[Mesa-dev] [PATCH] i965/tiled_memcpy: realign rgba8_copy_aligned_dst stack in 32-bit builds

[Eric Engestrom]

On Tuesday, 2018-03-20 13:39:25 -0700, Scott D Phillips wrote:
> When building intel_tiled_memcpy for i686, the stack will only be
> 4-byte aligned. This isn't sufficient for SSE temporaries which
> require 16-byte alignment.  Use the force_align_arg_pointer
> function attribute in that case to ensure sufficient alignment.
> ---
>  src/mesa/drivers/dri/i965/intel_tiled_memcpy.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c b/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
> index 69306828d72..bd8bafbd2d7 100644
> --- a/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
> +++ b/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
> @@ -42,6 +42,12 @@
>  #include <emmintrin.h>
>  #endif
>  
> +#if defined(__GNUC__) && defined(__i386__) && (defined(__SSSE3__) || defined(__SSE2__))

Is that a typo?  s/SSSE3/SSE3/ ?

> +#define REALIGN __attribute__((force_align_arg_pointer))
> +#else
> +#define REALIGN
> +#endif
> +
>  #define FILE_DEBUG_FLAG DEBUG_TEXTURE
>  
>  #define ALIGN_DOWN(a, b) ROUND_DOWN_TO(a, b)
> @@ -156,7 +162,7 @@ rgba8_copy_16_aligned_src(void *dst, const void *src)
>  /**
>   * Copy RGBA to BGRA - swap R and B, with the destination 16-byte aligned.
>   */
> -static inline void *
> +static REALIGN inline void *
>  rgba8_copy_aligned_dst(void *dst, const void *src, size_t bytes)
>  {
>     assert(bytes == 0 || !(((uintptr_t)dst) & 0xf));
> -- 
> 2.14.3
> 
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev