Commit graph

3043 commits

Author SHA1 Message Date
Kevin Lim
a9ca36639f Initial try of consolidating configuration files so they can be shared more easily, especially across regression tests and simple examples.
configs/test/fs.py:
    Pull a lot of the default options out of the config file now that they are in the Python objects themselves.  Also merge this file with the single_fs.py, allowing one file to be used for both.  Previously they differed only by the system they instantiated.
configs/test/test.py:
    Initial stab at consolidating configuration files so they aren't redundant between the regression tests and the simple examples.

--HG--
extra : convert_revision : e8ae3de5a6d8864831f21089d4fdb8ec690e4731
2006-07-12 17:17:17 -04:00
Kevin Lim
35ab2296d3 Push more default options to the Python object level as they are rarely changed. These are the changes that Steve was working on.
src/python/m5/objects/DiskImage.py:
src/python/m5/objects/Ethernet.py:
src/python/m5/objects/Ide.py:
src/python/m5/objects/Tsunami.py:
    Push more default options to the Python object level as they are rarely changed.

--HG--
extra : convert_revision : 963eb7a34cd04529b3c5f24b92904ab725c93efb
2006-07-12 17:16:00 -04:00
Kevin Lim
807afe8292 Updates for serialization. As long as the tickEvent doesn't need to be serialized (I don't believe it does because we drain all CPUs prior to checkpointing), it should be feasible to start up from other CPU's checkpoints.
src/cpu/simple/atomic.cc:
src/cpu/simple/atomic.hh:
src/cpu/simple/base.cc:
src/cpu/simple/timing.cc:
src/cpu/simple_thread.cc:
    Updates for serialization.

--HG--
extra : convert_revision : 0f150de75d4bc833e4c9b83568e7fd22688d5727
2006-07-12 17:11:57 -04:00
Kevin Lim
bbfe1db6b3 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem

src/cpu/o3/fetch_impl.hh:
    Hand merge.

--HG--
extra : convert_revision : 820dab2bc921cbadecaca51cd069327f984f5c74
2006-07-12 15:25:34 -04:00
Kevin Lim
6d120b7912 Track the PC of the cache data stored in fetch so it doesn't access memory multiple times if information is already in fetch.
--HG--
extra : convert_revision : 00b160b255e998cf99286bcc21894110c7642624
2006-07-12 15:24:27 -04:00
Nathan Binkert
bf4fdbe25a Add --pdb
src/python/m5/main.py:
    Add a command line option to invoke pdb on your script

--HG--
extra : convert_revision : ef5a2860bd3f6e479fa80eccaae0cb5541a20b50
2006-07-12 15:21:23 -04:00
Nathan Binkert
10a8769d6d Merge m5.eecs.umich.edu:/bk/newmem
into  iceaxe.:/Volumes/work/research/m5/current

--HG--
extra : convert_revision : 842a23da034c40c75364b76ca75de076da776ac6
2006-07-12 15:19:08 -04:00
Nathan Binkert
3218538740 Fix __file__ for scripts
src/python/m5/main.py:
    set __file__ to the script, not the m5 binary.

--HG--
extra : convert_revision : a0bbd059d2fd321ae8ff68225abc8a7bb5c410ed
2006-07-12 15:18:49 -04:00
Ron Dreslinski
f7216daee0 Add a cache version of FS (should really make this an option in original)
Now to work on caches in FS, first steps:
1) LL/SC support (Top Level Cache Hooks)
2) Snooping in the bus (CSHR's for DMA Invalidates)

--HG--
extra : convert_revision : b4e7984712f7dcd42649070c5ca538c87461e179
2006-07-11 16:03:42 -04:00
Ron Dreslinski
2b7e685268 Merge zizzer:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zazzer/z/rdreslin/m5bk/newmem

--HG--
extra : convert_revision : 3be1aa4892aa8bbd458bdc5538bbcbd6c1ebe299
2006-07-11 15:42:49 -04:00
Ron Dreslinski
6bcc65c1f8 Fix ordering issue with squashed Icache Fetches and Static data in packet.
Now hello world works with 2 levels of cache with O3 CPU(multiple outstanding requests).

src/cpu/o3/fetch_impl.hh:
    Fix ordering issue with squashed Icache Fetches and Static data in packet.

