diff --git a/dl/dl.gyp b/dl/dl.gyp
index 919eeb7..29e27b1 100644
--- a/dl/dl.gyp
+++ b/dl/dl.gyp
@@ -43,68 +43,71 @@
             'api/armOMX.h',
             'api/omxtypes_s.h',
             'sp/api/armSP.h',
+            # Common C code for NEON and non-NEON implementations.
+            'sp/src/arm/armSP_FFT_S32TwiddleTable.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_C_SC32.c',
+            'sp/src/arm/omxSP_FFTInit_C_SC32.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_R_S32.c',
+            'sp/src/arm/omxSP_FFTInit_R_S32.c',
+            'sp/src/arm/omxSP_FFTInit_C_SC16.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_C_SC16.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_R_S16.c',
+            'sp/src/arm/omxSP_FFTInit_R_S16.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_R_S16S32.c',
+            'sp/src/arm/omxSP_FFTInit_R_S16S32.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_C_FC32.c',
+            'sp/src/arm/omxSP_FFTInit_C_FC32.c',
+            'sp/src/arm/omxSP_FFTGetBufSize_R_F32.c',
+            'sp/src/arm/omxSP_FFTInit_R_F32.c',
+
+            # NEON-specific implementation
             # Complex 32-bit fixed-point FFT.
-            'sp/src/armSP_FFT_S32TwiddleTable.c',
-            'sp/src/omxSP_FFTGetBufSize_C_SC32.c',
-            'sp/src/omxSP_FFTInit_C_SC32.c',
-            'sp/src/armSP_FFT_CToC_SC32_Radix2_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix2_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix2_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix4_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix4_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix2_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix4_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC32_Radix8_fs_unsafe_s.S',
-            'sp/src/omxSP_FFTInv_CToC_SC32_Sfs_s.S',
-            'sp/src/omxSP_FFTFwd_CToC_SC32_Sfs_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix2_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix2_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix2_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix4_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix4_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix2_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix4_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC32_Radix8_fs_unsafe_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CToC_SC32_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_CToC_SC32_Sfs_s.S',
             # Real 32-bit fixed-point FFT
-            'sp/src/armSP_FFTInv_CCSToR_S32_preTwiddleRadix2_unsafe_s.S',
-            'sp/src/omxSP_FFTFwd_RToCCS_S32_Sfs_s.S',
-            'sp/src/omxSP_FFTGetBufSize_R_S32.c',
-            'sp/src/omxSP_FFTInit_R_S32.c',
-            'sp/src/omxSP_FFTInv_CCSToR_S32_Sfs_s.S',
+            'sp/src/arm/neon/armSP_FFTInv_CCSToR_S32_preTwiddleRadix2_unsafe_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_RToCCS_S32_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CCSToR_S32_Sfs_s.S',
             # Complex 16-bit fixed-point FFT
-            'sp/src/armSP_FFTInv_CCSToR_S16_preTwiddleRadix2_unsafe_s.S',
-            'sp/src/omxSP_FFTInit_C_SC16.c',
-            'sp/src/omxSP_FFTGetBufSize_C_SC16.c',
-            'sp/src/armSP_FFT_CToC_SC16_Radix2_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix2_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix2_ps_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix2_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S',
-            'sp/src/omxSP_FFTFwd_CToC_SC16_Sfs_s.S',
-            'sp/src/omxSP_FFTInv_CToC_SC16_Sfs_s.S',
+            'sp/src/arm/neon/armSP_FFTInv_CCSToR_S16_preTwiddleRadix2_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix2_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix2_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix2_ps_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix2_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_CToC_SC16_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CToC_SC16_Sfs_s.S',
             # Real 16-bit fixed-point FFT
-            'sp/src/omxSP_FFTFwd_RToCCS_S16_Sfs_s.S',
-            'sp/src/omxSP_FFTGetBufSize_R_S16.c',
-            'sp/src/omxSP_FFTInit_R_S16.c',
-            'sp/src/omxSP_FFTInv_CCSToR_S16_Sfs_s.S',
-            'sp/src/omxSP_FFTFwd_RToCCS_S16S32_Sfs_s.S',
-            'sp/src/omxSP_FFTGetBufSize_R_S16S32.c',
-            'sp/src/omxSP_FFTInit_R_S16S32.c',
-            'sp/src/omxSP_FFTInv_CCSToR_S32S16_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_RToCCS_S16_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CCSToR_S16_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_RToCCS_S16S32_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CCSToR_S32S16_Sfs_s.S',
             # Complex floating-point FFT
-            'sp/src/armSP_FFT_CToC_FC32_Radix2_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix2_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix2_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix4_fs_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix4_ls_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix2_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix4_unsafe_s.S',
-            'sp/src/armSP_FFT_CToC_FC32_Radix8_fs_unsafe_s.S',
-            'sp/src/omxSP_FFTGetBufSize_C_FC32.c',
-            'sp/src/omxSP_FFTInit_C_FC32.c',
-            'sp/src/omxSP_FFTInv_CToC_FC32_Sfs_s.S',
-            'sp/src/omxSP_FFTFwd_CToC_FC32_Sfs_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix2_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix2_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix2_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix4_fs_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix4_ls_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix2_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix4_unsafe_s.S',
+            'sp/src/arm/neon/armSP_FFT_CToC_FC32_Radix8_fs_unsafe_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CToC_FC32_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_CToC_FC32_Sfs_s.S',
             # Real floating-point FFT
-            'sp/src/armSP_FFTInv_CCSToR_F32_preTwiddleRadix2_unsafe_s.S',
-            'sp/src/omxSP_FFTFwd_RToCCS_F32_Sfs_s.S',
-            'sp/src/omxSP_FFTGetBufSize_R_F32.c',
-            'sp/src/omxSP_FFTInit_R_F32.c',
-            'sp/src/omxSP_FFTInv_CCSToR_F32_Sfs_s.S',
+            'sp/src/arm/neon/armSP_FFTInv_CCSToR_F32_preTwiddleRadix2_unsafe_s.S',
+            'sp/src/arm/neon/omxSP_FFTFwd_RToCCS_F32_Sfs_s.S',
+            'sp/src/arm/neon/omxSP_FFTInv_CCSToR_F32_Sfs_s.S',
           ],
         }],
         ['target_arch=="ia32" or target_arch=="x64"', {