gem5/src/arch/sparc/process.cc

560 lines
20 KiB
C++
Raw Normal View History

/*
* Copyright (c) 2003-2004 The Regents of The University of Michigan
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met: redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer;
* redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution;
* neither the name of the copyright holders nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Authors: Gabe Black
* Ali Saidi
*/
#include "arch/sparc/asi.hh"
#include "arch/sparc/handlers.hh"
#include "arch/sparc/isa_traits.hh"
#include "arch/sparc/process.hh"
2011-04-15 19:44:06 +02:00
#include "arch/sparc/registers.hh"
#include "arch/sparc/types.hh"
#include "base/loader/elf_object.hh"
2011-04-15 19:44:06 +02:00
#include "base/loader/object_file.hh"
#include "base/misc.hh"
Change ExecContext to ThreadContext. This is being renamed to differentiate between the interface used objects outside of the CPU, and the interface used by the ISA. ThreadContext is used by objects outside of the CPU and is specifically defined in thread_context.hh. ExecContext is more implicit, and is defined by files such as base_dyn_inst.hh or cpu/simple/base.hh. Further renames/reorganization will be coming shortly; what is currently CPUExecContext (the old ExecContext from m5) will be renamed to SimpleThread or something similar. src/arch/alpha/arguments.cc: src/arch/alpha/arguments.hh: src/arch/alpha/ev5.cc: src/arch/alpha/faults.cc: src/arch/alpha/faults.hh: src/arch/alpha/freebsd/system.cc: src/arch/alpha/freebsd/system.hh: src/arch/alpha/isa/branch.isa: src/arch/alpha/isa/decoder.isa: src/arch/alpha/isa/main.isa: src/arch/alpha/linux/process.cc: src/arch/alpha/linux/system.cc: src/arch/alpha/linux/system.hh: src/arch/alpha/linux/threadinfo.hh: src/arch/alpha/process.cc: src/arch/alpha/regfile.hh: src/arch/alpha/stacktrace.cc: src/arch/alpha/stacktrace.hh: src/arch/alpha/tlb.cc: src/arch/alpha/tlb.hh: src/arch/alpha/tru64/process.cc: src/arch/alpha/tru64/system.cc: src/arch/alpha/tru64/system.hh: src/arch/alpha/utility.hh: src/arch/alpha/vtophys.cc: src/arch/alpha/vtophys.hh: src/arch/mips/faults.cc: src/arch/mips/faults.hh: src/arch/mips/isa_traits.cc: src/arch/mips/isa_traits.hh: src/arch/mips/linux/process.cc: src/arch/mips/process.cc: src/arch/mips/regfile/float_regfile.hh: src/arch/mips/regfile/int_regfile.hh: src/arch/mips/regfile/misc_regfile.hh: src/arch/mips/regfile/regfile.hh: src/arch/mips/stacktrace.hh: src/arch/sparc/faults.cc: src/arch/sparc/faults.hh: src/arch/sparc/isa_traits.hh: src/arch/sparc/linux/process.cc: src/arch/sparc/linux/process.hh: src/arch/sparc/process.cc: src/arch/sparc/regfile.hh: src/arch/sparc/solaris/process.cc: src/arch/sparc/stacktrace.hh: src/arch/sparc/ua2005.cc: src/arch/sparc/utility.hh: src/arch/sparc/vtophys.cc: src/arch/sparc/vtophys.hh: src/base/remote_gdb.cc: src/base/remote_gdb.hh: src/cpu/base.cc: src/cpu/base.hh: src/cpu/base_dyn_inst.hh: src/cpu/checker/cpu.cc: src/cpu/checker/cpu.hh: src/cpu/checker/exec_context.hh: src/cpu/cpu_exec_context.cc: src/cpu/cpu_exec_context.hh: src/cpu/cpuevent.cc: src/cpu/cpuevent.hh: src/cpu/exetrace.hh: src/cpu/intr_control.cc: src/cpu/memtest/memtest.hh: src/cpu/o3/alpha_cpu.hh: src/cpu/o3/alpha_cpu_impl.hh: src/cpu/o3/alpha_dyn_inst_impl.hh: src/cpu/o3/commit.hh: src/cpu/o3/commit_impl.hh: src/cpu/o3/cpu.cc: src/cpu/o3/cpu.hh: src/cpu/o3/fetch_impl.hh: src/cpu/o3/regfile.hh: src/cpu/o3/thread_state.hh: src/cpu/ozone/back_end.hh: 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/inorder_back_end.hh: src/cpu/ozone/lw_back_end.hh: src/cpu/ozone/lw_back_end_impl.hh: src/cpu/ozone/lw_lsq.hh: src/cpu/ozone/lw_lsq_impl.hh: src/cpu/ozone/thread_state.hh: src/cpu/pc_event.cc: src/cpu/pc_event.hh: src/cpu/profile.cc: src/cpu/profile.hh: src/cpu/quiesce_event.cc: src/cpu/quiesce_event.hh: src/cpu/simple/atomic.cc: src/cpu/simple/base.cc: src/cpu/simple/base.hh: src/cpu/simple/timing.cc: src/cpu/static_inst.cc: src/cpu/static_inst.hh: src/cpu/thread_state.hh: src/dev/alpha_console.cc: src/dev/ns_gige.cc: src/dev/sinic.cc: src/dev/tsunami_cchip.cc: src/kern/kernel_stats.cc: src/kern/kernel_stats.hh: src/kern/linux/events.cc: src/kern/linux/events.hh: src/kern/system_events.cc: src/kern/system_events.hh: src/kern/tru64/dump_mbuf.cc: src/kern/tru64/tru64.hh: src/kern/tru64/tru64_events.cc: src/kern/tru64/tru64_events.hh: src/mem/vport.cc: src/mem/vport.hh: src/sim/faults.cc: src/sim/faults.hh: src/sim/process.cc: src/sim/process.hh: src/sim/pseudo_inst.cc: src/sim/pseudo_inst.hh: src/sim/syscall_emul.cc: src/sim/syscall_emul.hh: src/sim/system.cc: src/cpu/thread_context.hh: src/sim/system.hh: src/sim/vptr.hh: Change ExecContext to ThreadContext. --HG-- rename : src/cpu/exec_context.hh => src/cpu/thread_context.hh extra : convert_revision : 108bb97d15a114a565a2a6a23faa554f4e2fd77e
2006-06-06 23:32:21 +02:00
#include "cpu/thread_context.hh"
#include "debug/Stack.hh"
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
#include "mem/page_table.hh"
2011-04-15 19:44:06 +02:00
#include "sim/process_impl.hh"
#include "sim/system.hh"
using namespace std;
using namespace SparcISA;
static const int FirstArgumentReg = 8;
SparcLiveProcess::SparcLiveProcess(LiveProcessParams * params,
ObjectFile *objFile, Addr _StackBias)
: LiveProcess(params, objFile), StackBias(_StackBias)
{
// XXX all the below need to be updated for SPARC - Ali
brk_point = objFile->dataBase() + objFile->dataSize() + objFile->bssSize();
brk_point = roundUp(brk_point, VMPageSize);
// Set pointer for next thread stack. Reserve 8M for main stack.
next_thread_stack_base = stack_base - (8 * 1024 * 1024);
// Initialize these to 0s
fillStart = 0;
spillStart = 0;
}
void
SparcLiveProcess::handleTrap(int trapNum, ThreadContext *tc)
{
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
PCState pc = tc->pcState();
switch (trapNum) {
case 0x01: // Software breakpoint
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
warn("Software breakpoint encountered at pc %#x.\n", pc.pc());
break;
case 0x02: // Division by zero
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
warn("Software signaled a division by zero at pc %#x.\n", pc.pc());
break;
case 0x03: // Flush window trap
flushWindows(tc);
break;
case 0x04: // Clean windows
warn("Ignoring process request for clean register "
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
"windows at pc %#x.\n", pc.pc());
break;
case 0x05: // Range check
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
warn("Software signaled a range check at pc %#x.\n", pc.pc());
break;
case 0x06: // Fix alignment
warn("Ignoring process request for os assisted unaligned accesses "
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
"at pc %#x.\n", pc.pc());
break;
case 0x07: // Integer overflow
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
warn("Software signaled an integer overflow at pc %#x.\n", pc.pc());
break;
case 0x32: // Get integer condition codes
warn("Ignoring process request to get the integer condition codes "
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
"at pc %#x.\n", pc.pc());
break;
case 0x33: // Set integer condition codes
warn("Ignoring process request to set the integer condition codes "
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
"at pc %#x.\n", pc.pc());
break;
default:
panic("Unimplemented trap to operating system: trap number %#x.\n", trapNum);
}
}
void
SparcLiveProcess::initState()
{
LiveProcess::initState();
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
ThreadContext *tc = system->getThreadContext(contextIds[0]);
// From the SPARC ABI
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
// Setup default FP state
tc->setMiscRegNoEffect(MISCREG_FSR, 0);
