cpu/base_dyn_inst.hh:
Remove snoop function (did not mean to commit it).
cpu/ozone/back_end_impl.hh:
Set instruction as having its result ready, not completed.
cpu/ozone/cpu.hh:
Fixes for store conditionals. Use an additional lock addr list to make sure that the access is valid. I don't know if this is fully necessary, but it gives me a peace of mind (at some performance cost).
Make sure to schedule for cycles(1) and not just 1 cycle in the future as tick = 1ps.
Also support the new Checker.
cpu/ozone/cpu_builder.cc:
Add parameter for maxOutstandingMemOps so it can be set through the config.
Also add in the checker. Right now it's a BaseCPU simobject, but that may change in the future.
cpu/ozone/cpu_impl.hh:
Add support for the checker. For now there's a dynamic cast to convert the simobject passed back from the builder to the proper Checker type. It's ugly, but only happens at startup, and is probably a justified use of dynamic cast.
Support switching out/taking over from other CPUs.
Correct indexing problem for float registers.
cpu/ozone/dyn_inst.hh:
Add ability for instructions to wait on memory instructions in addition to source register instructions. This is needed for memory dependence predictors and memory barriers.
cpu/ozone/dyn_inst_impl.hh:
Support waiting on memory operations.
Use "resultReady" to differentiate an instruction having its registers produced vs being totally completed.
cpu/ozone/front_end.hh:
Support switching out.
Also record if an interrupt is pending.
cpu/ozone/front_end_impl.hh:
Support switching out. Also support stalling the front end if an interrupt is pending.
cpu/ozone/lw_back_end.hh:
Add checker in.
Support switching out.
Support memory barriers.
cpu/ozone/lw_back_end_impl.hh:
Lots of changes to get things to work right.
Faults, traps, interrupts all wait until all stores have written back (important).
Memory barriers are supported, as is the general ability for instructions to be dependent on other memory instructions.
cpu/ozone/lw_lsq.hh:
Support switching out.
Also use store writeback events in all cases, not just dcache misses.
cpu/ozone/lw_lsq_impl.hh:
Support switching out.
Also use store writeback events in all cases, not just dcache misses.
Support the checker CPU. Marks instructions as completed once the functional access is done (which has to be done for the checker to be able to verify results).
cpu/ozone/simple_params.hh:
Add max outstanding mem ops parameter.
python/m5/objects/OzoneCPU.py:
Add max outstanding mem ops, checker.
--HG--
extra : convert_revision : f4d408e1bb1f25836a097b6abe3856111e950c59
Intended Use:
A SimObject will call the serializer when it needs the state to be serializable (i.e. switchCPUs, checkpoint, switch memory access model). It will call the requestSeialization() function.
The Serializer will signal all the objects in its list to drain their state via the SimObject method drain(). Drain() has a default implementation to just signal done.
When each object is drained it will signal the Serializer that it has drained via the signalDrained() function.
The Serializer will collect these signals, when all have drained it will signal the initial requestor via serializationComplete() method in the SimObject.
Once that object is done, it will signal the Serializer to resumeExecution().
The Serializer will signal all the objects in its list to resume via the resume() method on the SimObject.
SConscript:
Add serializer object to build list
sim/sim_object.cc:
Add default behavior for drain (just signal finished, must be overided if you really must drain something)
sim/sim_object.hh:
Add functions for serializer
--HG--
extra : convert_revision : 15aa2d1b42010c2d703bef9114c11d079c216170
cpu/o3/alpha_dyn_inst.hh:
Set the instResult using a function on the base dyn inst.
cpu/o3/bpred_unit_impl.hh:
Don't need to reset the state.
cpu/o3/commit_impl.hh:
Mark instructions as completed.
Wait until all stores are written back to handle a fault.
cpu/o3/cpu.cc:
Clear instruction lists when switching out.
cpu/o3/lsq_unit.hh:
Allow wbEvent to be set externally.
cpu/o3/lsq_unit_impl.hh:
Mark instructions as completed properly. Also use events for writing back stores even if there is a hit in the dcache.
--HG--
extra : convert_revision : 172ad088b75ac31e848a5040633152b5c051444c
Now all the variations of FP should be implemented correctly in the decoder.
The new formats and functions supporting these functions need to be implemented for
some of the FP stuff but for the most part things are looking like their "supposed to"...
arch/mips/isa/decoder.isa:
Fixes for Paired-Single FP Compare Operations...
Now all the variations of FP should be implemented correctly in the decoder.
arch/mips/isa/formats/fp.isa:
Add new PS formats
arch/mips/isa_traits.cc:
Add skeleton overloaded round & truncate functions
arch/mips/isa_traits.hh:
declare overloaded functions
--HG--
extra : convert_revision : 15d5cf7b08ac2dc9ebcd6b268e92d4abffdd8597
Auxiliary Functions and Formats for FP in general
arch/mips/isa/decoder.isa:
ISA Parser doesnt accept operands of different types in one instruction so fix this for unorderedFP functions...
