gem5.git
13 years agopython: get rid of internal.enums package.
Steve Reinhardt [Wed, 22 Sep 2010 15:45:35 +0000 (08:45 -0700)]
python: get rid of internal.enums package.

Move generated enums into internal.params, which gets
imported into object.params, restoring backward
compatibility for scripts that expect to find them there.

13 years agostats: update stats for previous cset
Steve Reinhardt [Wed, 22 Sep 2010 06:07:35 +0000 (23:07 -0700)]
stats: update stats for previous cset
Coherence protocol change basically got rid
of UpgradeReqs in L2 caches, other minor
related cache stat changes.

13 years agocache: improve coherence handling of writebacks
Steve Reinhardt [Wed, 22 Sep 2010 06:07:34 +0000 (23:07 -0700)]
cache: improve coherence handling of writebacks
If we write back an exclusive copy, we now mark it
as such, so the cache receiving the writeback can
mark its copy as exclusive.  This avoids some
unnecessary upgrade requests when a cache later
tries to re-acquire exclusive access to the block.

13 years agodiff-out: clean up options
Steve Reinhardt [Wed, 22 Sep 2010 06:07:34 +0000 (23:07 -0700)]
diff-out: clean up options
Make diff-out sort stats changes by percentage
by default, with '-a' to use current alpha sort
(instead of requiring '-p' to sort by percentage).
Other minor options cleanup too.

13 years agotests: print if output files match
Steve Reinhardt [Wed, 22 Sep 2010 06:07:34 +0000 (23:07 -0700)]
tests: print if output files match
Add '-s' flag to diff command generating outdiff
file so we have positive confirmation when
outputs match.

13 years agoCPU: Fix O3 and possible InOrder segfaults in FS.
Gabe Black [Mon, 20 Sep 2010 09:46:42 +0000 (02:46 -0700)]
CPU: Fix O3 and possible InOrder segfaults in FS.

13 years agodevices: undo cset 017baf09599f that added timer drain functions.
Steve Reinhardt [Fri, 17 Sep 2010 03:24:05 +0000 (20:24 -0700)]
devices: undo cset 017baf09599f that added timer drain functions.
It's not the right fix for the checkpoint deadlock problem
Brad was having, and creates another bug where the system can
deadlock on restore.  Brad can't reproduce the original bug
right now, so we'll wait until it arises again and then try
to fix it the right way then.

13 years agoX86: Make the halt microop non-speculative.
Gabe Black [Tue, 14 Sep 2010 19:31:37 +0000 (12:31 -0700)]
X86: Make the halt microop non-speculative.

Executing this microop makes the CPU halt even if it was misspeculated.

13 years agoX86: Make unrecognized instructions behave better in x86.
Gabe Black [Tue, 14 Sep 2010 19:27:30 +0000 (12:27 -0700)]
X86: Make unrecognized instructions behave better in x86.

13 years agoCPU: Trim unnecessary includes from some common files.
Gabe Black [Tue, 14 Sep 2010 07:29:38 +0000 (00:29 -0700)]
CPU: Trim unnecessary includes from some common files.

This reduces the scope of those includes and makes it less likely for there to
be a dependency loop. This also moves the hashing functions associated with
ExtMachInst objects to be with the ExtMachInst definitions and out of
utility.hh.

13 years agoCPU: Get rid of the now unnecessary getInst/setInst family of functions.
Gabe Black [Tue, 14 Sep 2010 04:58:34 +0000 (21:58 -0700)]
CPU: Get rid of the now unnecessary getInst/setInst family of functions.

This code is no longer needed because of the preceeding change which adds a
StaticInstPtr parameter to the fault's invoke method, obviating the only use
for this pair of functions.

13 years agoFaults: Pass the StaticInst involved, if any, to a Fault's invoke method.
Gabe Black [Tue, 14 Sep 2010 02:26:03 +0000 (19:26 -0700)]
Faults: Pass the StaticInst involved, if any, to a Fault's invoke method.

