Re: [Gems-users] Completing SPLASH benchmarks with opal


Date: Wed, 31 Aug 2005 11:08:30 -0400 (EDT)
From: Sean Ryan Leventhal <sleventh@xxxxxxxxxxxx>
Subject: Re: [Gems-users] Completing SPLASH benchmarks with opal
The sample beginning with bash-2.03#... contains everything that prints out. Opal continues running without printing error messages. I had been using the default opal parameters, but I have copied them below in case there is a misunderstanding here:

# Configuration File named       : default
# system status: warmup          : 0
# system status: simulation      : 0
# system status: trace           : 0
# IWINDOW_ROB_SIZE:                128
# IWINDOW_WIN_SIZE:                64
# Cache model:
# RUBY_CLOCK_DIVISOR          1

# Branch Predictor model:
#   Predictor Type: YAGS
#   mispenalty    : 1

#   YAGS Predictor:
#     PHT table         : 1024 entries
#     Exception table   : 256 entries
#     Tag bits          : 6 bits
#     size = (2*PHT + 2*Except*(TAG+2))/8 = 768 bytes

# Cascaded Indirect Predictor:
#   simple size   : 64 entries
#   except size   : 64 entries
#   tag size      : 2 bits

# RAS size        : 64 entries
# RAS exception   : 64 entries

# CONFIG_IREG_PHYSICAL           : 224
# CONFIG_IREG_LOGICAL            : 160
# CONFIG_CCREG_PHYSICAL          : 69
# CONFIG_NUM_CONTROL_SETS        : 64

# Maximum trap level (MAXTL) : 5
# Number of windows (NUMWINDOWS): 8
# Num logical integer registers (IREG_LOGICAL) : 160
# Num logical fp registers (FPREG_LOGICAL) : 64
# Num logical cc registers (CCREG_LOGICAL) : 64
# Pipeline model:
# I$ latency : 1
# fetch pass cache line : 1
# fetch pass taken branch : 1
# fetch stages : 3
# decode stages : 4
# retire stages : 3
# max fetch / cycle: 4
# max decode / cycle: 4
# max dispatch / cycle: 4
# max execute / cycle: 4
# max retire / cycle: 4
# FU TYPE # LATENCY
# FU_NONE : 127 1
# FU_INTALU : 4 1
# FU_INTMULT : 2 4
# FU_INTDIV : 2 20
# FU_BRANCH : 4 1
# FU_FLOATADD : 2 2
# FU_FLOATCMP : 2 2
# FU_FLOATCVT : 2 2
# FU_FLOATMULT: 2 4
# FU_FLOATDIV : 0 12
# FU_FLOATSQRT: 0 24
# FU_RDPORT : 0 1
# FU_WRPORT : 0 1
# autogenerated listing of all parameters
FETCH_STAGES : 3
DECODE_STAGES : 4
RETIRE_STAGES : 3
MAX_FETCH : 4
MAX_DECODE : 4
MAX_DISPATCH : 4
MAX_EXECUTE : 4
MAX_RETIRE : 4
PSEQ_MAX_UNCHECKED : 1
CONFIG_IREG_PHYSICAL : 224
CONFIG_FPREG_PHYSICAL : 192
CONFIG_CCREG_PHYSICAL : 69
CONFIG_NUM_CONTROL_SETS : 64
CONFIG_ALU_MAPPING : (0, 1, 1, 2, 3, 4, 4, 4, 5, 6, 6, 7, 8)
CONFIG_NUM_ALUS : (127, 4, 2, 2, 4, 2, 2, 2, 2, 0, 0, 0, 0)
CONFIG_ALU_LATENCY : (1, 1, 4, 20, 1, 2, 2, 2, 4, 12, 24, 1, 1)
IL1_ASSOC : 1
IL1_SET_BITS : 8
IL1_BLOCK_BITS : 6
IL1_MSHR_ENTRIES : 256
IL1_STREAM_BUFFERS : 0
IL1_IDEAL : 0
IL1_NL_TYPE : 0
IL1_NL_LINES : 0
IL1_MSHR_QUEUE_SIZE : 32
IL1_MSHR_QUEUE_ISSUE_WIDTH : 2
DL1_ASSOC : 1
DL1_SET_BITS : 8
DL1_BLOCK_BITS : 6
DL1_MSHR_ENTRIES : 256
DL1_STREAM_BUFFERS : 0
DL1_IDEAL : 0
L2_ASSOC : 4
L2_SET_BITS : 12
L2_BLOCK_BITS : 6
L2_MSHR_ENTRIES : 128
L2_STREAM_BUFFERS : 0
L2_LATENCY : 6
L2_IDEAL : 0
MEMORY_DRAM_LATENCY : 80
MEMORY_OUTSTANDING_REQUESTS : 128
MEMOP_STALE_DATA : 0
L1_FILL_BUS_CYCLES : 1
L2_FILL_BUS_CYCLES : 11
RUBY_CLOCK_DIVISOR : 1
TLB_NUM_ENTRIES : 64
TLB_NUM_PAGE_SIZES : 4
TLB_PAGE_SIZES : (8192, 65536, 524288, 4194304)
TLB_IS_IDEAL : 0
ICACHE_CYCLE : 1
FETCH_PASS_CACHE_LINE : 1
FETCH_PASS_TAKEN_BRANCH : 1
BRANCHPRED_TYPE : YAGS
BRANCHPRED_PHT_BITS : 10
BRANCHPRED_EXCEPTION_BITS : 8
BRANCHPRED_TAG_BITS : 6
BRANCHPRED_MISPRED_PENALTY : 1
BRANCHPRED_USE_GLOBAL : 0
CAS_TABLE_BITS : 6
CAS_EXCEPT_BITS : 6
CAS_EXCEPT_SHIFT : 1
RAS_BITS : 6
RAS_EXCEPTION_TABLE_BITS : 6
IWINDOW_ROB_SIZE : 128
IWINDOW_WIN_SIZE : 64
STAT_EXPENSIVE_PROFILE : 0
CHAIN_IDEAL_WINDOW : 0
CHAIN_IDEAL_BRANCH_PRED : 0
CHAIN_ST_LD_FORWARDING : 1
CHAIN_MP_MODE : 1
MEMOP_BLOCK_MASK : 63
IL1_NUM_SETS : 256
DL1_NUM_SETS : 256
CAS_TABLE_SIZE : 64
CAS_TABLE_MASK : 63
CAS_EXCEPT_SIZE : 64
CAS_EXCEPT_MASK : 63



