2 * Copyright © 2013 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
24 * Eric Anholt <eric@anholt.net>
25 * Matt Turner <mattst88@gmail.com>
29 #include "main/macros.h"
30 #include "main/streaming-load-memcpy.h"
31 #include <smmintrin.h>
33 /* Copies memory from src to dst, using SSE 4.1's MOVNTDQA to get streaming
34 * read performance from uncached memory.
37 _mesa_streaming_load_memcpy(void *restrict dst
, void *restrict src
, size_t len
)
39 char *restrict d
= dst
;
40 char *restrict s
= src
;
42 /* If dst and src are not co-aligned, fallback to memcpy(). */
43 if (((uintptr_t)d
& 15) != ((uintptr_t)s
& 15)) {
48 /* memcpy() the misaligned header. At the end of this if block, <d> and <s>
49 * are aligned to a 16-byte boundary or <len> == 0.
51 if ((uintptr_t)d
& 15) {
52 uintptr_t bytes_before_alignment_boundary
= 16 - ((uintptr_t)d
& 15);
53 assert(bytes_before_alignment_boundary
< 16);
55 memcpy(d
, s
, MIN2(bytes_before_alignment_boundary
, len
));
57 d
= (char *)ALIGN((uintptr_t)d
, 16);
58 s
= (char *)ALIGN((uintptr_t)s
, 16);
59 len
-= MIN2(bytes_before_alignment_boundary
, len
);
66 __m128i
*dst_cacheline
= (__m128i
*)d
;
67 __m128i
*src_cacheline
= (__m128i
*)s
;
69 __m128i temp1
= _mm_stream_load_si128(src_cacheline
+ 0);
70 __m128i temp2
= _mm_stream_load_si128(src_cacheline
+ 1);
71 __m128i temp3
= _mm_stream_load_si128(src_cacheline
+ 2);
72 __m128i temp4
= _mm_stream_load_si128(src_cacheline
+ 3);
74 _mm_store_si128(dst_cacheline
+ 0, temp1
);
75 _mm_store_si128(dst_cacheline
+ 1, temp2
);
76 _mm_store_si128(dst_cacheline
+ 2, temp3
);
77 _mm_store_si128(dst_cacheline
+ 3, temp4
);
84 /* memcpy() the tail. */