Also move the "Fault" reference counted pointer type into a separate file,
sim/fault.hh. It would be better to name this less similarly to sim/faults.hh
to reduce confusion, but fault.hh matches the name of the type. We could change
Fault to FaultPtr to match other pointer types, and then changing the name of
the file would make more sense.

13 years agoswig: make all generated files go into the m5.internal package
Nathan Binkert [Sun, 12 Sep 2010 22:41:34 +0000 (15:41 -0700)]
swig: make all generated files go into the m5.internal package

This is necessary because versions of swig older than 1.3.39 fail to
do the right thing and try to do relative imports for everything (even
with the package= option to %module).  Instead of putting params in
the m5.internal.params package, put params in the m5.internal package
and make all param modules start with param_.  Same thing for
m5.internal.enums.

Also, stop importing all generated params into m5.objects.  They are
not necessary and now with everything using relative imports we wound
up with pollution of the namespace (where builtin-range got overridden).

--HG--
rename : src/python/m5/internal/enums/__init__.py => src/python/m5/internal/enums.py
rename : src/python/m5/internal/params/__init__.py => src/python/m5/internal/params.py

13 years agostyle: fix sorting of includes and whitespace in some files
Nathan Binkert [Fri, 10 Sep 2010 21:58:04 +0000 (14:58 -0700)]
style: fix sorting of includes and whitespace in some files

13 years agoscons: Stop building the big monolithic swigged params module
Nathan Binkert [Thu, 9 Sep 2010 21:26:29 +0000 (14:26 -0700)]
scons: Stop building the big monolithic swigged params module
kill params.i and create a separate .i for each object (param, enums, etc.)

13 years agoinit: don't build files that centralize python and swig code
Nathan Binkert [Thu, 9 Sep 2010 21:15:42 +0000 (14:15 -0700)]
init: don't build files that centralize python and swig code

Instead of putting all object files into m5/object/__init__.py, interrogate
the importer to find out what should be imported.
Instead of creating a single file that lists all of the embedded python
modules, use static object construction to put those objects onto a list.
Do something similar for embedded swig (C++) code.

13 years agoscons: use code_formatter wherever we can in the build system
Nathan Binkert [Thu, 9 Sep 2010 21:15:41 +0000 (14:15 -0700)]
scons: use code_formatter wherever we can in the build system

13 years agocode_formatter: make it easier to insert whitespace
Nathan Binkert [Thu, 9 Sep 2010 21:15:41 +0000 (14:15 -0700)]
code_formatter: make it easier to insert whitespace
a newline by just doing "code()". indent() and dedent() now take a
"count" parameter to indent/dedent multiple levels.

13 years agoswig: don't override SWIG_name anymore
Nathan Binkert [Thu, 9 Sep 2010 21:15:40 +0000 (14:15 -0700)]
swig: don't override SWIG_name anymore
It doesn't appear to be necessary and it is somewhat odd.  I'm pretty
sure that the package parameter to %module does whatever this might
have been before.  It's necessary in future revisions anyway.

13 years agostats: update stats for preceding coherence changes
Steve Reinhardt [Thu, 9 Sep 2010 18:40:19 +0000 (14:40 -0400)]
stats: update stats for preceding coherence changes
Because the handling of the E state for multilevel caches
has changed, stats are affected for any non-ruby config
with caches, even uniprocessor simple CPU.

13 years agocache: fail SC when invalidated while waiting for bus
Steve Reinhardt [Thu, 9 Sep 2010 18:40:19 +0000 (14:40 -0400)]
cache: fail SC when invalidated while waiting for bus
Corrects an oversight in cset f97b62be544f.  The fix there only
failed queued SCUpgradeReq packets that encountered an
invalidation, which meant that the upgrade had to reach the L2
cache.  To handle pending requests in the L1 we must similarly
fail StoreCondReq packets too.

13 years agomem: fix functional accesses to deal with coherence change
Steve Reinhardt [Thu, 9 Sep 2010 18:40:19 +0000 (14:40 -0400)]
mem: fix functional accesses to deal with coherence change
We can't just obliviously return the first valid cache block
we find any more... see comments for details.

