Performance Evaluation on SGI Altix 4700

1
Performance Evaluation on SGI Altix 4700

• Guangdeng Liu and Danny Guo

2
SGI Platform Intro

• Model Name SGI Altix 4700 Bandwidth System
  (1600MHz 24M L3, DC Itanium2 9050)
• CPU Intel DC Itanium2 Processor 9050 (533 MHz
  FSB)
• CPU(s) enabled32 cores, 16 chips, 2 cores/chip
  (Hyper-Threading Technology disabled)
• Memory128 GB (81GB PC2-3200 DIMMS per 1-chip
  module)
• Disk Subsystem1 x 147 GB SCSI (Seagate Cheetah
  10k rpm)
• 2 FPGAs on RASC blade

3
RASC Blade

• 2 FPGAs on board together with other 64
  processors on SGI Altix 4700
• 40MBytes of SRAM
• 200,448 logic cells
• 128bit/sec read and write

4
How to Login

• Client for remote login
• Putty
• Etc
• Type hostname, your account
• sgi-2.cs.ucr.edu
• your user name/password

5
SPLASH2

• Download
• http//www-flash.stanford.edu/apps/SPLASH/
• Compilation errors
• http//kbarr.net/splash2
• It gives a fix to most of the applications,
  admittedly, there are still some erroneous output
  that does not match the correct_out provided.
• http//www.capsl.udel.edu/splash/
• If you still dont want to patch the fix, simply
  download an updated version from the link above.

6
PAPI

• The Performance API (PAPI) project specifies a
  standard application programming interface (API)
  for accessing hardware performance counters
  available on most modern microprocessors, such as
  cache miss, retired instruction and cycles etc.
• PAPI provides three interfaces to the underlying
  counter hardware
• The low level interface manages hardware events
  in user defined groups called EventSets.
• The high level interface simply provides the
  ability to start, stop and read the counters for
  a specified list of events.
• Graphical tools to visualize information..

7
High Level Interface

• Meant for application programmers wanting
  coarse-grained measurements
• Not thread safe
• Calls the lower level API
• Allows only PAPI preset events
• Easier to use and less setup (additional code)
  than low-level

8
Example

• long long valuesNUM_EVENTS
• unsigned int EventsNUM_EVENTSPAPI_TOT_INS,PAP
  I_TOT_CYC
• / Start the counters /
• PAPI_start_counters((int)Events,NUM_EVENTS)
• / What we are monitoring? /
• do_work()
• / Stop the counters and store the results in
  values /
• retval PAPI_stop_counters(values,NUM_EVENTS)

9
Low Level Interface

• Increased efficiency and functionality over the
  high level PAPI interface
• About 40 functions
• Obtain information about the executable and the
  hardware
• Thread-safe
• Fully programmable
• Callbacks on counter overflow

10
Example

• include "papi.h
• define NUM_EVENTS 2
• int EventsNUM_EVENTSPAPI_FP_INS,PAPI_TOT_CYC,
  EventSetlong_long valuesNUM_EVENTS
• / Initialize the Library /
• retval PAPI_library_init(PAPI_VER_CURRENT)
• / Allocate space for the new eventset and do
  setup /
• retval PAPI_create_eventset(EventSet)
• / Add Flops and total cycles to the eventset /
• retval PAPI_add_events(EventSet,Events,NUM_EVEN
  TS)
• / Start the counters /
• retval PAPI_start(EventSet)
• do_work() / What we want to monitor SPLASH/
• /Stop counters and store results in values /
• retval PAPI_stop(EventSet,values)

11
More examples..

• From the PAPI source distribution
• tests/native.c