Add basic support for Paired Singled (PS) FP ops which happen to be part of the MIPS 32-ASE but turned out to
be included in the MIPS32ISA manual... The PS instructions allow SIMD in a pipeline...
arch/mips/isa/formats/fp.isa:
Add some more Formats for FP operation. I will add some auxiliary code through these formats
to alleviate code redundancy in the decoder.isa
arch/mips/isa/operands.isa:
Add operands for Paired Singles Ops
arch/mips/isa_traits.cc:
removed convert&round function and replace with fpConvert.
The whole "rounding mode" stuff is something that should be considered for full-system mode...
Also added skeletons for the unorderedFP,truncFP,and condition code funcs.
arch/mips/isa_traits.hh:
declare some Functions
arch/mips/types.hh:
add new conversion types
--HG--
extra : convert_revision : 79251d590a27b74a3d6a62a2fbb937df3e59963f
into zazzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem
--HG--
extra : convert_revision : 0a146eed200abd2c18f135b112987c5cf91a649b
There still needs to be a work around to handle the paired singles operations ...
arch/mips/isa/decoder.isa:
More revamping of the floating point ops in decoder.isa. Change all of the
"convert and round" functions to fpConvert. Also, the utility functions
roundFP, truncFP, and unorderedFP are in place everywhere. Things
have been set up to appropriately use the FP condition codes in the decoder.isa
The fp.isa format file and the isa_traits.cc file now needed to be updated
to implement the appropriate "backend" operations/functionality...
arch/mips/isa_traits.hh:
Remove convert & round functions
Add roundFP, truncFP,unorderedFP, and the get/setFPconditionCode
functions
arch/mips/isa_traits.cc:
Add utility functions
--HG--
extra : convert_revision : 3d6708388abae5b432467f528d52e6343afecd9c
Now handles instructions for FP compares in single or double recision
arch/mips/isa/decoder.isa:
Now handles instructions for FP compares in single or double recision
--HG--
extra : convert_revision : eb3a13616e6736bf2d1ead0b816dda8c6099b20f
Basic Code for Floating Point Compare with Single Precision Floats
Added.
arch/mips/isa/decoder.isa:
Basic Code for Floating Point Compare with Single Precision Floats
Added.
--HG--
extra : convert_revision : 56b14da1e9d987c2d2090fd2f79af8b12fe8d2ec
Added support for FP compare instructions. Somehow these flew beneath
my radar. Also, I start to use special FP utility functions in FP code.
Right now, they are defined in isa_traits.hh but may be moved in the
future
arch/mips/isa/decoder.isa:
Added support for FP compare instructions. Somehow these flew beneath
my radar. Also, I start to use special FP utility functions in FP code.
Right now, they are defined in isa_traits.hh but may be moved in the
future
--HG--
extra : convert_revision : 84a3b66882f3977ce9c1356cf466d62a7fd8bf19
Debug FP instructions to handle these FP insts
arch/mips/isa/bitfields.isa:
add Bitfield for Floating Point Condition Codes
arch/mips/isa/decoder.isa:
Follow instruction naming style with FP single insts
Send the float value to the convert&round functions in single FP
add ll inst support
add 'token' sc support
arch/mips/isa_traits.cc:
Add SINGLE->WORD, WORD->SINGLE, & WORD->DOUBLE conversions
arch/mips/regfile.hh:
update header files
arch/mips/regfile/float_regfile.hh:
Add more FP registers
--HG--
rename : arch/mips/int_regfile.hh => arch/mips/regfile/int_regfile.hh
rename : arch/mips/misc_regfile.hh => arch/mips/regfile/misc_regfile.hh
extra : convert_revision : 92faf0bfd8542ade762ac569ec158d198f6a9c7e
Now we are to the point where more benchmarks and instruction-coverage
is necessary to totally verify/validate correct operation across
all MIPS instructions
arch/mips/isa_traits.hh:
fix for reading double values ... must rearrange bits before using void* to read double.
configs/test/hello_mips:
real hello world MIPS binary
--HG--
extra : convert_revision : 153de1f8a830882c6972bd0bdb56da818f614def
register with the higher # contains the most significant bytes...
arch/mips/isa/decoder.isa:
divide instruction fixes
arch/mips/isa_traits.cc:
use double as argument to cvt & round function.
clean up cout statements in function.
arch/mips/isa_traits.hh:
In MIPS the higher # reg of a doubles pair is ALSO the most significant reg.
Once I switched this the basic MIPS FP test I had worked.
