Commit graph

82 commits

Author SHA1 Message Date
Ali Saidi
06a9f58c68 DMA: Add IOCache and fix bus bridge to optionally only send requests one
way so a cache can handle partial block requests for i/o devices.

--HG--
extra : convert_revision : a68b5ae826731bc87ed93eb7ef326a2393053964
2007-08-10 16:14:01 -04:00
Ali Saidi
5c38668ed6 Bus: Only call end() on an stl object once in a loop
--HG--
extra : convert_revision : 238dcd6da7577b533e52ada2107591c4e9168ebd
2007-08-10 16:14:01 -04:00
Vincentius Robby
1db9e1fb8f port: Implement cache for port interfaces and ranges
--HG--
extra : convert_revision : d7cbec7c277fb8f4d8846203caae36ce629602d5
2007-08-04 16:05:55 -04:00
Steve Reinhardt
4a7d0c4b79 bus: take out response prioritization (timing was messed up).
Also make express snoops not occupy bus (since they're magic).

--HG--
extra : convert_revision : 75aa5211a59380026d1e3f122778425e48e2edcd
2007-07-29 13:24:48 -07:00
Nathan Binkert
f0fef8f850 Merge python and x86 changes with cache branch
--HG--
extra : convert_revision : e06a950964286604274fba81dcca362d75847233
2007-07-26 23:15:49 -07:00
Steve Reinhardt
58250b8e5f bus: Fix default port handling.
--HG--
extra : convert_revision : 121b6e31cddff17c51fc4f3df20e7e2bde87d04f
2007-07-26 17:04:12 -07:00
Steve Reinhardt
de52eebd3b Integrate snoop loop functions into their respective call sites.
Also some additional cleanup of Bus::recvTiming().

--HG--
extra : convert_revision : 156814119f75d04c2e954aec2d7ed6fdc186c26f
2007-07-24 22:37:41 -07:00
Nathan Binkert
abc76f20cb Major changes to how SimObjects are created and initialized. Almost all
creation and initialization now happens in python.  Parameter objects
are generated and initialized by python.  The .ini file is now solely for
debugging purposes and is not used in construction of the objects in any
way.

--HG--
extra : convert_revision : 7e722873e417cb3d696f2e34c35ff488b7bff4ed
2007-07-23 21:51:38 -07:00
Steve Reinhardt
a67a0025b3 Make sure responses never get blocked.
--HG--
extra : convert_revision : 29f359d743994a94dc403aa0621ba72cd137d1a1
2007-07-17 08:15:23 -07:00
Steve Reinhardt
884807a68a Fix up a bunch of multilevel coherence issues.
Atomic mode seems to work.  Timing is closer but not there yet.

--HG--
extra : convert_revision : 0dea5c3d4b973d009e9d4a4c21b9cad15961d56f
2007-07-15 20:11:06 -07:00
Steve Reinhardt
f790f34fe3 Make Bus::findPort() a little more useful.
Move check for loops outside, since half the call sites
end up working around it anyway.  Return integer port ID
instead of port object pointer.

--HG--
extra : convert_revision : 4c31fe9930f4d1aa4919e764efb7c50d43792ea3
2007-07-15 20:09:03 -07:00
Steve Reinhardt
a51e16dc89 Merge of DPRINTF fixes from head.
--HG--
extra : convert_revision : f99a33b2df6a6c5592856d17d00e73ee83267442
2007-07-14 12:09:37 -07:00
Steve Reinhardt
7cd6c7ee05 Fix & tweak DPRINTFs for tracediff w/new cache code.
Note that we should *not* print pointer values in DPRINTFs as
these needlessly clutter tracediff output.

--HG--
extra : convert_revision : 25a448f1b3ac8d453a717a104ad6dc0112fb30bb
2007-07-14 11:48:30 -07:00
Steve Reinhardt
ffd697e149 bus.cc:
Fix atomic timing issue.

src/mem/bus.cc:
    Fix atomic timing issue.

--HG--
extra : convert_revision : a22ff80cd75f83c785b0604c2a4fde2e2e9f71ef
2007-07-02 01:02:35 -07:00
Steve Reinhardt
6ab53415ef Get rid of Packet result field. Error responses are
now encoded in cmd field.

