Sandy Bridge: Difference between revisions

No edit summary
Line 14: Line 14:
[[File:Sandybridgefrontend.jpg|thumb|right|Sandy Bridge frontend (contrast with [[Loop Stream Decoder]] of [[Nehalem]])]]
[[File:Sandybridgefrontend.jpg|thumb|right|Sandy Bridge frontend (contrast with [[Loop Stream Decoder]] of [[Nehalem]])]]
See my [[architecture]] page for background information.
See my [[architecture]] page for background information.
* As opposed to [[Nehalem]]'s [[Loop Stream Decoder]], there's a simple, 8-way/32-set/LRU 1536-μop cache (6 μops per line)
* As opposed to [[Nehalem]]'s [[Loop Stream Decoder]], there's a simple (ie always-used) 8-way/32-set/LRU 1536-μop cache (6 μops per line)
** Unlike the P4's trace cache, we're caching at instruction granularity, not trace (superblock)
** Unlike the P4's trace cache, we're caching at instruction granularity, not trace (superblock)
** Augments 32K L1 icache (completely distinct hardware)
** Augments 32K L1 icache (completely distinct hardware)