diff options
| author | Björn Gustavsson <[email protected]> | 2015-07-06 11:27:45 +0200 |
|---|---|---|
| committer | Björn Gustavsson <[email protected]> | 2015-07-06 13:38:18 +0200 |
| commit | 1ffb2647ffcc29170b461c31c018c5d2b046beae (patch) | |
| tree | f053f4c1b74d73b67d61e16bcc8ddef5923e3040 /erts/include | |
| parent | c5755ae5ded3ba27dc5d884303ead232abc77a40 (diff) | |
| download | otp-1ffb2647ffcc29170b461c31c018c5d2b046beae.tar.gz otp-1ffb2647ffcc29170b461c31c018c5d2b046beae.tar.bz2 otp-1ffb2647ffcc29170b461c31c018c5d2b046beae.zip | |
Improve unpacking performance on x86_64
When unpacking operands on 64-bit CPUs, use a smarter mask to
help the compiler optimize the code.
It turns out that on x86_64, if we use the mask 0xFFFFUL (selecting
the 16 least significant bits), the compiler can combine a move and
a mask operation into the single insruction 'movzwl', which will
eliminate one instruction.
Diffstat (limited to 'erts/include')
0 files changed, 0 insertions, 0 deletions