Implement PR/HPR/ASR for full system Rip out storage in miscreg file that will never store anything Add storage and defines for Priv and Hyperpriv registers Change defines to match the spec register numbers Change the way misc registers are named to match the spec with offsets to deal with ASR/PR/HPR/FSR. Change contextval to an int since both global registers and windowed registers are indexed by int in UA2005. Use bitfields for things that are rarely used in decoder Instead of decoding ASR/PR/HPR and having a specfic instruction, use a generic instruction instead Still todo: Protect rdpr, rdhpr, wrpr, wrhpr with checks that fault in insufficient privs Deal with signaling interrupts on timer expiration Deal with writes to softint/PIL generating interrupts how those are vectored to the CPU Other misc: Instruction decoding needs major help! src/arch/sparc/isa/decoder.isa: Remove tons of MISCREG_XXXX defines that weren't used and ControlRegs in that were never used. Ones that were used rarely changed to bitfields. src/arch/sparc/isa/formats/integerop.isa: These seems like a whole lot of overkill in printing, but i'll leave it the way it is for now. Allow Ccr to be set at once src/arch/sparc/isa/formats/priv.isa: PrivTick is handled by miscreg now, don't need a seperate class for it src/arch/sparc/isa/operands.isa: prune the number of control regs down to a reasonable amount src/arch/sparc/isa_traits.hh: Replace 8 defines with 1 and flick some bits src/arch/sparc/process.cc: Better to clean the entire registers that specific bits which leads to indetermanistic behavior. src/arch/sparc/regfile.hh: Rip out storage that will never be backed by anything Add storage for Priv and Hyperpriv registers change defines to match the spec change the way misc registers are named to match the spec with offsets to deal with ASR/PR/HPR/FSR. change contextval to an int since both global registers and windowed registers are indexed by int in UA2005. --HG-- extra : convert_revision : 64276a3ea884eea70112e721f85a515946ded4c2
2006-05-27 00:40:00 +02:00
tc->setMiscRegNoEffect(MISCREG_TICK, 0);
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
/*
* Register window management registers
*/
// No windows contain info from other programs
// tc->setMiscRegNoEffect(MISCREG_OTHERWIN, 0);
tc->setIntReg(NumIntArchRegs + 6, 0);
// There are no windows to pop
// tc->setMiscRegNoEffect(MISCREG_CANRESTORE, 0);
tc->setIntReg(NumIntArchRegs + 4, 0);
// All windows are available to save into
// tc->setMiscRegNoEffect(MISCREG_CANSAVE, NWindows - 2);
tc->setIntReg(NumIntArchRegs + 3, NWindows - 2);
// All windows are "clean"
// tc->setMiscRegNoEffect(MISCREG_CLEANWIN, NWindows);
tc->setIntReg(NumIntArchRegs + 5, NWindows);
// Start with register window 0
tc->setMiscReg(MISCREG_CWP, 0);
// Always use spill and fill traps 0
// tc->setMiscRegNoEffect(MISCREG_WSTATE, 0);
tc->setIntReg(NumIntArchRegs + 7, 0);
// Set the trap level to 0
tc->setMiscRegNoEffect(MISCREG_TL, 0);
// Set the ASI register to something fixed
tc->setMiscRegNoEffect(MISCREG_ASI, ASI_PRIMARY);
/*
* T1 specific registers
*/
// Turn on the icache, dcache, dtb translation, and itb translation.
tc->setMiscRegNoEffect(MISCREG_MMU_LSU_CTRL, 15);
}
void
Sparc32LiveProcess::initState()
{
SparcLiveProcess::initState();
ThreadContext *tc = system->getThreadContext(contextIds[0]);
// The process runs in user mode with 32 bit addresses
tc->setMiscReg(MISCREG_PSTATE, 0x0a);
argsInit(32 / 8, VMPageSize);
}
void
Sparc64LiveProcess::initState()
{
SparcLiveProcess::initState();
ThreadContext *tc = system->getThreadContext(contextIds[0]);
// The process runs in user mode
tc->setMiscReg(MISCREG_PSTATE, 0x02);
argsInit(sizeof(IntReg), VMPageSize);
}
template<class IntType>
void
SparcLiveProcess::argsInit(int pageSize)
{
int intSize = sizeof(IntType);
2008-12-07 21:07:42 +01:00
typedef AuxVector<IntType> auxv_t;
std::vector<auxv_t> auxv;
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
string filename;
if (argv.size() < 1)
filename = "";
else
filename = argv[0];
// Even for a 32 bit process, the ABI says we still need to
// maintain double word alignment of the stack pointer.
