1 | /*------------------------------------------------------------------------- |
2 | * |
3 | * pg_crc32c_sse42.c |
4 | * Compute CRC-32C checksum using Intel SSE 4.2 instructions. |
5 | * |
6 | * Portions Copyright (c) 1996-2019, PostgreSQL Global Development Group |
7 | * Portions Copyright (c) 1994, Regents of the University of California |
8 | * |
9 | * |
10 | * IDENTIFICATION |
11 | * src/port/pg_crc32c_sse42.c |
12 | * |
13 | *------------------------------------------------------------------------- |
14 | */ |
15 | #include "c.h" |
16 | |
17 | #include "port/pg_crc32c.h" |
18 | |
19 | #include <nmmintrin.h> |
20 | |
21 | pg_crc32c |
22 | pg_comp_crc32c_sse42(pg_crc32c crc, const void *data, size_t len) |
23 | { |
24 | const unsigned char *p = data; |
25 | const unsigned char *pend = p + len; |
26 | |
27 | /* |
28 | * Process eight bytes of data at a time. |
29 | * |
30 | * NB: We do unaligned accesses here. The Intel architecture allows that, |
31 | * and performance testing didn't show any performance gain from aligning |
32 | * the begin address. |
33 | */ |
34 | #ifdef __x86_64__ |
35 | while (p + 8 <= pend) |
36 | { |
37 | crc = (uint32) _mm_crc32_u64(crc, *((const uint64 *) p)); |
38 | p += 8; |
39 | } |
40 | |
41 | /* Process remaining full four bytes if any */ |
42 | if (p + 4 <= pend) |
43 | { |
44 | crc = _mm_crc32_u32(crc, *((const unsigned int *) p)); |
45 | p += 4; |
46 | } |
47 | #else |
48 | |
49 | /* |
50 | * Process four bytes at a time. (The eight byte instruction is not |
51 | * available on the 32-bit x86 architecture). |
52 | */ |
53 | while (p + 4 <= pend) |
54 | { |
55 | crc = _mm_crc32_u32(crc, *((const unsigned int *) p)); |
56 | p += 4; |
57 | } |
58 | #endif /* __x86_64__ */ |
59 | |
60 | /* Process any remaining bytes one at a time. */ |
61 | while (p < pend) |
62 | { |
63 | crc = _mm_crc32_u8(crc, *p); |
64 | p++; |
65 | } |
66 | |
67 | return crc; |
68 | } |
69 | |