--HG--
extra : convert_revision : a6adb87540b007ead0b4982cb3f31da8199fb5ca
2006-07-11 15:42:31 -04:00
Kevin Lim
0b0cb2bca7 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem

--HG--
extra : convert_revision : c565fd7cebaa4058ba510b3db50a9c76bf301228
2006-07-11 13:43:30 -04:00
Nathan Binkert
7078d8d1b4 Fix option parsing.
src/python/m5/main.py:
    Don't allow interspersed arguments, it messes things up

--HG--
extra : convert_revision : 8f1bcf4391f570741d92bf5420879862a48f6016
2006-07-11 11:28:59 -04:00
Nathan Binkert
55ea050d48 Migrate most of main() and and all option parsing to python
configs/test/fs.py:
configs/test/test.py:
    update for the new way that m5 deals with options
src/python/SConscript:
    Compile AUTHORS, LICENSE, README, and RELEASE_NOTES into the
    python stuff.
src/python/m5/__init__.py:
    redo the way options work.
    Move them all to main.py
src/sim/main.cc:
    Migrate more functionality for main() into python.
    Namely option parsing
src/python/m5/attrdict.py:
    A dictionary object that overrides attribute access to
    do item access.
src/python/m5/main.py:
    The new location for M5's option parsing, and the main()
    routine to set up the simulation.

--HG--
extra : convert_revision : c86b87a9f508bde1994088e23fd470c7753ee4c1
2006-07-10 23:00:13 -04:00
Ron Dreslinski
ad4374e0cd Merge zizzer:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zazzer/z/rdreslin/m5bk/newmem

--HG--
extra : convert_revision : c811eb9eacc480b14862f8074af80c56ec1e07f1
2006-07-10 17:19:54 -04:00
Ron Dreslinski
6592045cbc Some fixes so that MSHR's are matched and we don't issue overlapping requests with detailed cpu
src/mem/cache/base_cache.cc:
    If we still have outstanding requests, need to schedule event again
src/mem/cache/miss/miss_queue.cc:
    Need to use block size so overlapping requests match in the MSHR's
src/mem/cache/miss/mshr.cc:
    Actually save the address, otherwise we can't match MSHR's

--HG--
extra : convert_revision : f0f018b89c2fb99f3ce8d6eafc0712ee8edeeda8
2006-07-10 17:16:15 -04:00
Kevin Lim
185a5502b7 Minor fixes.
src/cpu/checker/thread_context.hh:
src/cpu/ozone/cpu.hh:
src/cpu/ozone/cpu_impl.hh:
    Change functions to match Korey's changes.
src/cpu/ozone/lw_back_end.hh:
    Fix compile error.

--HG--
extra : convert_revision : fb11ac2d6db3a75c1cdbad2c1c02f921ad7344a6
2006-07-10 16:31:42 -04:00
Kevin Lim
5dbd7a3f76 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem-merge

--HG--
extra : convert_revision : 0e4c7684879b8552908e0b64a00b4824de807244
2006-07-10 15:41:35 -04:00
Kevin Lim
f60d8217e3 Some minor cleanups.
src/cpu/SConscript:
    Change the error message to be slightly nicer.
src/cpu/o3/commit.hh:
    Remove old code.
src/cpu/o3/commit_impl.hh:
    Remove old unused code.

--HG--
extra : convert_revision : 48aa430e1f3554007dd5e4f3d9e89b5e4f124390
2006-07-10 15:41:28 -04:00
Kevin Lim
fcaafdc48c Add parameters for backwards and forwards sizes for time buffers.
src/base/timebuf.hh:
    Add a function to return the size of the time buffer.

--HG--
extra : convert_revision : 8ffacd8b9013eb76264df065244e00dc1460efd4
2006-07-10 15:40:28 -04:00
Ron Dreslinski
5584e2b26e Update config for a system with an L2
--HG--
extra : convert_revision : c73a532ad6ad8d5115bda81fa778a4b97fbab713
2006-07-10 12:42:35 -04:00
Ron Dreslinski
39ffd24b64 Fix offset calculation. Now L2's work with timing&atomic.
src/mem/packet.hh:
    Offset is based on packet, not request.