13 years agocache: coherence protocol enhancements & bug fixes
Steve Reinhardt [Thu, 9 Sep 2010 18:40:18 +0000 (14:40 -0400)]
cache: coherence protocol enhancements & bug fixes
Allow lower-level caches (e.g., L2 or L3) to pass exclusive
copies to higher levels (e.g., L1).  This eliminates a lot
of unnecessary upgrade transactions on read-write sequences
to non-shared data.

Also some cleanup of MSHR coherence handling and multiple
bug fixes.

13 years agoARM: Get rid of the checkFpEnableFault function in ARM.
Gabe Black [Tue, 31 Aug 2010 16:50:49 +0000 (09:50 -0700)]
ARM: Get rid of the checkFpEnableFault function in ARM.

13 years agoAlpha: Alpha's mt.hh was including mips header files.
Gabe Black [Tue, 31 Aug 2010 16:48:05 +0000 (09:48 -0700)]
Alpha: Alpha's mt.hh was including mips header files.

13 years agoCPU: Get rid of the unused ev5_trap function on the simple and checker CPUs.
Gabe Black [Tue, 31 Aug 2010 16:47:29 +0000 (09:47 -0700)]
CPU: Get rid of the unused ev5_trap function on the simple and checker CPUs.

13 years agoconfig: fixed numa high bit setting bug
Brad Beckmann [Mon, 30 Aug 2010 22:26:51 +0000 (15:26 -0700)]
config: fixed numa high bit setting bug

13 years agoconfig: minor fix to probe filter index setting
Brad Beckmann [Mon, 30 Aug 2010 19:07:21 +0000 (12:07 -0700)]
config: minor fix to probe filter index setting

13 years agoruby: None, not none
Nathan Binkert [Sun, 29 Aug 2010 14:02:42 +0000 (07:02 -0700)]
ruby: None, not none

13 years agoX86: Change the copyright holder to AMD.
Gabe Black [Fri, 27 Aug 2010 22:35:36 +0000 (15:35 -0700)]
X86: Change the copyright holder to AMD.

I accidentally left myself as a placeholder copyright holder on this file when
I checked it in. Copyright should be assigned to AMD.

13 years agomem: fix m5.fast compile bug in previous cset
Steve Reinhardt [Thu, 26 Aug 2010 15:03:20 +0000 (08:03 -0700)]
mem: fix m5.fast compile bug in previous cset

13 years agocache: fix a bug in atomic multilevel snoops
Steve Reinhardt [Thu, 26 Aug 2010 04:55:55 +0000 (21:55 -0700)]
cache: fix a bug in atomic multilevel snoops

13 years agomemtest: fix/cleanup functional access testing
Steve Reinhardt [Thu, 26 Aug 2010 04:55:44 +0000 (21:55 -0700)]
memtest: fix/cleanup functional access testing
Don't assert that the response packet is marked as a response
since it won't always be so for functional accesses.

Also cleanup code to refer to functional accesses rather
than "probes" (old terminology), and mention in the
DPRINTF which type of access we're doing.

13 years agomemtest: scale associativity and mshrs according to config
Steve Reinhardt [Thu, 26 Aug 2010 04:55:42 +0000 (21:55 -0700)]
memtest: scale associativity and mshrs according to config
Use the actual fanouts in the tree specification to scale
cache associativity and mshrs instead of dumb constants.

13 years agoCPU: Print out traces for faluting inst when the flag ExecFaulting is set
Ali Saidi [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
CPU: Print out traces for faluting inst when the flag ExecFaulting is set

13 years agoARM: Support unaligned memory access.
Min Kyu Jeong [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Support unaligned memory access.

Without this flag set, page-crossing requests were not split into two mem
request.

Depending on the alignment bit in the SCTLR, misaligned access could
raise a fault. However it seems unnecessary to implement that.

13 years agoARM: Seperate the queues of L1 and L2 walker states.
Gene WU [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Seperate the queues of L1 and L2 walker states.

