diff options
| author | Mika Lindqvist <postmaster@raasu.org> | 2026-03-06 14:24:43 +0200 |
|---|---|---|
| committer | Hans Kristian Rosbach <hk-github@circlestorm.org> | 2026-03-06 14:07:33 +0100 |
| commit | ced54ac89cb79d8df912d741c25ea7bce9061761 (patch) | |
| tree | e6887b51649b48a905fb1c3f527428da71c0487b | |
| parent | 493247c15b858c5f499804c366569dbf100857b2 (diff) | |
| download | Project-Tick-ced54ac89cb79d8df912d741c25ea7bce9061761.tar.gz Project-Tick-ced54ac89cb79d8df912d741c25ea7bce9061761.zip | |
Relax alignment requirement in NEON_accum32.
* UBSAN thinks only 16 byte alignment can be guaranteed.
| -rw-r--r-- | arch/arm/adler32_neon.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/arch/arm/adler32_neon.c b/arch/arm/adler32_neon.c index 5ba9b41178..9bbb293bd8 100644 --- a/arch/arm/adler32_neon.c +++ b/arch/arm/adler32_neon.c @@ -169,7 +169,7 @@ Z_FORCEINLINE static void NEON_accum32(uint32_t *s, const uint8_t *buf, size_t l int rem = len & 3; for (size_t i = 0; i < num_iter; ++i) { - uint8x16x4_t d0_d3 = vld1q_u8_x4_ex(buf, 256); + uint8x16x4_t d0_d3 = vld1q_u8_x4_ex(buf, 128); /* Unfortunately it doesn't look like there's a direct sum 8 bit to 32 * bit instruction, we'll have to make due summing to 16 bits first */ |