Port 1ed17cc2 to C_MUL and C_MUL4.

Measures a 0.1% speedup on 96 kbps stereo encode+decode on a
 Cortex A8.

celt/arm/kiss_fft_armv4.h

@@ -40,18 +40,18 @@
        int tt__; \
         __asm__ __volatile__( \
             "#C_MUL\n\t" \
-            "ldm %[ap], {r0,r1}\n\t" \
             "ldrsh %[br], [%[bp], #0]\n\t" \
+            "ldm %[ap], {r0,r1}\n\t" \
             "ldrsh %[bi], [%[bp], #2]\n\t" \
             "smull %[tt], %[mi], r1, %[br]\n\t" \
             "smlal %[tt], %[mi], r0, %[bi]\n\t" \
             "rsb %[bi], %[bi], #0\n\t" \
-            "smull r0, %[mr], %[br], r0\n\t" \
+            "smull %[br], %[mr], r0, %[br]\n\t" \
             "mov %[tt], %[tt], lsr #15\n\t" \
-            "smlal r0, %[mr], r1, %[bi]\n\t" \
+            "smlal %[br], %[mr], r1, %[bi]\n\t" \
             "orr %[mi], %[tt], %[mi], lsl #17\n\t" \
-            "mov r0, r0, lsr #15\n\t" \
-            "orr %[mr], r0, %[mr], lsl #17\n\t" \
+            "mov %[br], %[br], lsr #15\n\t" \
+            "orr %[mr], %[br], %[mr], lsl #17\n\t" \
             : [mr]"=r"((m).r), [mi]"=r"((m).i), \
               [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \
             : [ap]"r"(&(a)), [bp]"r"(&(b)) \
@@ -68,18 +68,18 @@
        int tt__; \
         __asm__ __volatile__( \
             "#C_MUL4\n\t" \
-            "ldm %[ap], {r0,r1}\n\t" \
             "ldrsh %[br], [%[bp], #0]\n\t" \
+            "ldm %[ap], {r0,r1}\n\t" \
             "ldrsh %[bi], [%[bp], #2]\n\t" \
             "smull %[tt], %[mi], r1, %[br]\n\t" \
             "smlal %[tt], %[mi], r0, %[bi]\n\t" \
             "rsb %[bi], %[bi], #0\n\t" \
-            "smull r0, %[mr], %[br], r0\n\t" \
+            "smull %[br], %[mr], r0, %[br]\n\t" \
             "mov %[tt], %[tt], lsr #17\n\t" \
-            "smlal r0, %[mr], r1, %[bi]\n\t" \
+            "smlal %[br], %[mr], r1, %[bi]\n\t" \
             "orr %[mi], %[tt], %[mi], lsl #15\n\t" \
-            "mov r0, r0, lsr #17\n\t" \
-            "orr %[mr], r0, %[mr], lsl #15\n\t" \
+            "mov %[br], %[br], lsr #17\n\t" \
+            "orr %[mr], %[br], %[mr], lsl #15\n\t" \
             : [mr]"=r"((m).r), [mi]"=r"((m).i), \
               [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \
             : [ap]"r"(&(a)), [bp]"r"(&(b)) \