x86: build: Only compile FDCT code if MMX is enabled

[ffmpeg] / libavcodec / x86 / cavsdsp.c

diff --git a/libavcodec/x86/cavsdsp.c b/libavcodec/x86/cavsdsp.c
index 37e7eec48336b43f5861e8ac835b49ebc064c5d1..f0e8cfcd17a4d54b96f62768c02585efc4c69c9e 100644 (file)
--- a/libavcodec/x86/cavsdsp.c
+++ b/libavcodec/x86/cavsdsp.c
@@ -28,7 +28,10 @@
 #include "libavutil/x86/asm.h"
 #include "libavutil/x86/cpu.h"
 #include "libavcodec/cavsdsp.h"
-#include "dsputil_mmx.h"
+#include "libavcodec/idctdsp.h"
+#include "constants.h"
+#include "fpel.h"
+#include "idctdsp.h"
 #include "config.h"
 
 #if HAVE_MMX_INLINE
@@ -122,6 +125,17 @@ static inline void cavs_idct8_1d(int16_t *block, uint64_t bias)
     );
 }
 
+#define SBUTTERFLY(a,b,t,n,m)\
+    "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
+    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
+    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
+
+#define TRANSPOSE4(a,b,c,d,t)\
+    SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
+    SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
+    SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
+    SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
+
 static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
 {
     int i;
@@ -446,13 +460,37 @@ static void OPNAME ## cavs_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *
 #endif /* (HAVE_MMXEXT_INLINE || HAVE_AMD3DNOW_INLINE) */
 
 #if HAVE_MMX_INLINE
+static void put_cavs_qpel8_mc00_mmx(uint8_t *dst, uint8_t *src,
+                                    ptrdiff_t stride)
+{
+    ff_put_pixels8_mmx(dst, src, stride, 8);
+}
+
+static void avg_cavs_qpel8_mc00_mmx(uint8_t *dst, uint8_t *src,
+                                    ptrdiff_t stride)
+{
+    ff_avg_pixels8_mmx(dst, src, stride, 8);
+}
+
+static void put_cavs_qpel16_mc00_mmx(uint8_t *dst, uint8_t *src,
+                                     ptrdiff_t stride)
+{
+    ff_put_pixels16_mmx(dst, src, stride, 16);
+}
+
+static void avg_cavs_qpel16_mc00_mmx(uint8_t *dst, uint8_t *src,
+                                     ptrdiff_t stride)
+{
+    ff_avg_pixels16_mmx(dst, src, stride, 16);
+}
+
 static av_cold void cavsdsp_init_mmx(CAVSDSPContext *c,
                                      AVCodecContext *avctx)
 {
-    c->put_cavs_qpel_pixels_tab[0][0] = ff_put_cavs_qpel16_mc00_mmx;
-    c->put_cavs_qpel_pixels_tab[1][0] = ff_put_cavs_qpel8_mc00_mmx;
-    c->avg_cavs_qpel_pixels_tab[0][0] = ff_avg_cavs_qpel16_mc00_mmx;
-    c->avg_cavs_qpel_pixels_tab[1][0] = ff_avg_cavs_qpel8_mc00_mmx;
+    c->put_cavs_qpel_pixels_tab[0][0] = put_cavs_qpel16_mc00_mmx;
+    c->put_cavs_qpel_pixels_tab[1][0] = put_cavs_qpel8_mc00_mmx;
+    c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_mmx;
+    c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmx;
 
     c->cavs_idct8_add = cavs_idct8_add_mmx;
     c->idct_perm      = FF_TRANSPOSE_IDCT_PERM;
@@ -506,17 +544,17 @@ static av_cold void cavsdsp_init_3dnow(CAVSDSPContext *c,
 av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c, AVCodecContext *avctx)
 {
 #if HAVE_MMX_INLINE
-    int mm_flags = av_get_cpu_flags();
+    int cpu_flags = av_get_cpu_flags();
 
-    if (mm_flags & AV_CPU_FLAG_MMX)
+    if (INLINE_MMX(cpu_flags))
         cavsdsp_init_mmx(c, avctx);
 #endif /* HAVE_MMX_INLINE */
-#if HAVE_MMXEXT_INLINE
-    if (mm_flags & AV_CPU_FLAG_MMXEXT)
-        cavsdsp_init_mmxext(c, avctx);
-#endif /* HAVE_MMXEXT_INLINE */
 #if HAVE_AMD3DNOW_INLINE
-    if (mm_flags & AV_CPU_FLAG_3DNOW)
+    if (INLINE_AMD3DNOW(cpu_flags))
         cavsdsp_init_3dnow(c, avctx);
 #endif /* HAVE_AMD3DNOW_INLINE */
+#if HAVE_MMXEXT_INLINE
+    if (INLINE_MMXEXT(cpu_flags))
+        cavsdsp_init_mmxext(c, avctx);
+#endif /* HAVE_MMXEXT_INLINE */
 }