uint64_t align = 16;
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
// load object file into target memory
objFile->loadSections(initVirtMem);
enum hardwareCaps
{
M5_HWCAP_SPARC_FLUSH = 1,
M5_HWCAP_SPARC_STBAR = 2,
M5_HWCAP_SPARC_SWAP = 4,
M5_HWCAP_SPARC_MULDIV = 8,
M5_HWCAP_SPARC_V9 = 16,
// This one should technically only be set
// if there is a cheetah or cheetah_plus tlb,
// but we'll use it all the time
M5_HWCAP_SPARC_ULTRA3 = 32
};
const int64_t hwcap =
M5_HWCAP_SPARC_FLUSH |
M5_HWCAP_SPARC_STBAR |
M5_HWCAP_SPARC_SWAP |
M5_HWCAP_SPARC_MULDIV |
M5_HWCAP_SPARC_V9 |
M5_HWCAP_SPARC_ULTRA3;
// Setup the auxilliary vectors. These will already have endian conversion.
// Auxilliary vectors are loaded only for elf formatted executables.
ElfObject * elfObject = dynamic_cast<ElfObject *>(objFile);
if (elfObject) {
// Bits which describe the system hardware capabilities
auxv.push_back(auxv_t(M5_AT_HWCAP, hwcap));
// The system page size
auxv.push_back(auxv_t(M5_AT_PAGESZ, SparcISA::VMPageSize));
// Defined to be 100 in the kernel source.
// Frequency at which times() increments
auxv.push_back(auxv_t(M5_AT_CLKTCK, 100));
// For statically linked executables, this is the virtual address of the
// program header tables if they appear in the executable image
auxv.push_back(auxv_t(M5_AT_PHDR, elfObject->programHeaderTable()));
// This is the size of a program header entry from the elf file.
auxv.push_back(auxv_t(M5_AT_PHENT, elfObject->programHeaderSize()));
// This is the number of program headers from the original elf file.
auxv.push_back(auxv_t(M5_AT_PHNUM, elfObject->programHeaderCount()));
// This is the address of the elf "interpreter", It should be set
// to 0 for regular executables. It should be something else
// (not sure what) for dynamic libraries.
auxv.push_back(auxv_t(M5_AT_BASE, 0));
// This is hardwired to 0 in the elf loading code in the kernel
auxv.push_back(auxv_t(M5_AT_FLAGS, 0));
// The entry point to the program
auxv.push_back(auxv_t(M5_AT_ENTRY, objFile->entryPoint()));
// Different user and group IDs
auxv.push_back(auxv_t(M5_AT_UID, uid()));
auxv.push_back(auxv_t(M5_AT_EUID, euid()));
auxv.push_back(auxv_t(M5_AT_GID, gid()));
auxv.push_back(auxv_t(M5_AT_EGID, egid()));
// Whether to enable "secure mode" in the executable
auxv.push_back(auxv_t(M5_AT_SECURE, 0));
}
// Figure out how big the initial stack needs to be
Fixes to SPARC syscall emulation mode. arch/sparc/isa/base.isa: Added a set of abbreviations for the different condition tests. arch/sparc/isa/decoder.isa: Fixes and additions to get syscall emulation closer to working. arch/sparc/isa/formats/branch.isa: Fixed branches so that the immediate version actually uses the immediate value arch/sparc/isa/formats/integerop.isa: Compute the condition codes -before- writing to the state of the machine. arch/sparc/isa/formats/mem.isa: An attempt to fix up the output of the disassembly of loads and stores. arch/sparc/isa/formats/trap.isa: Added code to disassemble a trap instruction. This probably needs to be fixed up so there are immediate and register versions. arch/sparc/isa/operands.isa: Added an R1 operand, and fixed up the numbering arch/sparc/isa_traits.hh: SyscallNumReg is no longer needed, the max number of sources and destinations are fixed up, and the syscall return uses xcc instead of icc. arch/sparc/linux/process.cc: arch/sparc/linux/process.hh: Added a getresuidFunc syscall implementation. This isn't actually used, but I thought it was and will leave it in. arch/sparc/process.cc: arch/sparc/process.hh: Fixed up how the initial stack frame is set up. arch/sparc/regfile.hh: Changed the number of windows from 6 to 32 so we don't have to worry about spill and fill traps for now, and commented out the register file setting itself up. cpu/cpu_exec_context.hh: cpu/exec_context.hh: cpu/simple/cpu.hh: sim/process.cc: sim/process.hh: Changed the syscall mechanism to pass down the syscall number directly. --HG-- extra : convert_revision : 15723b949a0ddb3d24e68c079343b4dba2439f43
2006-04-18 15:27:22 +02:00
// The unaccounted for 8 byte 0 at the top of the stack
int sentry_size = 8;
// This is the name of the file which is present on the initial stack
// It's purpose is to let the user space linker examine the original file.
