avcodec: Extend fft to size 2^17

Asked-for-by: durandal_1707
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/fft.h

@@ -134,7 +134,8 @@ extern COSTABLE(8192);
 extern COSTABLE(16384);
 extern COSTABLE(32768);
 extern COSTABLE(65536);
-extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[17];
+extern COSTABLE(131072);
+extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[18];
 
 #define ff_init_ff_cos_tabs FFT_NAME(ff_init_ff_cos_tabs)
 

libavcodec/fft_table.h

@@ -56,7 +56,7 @@
 
 #include "libavcodec/fft.h"
 
-#define MAX_LOG2_NFFT 16 //!< Specifies maximum allowed fft size
+#define MAX_LOG2_NFFT 17 //!< Specifies maximum allowed fft size
 #define MAX_FFT_SIZE (1 << MAX_LOG2_NFFT)
 
 extern const int32_t ff_w_tab_sr[];

libavcodec/fft_template.c

@@ -51,6 +51,7 @@ COSTABLE(8192);
 COSTABLE(16384);
 COSTABLE(32768);
 COSTABLE(65536);
+COSTABLE(131072);
 #endif
 COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = {
     NULL, NULL, NULL, NULL,
@@ -67,6 +68,7 @@ COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = {
     FFT_NAME(ff_cos_16384),
     FFT_NAME(ff_cos_32768),
     FFT_NAME(ff_cos_65536),
+    FFT_NAME(ff_cos_131072),
 };
 
 #endif /* FFT_FIXED_32 */
@@ -141,7 +143,7 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
 {
     int i, j, n;
 
-    if (nbits < 2 || nbits > 16)
+    if (nbits < 2 || nbits > 17)
         goto fail;
     s->nbits = nbits;
     n = 1 << nbits;
@@ -166,7 +168,7 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
 #if FFT_FIXED_32
     {
         int n=0;
-        ff_fft_lut_init(ff_fft_offsets_lut, 0, 1 << 16, &n);
+        ff_fft_lut_init(ff_fft_offsets_lut, 0, 1 << 17, &n);
     }
 #else /* FFT_FIXED_32 */
 #if FFT_FLOAT
@@ -515,10 +517,11 @@ DECL_FFT(8192,4096,2048)
 DECL_FFT(16384,8192,4096)
 DECL_FFT(32768,16384,8192)
 DECL_FFT(65536,32768,16384)
+DECL_FFT(131072,65536,32768)
 
 static void (* const fft_dispatch[])(FFTComplex*) = {
     fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024,
-    fft2048, fft4096, fft8192, fft16384, fft32768, fft65536,
+    fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, fft131072
 };
 
 static void fft_calc_c(FFTContext *s, FFTComplex *z)

libavcodec/mips/fft_mips.c

@@ -500,7 +500,10 @@ av_cold void ff_fft_init_mips(FFTContext *s)
 {
     int n=0;
 
-    ff_fft_lut_init(ff_fft_offsets_lut, 0, 1 << 16, &n);
+    if (s->nbits > 16)
+        return;
+
+    ff_fft_lut_init(ff_fft_offsets_lut, 0, 1 << 17, &n);
     ff_init_ff_cos_tabs(16);
 
 #if HAVE_INLINE_ASM

libavcodec/x86/fft.asm

@@ -73,7 +73,7 @@ ps_m1p1: dd 1<<31, 0
 cextern ps_neg
 
 %assign i 16
-%rep 13
+%rep 14
 cextern cos_ %+ i
 %assign i i<<1
 %endrep
@@ -756,7 +756,7 @@ DECL_PASS pass_interleave_3dnow, PASS_BIG 0
 %endif
 
 %assign n 1<<%1
-%rep 17-%1
+%rep 18-%1
 %assign n2 n/2
 %assign n4 n/4
 %xdefine list_of_fft list_of_fft, fft %+ n %+ fullsuffix SECTION_REL