Single-core multi-thread simulation in gem5 SE mode with O3CPU

[calofmijuck]

Is there a way to run the tests/test-progs/threads/src/threads.cpp program with multiple threads, using only a single O3CPU core in SE mode?

Specifically, I want to only use a single O3CPU core, but want to spawn 4 threads in that core.

I may be doing something wrong, but from this line in the threads.cpp program,

unsigned cpus = thread::hardware_concurrency();

the value of cpus is 1 when ran using gem5. Changing the line to cpus = 4 gives an error.

Here is my configuration file, which is a slightly modified version from configs/learning_gem5/part1/simple.py

import m5

from m5.objects import *

system = System()

system.clk_domain = SrcClockDomain()
system.clk_domain.clock = "1GHz"
system.clk_domain.voltage_domain = VoltageDomain()

system.mem_mode = "timing"
system.mem_ranges = [AddrRange("16384MiB")]

system.cpu = X86O3CPU() # modified to use O3CPU

system.membus = SystemXBar()

system.cpu.icache_port = system.membus.cpu_side_ports
system.cpu.dcache_port = system.membus.cpu_side_ports

system.cpu.createInterruptController()

system.cpu.interrupts[0].pio = system.membus.mem_side_ports
system.cpu.interrupts[0].int_requestor = system.membus.cpu_side_ports
system.cpu.interrupts[0].int_responder = system.membus.mem_side_ports

system.mem_ctrl = MemCtrl()
system.mem_ctrl.dram = DDR3_1600_8x8()
system.mem_ctrl.dram.range = system.mem_ranges[0]
system.mem_ctrl.port = system.membus.mem_side_ports

system.system_port = system.membus.cpu_side_ports

binary = "threads" # the threads binary, compiled for X86 architecture

system.workload = SEWorkload.init_compatible(binary)

process = Process()
process.cmd = [binary]
system.cpu.workload = process
system.cpu.createThreads()

root = Root(full_system=False, system=system)
m5.instantiate()

print(f"Beginning simulation!")
exit_event = m5.simulate()
print(f"Exiting @ tick {m5.curTick()} because {exit_event.getCause()}")

With the original threads.cpp file, the output is as follows:

Running on 1 cores. with 100 values
Waiting for other threads to complete
Validating...Success!

If I change the number of threads to 4, by setting cpu = 4, I get the following output

Running on 4 cores. with 100 values
src/sim/syscall_emul.cc:86: warn: ignoring syscall mprotect(...)
terminate called after throwing an instance of 'std::system_error'
  what():  Resource temporarily unavailable

---

Is there a guide or some reference that I can use to run multi-threaded simulations? Thanks in advance.

[powerjg]

The simple answer is "no". There is no operating system to do context switches in SE mode, so there's no way to "overcommit" the number of threads on a smaller number of cores. If you want to do this, you need to run a full OS.

Edit: Let me add that you need to have nthreads + 1 cores to get OpenMP to work since OpenMP uses a fork-join parallel model.

[calofmijuck]

Thank you for your comment. Can you please check if I understood correctly?

1. In SE mode, I can only run 1 thread per core.
2. I need to use FS mode to do what I want.

[powerjg]

That's correct.