x86: bswapdsp: Don't treat 32-bit integers as 64-bit

The upper halves are not guaranteed to be zero in x86-64.

Also use `test` instead of `and` when the result isn't used for anything other
than as a branch condition, this allows some register moves to be eliminated.
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

libavcodec/x86/bswapdsp.asm

@@ -28,8 +28,8 @@ SECTION_TEXT
 
 ; %1 = aligned/unaligned
 %macro BSWAP_LOOPS  1
-    mov      r3, r2
-    sar      r2, 3
+    mov      r3d, r2d
+    sar      r2d, 3
     jz       .left4_%1
 .loop8_%1:
     mov%1    m0, [r1 +  0]
@@ -57,11 +57,11 @@ SECTION_TEXT
 %endif
     add      r0, 32
     add      r1, 32
-    dec      r2
+    dec      r2d
     jnz      .loop8_%1
 .left4_%1:
-    mov      r2, r3
-    and      r3, 4
+    mov      r2d, r3d
+    test     r3d, 4
     jz       .left
     mov%1    m0, [r1]
 %if cpuflag(ssse3)
@@ -84,13 +84,11 @@ SECTION_TEXT
 %macro BSWAP32_BUF 0
 %if cpuflag(ssse3)
 cglobal bswap32_buf, 3,4,3
-    mov      r3, r1
     mova     m2, [pb_bswap32]
 %else
 cglobal bswap32_buf, 3,4,5
-    mov      r3, r1
 %endif
-    and      r3, 15
+    test     r1, 15
     jz       .start_align
     BSWAP_LOOPS  u
     jmp      .left
@@ -98,8 +96,7 @@ cglobal bswap32_buf, 3,4,5
     BSWAP_LOOPS  a
 .left:
 %if cpuflag(ssse3)
-    mov      r3, r2
-    and      r2, 2
+    test     r2d, 2
     jz       .left1
     movq     m0, [r1]
     pshufb   m0, m2
@@ -107,13 +104,13 @@ cglobal bswap32_buf, 3,4,5
     add      r1, 8
     add      r0, 8
 .left1:
-    and      r3, 1
+    test     r2d, 1
     jz       .end
     mov      r2d, [r1]
     bswap    r2d
     mov      [r0], r2d
 %else
-    and      r2, 3
+    and      r2d, 3
     jz       .end
 .loop2:
     mov      r3d, [r1]
@@ -121,7 +118,7 @@ cglobal bswap32_buf, 3,4,5
     mov      [r0], r3d
     add      r1, 4
     add      r0, 4
-    dec      r2
+    dec      r2d
     jnz      .loop2
 %endif
 .end: