Merge commit '9d218d573f8088c606d873e80df572582e6773ef'
[ffmpeg.git] / libavcodec / arm / synth_filter_neon.S
index 5fd90ae..5417be7 100644 (file)
@@ -18,9 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
-
-        preserve8
+#include "libavutil/arm/asm.S"
 
 function ff_synth_filter_float_neon, export=1
         push            {r3-r11,lr}
@@ -35,12 +33,14 @@ function ff_synth_filter_float_neon, export=1
         ldr             r2,  [sp, #12*4]        @ in
         mov             r9,  r1                 @ synth_buf
 
-        bl              ff_imdct_half_neon
+VFP     vpush           {d0}
+        bl              X(ff_imdct_half_neon)
+VFP     vpop            {d0}
         pop             {r3}
 
         ldr             r5,  [sp, #9*4]         @ window
         ldr             r2,  [sp, #10*4]        @ out
-        vldr            d0,  [sp, #12*4]        @ scale, bias
+NOVFP   vldr            s0,  [sp, #12*4]        @ scale
         add             r8,  r9,  #12*4
 
         mov             lr,  #64*4
@@ -88,10 +88,8 @@ function ff_synth_filter_float_neon, export=1
         sub             r11, r11, #512*4
         b               2b
 3:
-        vdup.32         q8,  d0[1]
-        vdup.32         q9,  d0[1]
-        vmla.f32        q8,  q10, d0[0]
-        vmla.f32        q9,  q1,  d0[0]
+        vmul.f32        q8,  q10, d0[0]
+        vmul.f32        q9,  q1,  d0[0]
         vst1.32         {q3},     [r3,:128]
         sub             r3,  r3,  #16*4
         vst1.32         {q2},     [r3,:128]
@@ -100,9 +98,11 @@ function ff_synth_filter_float_neon, export=1
         vst1.32         {q9},     [r2,:128]
 
         subs            r1,  r1,  #1
+        it              eq
         popeq           {r4-r11,pc}
 
         cmp             r4,  #0
+        itt             eq
         subeq           r8,  r8,  #512*4
         subeq           r9,  r9,  #512*4
         sub             r5,  r5,  #512*4