[Python-Dev] String hash function multiplier (original) (raw)
Jeff Epler jepler at unpythonic.net
Wed Apr 14 15:33:59 EDT 2004
- Previous message: [Python-Dev] String hash function multiplier
- Next message: [Python-Dev] String hash function multiplier
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Wed, Apr 14, 2004 at 11:34:13PM +1000, Andrew MacIntyre wrote:
Ummm... are you showing what you think you're showing here? As I recall, i386 gcc uses -mcpu and -march differently to most other architectures: - -mcpu just sets scheduling parameters, but not instruction set; - -march sets the instruction set.
I'm pretty sure I am. (I'm not sure why I used -march on the long-at-a-time version, though)
imul is on all x86 architectures, but whether to use it or not depends on the characteristics of the target CPU. With -mcpu=i386, imul is considered quite slow and a shift sequence is (almost?) always preferred when one operand is constant. With -mcpu=i686 and newer, imul seems to be preferred.
I did actually inspect the generated assembly with -mcpu=i386 and -mcpu=i686, though I didn't look at it for all the combinations I posted.
Jeff
- Previous message: [Python-Dev] String hash function multiplier
- Next message: [Python-Dev] String hash function multiplier
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]