--HG--
extra : convert_revision : d67819b7e3ee4b9a5bf08541104de0a89485e90b
2007-06-30 10:16:18 -07:00
Steve Reinhardt
eff122797b Merge vm1.(none):/home/stever/bk/newmem-head
into  vm1.(none):/home/stever/bk/newmem-cache2

--HG--
extra : convert_revision : 9002940097a166c8442ae1adf41b974227968920
2007-06-21 12:03:22 -07:00
Vincentius Robby
4a7bc06553 Minor error.
--HG--
extra : convert_revision : 514032e21c8861f20fcbcae7204e132088cc7dbc
2007-06-20 15:04:36 -04:00
Vincentius Robby
d540dde5b4 Removed "adding instead of dividing" trick.
Caused slowdown in performance instead of speeding up.

src/cpu/base.cc:
    Removed "adding instead of dividing" trick.
src/mem/bus.cc:
    Fixed spelling in comments.
    Removed "adding instead of dividing" trick.

--HG--
extra : convert_revision : 65a736f4f09a64e737dc7aeee53b117976330488
2007-06-20 14:54:17 -04:00
Steve Reinhardt
d69a763833 Merge vm1.(none):/home/stever/bk/newmem-head
into  vm1.(none):/home/stever/bk/newmem-cache2

configs/example/memtest.py:
    Hand merge redundant changes.

--HG--
extra : convert_revision : a2e36be254bf052024f37bcb23b5209f367d37e1
2007-06-17 17:30:24 -07:00
Steve Reinhardt
35cf19d441 More major reorg of cache. Seems to work for atomic mode now,
timing mode still broken.

configs/example/memtest.py:
    Revamp options.
src/cpu/memtest/memtest.cc:
    No need for memory initialization.
    No need to make atomic response... memory system should do that now.
src/cpu/memtest/memtest.hh:
    MemTest really doesn't want to snoop.
src/mem/bridge.cc:
    checkFunctional() cleanup.
src/mem/bus.cc:
src/mem/bus.hh:
src/mem/cache/base_cache.cc:
src/mem/cache/base_cache.hh:
src/mem/cache/cache.cc:
src/mem/cache/cache.hh:
src/mem/cache/cache_blk.hh:
src/mem/cache/cache_builder.cc:
src/mem/cache/cache_impl.hh:
src/mem/cache/coherence/coherence_protocol.cc:
src/mem/cache/coherence/coherence_protocol.hh:
src/mem/cache/coherence/simple_coherence.hh:
src/mem/cache/miss/SConscript:
src/mem/cache/miss/mshr.cc:
src/mem/cache/miss/mshr.hh:
src/mem/cache/miss/mshr_queue.cc:
src/mem/cache/miss/mshr_queue.hh:
src/mem/cache/prefetch/base_prefetcher.cc:
src/mem/cache/tags/fa_lru.cc:
src/mem/cache/tags/fa_lru.hh:
src/mem/cache/tags/iic.cc:
src/mem/cache/tags/iic.hh:
src/mem/cache/tags/lru.cc:
src/mem/cache/tags/lru.hh:
src/mem/cache/tags/split.cc:
src/mem/cache/tags/split.hh:
src/mem/cache/tags/split_lifo.cc:
src/mem/cache/tags/split_lifo.hh:
src/mem/cache/tags/split_lru.cc:
src/mem/cache/tags/split_lru.hh:
src/mem/packet.cc:
src/mem/packet.hh:
src/mem/physical.cc:
src/mem/physical.hh:
src/mem/tport.cc:
    More major reorg.  Seems to work for atomic mode now,
    timing mode still broken.

--HG--
extra : convert_revision : 7e70dfc4a752393b911880ff028271433855ae87
2007-06-17 17:27:53 -07:00
Nathan Binkert
fc4ab050b4 Add a startup function that will fast forward to the right clock edge
using a divide in order to not loop forever after resuming from a checkpoint

--HG--
extra : convert_revision : 4bbc70b1be4e5c4ed99d4f88418ab620d5ce475a
2007-06-09 23:01:47 -07:00
Steve Reinhardt
0484867d85 Merge vm1.(none):/home/stever/bk/newmem-head
into  vm1.(none):/home/stever/bk/newmem-cache2

src/mem/cache/base_cache.hh:
    Manual conflict resolution.