13 years agoARM: Adding a bogus fault that does nothing.
Min Kyu Jeong [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Adding a bogus fault that does nothing.
This fault can used to flush the pipe, not including the faulting instruction.

The particular case I needed this was for a self-modifying code. It needed to
drain the store queue and force the following instruction to refetch from
icache. DCCMVAC cp15 mcr instruction is modified to raise this fault.

13 years agoARM: Remove ALPHA KSeg functions.
William Wang [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Remove ALPHA KSeg functions.

These were erronously copied years ago into the ARM directory.

13 years agoARM: Limited implementation of dprintk.
Ali Saidi [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Limited implementation of dprintk.

Does not work with vfp arguments or arguments passed on the stack.

13 years agoARM: Fixed register flattening logic (FP_Base_DepTag was set too low)
Min Kyu Jeong [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Fixed register flattening logic (FP_Base_DepTag was set too low)

When decoding a srs instruction, invalid mode encoding returns invalid instruction.
This can happen when garbage instructions are fetched from mispredicted path

13 years agoARM: Make VMSR, RFE PC/LR etc non speculative, and serializing
Ali Saidi [Thu, 26 Aug 2010 00:10:43 +0000 (19:10 -0500)]
ARM: Make VMSR, RFE PC/LR etc non speculative, and serializing

13 years agoARM: Update regression tests for ldr/str microcode changes.
Ali Saidi [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Update regression tests for ldr/str microcode changes.

13 years agoARM: Use fewer micro-ops for register update loads if possible.
Gene WU [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Use fewer micro-ops for register update loads if possible.

Allow some loads that update the base register to use just two micro-ops. three
micro-ops are only used if the destination register matches the offset register
or the PC is the destination regsiter. If the PC is updated it needs to be
the last micro-op otherwise O3 will mispredict.

13 years agoARM: Set the high bits in the part number so it's considered new by some code.
Ali Saidi [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Set the high bits in the part number so it's considered new by some code.

13 years agoARM: Fix VFP enabled checks for mem instructions
Ali Saidi [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Fix VFP enabled checks for mem instructions

13 years agoARM: Seperate out the renamable bits in the FPSCR.
Gabe Black [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Seperate out the renamable bits in the FPSCR.

13 years agoARM: Eliminate some unused enums.
Gabe Black [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Eliminate some unused enums.

13 years agoARM: Fix type comparison warnings in Neon.
Gabe Black [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Fix type comparison warnings in Neon.

13 years agoARM: Implement CPACR register and return Undefined Instruction when FP access is...
Gabe Black [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Implement CPACR register and return Undefined Instruction when FP access is disabled.

13 years agoARM: Implement all ARM SIMD instructions.
Gabe Black [Thu, 26 Aug 2010 00:10:42 +0000 (19:10 -0500)]
ARM: Implement all ARM SIMD instructions.

13 years agoARM: Expand the mode checking utility functions.
Gabe Black [Thu, 26 Aug 2010 00:10:41 +0000 (19:10 -0500)]
ARM: Expand the mode checking utility functions.

inUserMode now can take either a threadcontext or a CPSR value directly. If
given a thread context it just extracts the CPSR and calls the other version.
An inPrivelegedMode function was also implemented which just returns the
opposite of inUserMode.

13 years agoTracing: Fix trace so 'Predicated False' doesn't show up
Ali Saidi [Thu, 26 Aug 2010 00:10:41 +0000 (19:10 -0500)]
Tracing: Fix trace so 'Predicated False' doesn't show up

13 years agomem: fix dumb typo in copyrights
Steve Reinhardt [Wed, 25 Aug 2010 21:08:27 +0000 (14:08 -0700)]
mem: fix dumb typo in copyrights

13 years agoconfig: changed ruby config file names to be consistent
Brad Beckmann [Tue, 24 Aug 2010 21:08:23 +0000 (14:08 -0700)]
config: changed ruby config file names to be consistent

13 years agoconfig: remove ruby's requirement on the timing cmd line param
Brad Beckmann [Tue, 24 Aug 2010 20:20:32 +0000 (13:20 -0700)]
config: remove ruby's requirement on the timing cmd line param

Since ruby only works in timing mode, explicitly requiring the timing cmd line
param to be specified is not necessary.

