diff options
| author | Jonathan Wright <jonathan.wright@arm.com> | 2019-05-20 15:13:57 +0100 |
|---|---|---|
| committer | Jonathan Wright <jonathan.wright@arm.com> | 2019-05-20 15:57:11 +0100 |
| commit | fdcaeede1eae1dc9898212fdb963b0ae98ec43c8 (patch) | |
| tree | 17ed58ab7070de403d5363f9bf995b89c07e4e71 /jmemmgr.c | |
| parent | 2de84a43e683c2c3c8ff4922da16b9053f024144 (diff) | |
Add memory alignment size check in jmemmgr.c
SIMD architectures have different memory alignment requirements which
must be complied with to prevent performance - and more importantly -
security issues.
This patch introduces an explicit compile-time check to make sure
that ALIGN_SIZE is a multiple of 32 bytes for AVX2 and a multiple of
16 bytes for all other architectures.
Future Arm NEON performance patches will rely on this behaviour in
the memory manager for correctness.
Bug: 922430
Change-Id: I8cf0503fba301baa6505459e3308ba89e670cc26
Diffstat (limited to 'jmemmgr.c')
| -rw-r--r-- | jmemmgr.c | 9 |
1 files changed, 9 insertions, 0 deletions
@@ -95,6 +95,15 @@ round_up_pow2(size_t a, size_t b) #endif #endif +#ifdef WITH_SIMD +#if (ALIGN_SIZE % 16) + #error "ALIGN_SIZE is not a multiple of 16 bytes - required for SIMD instructions." +#endif +#if defined(__AVX2__) && (ALIGN_SIZE % 32) + #error "AVX2 requires 32-byte alignment. ALIGN_SIZE is not a multiple of 32 bytes." +#endif +#endif + /* * We allocate objects from "pools", where each pool is gotten with a single * request to jpeg_get_small() or jpeg_get_large(). There is no per-object |