int file_name_size = filename.size() + 1;
int env_data_size = 0;
for (int i = 0; i < envp.size(); ++i) {
env_data_size += envp[i].size() + 1;
}
int arg_data_size = 0;
for (int i = 0; i < argv.size(); ++i) {
arg_data_size += argv[i].size() + 1;
}
// The info_block.
int base_info_block_size =
sentry_size + file_name_size + env_data_size + arg_data_size;
int info_block_size = roundUp(base_info_block_size, align);
int info_block_padding = info_block_size - base_info_block_size;
// Each auxilliary vector is two words
int aux_array_size = intSize * 2 * (auxv.size() + 1);
int envp_array_size = intSize * (envp.size() + 1);
int argv_array_size = intSize * (argv.size() + 1);
int argc_size = intSize;
int window_save_size = intSize * 16;
// Figure out the size of the contents of the actual initial frame
int frame_size =
aux_array_size +
envp_array_size +
argv_array_size +
argc_size +
window_save_size;
// There needs to be padding after the auxiliary vector data so that the
// very bottom of the stack is aligned properly.
int aligned_partial_size = roundUp(frame_size, align);
int aux_padding = aligned_partial_size - frame_size;
int space_needed =
info_block_size +
aux_padding +
frame_size;
stack_min = stack_base - space_needed;
stack_min = roundDown(stack_min, align);
stack_size = stack_base - stack_min;
// Allocate space for the stack
allocateMem(roundDown(stack_min, pageSize), roundUp(stack_size, pageSize));
// map out initial stack contents
IntType sentry_base = stack_base - sentry_size;
IntType file_name_base = sentry_base - file_name_size;
IntType env_data_base = file_name_base - env_data_size;
IntType arg_data_base = env_data_base - arg_data_size;
IntType auxv_array_base = arg_data_base -
info_block_padding - aux_array_size - aux_padding;
IntType envp_array_base = auxv_array_base - envp_array_size;
IntType argv_array_base = envp_array_base - argv_array_size;
IntType argc_base = argv_array_base - argc_size;
#if TRACING_ON
IntType window_save_base = argc_base - window_save_size;
#endif
DPRINTF(Stack, "The addresses of items on the initial stack:\n");
DPRINTF(Stack, "%#x - sentry NULL\n", sentry_base);
DPRINTF(Stack, "filename = %s\n", filename);
DPRINTF(Stack, "%#x - file name\n", file_name_base);
DPRINTF(Stack, "%#x - env data\n", env_data_base);
DPRINTF(Stack, "%#x - arg data\n", arg_data_base);
DPRINTF(Stack, "%#x - auxv array\n", auxv_array_base);
DPRINTF(Stack, "%#x - envp array\n", envp_array_base);
DPRINTF(Stack, "%#x - argv array\n", argv_array_base);
DPRINTF(Stack, "%#x - argc \n", argc_base);
DPRINTF(Stack, "%#x - window save\n", window_save_base);
DPRINTF(Stack, "%#x - stack min\n", stack_min);
assert(window_save_base == stack_min);
// write contents to stack
// figure out argc
IntType argc = argv.size();
IntType guestArgc = SparcISA::htog(argc);
// Write out the sentry void *
uint64_t sentry_NULL = 0;
initVirtMem->writeBlob(sentry_base,
(uint8_t*)&sentry_NULL, sentry_size);
// Write the file name
initVirtMem->writeString(file_name_base, filename.c_str());
// Copy the aux stuff
for (int x = 0; x < auxv.size(); x++) {
initVirtMem->writeBlob(auxv_array_base + x * 2 * intSize,
(uint8_t*)&(auxv[x].a_type), intSize);
initVirtMem->writeBlob(auxv_array_base + (x * 2 + 1) * intSize,
(uint8_t*)&(auxv[x].a_val), intSize);
}
// Write out the terminating zeroed auxilliary vector
const IntType zero = 0;
initVirtMem->writeBlob(auxv_array_base + intSize * 2 * auxv.size(),
(uint8_t*)&zero, intSize);
initVirtMem->writeBlob(auxv_array_base + intSize * (2 * auxv.size() + 1),
(uint8_t*)&zero, intSize);
copyStringArray(envp, envp_array_base, env_data_base, initVirtMem);
copyStringArray(argv, argv_array_base, arg_data_base, initVirtMem);
initVirtMem->writeBlob(argc_base, (uint8_t*)&guestArgc, intSize);
// Set up space for the trap handlers into the processes address space.
