From 36ea9922ada5ea99e54231697a4afb31d5f6b9bf Mon Sep 17 00:00:00 2001 From: Matt Turner Date: Thu, 1 Oct 2015 10:06:55 -0700 Subject: [PATCH] mesa: Add missing _mm_mfence() before streaming loads. According to the Intel Software Development Manual (Volume 1: Basic Architecture, 12.10.3 Streaming Load Hint Instruction): Streaming loads may be weakly ordered and may appear to software to execute out of order with respect to other memory operations. Software must explicitly use fences (e.g. MFENCE) if it needs to preserve order among streaming loads or between streaming loads and other memory operations. That is, a memory fence is needed to preserve the order between the GPU writing the buffer and the streaming loads reading it back. Reported-by: Joseph Nuzman Reviewed-by: Jordan Justen --- src/mesa/main/streaming-load-memcpy.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/mesa/main/streaming-load-memcpy.c b/src/mesa/main/streaming-load-memcpy.c index d7147afdc5c..32854b60eb2 100644 --- a/src/mesa/main/streaming-load-memcpy.c +++ b/src/mesa/main/streaming-load-memcpy.c @@ -59,6 +59,9 @@ _mesa_streaming_load_memcpy(void *restrict dst, void *restrict src, size_t len) len -= MIN2(bytes_before_alignment_boundary, len); } + if (len >= 64) + _mm_mfence(); + while (len >= 64) { __m128i *dst_cacheline = (__m128i *)d; __m128i *src_cacheline = (__m128i *)s; -- 2.30.2