Commit graph

733 commits

Author SHA1 Message Date
Nathan Binkert
5e8631fd16 stoopid
--HG--
extra : convert_revision : ccd87a2de7f684bb3a4a448a307eb30afd363e12
2004-02-26 14:16:18 -05:00
Steve Reinhardt
40c350fcfe Merge.
--HG--
extra : convert_revision : e83f6895bceb0ced656e90275ec1a93d0af5498d
2004-02-26 07:08:45 -08:00
Steve Reinhardt
6f5e104fc5 Make SW prefetch flag a parameter again, and add code to make
it actually do something on FullCPU.  Still disabled, as it
causes detailed-boot to hang when you turn it on.

arch/alpha/isa_desc:
    Add EAComp and MemAcc pseudo-instructions to prefetch StaticInst.
cpu/simple_cpu/simple_cpu.hh:
    Changed prefetch() return type from Fault to void.

--HG--
extra : convert_revision : c7cb42682bfea6af117c87d4dfdb06176b6fe6b7
2004-02-26 07:05:36 -08:00
Nathan Binkert
f8ad539aa5 Save the status of faulting write-hints so that they
won't accidentally be issued to the memory system.

--HG--
extra : convert_revision : 5696a09abcdee54c8bec72d9374f7944fb136740
2004-02-26 01:18:13 -05:00
Erik Hallnor
f93e9e063a Make this compile under non full system.
--HG--
extra : convert_revision : 6ae9b4af78cff3c8e5dc367fdbefad496a28857d
2004-02-25 21:32:37 -05:00
Lisa Hsu
752b6cf7c6 delay execContext registration from create() to init() callback in the sampling cpu also.
--HG--
extra : convert_revision : d3947c55f95bb03e73a815188a517043f39925d4
2004-02-25 18:25:20 -05:00
Nathan Binkert
a253251ae0 prevent bogus addresses from being sent to the memory system
--HG--
extra : convert_revision : 4f08f8c7ab380436b2ad2c30ba53d9fe305e4496
2004-02-25 16:16:17 -05:00
Nathan Binkert
087a2ed6a1 Merge zizzer.eecs.umich.edu:/bk/m5
into ziff.eecs.umich.edu:/z/binkertn/research/m5/latest

--HG--
extra : convert_revision : 2a258d16ba51ce34f5748825a94b456abfb577cc
2004-02-25 16:13:00 -05:00
Nathan Binkert
b678152277 Fix stats debugging and always compile it in for the
debug target

--HG--
extra : convert_revision : aa16e6256a056e6df9affec6fd973e62e812e23c
2004-02-25 16:12:48 -05:00
Lisa Hsu
6a306d4caf add in an init() callback for CPU's so that no stats are accessed prior to the end of the build process. (Done by doing the registerExecContext() calling sequence in the init() process rather than the create() process).
cpu/simple_cpu/simple_cpu.cc:
cpu/simple_cpu/simple_cpu.hh:
    same thing for simple cpu's.

--HG--
extra : convert_revision : aac9f91742866fb26f8cace622f9b88454a69662
2004-02-24 14:59:25 -05:00
Steve Reinhardt
f9c91277f3 Quote args properly in tracediff.
util/tracediff:
    Quote simulator args so args with spaces get handled correctly.

--HG--
extra : convert_revision : b48677bc712be17e6e50ca35680e757ba9364692
2004-02-23 08:07:32 -08:00
Nathan Binkert
6d574e9cff Only one request may be on the dma interface at a time,
so add some states to indicate that the dma read/write
hasn't issued yet, and add some support to reissue the
request when the interface becomes unblocked.