13 years agoconfig: fixed ruby dma device connections
Brad Beckmann [Tue, 24 Aug 2010 20:20:31 +0000 (13:20 -0700)]
config: fixed ruby dma device connections

13 years agotesters: move testers to a new directory
Brad Beckmann [Tue, 24 Aug 2010 19:07:22 +0000 (12:07 -0700)]
testers: move testers to a new directory

This patch moves the testers to a new subdirectory under src/cpu and includes
the necessary fixes to work with latest m5 initialization patches.

--HG--
rename : configs/example/determ_test.py => configs/example/ruby_direct_test.py
rename : src/cpu/directedtest/DirectedGenerator.cc => src/cpu/testers/directedtest/DirectedGenerator.cc
rename : src/cpu/directedtest/DirectedGenerator.hh => src/cpu/testers/directedtest/DirectedGenerator.hh
rename : src/cpu/directedtest/InvalidateGenerator.cc => src/cpu/testers/directedtest/InvalidateGenerator.cc
rename : src/cpu/directedtest/InvalidateGenerator.hh => src/cpu/testers/directedtest/InvalidateGenerator.hh
rename : src/cpu/directedtest/RubyDirectedTester.cc => src/cpu/testers/directedtest/RubyDirectedTester.cc
rename : src/cpu/directedtest/RubyDirectedTester.hh => src/cpu/testers/directedtest/RubyDirectedTester.hh
rename : src/cpu/directedtest/RubyDirectedTester.py => src/cpu/testers/directedtest/RubyDirectedTester.py
rename : src/cpu/directedtest/SConscript => src/cpu/testers/directedtest/SConscript
rename : src/cpu/directedtest/SeriesRequestGenerator.cc => src/cpu/testers/directedtest/SeriesRequestGenerator.cc
rename : src/cpu/directedtest/SeriesRequestGenerator.hh => src/cpu/testers/directedtest/SeriesRequestGenerator.hh
rename : src/cpu/memtest/MemTest.py => src/cpu/testers/memtest/MemTest.py
rename : src/cpu/memtest/SConscript => src/cpu/testers/memtest/SConscript
rename : src/cpu/memtest/memtest.cc => src/cpu/testers/memtest/memtest.cc
rename : src/cpu/memtest/memtest.hh => src/cpu/testers/memtest/memtest.hh
rename : src/cpu/rubytest/Check.cc => src/cpu/testers/rubytest/Check.cc
rename : src/cpu/rubytest/Check.hh => src/cpu/testers/rubytest/Check.hh
rename : src/cpu/rubytest/CheckTable.cc => src/cpu/testers/rubytest/CheckTable.cc
rename : src/cpu/rubytest/CheckTable.hh => src/cpu/testers/rubytest/CheckTable.hh
rename : src/cpu/rubytest/RubyTester.cc => src/cpu/testers/rubytest/RubyTester.cc
rename : src/cpu/rubytest/RubyTester.hh => src/cpu/testers/rubytest/RubyTester.hh
rename : src/cpu/rubytest/RubyTester.py => src/cpu/testers/rubytest/RubyTester.py
rename : src/cpu/rubytest/SConscript => src/cpu/testers/rubytest/SConscript

13 years agoMOESI_hammer: fixed bug for dma reads in single cpu systems
Brad Beckmann [Tue, 24 Aug 2010 19:06:53 +0000 (12:06 -0700)]
MOESI_hammer: fixed bug for dma reads in single cpu systems

13 years agoFaults: Get rid of some commented out code in sim/faults.hh.
Gabe Black [Mon, 23 Aug 2010 23:23:47 +0000 (16:23 -0700)]
Faults: Get rid of some commented out code in sim/faults.hh.

13 years agoX86: Create a directory for files that define register indexes.
Gabe Black [Mon, 23 Aug 2010 23:14:24 +0000 (16:14 -0700)]
X86: Create a directory for files that define register indexes.