--HG--
extra : convert_revision : 5ebfd7abb4f978caa88bf43d25935869edfc6b9f
2007-05-22 06:36:01 -07:00
Steve Reinhardt
41241799ae Change getDeviceAddressRanges to use bool for snoop arg.
--HG--
extra : convert_revision : 832e52ba80cbab2f5bb6d5b5977a499d41b4d638
2007-05-21 23:36:09 -07:00
Steve Reinhardt
792d5b9e5e First set of changes for reorganized cache coherence support.
Compiles but doesn't work... committing just so I can merge
(stupid bk!).

src/mem/bridge.cc:
    Get rid of SNOOP_COMMIT.
src/mem/bus.cc:
src/mem/packet.hh:
    Get rid of SNOOP_COMMIT & two-pass snoop.
    First bits of EXPRESS_SNOOP support.
src/mem/cache/base_cache.cc:
src/mem/cache/base_cache.hh:
src/mem/cache/cache.hh:
src/mem/cache/cache_impl.hh:
src/mem/cache/miss/blocking_buffer.cc:
src/mem/cache/miss/miss_queue.cc:
src/mem/cache/prefetch/base_prefetcher.cc:
    Big reorg of ports and port-related functions & events.
src/mem/cache/cache.cc:
src/mem/cache/cache_builder.cc:
src/mem/cache/coherence/SConscript:
    Get rid of UniCoherence object.

--HG--
extra : convert_revision : 7672434fa3115c9b1c94686f497e57e90413b7c3
2007-05-18 22:35:04 -07:00
Ali Saidi
ff55888575 undo my previous bus change, it can make the bus deadlock.. so it still constantly reschedules itself
--HG--
extra : convert_revision : b5ef1aa0a6a2e32bd775d2dbcad9cd9505ad9b78
2007-05-09 22:23:01 -04:00
Ali Saidi
0dfc29a023 fix partial writes with a functional memory hack
figure out the block size from devices attached to the bus otherwise use a default block size when no devices that care are attached

configs/common/FSConfig.py:
src/mem/bridge.cc:
src/mem/bridge.hh:
src/python/m5/objects/Bridge.py:
    fix partial writes with a functional memory hack
src/mem/bus.cc:
src/mem/bus.hh:
src/python/m5/objects/Bus.py:
    figure out the block size from devices attached to the bus otherwise use a default block size when no devices that care are attached
src/mem/packet.cc:
    fix WriteInvalidateResp to not be a request that needs a response since it isn't
src/mem/port.hh:
    by default return 0 for deviceBlockSize instead of panicing. This makes finding the block size the bus should use easier

--HG--
extra : convert_revision : 3fcfe95f9f392ef76f324ee8bd1d7f6de95c1a64
2007-05-07 14:42:03 -04:00
Ali Saidi
d0ea8ff088 The MemoryObject tha owns a port should delete it if it so chooses when deletePortRefs() is called on it with that port as a parameter.
In this way a MemoryObject can keep a functional port around and give it to anyone who wants to do functional accesses rather
than creating a new one each time.

src/mem/bus.cc:
src/mem/bus.hh:
src/mem/cache/cache_impl.hh:
    only keep around one func port we give to anyone who wants it. Otherwise we can run out of port ids reasonably quickly if
    a lot of functional accesses are happening (e.g. remote debugging, dprintk, etc)

--HG--
extra : convert_revision : 6a9e3e96f51cedaab6de1b36cf317203899a3716
2007-04-04 13:56:38 -04:00
Kevin Lim
047f77102b Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/tmp/clean2

src/cpu/base_dyn_inst.hh:
    Hand merge.  Line is no longer needed because it's handled in the ISA.

--HG--
extra : convert_revision : 0be4067aa38759a5631c6940f0167d48fde2b680
2007-03-23 13:20:19 -04:00
Kevin Lim
78de00091b 3 memory system fixes:
1. Update packet's flags properly when a snoop happens
2. Don't allow accesses to read a block's data if the block has outstanding MSHRs.  This avoids a RAW hazard in MP systems that the memory system was not detecting properly earlier (a write required a block to upgrade, and while the upgrade was outstanding, a read came along and read old data).
3. Update MSHR's request upon a response being handled.  If the MSHR has more targets than it can respond to in one cycle, then its request must be properly updated to the new head of the targets list.

src/mem/bus.cc:
    Update packet's flags properly upon snoop.
src/mem/cache/cache_impl.hh:
    Be sure to not allow accesses to a block with outstanding MSHRs.
src/mem/cache/miss/miss_queue.cc:
    Update MSHR's request upon a response being handled.