--HG--
extra : convert_revision : 448695a8eb2e9f98554769f3165df1f796adf44a
2004-02-23 02:49:09 -05:00
Nathan Binkert
cc863129b8 extend the hack by turning the address into a real
physical address

--HG--
extra : convert_revision : 9f752d5ab6def492056d4dfe0bda5eecb022ba80
2004-02-23 02:46:47 -05:00
Nathan Binkert
220952d592 Don't issue memory requests for bogus addresses
--HG--
extra : convert_revision : 786b7fa503436f9a269ebeaedf152c6364874ab5
2004-02-23 02:46:04 -05:00
Nathan Binkert
052e0e3116 Make this like PCI-X
--HG--
extra : convert_revision : 59eaf2c548a88df075b13e78aaa7a4377f20e6c8
2004-02-23 02:44:22 -05:00
Nathan Binkert
bc5ed80b32 configurable latency for programmed IO
fix serialization

dev/etherlink.cc:
    fix serialization

--HG--
extra : convert_revision : 6739001f3f97b745d9874bb4c5729cc4909625c2
2004-02-22 18:09:11 -05:00
Nathan Binkert
e067bc751b Change order of serialization
sim/sim_object.cc:
    serialize objects in the reverse order of their creation.
    This causes a SimObject that is dependent on another
    to be serialized first.  For example, the ethernet device
    writes back some data to physical memory while serializing,
    so this will cause the physical memory to be serialized
    after that, preserving the data written back.

--HG--
extra : convert_revision : f9a37b63ce777df1cfecefa80f94f8fc69e42448
2004-02-21 10:46:31 -05:00
Nathan Binkert
610081079c New Ethernet device. Descriptors are now done via DMA instead
of faking it.  Renamed stuff to follow our style.  Lots of
general cleanup.

dev/etherpkt.hh:
    fix up includes

--HG--
extra : convert_revision : fb3a21466cdae00f8747d6e3027c9f2c63569c48
2004-02-21 10:43:37 -05:00
Nathan Binkert
a1259a4fcf Add a simple event wrapper class that takes a class pointer
and member function and will schedule it for the future.

--HG--
extra : convert_revision : f5c5a8df0839e1e10716850c2086862c4a5bc499
2004-02-20 15:24:21 -05:00
Nathan Binkert
ec06c63cc7 make the dma interface useable.
make it so that pio devices must respond with some delay.

dev/io_device.cc:
    don't forget to include dma_interface.hh so we could use it.
dev/io_device.hh:
    the generic BusInterface isn't enough for doing DMA
    we need the actual DMAInterface

--HG--
extra : convert_revision : 70298d33c8520a3f4ad11aa600825a8cec7e44bf
2004-02-20 15:23:27 -05:00
Nathan Binkert
695d51e513 make uncacheable stuff happen again
cpu/simple_cpu/simple_cpu.cc:
    Allow requests to be uncacheable

--HG--
extra : convert_revision : 7ab1442f2eec3763d5bc6a6f37b11f663851b12c
2004-02-20 15:22:41 -05:00
Nathan Binkert
38e8cb8df1 make the devices default to being attached to the memory bus
but provide a #define that can be used to put them on
a separate IO bus

--HG--
extra : convert_revision : 75d48f312c503ad67d22319db691ef5a0de12996
2004-02-20 15:22:02 -05:00
Nathan Binkert
aa483a0026 Don't need to specify two etherdump devices
--HG--
extra : convert_revision : 62e8d6b218909c00a97d71e0d3b0f11816499fbf
2004-02-20 15:20:37 -05:00
Nathan Binkert
a51cec6669 Make it so dump takes a void *
base/trace.cc:
base/trace.hh:
    take a void * for the raw data

--HG--
extra : convert_revision : fc336dc82b4d533c3a0f319977074f26342445ea
2004-02-20 15:19:55 -05:00
Nathan Binkert
ee8c8cbcd8 make etherdump work again
dev/etherdump.cc:
    now that init is automatically called by the framework, don't
    init twice.

