From 42f99bb55bf5bded00493feda8dfc9668c1e2e5a Mon Sep 17 00:00:00 2001 From: Martin Vignali Date: Sat, 21 Oct 2017 23:52:23 +0200 Subject: [PATCH 6/6] libavcodec/huffyuvdspenc : add diff_int16 AVX2 func --- libavcodec/x86/huffyuvencdsp.asm | 2 ++ libavcodec/x86/huffyuvencdsp_init.c | 6 ++++++ 2 files changed, 8 insertions(+) diff --git a/libavcodec/x86/huffyuvencdsp.asm b/libavcodec/x86/huffyuvencdsp.asm index 3f5a7df7ff..f0686e31a1 100644 --- a/libavcodec/x86/huffyuvencdsp.asm +++ b/libavcodec/x86/huffyuvencdsp.asm @@ -59,6 +59,8 @@ DIFF_INT16 INIT_XMM sse2 DIFF_INT16 +INIT_YMM avx2 +DIFF_INT16 INIT_MMX mmxext cglobal sub_hfyu_median_pred_int16, 7,7,0, dst, src1, src2, mask, w, left, left_top diff --git a/libavcodec/x86/huffyuvencdsp_init.c b/libavcodec/x86/huffyuvencdsp_init.c index f66bc8c4f0..6c6e068cf8 100644 --- a/libavcodec/x86/huffyuvencdsp_init.c +++ b/libavcodec/x86/huffyuvencdsp_init.c @@ -32,6 +32,8 @@ void ff_diff_int16_mmx (uint16_t *dst, const uint16_t *src1, const uint16_t *src unsigned mask, int w); void ff_diff_int16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w); +void ff_diff_int16_avx2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, + unsigned mask, int w); void ff_sub_hfyu_median_pred_int16_mmxext(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top); @@ -51,4 +53,8 @@ av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c, AVCodecContext * if (EXTERNAL_SSE2(cpu_flags)) { c->diff_int16 = ff_diff_int16_sse2; } + + if (EXTERNAL_AVX2_FAST(cpu_flags)) { + c->diff_int16 = ff_diff_int16_avx2; + } } -- 2.11.0 (Apple Git-81)