--HG--
extra : convert_revision : d85af5838370541328ca35072c612d8198020625
2006-07-10 12:35:18 -04:00
Ron Dreslinski
aef232a942 Update FS configs to use cpu connectors for ports
--HG--
extra : convert_revision : 1e2e503401f92c1f30e2e487d7aeed1c7c5b7ee4
2006-07-10 12:07:21 -04:00
Ron Dreslinski
6c9bde608b Fix cpu in full system to match SE.
--HG--
extra : convert_revision : 95e422221ff5bab6104925d50a8882d31729b0f5
2006-07-10 12:03:13 -04:00
Korey Sewell
cdf27a0a86 Merge zizzer.eecs.umich.edu:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem-o3

--HG--
extra : convert_revision : 9098d989832e2a5818b80771e3c02170c5c8cd5b
2006-07-07 19:02:12 -04:00
Kevin Lim
43245d9c2f Support for recent port changes.
src/cpu/ozone/cpu.hh:
src/cpu/ozone/cpu_impl.hh:
src/cpu/ozone/front_end.hh:
src/cpu/ozone/front_end_impl.hh:
src/cpu/ozone/lw_back_end.hh:
src/cpu/ozone/lw_lsq.hh:
src/cpu/ozone/lw_lsq_impl.hh:
src/python/m5/objects/OzoneCPU.py:
    Support Ron's recent port changes.
src/cpu/ozone/lw_back_end_impl.hh:
    Support Ron's recent port changes.  Also support handling faults in SE.

--HG--
extra : convert_revision : aa1ba5111b70199c052da3e13bae605525a69891
2006-07-07 18:24:13 -04:00
Kevin Lim
8ade33d324 Support Ron's changes for hooking up ports.
src/cpu/checker/cpu.hh:
    Now that BaseCPU is a MemObject, the checker must define this function.
src/cpu/o3/cpu.cc:
src/cpu/o3/cpu.hh:
src/cpu/o3/fetch.hh:
src/cpu/o3/iew.hh:
src/cpu/o3/lsq.hh:
src/cpu/o3/lsq_unit.hh:
    Implement getPort function so the connector can connect the ports properly.
src/cpu/o3/fetch_impl.hh:
src/cpu/o3/lsq_unit_impl.hh:
    The connector handles connecting the ports now.
src/python/m5/objects/O3CPU.py:
    Add ports to the parameters.

--HG--
extra : convert_revision : 0b1a216b9a5d0574e62165d7c6c242498104d918
2006-07-07 17:33:24 -04:00
Kevin Lim
744e0055b7 Fix for bug when draining and a memory access is outstanding.
--HG--
extra : convert_revision : 1af782cf023ae74c2a3ff9f7aefcea880bc87936
2006-07-07 16:48:44 -04:00
Kevin Lim
b2a479cfc8 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem-merge

--HG--
extra : convert_revision : e8933f852352164f4e50444f94cc6ee260e06766
2006-07-07 16:47:28 -04:00
Kevin Lim
1faada9bd9 Take the name of the checkpoint directory in when calling checkpoint() or restoreCheckpoint().
src/sim/main.cc:
src/sim/serialize.cc:
src/sim/serialize.hh:
    Take in the directory name when checkpointing.

--HG--
extra : convert_revision : 040e828622480f1051e2156f4439e24864c38d45
2006-07-07 16:46:08 -04:00
Korey Sewell
e5aff1a7f3 Merge zizzer.eecs.umich.edu:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem-o3

--HG--
extra : convert_revision : be8b295ebf54a7c6bf720a20ab6aa9f02aee8060
2006-07-07 16:19:13 -04:00
Ron Dreslinski
7811500eef Fix address range calculation. Still need bus to handle snoop ranges.
On the way towards multi-level caches (L2)

src/mem/cache/base_cache.cc:
src/mem/cache/base_cache.hh:
    Fix address range calculation.  Still need bus to handle snoop ranges.

--HG--
extra : convert_revision : 800078d88aab5e563f4a9bb599f91cd44f36e625
2006-07-07 16:02:22 -04:00
Korey Sewell
743737c28b Merge zizzer.eecs.umich.edu:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem-o3

--HG--
extra : convert_revision : f97469b7d19c82deb3d068f80546d729757c25e3
2006-07-07 15:58:22 -04:00
Korey Sewell
74d4d67138 Minor fix for SMT Hello Worlds to finish correctly.
Still, there is a problem with the LSQ and indexing out of range in the buffer.
I havent nailed down the fix yet, but it's coming ...

src/cpu/o3/commit_impl.hh:
    add space to DPRINT
src/cpu/o3/cpu.cc:
    add newline to DPRINT
src/cpu/o3/rob.hh:
src/cpu/o3/rob_impl.hh:
    Each thread needs it's own squashedSeqNum for the case where they are both squashing at the same time and they dont
    write over each other's squash number.

