Right, that's what I though, although I tried to change the type to uint32_t and it still didn't work.
Anyway, I avoided this by using a topology which does not require that much directories such as MeshDirCorners_XY. Using that topology, I was able to simulate 1024 in a 32x32 mesh. Going for 4096 (64x64) now, hope it works is well. Thank you for your answer. ________________________________ From: gem5-dev <[email protected]> on behalf of Andreas Sandberg <[email protected]> Sent: Tuesday, June 13, 2017 12:28:03 PM To: gem5 Developer List Subject: Re: [gem5-dev] Garnet standalone > 256 cores Hi Juan, There is probably something from with a configuration script somewhere that tries to set the interleaving match bits to start at bit 256. PyBind enforces that the value is small enough to fit in the target C++ type (uint8_t), which 256 clearly doesn't. This is why you get the somewhat cryptic error message. I'm not familiar enough with Garnet/Ruby to help you with the configuration scripts though. Cheers, Andreas On 13/06/2017 09:22, JUAN JOSE GARCIA-CASTRO CRESPO wrote: > I'm trying to launch simulations with more than 256 cores, this is my > simulation parameters and the output: > > ./build/NULL/gem5.opt configs/example/garnet_synth_traffic.py > --num-cpus=512 --num-dirs=512 --mem-size=8192MB --network=garnet2.0 > --topology=Mesh_XY --mesh-rows=1 --sim-cycles=1000 > --synthetic=uniform_random --injectionrate=0.01 > > The output (crash) after initialization and so on: > > Traceback (most recent call last): > File "<string>", line 1, in <module> > File "/home/jjgcc/dev/gem5/src/python/m5/main.py", line 437, in main > exec filecode in scope > File "configs/example/garnet_synth_traffic.py", line 152, in <module> > m5.instantiate() > File "/home/jjgcc/dev/gem5/src/python/m5/simulate.py", line 115, in > instantiate > for obj in root.descendants(): obj.createCCObject() > File "/home/jjgcc/dev/gem5/src/python/m5/SimObject.py", line 1463, in > createCCObject > self.getCCParams() > File "/home/jjgcc/dev/gem5/src/python/m5/SimObject.py", line 1411, in > getCCParams > value = value.getValue() > File "/home/jjgcc/dev/gem5/src/python/m5/params.py", line 245, in getValue > return [ v.getValue() for v in self ] > File "/home/jjgcc/dev/gem5/src/python/m5/SimObject.py", line 1467, in > getValue > return self.getCCObject() > File "/home/jjgcc/dev/gem5/src/python/m5/SimObject.py", line 1445, in > getCCObject > params = self.getCCParams() > File "/home/jjgcc/dev/gem5/src/python/m5/SimObject.py", line 1411, in > getCCParams > value = value.getValue() > File "/home/jjgcc/dev/gem5/src/python/m5/params.py", line 801, in getValue > int(self.intlvBits), int(self.intlvMatch)) > TypeError: __init__(): incompatible constructor arguments. The following > argument types are supported: > 1. _m5.range.AddrRange() > 2. _m5.range.AddrRange(arg0: int, arg1: int) > 3. _m5.range.AddrRange(arg0: std::vector<AddrRange, > std::allocator<AddrRange> >) > 4. _m5.range.AddrRange(arg0: int, arg1: int, arg2: int, arg3: int, > arg4: int, arg5: int) > > Invoked with: 0L, 8589934591L, 14, 28, 9, 256 > > Looks like something with interleaving address ranges is wrong, I've > followed the wiki to build in Garnet standalone mode and I'm using the > example script with no modifications at all. All I've changed is the > --num-cpus=512, --num-dirs=512 and --mem-size=8192 command line options to > simulate a network of 512 cpus. > > Does anyone have any idea about what is going on here? Is there any way to > disable interleaving address ranges? > _______________________________________________ > gem5-dev mailing list > [email protected] > http://m5sim.org/mailman/listinfo/gem5-dev IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