This is to help tidy up arch/x86. These files should not be used external to
the ISA.

--HG--
rename : src/arch/x86/apicregs.hh => src/arch/x86/regs/apic.hh
rename : src/arch/x86/floatregs.hh => src/arch/x86/regs/float.hh
rename : src/arch/x86/intregs.hh => src/arch/x86/regs/int.hh
rename : src/arch/x86/miscregs.hh => src/arch/x86/regs/misc.hh
rename : src/arch/x86/segmentregs.hh => src/arch/x86/regs/segment.hh

13 years agoPower: Get rid of unused checkFpEnableFault.
Gabe Black [Mon, 23 Aug 2010 23:14:23 +0000 (16:14 -0700)]
Power: Get rid of unused checkFpEnableFault.

This function was brought in from another ISA and doesn't actually do anything
or get used.

13 years agoISA: Get rid of old, unused utility functions cluttering up the ISAs.
Gabe Black [Mon, 23 Aug 2010 23:14:20 +0000 (16:14 -0700)]
ISA: Get rid of old, unused utility functions cluttering up the ISAs.

13 years agoX86: Get rid of the flagless microop constructor.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Get rid of the flagless microop constructor.

This will reduce clutter in the source and hopefully speed up compilation.

13 years agoX86: Make the TLB fault instead of panic when something is unmapped in SE mode.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Make the TLB fault instead of panic when something is unmapped in SE mode.

The fault object, if invoked, would then panic. This is a bit less direct, but
it means speculative execution won't panic the simulator.

13 years agoX86: Make the x86 ExtMachInst serializable with (UN)SERIALIZE_SCALAR.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Make the x86 ExtMachInst serializable with (UN)SERIALIZE_SCALAR.

--HG--
rename : src/arch/x86/types.hh => src/arch/x86/types.cc

13 years agoX86: Define a noop ExtMachInst.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Define a noop ExtMachInst.

13 years agoX86: Mark serializing macroops and regular instructions as such.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Mark serializing macroops and regular instructions as such.

13 years agoX86: Add a .serializing directive that makes a macroop serializing.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Add a .serializing directive that makes a macroop serializing.

This directive really just tells the macroop to set IsSerializing and
IsSerializeAfter on its final microop.

13 years agoX86: Consolidate extra microop flags into one parameter.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
X86: Consolidate extra microop flags into one parameter.

This single parameter replaces the collection of bools that set up various
flavors of microops. A flag parameter also allows other flags to be set like
the serialize before/after flags, etc., without having to change the
constructor.

13 years agoCPU: Make the constants for StaticInst flags visible outside the class.
Gabe Black [Mon, 23 Aug 2010 16:44:19 +0000 (09:44 -0700)]
CPU: Make the constants for StaticInst flags visible outside the class.

13 years agoBUILD: GCC 4.4.1/2 have a bug in their auto-vectorizer that we trip on
Ali Saidi [Mon, 23 Aug 2010 16:18:42 +0000 (11:18 -0500)]
BUILD: GCC 4.4.1/2 have a bug in their auto-vectorizer that we trip on

13 years agoALPHA: The previous O3 patch causes a slight stats change with fullsys.
Ali Saidi [Mon, 23 Aug 2010 16:18:42 +0000 (11:18 -0500)]
ALPHA: The previous O3 patch causes a slight stats change with fullsys.

13 years agoO3: Skipping mem-order violation check for uncachable loads.
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:42 +0000 (11:18 -0500)]
O3: Skipping mem-order violation check for uncachable loads.
Uncachable load is not executed until it reaches the head of the ROB,
hence cannot cause one.

13 years agoARM: Improve printing of uop disassembly.
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:42 +0000 (11:18 -0500)]
ARM: Improve printing of uop disassembly.

13 years agoARM: Clean up flattening for SPSR adding
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Clean up flattening for SPSR adding

13 years agoARM: Implement DBG instruction that doesn't do much for now.
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Implement DBG instruction that doesn't do much for now.

