gpu-compute: call createThreads() on cpu objs in apu_se.py
authorTony Gutierrez <anthony.gutierrez@amd.com>
Fri, 5 Jan 2018 23:10:47 +0000 (18:10 -0500)
committerAnthony Gutierrez <anthony.gutierrez@amd.com>
Mon, 8 Jan 2018 19:30:32 +0000 (19:30 +0000)
commit 8ad26e2688b8736f9290086bb4026cc7500429e9
cpu: Don't override ISA if provided by user

removed the default ISA from the BaseCPU, and instead relies on
createThreads() to initiate a default ISA if none is specified. the apu_se.py
script, however does not call creatThreads() leading to a fatal when
constructing CPU objects. this patch adds the appropriate calls to
createThreads() inside apu_se.py.

Change-Id: I16a5929454c59d68a3f1b7b3858c48a70cb76412
Reviewed-on: https://gem5-review.googlesource.com/7101
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Michael LeBeane <Michael.Lebeane@amd.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>

configs/example/apu_se.py

index f93442d62672d09de9908c1b47205f9b2d8a2e3d..012fdd36baa1816ea660090823918d235faac257 100644 (file)
@@ -392,6 +392,7 @@ else:
 # OpenCL driver
 driver = ClDriver(filename="hsa", codefile=kernel_files)
 for cpu in cpu_list:
+    cpu.createThreads()
     cpu.workload = Process(executable = executable,
                            cmd = [options.cmd] + options.options.split(),
                            drivers = [driver])
@@ -401,6 +402,7 @@ for cp in cp_list:
 if fast_forward:
     for i in xrange(len(future_cpu_list)):
         future_cpu_list[i].workload = cpu_list[i].workload
+        future_cpu_list[i].createThreads()
 
 ########################## Create the overall system ########################
 # List of CPUs that must be switched when moving between KVM and simulation