x86: dsputil hpeldsp: Move shared template functions into separate object
authorDiego Biurrun <diego@biurrun.de>
Mon, 6 May 2013 22:42:22 +0000 (00:42 +0200)
committerDiego Biurrun <diego@biurrun.de>
Tue, 7 May 2013 16:03:34 +0000 (18:03 +0200)
libavcodec/x86/Makefile
libavcodec/x86/dsputil_mmx.c
libavcodec/x86/dsputil_mmx.h
libavcodec/x86/hpeldsp_init.c
libavcodec/x86/hpeldsp_rnd_template.c
libavcodec/x86/rnd_mmx.c [new file with mode: 0644]
libavcodec/x86/rnd_template.c

index da4798f..f13f3e0 100644 (file)
@@ -36,12 +36,14 @@ MMX-OBJS-$(CONFIG_DSPUTIL)             += x86/dsputil_mmx.o             \
                                           x86/fpel_mmx.o                \
                                           x86/idct_mmx_xvid.o           \
                                           x86/idct_sse2_xvid.o          \
+                                          x86/rnd_mmx.o                 \
                                           x86/simple_idct.o             \
 
 MMX-OBJS-$(CONFIG_ENCODERS)            += x86/dsputilenc_mmx.o          \
                                           x86/motion_est.o
 MMX-OBJS-$(CONFIG_HPELDSP)             += x86/fpel_mmx.o                \
-                                          x86/hpeldsp_mmx.o
+                                          x86/hpeldsp_mmx.o             \
+                                          x86/rnd_mmx.o
 MMX-OBJS-$(CONFIG_VC1_DECODER)         += x86/vc1dsp_mmx.o
 
 YASM-OBJS-$(CONFIG_AAC_DECODER)        += x86/sbrdsp.o
index 46859a6..bd19947 100644 (file)
@@ -99,21 +99,6 @@ void ff_put_no_rnd_mpeg4_qpel8_v_lowpass_mmxext(uint8_t *dst, uint8_t *src,
 
 #if HAVE_INLINE_ASM
 
-/***********************************/
-/* MMX rounding */
-
-#define DEF(x, y) x ## _ ## y ## _mmx
-#define SET_RND  MOVQ_WTWO
-#define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX(a, b, c, d, e, f)
-#define PAVGB(a, b, c, e)               PAVGB_MMX(a, b, c, e)
-
-#include "rnd_template.c"
-
-#undef DEF
-#undef SET_RND
-#undef PAVGBP
-#undef PAVGB
-
 /***********************************/
 /* standard MMX */
 
@@ -877,19 +862,19 @@ QPEL_OP(put_no_rnd_,   ff_pw_15, _no_rnd_, mmxext)
 #if HAVE_INLINE_ASM
 void ff_put_rv40_qpel8_mc33_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
 {
-  put_pixels8_xy2_mmx(dst, src, stride, 8);
+    ff_put_pixels8_xy2_mmx(dst, src, stride, 8);
 }
 void ff_put_rv40_qpel16_mc33_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
 {
-  put_pixels16_xy2_mmx(dst, src, stride, 16);
+    ff_put_pixels16_xy2_mmx(dst, src, stride, 16);
 }
 void ff_avg_rv40_qpel8_mc33_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
 {
-  avg_pixels8_xy2_mmx(dst, src, stride, 8);
+    ff_avg_pixels8_xy2_mmx(dst, src, stride, 8);
 }
 void ff_avg_rv40_qpel16_mc33_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
 {
-  avg_pixels16_xy2_mmx(dst, src, stride, 16);
+    ff_avg_pixels16_xy2_mmx(dst, src, stride, 16);
 }
 
 static void gmc_mmx(uint8_t *dst, uint8_t *src,
index 5fe7366..28cc07b 100644 (file)
@@ -176,6 +176,16 @@ void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
 void ff_avg_pixels8_x2_mmx(uint8_t *block, const uint8_t *pixels,
                            ptrdiff_t line_size, int h);
 
