2D_8TAP_REGULAR_SHARP,  8tap_regular_sharp,  avx2);
+    init_mct_fn(FILTER_2D_8TAP_SMOOTH_REGULAR, 8tap_smooth_regular, avx2);
+    init_mct_fn(FILTER_2D_8TAP_SMOOTH,         8tap_smooth,         avx2);
+    init_mct_fn(FILTER_2D_8TAP_SMOOTH_SHARP,   8tap_smooth_sharp,   avx2);
+    init_mct_fn(FILTER_2D_8TAP_SHARP_REGULAR,  8tap_sharp_regular,  avx2);
+    init_mct_fn(FILTER_2D_8TAP_SHARP_SMOOTH,   8tap_sharp_smooth,   avx2);
+    init_mct_fn(FILTER_2D_8TAP_SHARP,          8tap_sharp,          avx2);
+    init_mct_fn(FILTER_2D_BILINEAR,            bilin,               avx2);
+
+    c->avg = dav1d_avg_avx2;
+    c->w_avg = dav1d_w_avg_avx2;
+    c->mask = dav1d_mask_avx2;
+    c->w_mask[2] = dav1d_w_mask_420_avx2;
+#endif
+}