--HG--
extra : convert_revision : 45c80df229e6174d0b52fc7cfb530642b1f1fc35
into zizzer.eecs.umich.edu:/.automount/zooks/y/ksewell/research/m5-sim/newmem
--HG--
extra : convert_revision : c48a8857f5a520ff8061eb3d8f08dcd43661e68c
cpu/simple/cpu.cc:
Sampler fixes. The status may be switched out when calling activate or suspend if there is a switchover during a quiesce.
--HG--
extra : convert_revision : da026e75dfb86289484cf01c5b1ecd9b03a72bd3
Edit the convert_and_round function which access FloatRegFile
arch/isa_parser.py:
recognize when we are writing a 'uint64_t' FloatReg and set the width appropriately
arch/mips/isa/decoder.isa:
Send a 'float' to the convert function instead of a unsigned word. Do this so we dont have to worry about the
bit manipulation ourselves. We can just concern ourselves with values.
Use unsigned double to get movd...
arch/mips/isa/formats/fp.isa:
float debug statement
arch/mips/isa_traits.cc:
add different versions of convert_and_round functions
arch/mips/isa_traits.hh:
Use an array of uint32_t unsigned integers to represent the Floating Point Regfile
configs/test/hello_mips:
basic FP program
cpu/simple/cpu.hh:
spacing
--HG--
extra : convert_revision : a6fca91ad6365c83025f1131d71fa1b8ee76d7bc
util/stats/barchart.py:
- there is no self.inner_axes
- don't append an empty value to self.xsubticks, otherwise
subsequent calls will get extra empty ticks
- rotate labels 30 degrees instead of 90 so it looks better
--HG--
extra : convert_revision : 1cbac6d1f92bfc6b2c1e886ad5f9d4c78a2b3820
fix very annoying not-compiler bug
arch/sparc/regfile.hh:
You have not included an out-of-class definition of your static members. See [9.4.2]/4 and about a billion gcc bug reports.
If statements get around the problem through some magic, and than seems nicer that putting a definition of them in a c file
somewhere.
cpu/simple/cpu.cc:
get() and set() do the conversion now
dev/io_device.hh:
need get() and set() defentions in all the devices
mem/packet.cc:
mem/packet.hh:
move code from packet.hh to packet.cc
mem/physical.cc:
packet_impl needed for templated packet functions
--HG--
extra : convert_revision : 6c11842aa928d9af7b4cabe826306fe1fe09e693
arch/sparc/isa/decoder.isa:
Made sure if a register was assigned to along some control path, then all paths on which no exception would block commit set a value as well. Also, Rs1 is treated as signed for bpr instructions.
arch/sparc/isa/formats/integerop.isa:
Added an IntOpImm11 class which sign extends the SIMM11 immediate field.
arch/sparc/isa/formats/mem.isa:
Fixed how offsets are used, and how disassembly is generated.
arch/sparc/linux/process.cc:
Added fstat and exit_group syscalls.
--HG--
extra : convert_revision : 3b4427d239d254a92179a4137441125b8a364264
configs/test/fs.py:
update fs.py to use a bus bridge
cpu/simple/cpu.hh:
cpu should just return that it doesn't snoop any address ranges
python/m5/objects/System.py:
move boot_osflags to system
--HG--
extra : convert_revision : b4256df7eada7e65b69513361de8bffc3fdd680b
SConscript:
add new cc files to scons
mem/bus.cc:
mem/bus.hh:
implement addressRanges() on the bus.
propigate address ranges to anyone who is interested stripping out ranges of who your propigating to (to avoid livelock)
mem/packet.hh:
add intersect function that returns true if two packets touch at least one byte of the same data (for functional access)
add fixPacket() that will eventually take the correct action giving a timing and functional packet, right now it panics
mem/physical.cc:
Don't panic if the physical memory recieves a status change, just ignore.
--HG--
extra : convert_revision : d470d51f2fb1db2700ad271e09792315ef33ba01
arch/mips/isa/decoder.isa:
Fix Reg. Operands for FP Conversion Instructions - Must Make Sure That You use 'uw' or 'ud' as needed.
arch/mips/isa_traits.cc:
if a conversion function isnt implemented yet, than have M5 panic...
(plan to implement SINGLE_TO_DOUBLE first)
--HG--
extra : convert_revision : 6a7f703a5d65139d3981a8753c31fc8f5bf313cf
Have FP conversion instructions use re-defined convert_and_round() function
arch/mips/isa/decoder.isa:
Use lower two bits of FCSR reg to determine rounding mode (may want to move this out of decoder.isa and into a template)
Have FP conversion instructions to use re-defined convert_and_round() function
arch/mips/isa/formats/util.isa:
Remove convert_and_round function from here
arch/mips/isa_traits.cc:
Define convert_and_round function here
arch/mips/isa_traits.hh:
Use "enums" to define FP conversion types & Round Modes
Declare convert_and_round function here
--HG--
extra : convert_revision : 0f4f8c1732a53b277361559ea71af2a1feb4fc64