// Since the stack grows down and there is reserved address space abov
// it, we can put stuff above it and stay out of the way.
fillStart = stack_base;
spillStart = fillStart + sizeof(MachInst) * numFillInsts;
ThreadContext *tc = system->getThreadContext(contextIds[0]);
// Set up the thread context to start running the process
// assert(NumArgumentRegs >= 2);
// tc->setIntReg(ArgumentReg[0], argc);
// tc->setIntReg(ArgumentReg[1], argv_array_base);
tc->setIntReg(StackPointerReg, stack_min - StackBias);
// %g1 is a pointer to a function that should be run at exit. Since we
// don't have anything like that, it should be set to 0.
tc->setIntReg(1, 0);
ISA,CPU,etc: Create an ISA defined PC type that abstracts out ISA behaviors. This change is a low level and pervasive reorganization of how PCs are managed in M5. Back when Alpha was the only ISA, there were only 2 PCs to worry about, the PC and the NPC, and the lsb of the PC signaled whether or not you were in PAL mode. As other ISAs were added, we had to add an NNPC, micro PC and next micropc, x86 and ARM introduced variable length instruction sets, and ARM started to keep track of mode bits in the PC. Each CPU model handled PCs in its own custom way that needed to be updated individually to handle the new dimensions of variability, or, in the case of ARMs mode-bit-in-the-pc hack, the complexity could be hidden in the ISA at the ISA implementation's expense. Areas like the branch predictor hadn't been updated to handle branch delay slots or micropcs, and it turns out that had introduced a significant (10s of percent) performance bug in SPARC and to a lesser extend MIPS. Rather than perpetuate the problem by reworking O3 again to handle the PC features needed by x86, this change was introduced to rework PC handling in a more modular, transparent, and hopefully efficient way. PC type: Rather than having the superset of all possible elements of PC state declared in each of the CPU models, each ISA defines its own PCState type which has exactly the elements it needs. A cross product of canned PCState classes are defined in the new "generic" ISA directory for ISAs with/without delay slots and microcode. These are either typedef-ed or subclassed by each ISA. To read or write this structure through a *Context, you use the new pcState() accessor which reads or writes depending on whether it has an argument. If you just want the address of the current or next instruction or the current micro PC, you can get those through read-only accessors on either the PCState type or the *Contexts. These are instAddr(), nextInstAddr(), and microPC(). Note the move away from readPC. That name is ambiguous since it's not clear whether or not it should be the actual address to fetch from, or if it should have extra bits in it like the PAL mode bit. Each class is free to define its own functions to get at whatever values it needs however it needs to to be used in ISA specific code. Eventually Alpha's PAL mode bit could be moved out of the PC and into a separate field like ARM. These types can be reset to a particular pc (where npc = pc + sizeof(MachInst), nnpc = npc + sizeof(MachInst), upc = 0, nupc = 1 as appropriate), printed, serialized, and compared. There is a branching() function which encapsulates code in the CPU models that checked if an instruction branched or not. Exactly what that means in the context of branch delay slots which can skip an instruction when not taken is ambiguous, and ideally this function and its uses can be eliminated. PCStates also generally know how to advance themselves in various ways depending on if they point at an instruction, a microop, or the last microop of a macroop. More on that later. Ideally, accessing all the PCs at once when setting them will improve performance of M5 even though more data needs to be moved around. This is because often all the PCs need to be manipulated together, and by getting them all at once you avoid multiple function calls. Also, the PCs of a particular thread will have spatial locality in the cache. Previously they were grouped by element in arrays which spread out accesses. Advancing the PC: The PCs were previously managed entirely by the CPU which had to know about PC semantics, try to figure out which dimension to