--HG--
extra : convert_revision : 2155421a8b5b20e4544eea3d3c53d3e715465fa6
2006-07-07 15:58:03 -04:00
Kevin Lim
018ba50f2c Switch out fixes for CPUs.
src/cpu/o3/cpu.cc:
    Fix up keeping proper state when switched out and drained.
src/cpu/simple/timing.cc:
src/cpu/simple/timing.hh:
    Keep track of the event we use to schedule fetch initially and upon resume.  We may have to cancel the event if the CPU is switched out.

--HG--
extra : convert_revision : 60a2a1bd2cdc67bd53ca4a67aa77166c826a4c8c
2006-07-07 15:38:15 -04:00
Ron Dreslinski
76c110d924 Remove hack now that ports work properly
--HG--
extra : convert_revision : 43c22294867d7cbbc67ae66ec41a1d1c89f5a59d
2006-07-07 15:16:41 -04:00
Ron Dreslinski
ea11c7bdbe Update cpus to use the getPort function to use a connector object to connect the I/D cache ports to memory
configs/test/test.py:
    Update to use new cpu getPort functionality
src/cpu/base.cc:
    Make cpu's a memObject to expose getPort interface
src/cpu/base.hh:
    Make cpu's a memObject to export getPort interface
src/cpu/simple/atomic.cc:
src/cpu/simple/atomic.hh:
src/cpu/simple/timing.cc:
src/cpu/simple/timing.hh:
    Now use the connector via getPort interface
src/mem/cache/base_cache.cc:
    Make sure the cache recognizes all port names

--HG--
extra : convert_revision : dbfefa978ec755bc8aa6f962ae158acf32dafe61
2006-07-07 15:15:11 -04:00
Korey Sewell
161e8bf874 Merge zizzer.eecs.umich.edu:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem-o3

--HG--
extra : convert_revision : 90717b492139428e0c48be35a6bda45960c61086
2006-07-07 04:07:00 -04:00
Korey Sewell
c355df5bfe Fix so that O3CPU doesnt segfault on exit.
Major thing was to not execute commit if there are no active threads in CPU.

src/cpu/o3/alpha/thread_context.hh:
    call deallocate instead of deallocateContext
src/cpu/o3/commit_impl.hh:
    dont run commit stage if there are no instructions
src/cpu/o3/cpu.cc:
    add deallocate event, deactivateThread function, and edit deallocateContext.
src/cpu/o3/cpu.hh:
    add deallocate event and add optional delay to deallocateContext
src/cpu/o3/thread_context.hh:
    optional delay for deallocate
src/cpu/o3/thread_context_impl.hh:
    edit DPRINTFs to say Thread Context instead of Alpha TC
src/cpu/thread_context.hh:
    optional delay
src/sim/syscall_emul.hh:
    name stuff

--HG--
extra : convert_revision : f4033e1f66b3043d30ad98dcc70d8b193dea70b6
2006-07-07 04:06:26 -04:00
Kevin Lim
55e59e26c1 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem-merge

--HG--
extra : convert_revision : 942c43e2fdd68cde7aaaba5e88a667f80feab162
2006-07-06 23:20:44 -04:00
Kevin Lim
6872b99c29 Be sure to call resume after restoring from a checkpoint.
--HG--
extra : convert_revision : 4d672917038779a23f4ce7eb5d4e3039c1f5d726
2006-07-06 23:16:22 -04:00
Kevin Lim
fff7531677 Support serializing and unserializing in the O3 CPU. Also a few small fixes for draining/switching CPUs.
src/cpu/o3/commit_impl.hh:
    Fix to clear drainPending variable on call to resume.
src/cpu/o3/cpu.cc:
src/cpu/o3/cpu.hh:
    Support serializing and unserializing in the O3 CPU.
src/cpu/o3/lsq_impl.hh:
    Be sure to say we have no stores to write back if the active thread list is empty.
src/cpu/simple_thread.cc:
src/cpu/simple_thread.hh:
    Slightly change how SimpleThread is used to copy from other ThreadContexts.

