Hi Eliot,
On 27.10.2016, at 02:28, commits@source.squeak.org wrote:
the least significant 32-bits we could use long32At: to access these, in the hope that the
32-bit access will be quicker on 64-bits by virtue of fetching half the bits. But experiments
Do you want to explain why a half-word access should be faster than a full-word access on current Hardware? What I heard is that modern cpu's _never_ fetch below word-size but only mask afterwards. This would match you experiment results.
Best regards -Tobias
vm-dev@lists.squeakfoundation.org