diff mbox

[FFmpeg-devel,4/8] avutil: add alignment needed for AVX-512

Message ID 20171030130835.28327-5-jdarnley@obe.tv
State Superseded
Headers show

Commit Message

James Darnley Oct. 30, 2017, 1:08 p.m. UTC 
---
 libavutil/x86/cpu.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

James Almer Oct. 30, 2017, 11:50 p.m. UTC | #1 
On 10/30/2017 10:08 AM, James Darnley wrote:
> ---
>  libavutil/x86/cpu.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c
> index b22a950b79..c56f2a8754 100644
> --- a/libavutil/x86/cpu.c
> +++ b/libavutil/x86/cpu.c
> @@ -247,6 +247,8 @@ size_t ff_get_cpu_max_align_x86(void)
>  {
>      int flags = av_get_cpu_flags();
>  
> +    if (flags & AV_CPU_FLAG_AVX512)
> +        return 64;
>      if (flags & (AV_CPU_FLAG_AVX2      |
>                   AV_CPU_FLAG_AVX       |
>                   AV_CPU_FLAG_XOP       |
> 

av_get_cpu_max_align() is not currently being used to align buffers in
libavutil/libavcodec, so you also need to change the STRIDE_ALIGN
constant to 64 before you can feasibly start using avx512 functions.
diff mbox

Patch

diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c
index b22a950b79..c56f2a8754 100644
--- a/libavutil/x86/cpu.c
+++ b/libavutil/x86/cpu.c
@@ -247,6 +247,8 @@  size_t ff_get_cpu_max_align_x86(void)
 {
     int flags = av_get_cpu_flags();
 
+    if (flags & AV_CPU_FLAG_AVX512)
+        return 64;
     if (flags & (AV_CPU_FLAG_AVX2      |
                  AV_CPU_FLAG_AVX       |
                  AV_CPU_FLAG_XOP       |