INIT_YMM
+%ifdef HAVE_AVX
align 16
fft8_avx:
mova m0, Z(0)
sub r2d, mmsize/4
jg .deint_loop
ret
+%endif
INIT_XMM
%define movdqa movaps
INIT_YMM
+%ifdef HAVE_AVX
%macro INTERL_AVX 5
vunpckhps %3, %2, %1
vunpcklps %2, %2, %1
DECL_PASS pass_avx, PASS_BIG 1
DECL_PASS pass_interleave_avx, PASS_BIG 0
+%endif
INIT_XMM
RET
%endmacro ; DECL_FFT
+%ifdef HAVE_AVX
DECL_FFT 6, _avx
DECL_FFT 6, _avx, _interleave
+%endif
DECL_FFT 5, _sse
DECL_FFT 5, _sse, _interleave
DECL_FFT 4, _3dn
INIT_YMM
+%ifdef HAVE_AVX
DECL_IMDCT _avx, POSROTATESHUF_AVX
+%endif
#include "libavutil/x86_cpu.h"
#include "libavcodec/dsputil.h"
#include "fft.h"
+#include "config.h"
DECLARE_ASM_CONST(16, int, ff_m1m1m1m1)[4] =
{ 1 << 31, 1 << 31, 1 << 31, 1 << 31 };
void ff_fft_dispatch_interleave_sse(FFTComplex *z, int nbits);
void ff_fft_dispatch_interleave_avx(FFTComplex *z, int nbits);
+#if HAVE_AVX
void ff_fft_calc_avx(FFTContext *s, FFTComplex *z)
{
ff_fft_dispatch_interleave_avx(z, s->nbits);
}
+#endif
void ff_fft_calc_sse(FFTContext *s, FFTComplex *z)
{