+void ff_avg_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
+                            ptrdiff_t line_size, int h);
+void ff_avg_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
+                             ptrdiff_t line_size, int h);
+
+void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
+                            ptrdiff_t line_size, int h);
+void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
+                             ptrdiff_t line_size, int h);
+
 void ff_put_rv40_qpel8_mc33_mmx(uint8_t *block, uint8_t *pixels, ptrdiff_t stride);
 void ff_put_rv40_qpel16_mc33_mmx(uint8_t *block, uint8_t *pixels, ptrdiff_t stride);
 void ff_avg_rv40_qpel8_mc33_mmx(uint8_t *block, uint8_t *pixels, ptrdiff_t stride);
index 0f30517..739099e 100644 (file)
@@ -77,8 +77,12 @@ void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels,
 #define avg_pixels8_mmx         ff_avg_pixels8_mmx
 #define avg_pixels8_x2_mmx      ff_avg_pixels8_x2_mmx
 #define avg_pixels16_mmx        ff_avg_pixels16_mmx
+#define avg_pixels8_xy2_mmx     ff_avg_pixels8_xy2_mmx
+#define avg_pixels16_xy2_mmx    ff_avg_pixels16_xy2_mmx
 #define put_pixels8_mmx         ff_put_pixels8_mmx
 #define put_pixels16_mmx        ff_put_pixels16_mmx
+#define put_pixels8_xy2_mmx     ff_put_pixels8_xy2_mmx
+#define put_pixels16_xy2_mmx    ff_put_pixels16_xy2_mmx
 #define avg_no_rnd_pixels16_mmx ff_avg_pixels16_mmx
 #define put_no_rnd_pixels8_mmx  ff_put_pixels8_mmx
 #define put_no_rnd_pixels16_mmx ff_put_pixels16_mmx
@@ -91,13 +95,16 @@ void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels,
 #define SET_RND  MOVQ_WONE
 #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
 #define PAVGB(a, b, c, e)               PAVGB_MMX_NO_RND(a, b, c, e)
+#define STATIC static
 
+#include "rnd_template.c"
 #include "hpeldsp_rnd_template.c"
 
 #undef DEF
 #undef SET_RND
 #undef PAVGBP
 #undef PAVGB
+#undef STATIC
 /***********************************/
 /* MMX rounding */
 
index 90619b4..7ad54a7 100644 (file)
@@ -24,8 +24,6 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "rnd_template.c"
-
 // put_pixels
 static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
 {
diff --git a/libavcodec/x86/rnd_mmx.c b/libavcodec/x86/rnd_mmx.c
new file mode 100644 (file)
index 0000000..791b911
--- /dev/null
@@ -0,0 +1,32 @@
+/*
+ * 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 "config.h"
+#include "dsputil_mmx.h"
+
+#if HAVE_INLINE_ASM
+
+#define DEF(x, y) ff_ ## x ## _ ## y ## _mmx
+#define SET_RND  MOVQ_WTWO
+#define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX(a, b, c, d, e, f)
+#define PAVGB(a, b, c, e)               PAVGB_MMX(a, b, c, e)
+#define STATIC
+
+#include "rnd_template.c"
+
+#endif /* HAVE_INLINE_ASM */
index 03f2a0d..fb275f7 100644 (file)
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <stddef.h>
+#include <stdint.h>
+
 // put_pixels
-static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
+STATIC void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels,
+                                  ptrdiff_t line_size, int h)
 {
     MOVQ_ZERO(mm7);
     SET_RND(mm6); // =2 for rnd  and  =1 for no_rnd version
@@ -93,7 +97,8 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff
 
 // avg_pixels
 // this routine is 'slightly' suboptimal but mostly unused
-static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
+STATIC void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels,
+                                  ptrdiff_t line_size, int h)
 {
     MOVQ_ZERO(mm7);
     SET_RND(mm6); // =2 for rnd  and  =1 for no_rnd version
@@ -168,12 +173,16 @@ static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff
 }
 
 //FIXME optimize
-static void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h){
+STATIC void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels,
+                                   ptrdiff_t line_size, int h)
+{
     DEF(put, pixels8_xy2)(block  , pixels  , line_size, h);
     DEF(put, pixels8_xy2)(block+8, pixels+8, line_size, h);
 }
 
-static void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h){
+STATIC void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels,
+                                   ptrdiff_t line_size, int h)
+{
     DEF(avg, pixels8_xy2)(block  , pixels  , line_size, h);
     DEF(avg, pixels8_xy2)(block+8, pixels+8, line_size, h);
 }