diff --git a/audio/sinc.c b/audio/sinc.c
index 145a69ddc3..fb912a3371 100644
--- a/audio/sinc.c
+++ b/audio/sinc.c
@@ -224,15 +224,13 @@ static void process_sinc(rarch_sinc_resampler_t *resamp, float *out_buffer)
 }
 #elif defined(HAVE_NEON)
 
-#if TAPS != 16
-#error "NEON sinc is for now only implemented with 16 taps. Cannot continue."
-#endif
-
 // Need to make this function pointer as Android doesn't have built-in targets
 // for NEON and plain ARMv7a.
 static void (*process_sinc_func)(rarch_sinc_resampler_t *resamp, float *out_buffer);
 
-void process_sinc_neon_asm(float *out, const float *left, const float *right, const float *coeff);
+// Assumes that taps >= 8, and that taps is a multiple of 8.
+void process_sinc_neon_asm(float *out, const float *left, const float *right, const float *coeff, unsigned taps);
+
 static void process_sinc_neon(rarch_sinc_resampler_t *resamp, float *out_buffer)
 {
    const float *buffer_l = resamp->buffer_l + resamp->ptr;
@@ -241,7 +239,7 @@ static void process_sinc_neon(rarch_sinc_resampler_t *resamp, float *out_buffer)
    unsigned phase = resamp->time >> SUBPHASE_BITS;
    const float *phase_table = resamp->phase_table[phase];
 
-   process_sinc_neon_asm(out_buffer, buffer_l, buffer_r, phase_table);
+   process_sinc_neon_asm(out_buffer, buffer_l, buffer_r, phase_table, TAPS);
 }
 #else // Plain ol' C99
 #define process_sinc_func process_sinc_C
diff --git a/audio/sinc_neon.S b/audio/sinc_neon.S
index 0db4a4e3d0..d32ebf7a7b 100644
--- a/audio/sinc_neon.S
+++ b/audio/sinc_neon.S
@@ -1,5 +1,5 @@
 /*  RetroArch - A frontend for libretro.
- *  Copyright (C) 2010-2012 - Hans-Kristian Arntzen
+ *  Copyright (C) 2010-2013 - Hans-Kristian Arntzen
  * 
  *  RetroArch is free software: you can redistribute it and/or modify it under the terms
  *  of the GNU General Public License as published by the Free Software Found-
@@ -16,37 +16,39 @@
 .arm
 .align 4
 .global process_sinc_neon_asm
-# void process_sinc_neon(float *out, const float *left, const float *right, const float *coeff)
-# Hardcoded to 16 taps.
+# void process_sinc_neon(float *out, const float *left, const float *right, const float *coeff, unsigned taps)
+# Assumes taps is >= 8, and a multiple of 8.
 process_sinc_neon_asm:
+
+   push {r4, lr}   
+   vmov.f32 q0, #0.0
+   vmov.f32 q8, #0.0
+
+   # Taps argument (r4) goes on stack in armeabi.
+   ldr r4, [sp, #8]
+
+1:
    # Left
-   vld1.f32 {q0-q1},   [r1]!
-   vld1.f32 {q2-q3},   [r1]!
+   vld1.f32 {q2-q3}, [r1]!
    # Right
-   vld1.f32 {q8-q9},   [r2]!
    vld1.f32 {q10-q11}, [r2]!
    # Coeff
    vld1.f32 {q12-q13}, [r3, :128]!
-   vld1.f32 {q14-q15}, [r3, :128]!
 
-   # Left
-   vmul.f32 q0, q0, q12
-   vmul.f32 q1, q1, q13
-   vmla.f32 q0, q2, q14
-   vmla.f32 q1, q3, q15
+   # Left / Right
+   vmla.f32 q0, q2, q12
+   vmla.f32 q8, q10, q12
+   vmla.f32 q0, q3, q13
+   vmla.f32 q8, q11, q13
 
-   # Right
-   vmul.f32 q8, q8, q12
-   vmul.f32 q9, q9, q13
-   vmla.f32 q8, q10, q14
-   vmla.f32 q9, q11, q15
+   subs r4, r4, #8
+   bne 1b
 
    # Add everything together
-   vadd.f32 q0, q0, q1
-   vadd.f32 q8, q8, q9
    vadd.f32 d0, d0, d1
    vadd.f32 d16, d16, d17
    vpadd.f32 d0, d0, d16
    vst1.f32 d0, [r0]
    
-   bx lr
+   pop {r4, pc}
+
diff --git a/audio/utils_neon.S b/audio/utils_neon.S
index 4c45cd5235..4c7ef7c6f2 100644
--- a/audio/utils_neon.S
+++ b/audio/utils_neon.S
@@ -1,5 +1,5 @@
 /*  RetroArch - A frontend for libretro.
- *  Copyright (C) 2010-2012 - Hans-Kristian Arntzen
+ *  Copyright (C) 2010-2013 - Hans-Kristian Arntzen
  * 
  *  RetroArch is free software: you can redistribute it and/or modify it under the terms
  *  of the GNU General Public License as published by the Free Software Found-