The new CLIENT_JOB_INITIAL configuration option allows to configure
the client to request more than one job in the first request round.
If a reasonable initial value is chosen, this removes the job ramp-up
after each fail-client restart, and slightly improves overall
throughput.
Change-Id: Idac2721264ec264c520d341fac64a8311a974708
Bounding the outgoing queue is always a good idea: If the campaign has
separate threads for outgoing and incoming jobs (true for the
DatabaseCampaign), this keeps memory requirements reasonable. If the
campaign works in a single thread, this is not disadvantageous either.
Change-Id: Ic75272daa8266f051adf7b23e2ffe87f5c965b86
The LLVM Disassembler infrastructure can be used to analyze many kinds
of ELF Binaries. For every instruction the used and defined registers
is available as well as information about the instruction itself.
Change-Id: I9cc89b6c116ceff7b5143a6f179ae31c4e994d2d
The disassembler disassembles an elf file with
an external objdump tool.
The architecture specific objdump must be configured
via cmake (ARCH_TOOL_PREFIX), e.g. arm-none-eabi- for
arm-none-eabi-objdump.
Currently working:
- Connect/Disconnect, Read CPU info
- CMM Script generation and T32 startup via cmake (make runt32)
- Read/Write Register, Read Program Pointer
- Read/Write Memory
- Single Breakpoint
- Setting Memory Breakpoint
TODO:
- Fix mock aspect for T32_GetRam.
- Fix Thumb2 bit in function addresses from ELFReader
- Evaluate memory breakpoint hit
- Gem5 now has two different implementation for breakpoints.
- If only BPSingleListener are used, gem5 Breakpoints are used
- If BPRangeListener are used, gem5 calls onBreakpoint() in every simulated instruction
git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@2003 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
- The register manager is gone. It's functionality is now encapsulated in the
CPU classes.
- For the client, there is the ConcreteCPU class that encapsulates the access
to the CPU state (including registers) and architecture details. The
correspondig objects for the CPUs inside the simulator can be accessed
through the SimulatorController.getCPU() function.
- Listener got a new ConcreteCPU* member to filter for which CPU the events
should fire. The default NULL is used as wildcard for all aviable CPUs. The
events respectively got a ConcreteCPU* member to indicate which CPU really
fired the event.
- For the server, there is CPUArchitecture to access the architecture details.
git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@1966 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
- This commit only enables linking against QEMU. The abstraction layer is
completely dysfunctional at this time.
- QEMU's build system needs to be patched in order to create a static
library. This patch is currently not included in the Fail* repository.
- QEMU's JIT compilation may complicate or even preclude the implementation
of some of Fail*'s backend abstractions. Only a minimal subset (serial
I/O, memory, memory writes, save/restore) is planned for the first phase.
git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@1615 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
