[Mesa-stable] [PATCH] swr: modifications to allow gcc-4.8 compilation

[Andres Gomez]

Tim, I'm a bit confused with this patch.

You landed in master:
f0a22956be4802e01f2b4f3244f011212626f12d "swr/rast: _mm*_undefined_* implementations for gcc<4.9"

And marked it for -stable picking but the patch in this mail also has
changes that only apply to the stable branch.

Is this a backport and I should use it instead of the patch which
landed in master?


On Wed, 2017-07-05 at 13:26 -0500, Tim Rowley wrote:
> gcc-4.8 was missing the _mm*_undefined_* intrinsics and code
> unconditionally used avx2 intrinsics in the avx compilation.
> ---
>  src/gallium/drivers/swr/rasterizer/common/os.h           | 6 ++++++
>  src/gallium/drivers/swr/rasterizer/common/simd16intrin.h | 4 ++--
>  2 files changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/src/gallium/drivers/swr/rasterizer/common/os.h b/src/gallium/drivers/swr/rasterizer/common/os.h
> index ef00a25..0b8f273 100644
> --- a/src/gallium/drivers/swr/rasterizer/common/os.h
> +++ b/src/gallium/drivers/swr/rasterizer/common/os.h
> @@ -171,6 +171,12 @@ void _mm256_storeu2_m128i(__m128i *hi, __m128i *lo, __m256i a)
>      _mm_storeu_si128((__m128i*)lo, _mm256_castsi256_si128(a));
>      _mm_storeu_si128((__m128i*)hi, _mm256_extractf128_si256(a, 0x1));
>  }
> +
> +// gcc prior to 4.9 doesn't have _mm*_undefined_*
> +#if (__GNUC__) && (GCC_VERSION < 409000)
> +#define _mm_undefined_si128 _mm_setzero_si128
> +#define _mm256_undefined_ps _mm256_setzero_ps
> +#endif
>  #endif
>  
>  inline
> diff --git a/src/gallium/drivers/swr/rasterizer/common/simd16intrin.h b/src/gallium/drivers/swr/rasterizer/common/simd16intrin.h
> index aa47574..90c903d 100644
> --- a/src/gallium/drivers/swr/rasterizer/common/simd16intrin.h
> +++ b/src/gallium/drivers/swr/rasterizer/common/simd16intrin.h
> @@ -405,8 +405,8 @@ INLINE simd16mask SIMDAPI _simd16_movemask_pd(simd16scalard a)
>  
>  INLINE uint64_t SIMDAPI _simd16_movemask_epi8(simd16scalari a)
>  {
> -    uint32_t mask_lo = _mm256_movemask_epi8(a.lo);
> -    uint32_t mask_hi = _mm256_movemask_epi8(a.hi);
> +    uint32_t mask_lo = _simd_movemask_epi8(a.lo);
> +    uint32_t mask_hi = _simd_movemask_epi8(a.hi);
>  
>      return static_cast<uint64_t>(mask_lo) | (static_cast<uint64_t>(mask_hi) << 32);
>  }
-- 
Br,

Andres