increment the PC in, what to set NPC/NNPC, etc. These decisions are best left to the ISA in conjunction with the PC type itself. Because most of the information about how to increment the PC (mainly what type of instruction it refers to) is contained in the instruction object, a new advancePC virtual function was added to the StaticInst class. Subclasses provide an implementation that moves around the right element of the PC with a minimal amount of decision making. In ISAs like Alpha, the instructions always simply assign NPC to PC without having to worry about micropcs, nnpcs, etc. The added cost of a virtual function call should be outweighed by not having to figure out as much about what to do with the PCs and mucking around with the extra elements. One drawback of making the StaticInsts advance the PC is that you have to actually have one to advance the PC. This would, superficially, seem to require decoding an instruction before fetch could advance. This is, as far as I can tell, realistic. fetch would advance through memory addresses, not PCs, perhaps predicting new memory addresses using existing ones. More sophisticated decisions about control flow would be made later on, after the instruction was decoded, and handed back to fetch. If branching needs to happen, some amount of decoding needs to happen to see that it's a branch, what the target is, etc. This could get a little more complicated if that gets done by the predecoder, but I'm choosing to ignore that for now. Variable length instructions: To handle variable length instructions in x86 and ARM, the predecoder now takes in the current PC by reference to the getExtMachInst function. It can modify the PC however it needs to (by setting NPC to be the PC + instruction length, for instance). This could be improved since the CPU doesn't know if the PC was modified and always has to write it back. ISA parser: To support the new API, all PC related operand types were removed from the parser and replaced with a PCState type. There are two warts on this implementation. First, as with all the other operand types, the PCState still has to have a valid operand type even though it doesn't use it. Second, using syntax like PCS.npc(target) doesn't work for two reasons, this looks like the syntax for operand type overriding, and the parser can't figure out if you're reading or writing. Instructions that use the PCS operand (which I've consistently called it) need to first read it into a local variable, manipulate it, and then write it back out. Return address stack: The return address stack needed a little extra help because, in the presence of branch delay slots, it has to merge together elements of the return PC and the call PC. To handle that, a buildRetPC utility function was added. There are basically only two versions in all the ISAs, but it didn't seem short enough to put into the generic ISA directory. Also, the branch predictor code in O3 and InOrder were adjusted so that they always store the PC of the actual call instruction in the RAS, not the next PC. If the call instruction is a microop, the next PC refers to the next microop in the same macroop which is probably not desirable. The buildRetPC function advances the PC intelligently to the next macroop (in an ISA specific way) so that that case works. Change in stats: There were no change in stats except in MIPS and SPARC in the O3 model. MIPS runs in about 9% fewer ticks. SPARC runs with 30%-50% fewer ticks, which could likely be improved further by setting call/return instruction flags and taking advantage of the RAS. TODO: Add != operators to the PCState classes, defined trivially to be !(a==b). Smooth out places where PCs are split apart, passed around, and put back together later. I think this might happen in SPARC's fault code. Add ISA specific constructors that allow setting PC elements without calling a bunch of accessors. Try to eliminate the need for the branching() function. Factor out Alpha's PAL mode pc bit into a separate flag field, and eliminate places where it's blindly masked out or tested in the PC.
2010-10-31 08:07:20 +01:00
tc->pcState(objFile->entryPoint());
// Align the "stack_min" to a page boundary.
stack_min = roundDown(stack_min, pageSize);
// num_processes++;
}
void