--HG--
extra : convert_revision : 76a9abc610ca3f1904f075ad21637148a41982d6
2007-03-23 13:09:37 -04:00
Ali Saidi
027dfa01e6 stop m5 from leaking like a sieve
don't create a new physPort/virtPort every time activateContext() is called
add the ability to tell a memory object to delete it's reference to a port and a method to have a port call deletePortRefs()
on the port owner as well as delete it's peer
still need to stop calling connectMemoPorts() every time activateContext() is called or we'll overflow the bus id and panic

src/cpu/thread_state.cc:
    if we hav ea (phys|virt)Port don't create a new on, have it delete it's peer and then reuse it
src/mem/bus.cc:
src/mem/bus.hh:
    add ability to delete a port by usig a hash_map instead of an array to store port ids
    add a function to do deleting
src/mem/cache/cache.hh:
src/mem/cache/cache_impl.hh:
src/mem/mem_object.cc:
src/mem/mem_object.hh:
    adda function to delete port references from a memory object
src/mem/port.cc:
src/mem/port.hh:
    add a removeConn function that tell the owener to delete any references to the port and then deletes its peer

--HG--
extra : convert_revision : 272f0c8f80e1cf1ab1750d8be5a6c9aa110b06a4
2007-03-08 18:57:15 -05:00
Steve Reinhardt
23d970e6b9 More DPRINTF cleanup.
--HG--
extra : convert_revision : db89cea42b46476d19333038522a6c144eafdab1
2007-02-06 23:53:48 -08:00
Steve Reinhardt
51e54f519d Minor DPRINTF fixes.
--HG--
extra : convert_revision : 41956c9a480163ecac7807982215027e8ff1a4a9
2007-02-06 21:53:05 -08:00
Ali Saidi
4943d58272 Use my range_map to speed up findPort() in the bus. The snoop code could still use some work.
--HG--
extra : convert_revision : ba0a68bd378d68e4ebd80a101b965d36c8be1db9
2006-12-15 01:49:41 -05:00
Ron Dreslinski
28fd4ab39f Do a functional access to levels above on a read as a temporary solution for L2's in FS
Fix a small writeback bug when missing in the L2 in atomic mode

src/mem/bus.cc:
    Fix a comment to make sense
src/mem/cache/cache_impl.hh:
    Do a functional access to levels above on a read as a temporary solution for L2's in FS
    Also fix a small writeback miss in L2 issue
src/mem/cache/coherence/simple_coherence.hh:
src/mem/cache/coherence/uni_coherence.cc:
src/mem/cache/coherence/uni_coherence.hh:
    Do a functional access to levels above on a read as a temporary solution for L2's in FS
tests/quick/00.hello/ref/alpha/linux/o3-timing/m5stats.txt:
tests/quick/00.hello/ref/alpha/linux/simple-timing/m5stats.txt:
tests/quick/01.hello-2T-smt/ref/alpha/linux/o3-timing/m5stats.txt:
    Update ref's for writeback changes

--HG--
extra : convert_revision : 937febd577b16b7fd97a5a68acaf53541828a251
2006-11-22 20:20:38 -05:00
Ron Dreslinski
c32f3056f9 Fix bugs around uni-coherence invalidates being propogated properly.
src/mem/bus.cc:
    Make it so that invalidates being sent from the responder up don't call the responder
    but they should also not Panic.
src/mem/packet.hh:
    If we don't have data in the packet, don't call deleteData:
    Example: InvalidateRequests never have data.

--HG--
extra : convert_revision : 18766bc9f3bb4d852ac651d094254d347abd1634
2006-11-14 17:15:05 -05:00
Ron Dreslinski
8155e61a60 Update atomic and functional paths for snoops as well
--HG--
extra : convert_revision : 566d73438efb87ca683e4dee23454d880db3dfc7
2006-11-14 01:38:42 -05:00
Ron Dreslinski
4135dd48ed Update bus bridges now that snoop ranges are passed properly
src/mem/bridge.cc:
    Update brdiges, now that snoop addresses are properly forwarded.
    Bus bridge should only handle snoops on the second phase (SNOOP_COMMIT)
src/mem/bus.cc:
src/mem/bus.hh:
    Make sure if a busBridge has access to both things that snoop and things that respond it only takes the request once

--HG--
extra : convert_revision : 26cc9ee4429be45d4476fa435e0e9a54843c2509
2006-11-14 01:12:52 -05:00
Ron Dreslinski
9b6fd56dd5 Fix some errors related to snooping and functional access in the bus
src/mem/bus.cc:
    Only call snoop once per port, need to fix it so snoop ranges that overlap aren't added to list
    Functional accesses that call snoop and it goes to a higher bus may change the src, reset it after each snoop.