--HG--
extra : convert_revision : 92a5109b3783a989d5b451036061ef82c56d3121
2006-07-06 23:13:38 -04:00
Kevin Lim
fbe3e22474 Fix the O3CPU to support the multi-pass method for checking if the system has fully drained.
src/cpu/o3/commit.hh:
src/cpu/o3/commit_impl.hh:
src/cpu/o3/decode.hh:
src/cpu/o3/decode_impl.hh:
src/cpu/o3/fetch.hh:
src/cpu/o3/fetch_impl.hh:
src/cpu/o3/iew.hh:
src/cpu/o3/iew_impl.hh:
src/cpu/o3/rename.hh:
src/cpu/o3/rename_impl.hh:
    Return a value so that the CPU can instantly return from draining if the pipeline is already drained.
src/cpu/o3/cpu.cc:
    Use values returned from pipeline stages so that the CPU can instantly return from draining if the pipeline is already drained.

--HG--
extra : convert_revision : d8ef6b811644ea67c8b40c4719273fa224105811
2006-07-06 17:57:20 -04:00
Kevin Lim
e7ccc94ea3 Various serialization changes to make it possible for the O3CPU to checkpoint.
src/arch/alpha/regfile.hh:
    Define serialize/unserialize functions on MiscRegFile itself.
src/cpu/o3/regfile.hh:
    Remove old commented code.
src/cpu/simple_thread.cc:
src/cpu/simple_thread.hh:
    Push common serialization code to ThreadState level.  Also allow the SimpleThread to be used for checkpointing by other models.
src/cpu/thread_state.cc:
src/cpu/thread_state.hh:
    Move common serialization code into ThreadState.

--HG--
extra : convert_revision : ef64ef515355437439af967eda2e610e8c1b658b
2006-07-06 17:53:26 -04:00
Ron Dreslinski
1ccfdb442f Timing cache works for hello world test.
Still need
1) detailed CPU (blocking ability in cache)
1a) Multiple outstanding requests (need to keep track of times for events)
2)Multi-level support
3)MP coherece support
4)LL/SC support
5)Functional path needs to be correctly implemented (temporarily works without multiple outstanding requests (simple cpu))

src/cpu/simple/timing.cc:
    Temp hack because timing cpu doesn't export ports properly so single I/D cache communicates only through the Icache port.
src/mem/cache/base_cache.cc:
    Handle marking MSHR's in service
    Add support for getting CSHR's
src/mem/cache/base_cache.hh:
    Make these functions visible at the base cache level
src/mem/cache/cache.hh:
    make the functions virtual
src/mem/cache/cache_impl.hh:
    Rename the function to make sense
src/mem/packet.hh:
    Accidentally clearing the needsResponse field when sending a response back.

--HG--
extra : convert_revision : 2325d4e0b77e470fa9da91490317dc8ed88b17e2
2006-07-06 16:52:05 -04:00
Kevin Lim
b55cda163e Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem

--HG--
extra : convert_revision : 7c7fc8a2f9579d443786e86dbcf906b355de69fc
2006-07-06 16:51:50 -04:00
Kevin Lim
8bf9709d91 Two minor FS compile fixes.
src/dev/tsunami_pchip.hh:
    Need ULL() for 32-bit hosts.
src/sim/pseudo_inst.cc:
    Forgot to remove sampler include from here.

--HG--
extra : convert_revision : 6ab6bdc721290167b4c2b78da3d28a4992eb24d5
2006-07-06 16:26:44 -04:00
Kevin Lim
8ae4f45bc4 Fixes for draining.
src/cpu/simple/timing.cc:
    Update for changed return values.
src/python/m5/__init__.py:
    Loop in order to make sure all objects are really drained.  Objects may become undrained as other objects become drained (e.g. a bus-bridge has a packet, while a bus is empty, and the first drain() will cause the bus-bridge to give the packet to the bus).

    The only case we know every object is actually drained is if they all return immediately that they are drained.

--HG--
extra : convert_revision : 80057a1d6d30381bd0b67b23549bd202f447c5cb
2006-07-06 16:06:00 -04:00
Ron Dreslinski
a1d208a65d Merge zizzer:/z/m5/Bitkeeper/newmem
into  zizzer.eecs.umich.edu:/.automount/zazzer/z/rdreslin/m5bk/newmem

--HG--
extra : convert_revision : 507eefde3514c35ca8420408cc89590d83cc6fc6
2006-07-06 15:16:15 -04:00
Ron Dreslinski
329e32f8c6 Now timing reads work in single level of cache with simple cpu
src/mem/cache/base_cache.cc:
src/mem/cache/base_cache.hh:
src/mem/cache/cache.hh:
    Changes to handle timing reads in Simple CPU (blocking buffers)

--HG--
extra : convert_revision : a2e7d4287d7cdfd1bbf9c929ecbeafde499a5b9f
2006-07-06 15:15:37 -04:00