Update to opus-HEAD-66611f1.

celt/arm/celt_pitch_xcorr_arm.s

 ; Copyright (c) 2007-2008 CSIRO
 ; Copyright (c) 2007-2009 Xiph.Org Foundation
 ; Copyright (c) 2013      Parrot
 ; Written by Aurélien Zanelli
 ;
 ; Redistribution and use in source and binary forms, with or without
 ; modification, are permitted provided that the following conditions
 ; are met:
 ;
 ; - Redistributions of source code must retain the above copyright
@@ skipping 24 matching lines @@
 ENDIF
 
 IF OPUS_ARM_MAY_HAVE_NEON
   EXPORT celt_pitch_xcorr_neon
 ENDIF
 
 IF OPUS_ARM_MAY_HAVE_NEON
 
 ; Compute sum[k]=sum(x[j]*y[j+k],j=0...len-1), k=0...3
 xcorr_kernel_neon PROC
+xcorr_kernel_neon_start
   ; input:
   ;   r3     = int         len
   ;   r4     = opus_val16 *x
   ;   r5     = opus_val16 *y
   ;   q0     = opus_val32  sum[4]
   ; output:
   ;   q0     = opus_val32  sum[4]
   ; preserved: r0-r3, r6-r11, d2, q4-q7, q9-q15
   ; internal usage:
   ;   r12 = int j
@@ skipping 119 matching lines @@
   SUBS         r6, r6, #4
   BLT celt_pitch_xcorr_neon_process4_done
 celt_pitch_xcorr_neon_process4
   ; xcorr_kernel_neon parameters:
   ; r3 = len, r4 = _x, r5 = _y, q0 = {0, 0, 0, 0}
   MOV          r4, r0
   MOV          r5, r1
   VEOR         q0, q0, q0
   ; xcorr_kernel_neon only modifies r4, r5, r12, and q0...q3.
   ; So we don't save/restore any other registers.
-  BL xcorr_kernel_neon
+  BL xcorr_kernel_neon_start
   SUBS         r6, r6, #4
   VST1.32      {q0}, [r2]!
   ; _y += 4
   ADD          r1, r1, #8
   VMAX.S32     q15, q15, q0
   ; if (max_pitch < 4) goto celt_pitch_xcorr_neon_process4_done
   BGE celt_pitch_xcorr_neon_process4
 ; We have less than 4 sums left to compute.
 celt_pitch_xcorr_neon_process4_done
   ADDS         r6, r6, #4
@@ skipping 55 matching lines @@
   LDMFD        sp!, {r4-r6, pc}
   ENDP
 
 ENDIF
 
 IF OPUS_ARM_MAY_HAVE_EDSP
 
 ; This will get used on ARMv7 devices without NEON, so it has been optimized
 ; to take advantage of dual-issuing where possible.
 xcorr_kernel_edsp PROC
+xcorr_kernel_edsp_start
   ; input:
   ;   r3      = int         len
   ;   r4      = opus_val16 *_x (must be 32-bit aligned)
   ;   r5      = opus_val16 *_y (must be 32-bit aligned)
   ;   r6...r9 = opus_val32  sum[4]
   ; output:
   ;   r6...r9 = opus_val32  sum[4]
   ; preserved: r0-r5
   ; internal usage
   ;   r2      = int         j
@@ skipping 139 matching lines @@
   ; if (max_pitch < 4) goto celt_pitch_xcorr_edsp_process2
   SUBS         r1, r1, #4
   BLT celt_pitch_xcorr_edsp_process2
 celt_pitch_xcorr_edsp_process4
   ; xcorr_kernel_edsp parameters:
   ; r3 = len, r4 = _x, r5 = _y, r6...r9 = sum[4] = {0, 0, 0, 0}
   MOV          r6, #0
   MOV          r7, #0
   MOV          r8, #0
   MOV          r9, #0
-  BL xcorr_kernel_edsp  ; xcorr_kernel_edsp(_x, _y+i, xcorr+i, len)
+  BL xcorr_kernel_edsp_start  ; xcorr_kernel_edsp(_x, _y+i, xcorr+i, len)
   ; maxcorr = max(maxcorr, sum0, sum1, sum2, sum3)
   CMP          r0, r6
   ; _y+=4
   ADD          r5, r5, #8
   MOVLT        r0, r6
   CMP          r0, r7
   MOVLT        r0, r7
   CMP          r0, r8
   MOVLT        r0, r8
   CMP          r0, r9
@@ skipping 106 matching lines @@
   ; xcorr[i] = sum
   STR          r14, [r2], #4
   MOVLT        r0, r14
 celt_pitch_xcorr_edsp_done
   LDMFD        sp!, {r4-r11, pc}
   ENDP
 
 ENDIF
 
 END