Hello all,I do per-core DVFS with detailed (ALPHA) core and ruby infull system
mode. I want to have the same clock frequency for a core and itsattached L1
cache without changing clock of L2 cache. The commandline option --ruby-clock
does not work for me because it sets the same clock forL1 and L2 caches. I
found that the following code in configs/ruby/MESI_CMP_directory.pysets the
clock of L1 cache the same as itsattached core:
l1_cntrl = L1Cache_Controller(version = i, cntrl_id =cntrl_count, L1Icache =
l1i_cache, L1Dcache = l1d_cache, l2_select_num_bits =l2_bits, send_evictions =
(options.cpu_type == "detailed"), prefetcher= prefetcher, ruby_system =
ruby_system, transitions_per_cycle=options.ports, clk_domain=
system.cpu[i].clk_domain, enable_prefetch = False)
cpu_seq = RubySequencer(version = i, icache = l1i_cache, dcache= l1d_cache,
clk_domain = system.cpu[i].clk_domain, ruby_system = ruby_system)
However, when I set the clock of L1 according to the codeabove, I get the
following error:
panic:Possible Deadlock detected. Aborting!
version: 63request.paddr: 0x[0x6ac0, line 0x6ac0] m_readRequestTable: 1 current
time:1770005600 issue_time: 1566886800 difference: 203118800
@ cycle 1770005600
[wakeup:build/ALPHA/mem/ruby/system/Sequencer.cc,line 107]
MemoryUsage: 1211080 KBytes
Programaborted at cycle 1770005600
Can anybody pleasehelp me in finding out how I can set the clock of each core
and its L1 cachecorrectly?
Thanks,Shervin
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
