avcodec/x86/hevc: add avx2 dc idct

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

1 files changed, 18 insertions, 0 deletions

diff --git a/libavcodec/x86/hevcdsp_init.c b/libavcodec/x86/hevcdsp_init.c
index 58a0891e5b..cad236ddad 100644
--- a/libavcodec/x86/hevcdsp_init.c
+++ b/libavcodec/x86/hevcdsp_init.c
@@ -92,6 +92,17 @@ void ff_hevc_idct32_dc_add_10_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t strid
 }
 #endif //HAVE_AVX_EXTERNAL
 
+#if HAVE_AVX2_EXTERNAL
+
+void ff_hevc_idct32_dc_add_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
+{
+    ff_hevc_idct16_dc_add_10_avx2(dst, coeffs, stride);
+    ff_hevc_idct16_dc_add_10_avx2(dst+32, coeffs, stride);
+    ff_hevc_idct16_dc_add_10_avx2(dst+16*stride, coeffs, stride);
+    ff_hevc_idct16_dc_add_10_avx2(dst+16*stride+32, coeffs, stride);
+}
+#endif //HAVE_AVX2_EXTERNAL
+
 #define mc_rep_func(name, bitd, step, W, opt) \
 void ff_hevc_put_hevc_##name##W##_##bitd##_##opt(int16_t *_dst, ptrdiff_t dststride,                            \
                                                 uint8_t *_src, ptrdiff_t _srcstride, int height,                \
@@ -438,6 +449,9 @@ void ff_hevcdsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v,     8, sse4);
             QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv,    8, sse4);
         }
+        if (EXTERNAL_AVX2(mm_flags)) {
+            c->transform_dc_add[3]    =  ff_hevc_idct32_dc_add_8_avx2;
+        }
     } else if (bit_depth == 10) {
         if (EXTERNAL_MMXEXT(mm_flags)) {
                 c->transform_dc_add[0]    =  ff_hevc_idct4_dc_add_10_mmxext;
@@ -473,6 +487,10 @@ void ff_hevcdsp_init_x86(HEVCDSPContext *c, const int bit_depth)
             c->transform_dc_add[2]    =  ff_hevc_idct16_dc_add_10_avx;
             c->transform_dc_add[3]    =  ff_hevc_idct32_dc_add_10_avx;
         }
+        if (EXTERNAL_AVX2(mm_flags)) {
+            c->transform_dc_add[2]    =  ff_hevc_idct16_dc_add_10_avx2;
+            c->transform_dc_add[3]    =  ff_hevc_idct32_dc_add_10_avx2;
 
+        }
     }
 }