dsputil: Split clear_block*/fill_block* off into a separate context

5 files changed, 121 insertions, 78 deletions

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 5fddf3fb83..222a0ff9eb 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -44,6 +44,7 @@ OBJS-$(CONFIG_VP7_DECODER)             += x86/vp8dsp_init.o
 OBJS-$(CONFIG_VP8_DECODER)             += x86/vp8dsp_init.o
 OBJS-$(CONFIG_VP9_DECODER)             += x86/vp9dsp_init.o
 
+MMX-OBJS-$(CONFIG_BLOCKDSP)            += x86/blockdsp_mmx.o
 MMX-OBJS-$(CONFIG_DSPUTIL)             += x86/dsputil_mmx.o             \
                                           x86/idct_mmx_xvid.o           \
                                           x86/idct_sse2_xvid.o          \
diff --git a/libavcodec/x86/blockdsp_mmx.c b/libavcodec/x86/blockdsp_mmx.c
new file mode 100644
index 0000000000..b5294242ab
--- /dev/null
+++ b/libavcodec/x86/blockdsp_mmx.c
@@ -0,0 +1,120 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "config.h"
+#include "libavutil/attributes.h"
+#include "libavutil/internal.h"
+#include "libavutil/cpu.h"
+#include "libavutil/x86/asm.h"
+#include "libavutil/x86/cpu.h"
+#include "libavcodec/blockdsp.h"
+#include "libavcodec/version.h"
+
+#if HAVE_INLINE_ASM
+
+#define CLEAR_BLOCKS(name, n)                           \
+static void name(int16_t *blocks)                       \
+{                                                       \
+    __asm__ volatile (                                  \
+        "pxor %%mm7, %%mm7              \n\t"           \
+        "mov     %1,        %%"REG_a"   \n\t"           \
+        "1:                             \n\t"           \
+        "movq %%mm7,   (%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7,  8(%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7, 16(%0, %%"REG_a")  \n\t"           \
+        "movq %%mm7, 24(%0, %%"REG_a")  \n\t"           \
+        "add    $32, %%"REG_a"          \n\t"           \
+        "js      1b                     \n\t"           \
+        :: "r"(((uint8_t *) blocks) + 128 * n),         \
+           "i"(-128 * n)                                \
+        : "%"REG_a);                                    \
+}
+CLEAR_BLOCKS(clear_blocks_mmx, 6)
+CLEAR_BLOCKS(clear_block_mmx, 1)
+
+static void clear_block_sse(int16_t *block)
+{
+    __asm__ volatile (
+        "xorps  %%xmm0, %%xmm0          \n"
+        "movaps %%xmm0,    (%0)         \n"
+        "movaps %%xmm0,  16(%0)         \n"
+        "movaps %%xmm0,  32(%0)         \n"
+        "movaps %%xmm0,  48(%0)         \n"
+        "movaps %%xmm0,  64(%0)         \n"
+        "movaps %%xmm0,  80(%0)         \n"
+        "movaps %%xmm0,  96(%0)         \n"
+        "movaps %%xmm0, 112(%0)         \n"
+        :: "r" (block)
+        : "memory");
+}
+
+static void clear_blocks_sse(int16_t *blocks)
+{
+    __asm__ volatile (
+        "xorps  %%xmm0, %%xmm0              \n"
+        "mov        %1,         %%"REG_a"   \n"
+        "1:                                 \n"
+        "movaps %%xmm0,    (%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  16(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  32(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  48(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  64(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  80(%0, %%"REG_a")  \n"
+        "movaps %%xmm0,  96(%0, %%"REG_a")  \n"
+        "movaps %%xmm0, 112(%0, %%"REG_a")  \n"
+        "add      $128,         %%"REG_a"   \n"
+        "js         1b                      \n"
+        :: "r"(((uint8_t *) blocks) + 128 * 6), "i"(-128 * 6)
+        : "%"REG_a);
+}
+
+#endif /* HAVE_INLINE_ASM */
+
+#if FF_API_XVMC
+av_cold void ff_blockdsp_init_x86(BlockDSPContext *c, unsigned high_bit_depth,
+                                  AVCodecContext *avctx)
+#else
+av_cold void ff_blockdsp_init_x86(BlockDSPContext *c, unsigned high_bit_depth)
+#endif /* FF_API_XVMC */
+{
+#if HAVE_INLINE_ASM
+    int cpu_flags = av_get_cpu_flags();
+
+    if (!high_bit_depth) {
+        if (INLINE_MMX(cpu_flags)) {
+            c->clear_block  = clear_block_mmx;
+            c->clear_blocks = clear_blocks_mmx;
+        }
+
+#if FF_API_XVMC
+FF_DISABLE_DEPRECATION_WARNINGS
+    /* XvMCCreateBlocks() may not allocate 16-byte aligned blocks */
+    if (CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)
+        return;
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif /* FF_API_XVMC */
+
+        if (INLINE_SSE(cpu_flags)) {
+            c->clear_block  = clear_block_sse;
+            c->clear_blocks = clear_blocks_sse;
+        }
+    }
+#endif /* HAVE_INLINE_ASM */
+}
diff --git a/libavcodec/x86/dsputil_init.c b/libavcodec/x86/dsputil_init.c
index 389e7634dd..a19b83d83c 100644
--- a/libavcodec/x86/dsputil_init.c
+++ b/libavcodec/x86/dsputil_init.c
@@ -19,12 +19,10 @@
 #include "config.h"
 #include "libavutil/attributes.h"
 #include "libavutil/cpu.h"
-#include "libavutil/internal.h"
 #include "libavutil/x86/cpu.h"
 #include "libavcodec/avcodec.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/simple_idct.h"
-#include "libavcodec/version.h"
 #include "dsputil_x86.h"
 #include "idct_xvid.h"
 