On Wed, 31 Aug 2005, Luke Yen wrote:

Sean,

  It seems a number of people have been experiencing a similar problem.
I noticed you have the Simics output but not the Opal output.  Are there any
messages that Opal dumps before it crashes?  Also, can you give me some
info on the Opal configuration parameters you are using?

 Thanks,
   Luke

On Wed, 31 Aug 2005, Sean Ryan Leventhal wrote:

I am currently having difficulty getting benchmarks to complete while
running opal.  All of them complete on a 4-processor simulation using
ruby.  However, when I add opal they do not (except FFT).  They all have
to be left for a fair amount of time, and invariably get one of the
following:

"Bus Error" (simics crashes)
  or something of the following form:
bash-2.03# ./OCEANCON -p4

Ocean simulation with W-cycle multigrid solver
    Processors                         : 4
    Grid size                          : 258 x 258
    Grid resolution (meters)           : 20000.00
    Time between relaxations (seconds) : 28800
    Error tolerance                    : 1e-07

panic: failed to stop cpu1
panic: failed to stop cpu3

panic[cpu2]/thread=2a1001f7d20: send mondo timeout (target 0x3) [6934
NACK 0 BUSY]

000002a1001f7680 SUNW,UltraSPARC-III+:send_one_mondo+11c (104588e8,
104583c0, 0, 3, 10464958, 3)
  %l0-3: 00000000cb066647 0000000000001b16 0000000000000001
00000000cb06768e
  %l4-7: 0000000000000000 00000000cb067db7 0000000000000002
000002a1006912c0
000002a1001f7730 unix:xt_one_unchecked+f4 (1045cec4, 8, 2a1001f77e8, 1, 0,
0)
  %l0-3: 0000000010035770 0000000000000002 0000000000000003
000000000003c800
  %l4-7: 000000000002b940 000000000003cb10 0000000000000000
0000000000000000
000002a1001f7830 unix:xt_one+40 (3, 10035770, 0, 0, 2, 3000093b260)
  %l0-3: 0000000000000008 0000000000000002 0000000000000001
000003000126ce70
  %l4-7: 000003000126ca98 0000000000000016 000003000126d270
000002a1006914c0
000002a1001f78e0 SUNW,UltraSPARC-III+:cheetah_nudge_buddy+54 (0,
104aa6b0, 1e806c0ee4, 619af362, 264800, 0)
  %l0-3: 0000000000000008 0000000000000002 0000000000000001
000003000126ce90
  %l4-7: 000003000126ca98 0000000000000016 000003000126d270
0000030001255b98
000002a1001f79a0 genunix:cyclic_softint+a4 (3000126ca98, 300008175f0,
0, 1, 3000093b258, 1014e10c)
  %l0-3: 0000030000817610 0000000000000089 0000000000000000
000003000093b25c
  %l4-7: 00000300008175c8 0000030001255b98 00000300018ccb00
0000030001255b98
000002a1001f7a60 unix:cbe_level1+8 (0, 1001b, 3000126ca98,
2a1001f7d20, 200360, 1000ba78)
  %l0-3: 0000000010563260 0000000000010000 0000000000000001
000000001000a408
  %l4-7: 00000000107b0810 000003000080fef8 0000000000000000
000002a100157790

syncing file systems...
panic[cpu2]/thread=2a1001f7d20: panic sync timeout
dumping to /dev/dsk/c0t0d0s1, offset 54525952


Has anyone seen this. I have gotten nothing like it when running with only ruby, or running just simics. _______________________________________________ Gems-users mailing list Gems-users@xxxxxxxxxxx https://lists.cs.wisc.edu/mailman/listinfo/gems-users

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users

[← Prev in Thread] Current Thread [Next in Thread→]