--HG--
extra : convert_revision : 16dcdef67aa193ed71ff546e7b255d60f39bf13d
2004-02-20 15:19:19 -05:00
Nathan Binkert
7ad5b6ee22 Merge zizzer.eecs.umich.edu:/bk/m5
into ziff.eecs.umich.edu:/z/binkertn/research/m5/memory

--HG--
extra : convert_revision : 691c89ebacb3c5a5eb67359ed102646b8547dd17
2004-02-20 15:08:02 -05:00
Ron Dreslinski
ee799f47d5 Merge zizzer:/bk/m5 into zamp.eecs.umich.edu:/z/rdreslin/m5bk/clean
--HG--
extra : convert_revision : b8e4d9c2901c4c784ecca351fd34a3f5a2845473
2004-02-20 12:45:11 -05:00
Ron Dreslinski
373d70980e Add serialization for packets on the ethernet link,
and for link events

--HG--
extra : convert_revision : 0054dbc4a42dd38ff8bbf64af303bc509dd5aa8a
2004-02-20 12:44:41 -05:00
Lisa Hsu
5e494f0c59 put back $SETUPDIR = $ENV{PWD} for amd purposes.
--HG--
extra : convert_revision : 115a9ea8d08bf3e32ecb689962c8a12c9a7dba26
2004-02-20 12:37:52 -05:00
Nathan Binkert
cb928f0391 Merge zizzer.eecs.umich.edu:/bk/m5
into ziff.eecs.umich.edu:/z/binkertn/research/m5/memory

--HG--
extra : convert_revision : 626baf73d2e26201088b05b4e5fac19292abcb32
2004-02-20 11:29:39 -05:00
Lisa Hsu
4ec55bef11 interestingly, when setup is called from a perl script, when you chdir in the perl script, the ENV{PWD} does not change.
--HG--
extra : convert_revision : e1b2fb3e3b791921b9c3f4bb3b7417ff047bbcbf
2004-02-19 17:58:17 -05:00
Andrew Schultz
12747d3084 Change the physical memory logic, and also add misspeculation fix to
tlb index calls that are called from ExecContext::readIpr

arch/alpha/ev5.cc:
    Fix misspeculation bugs for misspeculated IPR accesses

--HG--
extra : convert_revision : c9ffcf9ef8123dfcaee1606c05aee8ad60d893d7
2004-02-18 21:38:55 -05:00
Nathan Binkert
26d0cf70ea Merge zizzer.eecs.umich.edu:/bk/m5
into ziff.eecs.umich.edu:/z/binkertn/research/m5/memory

--HG--
extra : convert_revision : 20e6f7edd432952b3c30d4f017058dfd52456fda
2004-02-18 14:17:44 -05:00
Nathan Binkert
2d5ef88e1c couple fixes
--HG--
extra : convert_revision : bd0948bc831610f442d5ff6893f0e2a68e3aa83f
2004-02-18 14:17:13 -05:00
Nathan Binkert
3bf012fe2f Merge
--HG--
extra : convert_revision : b29c9bbb50994e2567fd58deedb853123d2d69a0
2004-02-18 12:01:04 -05:00
Erik Hallnor
6610699987 Add COW support to the IIC.
cpu/memtest/memtest.cc:
    Move the trace Addr to the end of the printouts

--HG--
extra : convert_revision : 875f0a3f65e07f531a23fea6be07fbf3239ec2c4
2004-02-17 22:53:15 -05:00
Erik Hallnor
4bb902328a Merge ehallnor@zizzer:/bk/m5 into zazzer.eecs.umich.edu:/z/ehallnor/m5
--HG--
extra : convert_revision : 19800f3879c943addbf4bca747b357a94352b6a8
2004-02-17 03:14:21 -05:00
Erik Hallnor
5c77989c9b Change the way data is handled in the cache blocks.
--HG--
extra : convert_revision : 8c071c4f48581a398bebc4f033e18580dca0897f
2004-02-17 03:05:19 -05:00
Andrew Schultz
a73e263f94 Change logic for translating a memory addresses, extra checks for invalid
physical addresses.

--HG--
extra : convert_revision : efb4a5229d88cb3c024e0b24f5916048bd42d589
2004-02-16 21:56:38 -05:00
Erik Hallnor
b797f99213 Rework BusBridge to have a latency and handle bus IDs correctly.
--HG--
extra : convert_revision : 901ef4782d0c9f0d842deba7acdba7896da7e41c
2004-02-16 16:28:11 -05:00
Erik Hallnor
7fd91f7c4c Add full copy support.
cpu/memtest/memtest.cc:
cpu/memtest/memtest.hh:
    Add 2 new parameters to control the percentage of unaligned copy sources and destinations.