@@ -54,8 +52,6 @@ static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx,
     c->add_pixels_clamped        = ff_add_pixels_clamped_mmx;
 
     if (!high_bit_depth) {
-        c->clear_block  = ff_clear_block_mmx;
-        c->clear_blocks = ff_clear_blocks_mmx;
         c->draw_edges   = ff_draw_edges_mmx;
 
         switch (avctx->idct_algo) {
@@ -103,19 +99,6 @@ static av_cold void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx,
 {
 #if HAVE_SSE_INLINE
     c->vector_clipf = ff_vector_clipf_sse;
-
-#if FF_API_XVMC
-FF_DISABLE_DEPRECATION_WARNINGS
-    /* XvMCCreateBlocks() may not allocate 16-byte aligned blocks */
-    if (CONFIG_MPEG_XVMC_DECODER && avctx->xvmc_acceleration > 1)
-        return;
-FF_ENABLE_DEPRECATION_WARNINGS
-#endif /* FF_API_XVMC */
-
-    if (!high_bit_depth) {
-        c->clear_block  = ff_clear_block_sse;
-        c->clear_blocks = ff_clear_blocks_sse;
-    }
 #endif /* HAVE_SSE_INLINE */
 }
 
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index c17f8d00d5..fd74efeb3d 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -166,62 +166,6 @@ void ff_add_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
     } while (--i);
 }
 
-#define CLEAR_BLOCKS(name, n)                           \
-void name(int16_t *blocks)                              \
-{                                                       \
-    __asm__ volatile (                                  \
-        "pxor %%mm7, %%mm7              \n\t"           \
-        "mov     %1,        %%"REG_a"   \n\t"           \
-        "1:                             \n\t"           \
-        "movq %%mm7,   (%0, %%"REG_a")  \n\t"           \
-        "movq %%mm7,  8(%0, %%"REG_a")  \n\t"           \
-        "movq %%mm7, 16(%0, %%"REG_a")  \n\t"           \
-        "movq %%mm7, 24(%0, %%"REG_a")  \n\t"           \
-        "add    $32, %%"REG_a"          \n\t"           \
-        "js      1b                     \n\t"           \
-        :: "r"(((uint8_t *) blocks) + 128 * n),         \
-           "i"(-128 * n)                                \
-        : "%"REG_a);                                    \
-}
-CLEAR_BLOCKS(ff_clear_blocks_mmx, 6)
-CLEAR_BLOCKS(ff_clear_block_mmx, 1)
-
-void ff_clear_block_sse(int16_t *block)
-{
-    __asm__ volatile (
-        "xorps  %%xmm0, %%xmm0          \n"
-        "movaps %%xmm0,    (%0)         \n"
-        "movaps %%xmm0,  16(%0)         \n"
-        "movaps %%xmm0,  32(%0)         \n"
-        "movaps %%xmm0,  48(%0)         \n"
-        "movaps %%xmm0,  64(%0)         \n"
-        "movaps %%xmm0,  80(%0)         \n"
-        "movaps %%xmm0,  96(%0)         \n"
-        "movaps %%xmm0, 112(%0)         \n"
-        :: "r" (block)
-        : "memory");
-}
-
-void ff_clear_blocks_sse(int16_t *blocks)
-{
-    __asm__ volatile (
-        "xorps  %%xmm0, %%xmm0              \n"
-        "mov        %1,         %%"REG_a"   \n"
-        "1:                                 \n"
-        "movaps %%xmm0,    (%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  16(%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  32(%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  48(%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  64(%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  80(%0, %%"REG_a")  \n"
-        "movaps %%xmm0,  96(%0, %%"REG_a")  \n"
-        "movaps %%xmm0, 112(%0, %%"REG_a")  \n"
-        "add      $128,         %%"REG_a"   \n"
-        "js         1b                      \n"
-        :: "r"(((uint8_t *) blocks) + 128 * 6), "i"(-128 * 6)
-        : "%"REG_a);
-}
-
 /* Draw the edges of width 'w' of an image of size width, height
  * this MMX version can only handle w == 8 || w == 16. */
 void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
diff --git a/libavcodec/x86/dsputil_x86.h b/libavcodec/x86/dsputil_x86.h
index a4bc8c2730..e99b6b7630 100644
--- a/libavcodec/x86/dsputil_x86.h
+++ b/libavcodec/x86/dsputil_x86.h
@@ -38,11 +38,6 @@ void ff_put_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
 void ff_put_signed_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
                                       int line_size);
 
-void ff_clear_block_mmx(int16_t *block);
-void ff_clear_block_sse(int16_t *block);
-void ff_clear_blocks_mmx(int16_t *blocks);
-void ff_clear_blocks_sse(int16_t *blocks);
-
 void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
                        int w, int h, int sides);