13 years agoMEM: Make CLREX a first class request operation and clear locks in caches when it...
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
MEM: Make CLREX a first class request operation and clear locks in caches when it in received

13 years agoARM: Make sure that software prefetch instructions can't change the state of the TLB
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Make sure that software prefetch instructions can't change the state of the TLB

13 years agoARM: Don't write tracedata on writes, it might have been freed already.
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Don't write tracedata on writes, it might have been freed already.

13 years agoARM: Implement CLREX init/complete acc methods
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Implement CLREX init/complete acc methods

13 years agoARM: Fix Uncachable TLB requests and decoding of xn bit
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Fix Uncachable TLB requests and decoding of xn bit

13 years agoDevices: Allow a device to specify that a request is uncachable.
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
Devices: Allow a device to specify that a request is uncachable.

13 years agoARM: For non-cachable accesses set the UNCACHABLE flag
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: For non-cachable accesses set the UNCACHABLE flag

13 years agoARM: Implement DSB, DMB, ISB
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Implement DSB, DMB, ISB

13 years agoARM: Get SCTLR TE bit from reset SCTLR
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Get SCTLR TE bit from reset SCTLR

13 years agoARM: Implement CLREX
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: Implement CLREX

13 years agoARM: BX instruction can be contitional if last instruction in a IT block
Gene Wu [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: BX instruction can be contitional if last instruction in a IT block

Branches are allowed to be the last instuction in an IT block. Before it was
assumed that they could not. So Branches in thumb2 were Uncond.

13 years agoCPU: Print out flatten-out register index as with IntRegs/FloatRegs traceflag
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
CPU: Print out flatten-out register index as with IntRegs/FloatRegs traceflag

13 years agoCPU: Make Exec trace to print predication result (if false) for memory instructions
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
CPU: Make Exec trace to print predication result (if false) for memory instructions

13 years agoARM: mark msr/mrs instructions as SerializeBefore/After
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:41 +0000 (11:18 -0500)]
ARM: mark msr/mrs instructions as SerializeBefore/After
Since miscellaneous registers bypass wakeup logic, force serialization
to resolve data dependencies through them
* * *
ARM: adding non-speculative/serialize flags for instructions change CPSR

13 years agoO3: Handle loads when the destination is the PC.
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
O3: Handle loads when the destination is the PC.
For loads that PC is the destination, check if the load
was mispredicted again when the value being loaded returns from memory

13 years agoARM/O3: store the result of the predicate evaluation in DynInst or Threadstate.
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM/O3: store the result of the predicate evaluation in DynInst or Threadstate.
THis allows the CPU to handle predicated-false instructions accordingly.
This particular patch makes loads that are predicated-false to be sent
straight to the commit stage directly, not waiting for return of the data
that was never requested since it was predicated-false.

13 years agoARM: adding genMachineCheckFault() stub for ARM that doesn't panic
Min Kyu Jeong [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: adding genMachineCheckFault() stub for ARM that doesn't panic

13 years agoARM: DFSR status value for sync external data abort is expected to be 0x8 in ARMv7
Gene Wu [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: DFSR status value for sync external data abort is expected to be 0x8 in ARMv7

13 years agoARM: Temporary local variables can't conflict with isa parser operands.
Gene Wu [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: Temporary local variables can't conflict with isa parser operands.
PC is an operand, so we can't have a temp called PC

13 years agoARM: Exclusive accesses must be double word aligned
Ali Saidi [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: Exclusive accesses must be double word aligned

13 years agoARM: Add some registers for big loads/stores to support neon.
Ali Saidi [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: Add some registers for big loads/stores to support neon.

13 years agoARM: Decode neon memory instructions.
Ali Saidi [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: Decode neon memory instructions.

13 years agoARM: Clean up the ISA desc portion of the ARM memory instructions.
Gabe Black [Mon, 23 Aug 2010 16:18:40 +0000 (11:18 -0500)]
ARM: Clean up the ISA desc portion of the ARM memory instructions.