--HG--
extra : convert_revision : 2646ee2f195e9f3e76bc257b8716163ef63a9f40
2004-02-14 03:25:39 -05:00
Nathan Binkert
8c89212fa3 Merge zizzer.eecs.umich.edu:/m5/Bitkeeper/m5
into zizzer.eecs.umich.edu:/.automount/ziff/z/binkertn/research/m5/memory

--HG--
extra : convert_revision : 99f5289f4143fb09c84694a51b3394d6699d927a
2004-02-13 15:44:46 -05:00
Nathan Binkert
8232c9743d fix up vtophys a bit
arch/alpha/vtophys.cc:
    fix up vtophys to deal with translations if there
    is no ptbr, and to deal with PAL addresses
    add ptomem which is just a wrapper for dma_addr
arch/alpha/vtophys.hh:
    add ptomem which is a wrapper for dma_addr with the
    same usage as vtomem

--HG--
extra : convert_revision : 1ae22073d400e87b708a4a7ef501124227fc6c39
2004-02-13 15:36:21 -05:00
Nathan Binkert
2300633c0a better debugging
--HG--
extra : convert_revision : 504fc7174403aaeed0574a20587d4651c7351671
2004-02-12 13:59:56 -05:00
Nathan Binkert
cc53465cd5 Add support for all devices to get requests from a timing memory bus.
In the future, this can be used for actual data, but for now, it's
so that devices can respond to timing accesses properly.  This way,
an uncached access on a bus further away will take longer to respond.

dev/alpha_console.cc:
dev/alpha_console.hh:
    suport the separate IO bus

--HG--
extra : convert_revision : ececb70f5febfd00231f6e406f93b2a79be01261
2004-02-11 16:07:55 -05:00
Steve Reinhardt
5e82f8d84c Merge zizzer:/bk/m5 into isabel.reinhardt.house:/z/stever/bk/m5
--HG--
extra : convert_revision : 964126edcf72faf572a9272599264ba2f5cd7aa1
2004-02-10 21:44:08 -08:00
Steve Reinhardt
d38f995aee Fixes for Linux syscall emulation.
arch/alpha/alpha_linux_process.cc:
    Fixes for Linux emulation:
    - stat struct alignment
    - osf_{get,set}sysinfo return values
    - additional syscall numbers
    - initialize $r0 to 0
sim/syscall_emul.cc:
    brk(0) just returns brk value (don't update it!)

--HG--
extra : convert_revision : 78e22458321c81e81540d101c9e65e2e4b0ad117
2004-02-10 21:43:57 -08:00
Erik Hallnor
14c11155be Add support for multiple outstanding aligned copies
cpu/memtest/memtest.cc:
    Add traceBlockAddr to trace outputs.

--HG--
extra : convert_revision : e4ebaf9647fb393448367f4f4af3bae566f51c24
2004-02-10 20:37:38 -05:00
Erik Hallnor
4478bcb7c7 Merge ehallnor@zizzer:/bk/m5 into zazzer.eecs.umich.edu:/z/ehallnor/m5
--HG--
extra : convert_revision : 0c018b88d6ca80b1690ec99d795014848e375e44
2004-02-09 17:38:41 -05:00
Erik Hallnor
f0abedf769 Add LRU aligned copies to the hierarchy, with only one outstanding copy. Aligned copies now fully work in LRU (just need to write the IIC doCopy call). At the moment they are slow since a stalled copy stalls the entire cache.
cpu/memtest/memtest.cc:
cpu/memtest/memtest.hh:
    Add aligned copy tests, percent of copies is specified by percent_copies

--HG--
extra : convert_revision : eaf1900fcb8832db98249e94e3472ebfb049eb48
2004-02-09 17:37:27 -05:00