--HG--
extra : convert_revision : 7276059c798a85cb9d138ccc5531298ecd055c13
2006-11-13 21:33:01 -05:00
Ron Dreslinski
6c5c51338d Fix problems with snoop ranges not working properly on functional accesses
src/mem/bus.cc:
    Actually return the snoop list when asked for it.
    Don't get stuck in infinite functional loops

--HG--
extra : convert_revision : 8e6dafbd10b30d48d28b6b5d4b464e8e8f6a3ddc
2006-11-13 19:56:34 -05:00
Kevin Lim
21f43bfc4b Be sure to populate the packet's finishTime field in the atomic timing case.
--HG--
extra : convert_revision : ef34818eb2dea5b3a8e754bf56745a7cd2497bf0
2006-11-09 11:37:26 -05:00
Kevin Lim
0b0d5a282a Draining fixes.
src/cpu/o3/cpu.cc:
    Handle draining properly when CPU isn't actually being used.
src/cpu/simple/atomic.cc:
    Be sure to set status properly when draining.
src/mem/bus.cc:
    Fix for draining.

--HG--
extra : convert_revision : d9796e6693e974f022159029fc9743c49a970c8f
2006-11-09 11:33:44 -05:00
Kevin Lim
8ba73da056 Fix up bus draining and add draining to the caches.
src/mem/bus.cc:
    Fix up draining to work properly.
src/mem/bus.hh:
    Initialize drainEvent to NULL.
src/mem/cache/base_cache.cc:
src/mem/cache/base_cache.hh:
    Add draining to the caches.

--HG--
extra : convert_revision : 3082220a75d50876f10909f9f99bec535889f818
2006-11-07 14:25:54 -05:00
Kevin Lim
e71ccde663 Merge ktlim@zizzer:/bk/newmem
into  zamp.eecs.umich.edu:/z/ktlim2/clean/newmem-busfix

--HG--
extra : convert_revision : a9a41e2c292bd95aa148e1cf4d9a77c0622a462b
2006-11-02 15:20:47 -05:00
Kevin Lim
45363ea658 Have bus use the BadAddress device to handle bad addresses. The O3 CPU should be able to boot into Linux with caches on after this change.
src/mem/bus.cc:
src/mem/bus.hh:
    Bus now will be setup with a default responder, unless the user overrides it.  This default responder should return BadAddress if no matching port is found.
src/python/m5/objects/Bus.py:
    Bus now has a default responder for FS mode if the user doesn't override it.  It returns BadAddress if no matching port is found.
src/python/m5/objects/Tsunami.py:
    Add bad address device.  Also record when the user has specified their own default responder.

--HG--
extra : convert_revision : 59070477ae313ee711b2d59baa2369c9a91c5b85
2006-11-02 15:20:37 -05:00
Gabe Black
8dbab9f701 Added code to handle draining.
--HG--
extra : convert_revision : 3861f553bde5865cd21a8a58a4c410896726f0a3
2006-11-01 19:00:49 -05:00
Nathan Binkert
a4c6f0d69e Use PacketPtr everywhere
--HG--
extra : convert_revision : d9eb83ab77ffd2d725961f295b1733137e187711
2006-10-20 00:10:12 -07:00
Nathan Binkert
b937d54510 Fix compile of m5.fast
--HG--
extra : convert_revision : a8a37c318e55e48e697e4aaba339328f000b3f60
2006-10-19 23:34:59 -07:00
Ali Saidi
0615d92d33 small bus updates for functional accesses
--HG--
extra : convert_revision : c7a6b199c74ed4b4ffab14bbffb51e72d75b7742
2006-10-12 15:02:25 -04:00
Ron Dreslinski
388d484269 Make default ID unique (not broadcast)
Fix a segfault associated with DefaultId

src/mem/bus.cc:
    Handle a segfault in the bus when DefaultPort was being used
src/mem/bus.hh:
    Make the Default ID more unique (it overlapped with Broadcast ID)

--HG--
extra : convert_revision : 9182805c5cf4d9fe004e6c5be8547a8f41ed7bfe
2006-10-11 20:54:06 -04:00
Ron Dreslinski
3c7e0ec752 Fix bus in FS mode.
src/mem/bus.cc:
    Add debugging statement
src/mem/bus.hh:
    Fix implementation of bus for subsequent recvTimings while handling a retry request.
src/mem/tport.cc:
    Rework timing port to retry properly

--HG--
extra : convert_revision : fbfb5e8b4a625e49c6cd764da1df46a4f336b1b2
2006-10-11 19:25:48 -04:00