Sparc64LiveProcess::argsInit(int intSize, int pageSize)
{
SparcLiveProcess::argsInit<uint64_t>(pageSize);
// Stuff the trap handlers into the process address space
initVirtMem->writeBlob(fillStart,
(uint8_t*)fillHandler64, sizeof(MachInst) * numFillInsts);
initVirtMem->writeBlob(spillStart,
(uint8_t*)spillHandler64, sizeof(MachInst) * numSpillInsts);
}
void
Sparc32LiveProcess::argsInit(int intSize, int pageSize)
{
SparcLiveProcess::argsInit<uint32_t>(pageSize);
// Stuff the trap handlers into the process address space
initVirtMem->writeBlob(fillStart,
(uint8_t*)fillHandler32, sizeof(MachInst) * numFillInsts);
initVirtMem->writeBlob(spillStart,
(uint8_t*)spillHandler32, sizeof(MachInst) * numSpillInsts);
}
void Sparc32LiveProcess::flushWindows(ThreadContext *tc)
{
IntReg Cansave = tc->readIntReg(NumIntArchRegs + 3);
IntReg Canrestore = tc->readIntReg(NumIntArchRegs + 4);
IntReg Otherwin = tc->readIntReg(NumIntArchRegs + 6);
MiscReg CWP = tc->readMiscReg(MISCREG_CWP);
MiscReg origCWP = CWP;
CWP = (CWP + Cansave + 2) % NWindows;
while (NWindows - 2 - Cansave != 0) {
if (Otherwin) {
panic("Otherwin non-zero.\n");
} else {
tc->setMiscReg(MISCREG_CWP, CWP);
// Do the stores
IntReg sp = tc->readIntReg(StackPointerReg);
for (int index = 16; index < 32; index++) {
uint32_t regVal = tc->readIntReg(index);
regVal = htog(regVal);
if (!tc->getMemProxy()->tryWriteBlob(
sp + (index - 16) * 4, (uint8_t *)&regVal, 4)) {
warn("Failed to save register to the stack when "
"flushing windows.\n");
}
}
Canrestore--;
Cansave++;
CWP = (CWP + 1) % NWindows;
}
}
tc->setIntReg(NumIntArchRegs + 3, Cansave);
tc->setIntReg(NumIntArchRegs + 4, Canrestore);
tc->setMiscReg(MISCREG_CWP, origCWP);
}
void
Sparc64LiveProcess::flushWindows(ThreadContext *tc)
{
IntReg Cansave = tc->readIntReg(NumIntArchRegs + 3);
IntReg Canrestore = tc->readIntReg(NumIntArchRegs + 4);
IntReg Otherwin = tc->readIntReg(NumIntArchRegs + 6);
MiscReg CWP = tc->readMiscReg(MISCREG_CWP);
MiscReg origCWP = CWP;
CWP = (CWP + Cansave + 2) % NWindows;
while (NWindows - 2 - Cansave != 0) {
if (Otherwin) {
panic("Otherwin non-zero.\n");
} else {
tc->setMiscReg(MISCREG_CWP, CWP);
// Do the stores
IntReg sp = tc->readIntReg(StackPointerReg);
for (int index = 16; index < 32; index++) {
IntReg regVal = tc->readIntReg(index);
regVal = htog(regVal);
if (!tc->getMemProxy()->tryWriteBlob(
sp + 2047 + (index - 16) * 8, (uint8_t *)&regVal, 8)) {
warn("Failed to save register to the stack when "
"flushing windows.\n");
}
}
Canrestore--;
Cansave++;
CWP = (CWP + 1) % NWindows;
}
}
tc->setIntReg(NumIntArchRegs + 3, Cansave);
tc->setIntReg(NumIntArchRegs + 4, Canrestore);
tc->setMiscReg(MISCREG_CWP, origCWP);
}
IntReg
Sparc32LiveProcess::getSyscallArg(ThreadContext *tc, int &i)
{
assert(i < 6);
return bits(tc->readIntReg(FirstArgumentReg + i++), 31, 0);
}
void
Sparc32LiveProcess::setSyscallArg(ThreadContext *tc, int i, IntReg val)
{
assert(i < 6);
tc->setIntReg(FirstArgumentReg + i, bits(val, 31, 0));
}
IntReg
Sparc64LiveProcess::getSyscallArg(ThreadContext *tc, int &i)
{
assert(i < 6);
return tc->readIntReg(FirstArgumentReg + i++);
}
void
Sparc64LiveProcess::setSyscallArg(ThreadContext *tc, int i, IntReg val)
{
assert(i < 6);
tc->setIntReg(FirstArgumentReg + i, val);
}
void
SparcLiveProcess::setSyscallReturn(ThreadContext *tc,
SyscallReturn return_value)
{
// check for error condition. SPARC syscall convention is to
// indicate success/failure in reg the carry bit of the ccr
// and put the return value itself in the standard return value reg ().
if (return_value.successful()) {
// no error, clear XCC.C
tc->setIntReg(NumIntArchRegs + 2,
tc->readIntReg(NumIntArchRegs + 2) & 0xEE);
// tc->setMiscRegNoEffect(MISCREG_CCR, tc->readMiscRegNoEffect(MISCREG_CCR) & 0xEE);
IntReg val = return_value.value();
if (bits(tc->readMiscRegNoEffect(
SparcISA::MISCREG_PSTATE), 3, 3)) {
val = bits(val, 31, 0);
}
tc->setIntReg(ReturnValueReg, val);
} else {
// got an error, set XCC.C
tc->setIntReg(NumIntArchRegs + 2,
tc->readIntReg(NumIntArchRegs + 2) | 0x11);
// tc->setMiscRegNoEffect(MISCREG_CCR, tc->readMiscRegNoEffect(MISCREG_CCR) | 0x11);
IntReg val = -return_value.value();
if (bits(tc->readMiscRegNoEffect(
SparcISA::MISCREG_PSTATE), 3, 3)) {
val = bits(val, 31, 0);
}
tc->setIntReg(ReturnValueReg, val);
}
}