gem5/src/arch/alpha/isa/fp.isa

306 lines
11 KiB
Plaintext
Raw Normal View History

// -*- mode:c++ -*-
// Copyright (c) 2003-2005 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: Steve Reinhardt
////////////////////////////////////////////////////////////////////
//
// Floating-point instructions
//
2010-12-20 22:24:40 +01:00
// Note that many FP-type instructions which do not support all the
// various rounding & trapping modes use the simpler format
// BasicOperateWithNopCheck.
//
output exec {{
/// Check "FP enabled" machine status bit. Called when executing any FP
/// instruction in full-system mode.
Changed the fault enum into a class, and fixed everything up to work with it. Next, the faults need to be pulled out of all the other code so that they are only used to communicate between the CPU and the ISA. SConscript: The new faults.cc file in sim allocates the system wide faults. When these faults are generated through a function interface in the ISA, this file may go away. arch/alpha/alpha_memory.cc: Changed Fault to Fault * and took the underscores out of fault names. arch/alpha/alpha_memory.hh: Changed Fault to Fault *. Also, added an include for the alpha faults. arch/alpha/ev5.cc: Changed the fault_addr array into a fault_addr function. Once all of the faults can be expected to have the same type, fault_addr can go away completely and the info it provided will come from the fault itself. Also, Fault was changed to Fault *, and underscores were taken out of fault names. arch/alpha/isa/decoder.isa: Changed Fault to Fault * and took the underscores out fault names. arch/alpha/isa/fp.isa: Changed Fault to Fault *, and took the underscores out of fault names. arch/alpha/isa/main.isa: Changed Fault to Fault *, removed underscores from fault names, and made an include of the alpha faults show up in all the generated files. arch/alpha/isa/mem.isa: Changed Fault to Fault * and removed underscores from fault names. arch/alpha/isa/unimp.isa: arch/alpha/isa/unknown.isa: cpu/exec_context.hh: cpu/ozone/cpu.hh: cpu/simple/cpu.cc: dev/alpha_console.cc: dev/ide_ctrl.cc: dev/isa_fake.cc: dev/pciconfigall.cc: dev/pcidev.cc: dev/pcidev.hh: dev/tsunami_cchip.cc: dev/tsunami_io.cc: dev/tsunami_pchip.cc: Changed Fault to Fault *, and removed underscores from fault names. arch/alpha/isa_traits.hh: Changed the include of arch/alpha/faults.hh to sim/faults.hh, since the alpha faults weren't needed. cpu/base_dyn_inst.cc: Changed Fault to Fault *, and removed underscores from fault names. This file probably shouldn't use the Unimplemented Opcode fault. cpu/base_dyn_inst.hh: Changed Fault to Fault * and took the underscores out of the fault names. cpu/exec_context.cc: cpu/o3/alpha_dyn_inst.hh: cpu/o3/alpha_dyn_inst_impl.hh: cpu/o3/fetch.hh: dev/alpha_console.hh: dev/baddev.hh: dev/ide_ctrl.hh: dev/isa_fake.hh: dev/ns_gige.hh: dev/pciconfigall.hh: dev/sinic.hh: dev/tsunami_cchip.hh: dev/tsunami_io.hh: dev/tsunami_pchip.hh: dev/uart.hh: dev/uart8250.hh: Changed Fault to Fault *. cpu/o3/alpha_cpu.hh: Changed Fault to Fault *, removed underscores from fault names. cpu/o3/alpha_cpu_impl.hh: Changed Fault to Fault *, removed underscores from fault names, and changed the fault_addr array to the fault_addr function. Once all faults are from the ISA, this function will probably go away. cpu/o3/commit_impl.hh: cpu/o3/fetch_impl.hh: dev/baddev.cc: Changed Fault to Fault *, and removed underscores from the fault names. cpu/o3/regfile.hh: Added an include for the alpha specific faults which will hopefully go away once the ipr stuff is moved, changed Fault to Fault *, and removed the underscores from fault names. cpu/simple/cpu.hh: Changed Fault to Fault * dev/ns_gige.cc: Changed Fault to Fault *, and removdd underscores from fault names. dev/sinic.cc: Changed Fault to Fault *, and removed the underscores from fault names. dev/uart8250.cc: Chanted Fault to Fault *, and removed underscores from fault names. kern/kernel_stats.cc: Removed underscores from fault names, and from NumFaults. kern/kernel_stats.hh: Changed the predeclaration of Fault from an enum to a class, and changd the "fault" function to work with the classes instead of the enum. Once there are no system wide faults anymore, this code will simplify back to something like it was originally. sim/faults.cc: This allocates the system wide faults. sim/faults.hh: This declares the system wide faults. sim/syscall_emul.cc: sim/syscall_emul.hh: Removed the underscores from fault names. --HG-- rename : arch/alpha/faults.cc => sim/faults.cc rename : arch/alpha/faults.hh => sim/faults.hh extra : convert_revision : 253d39258237333ae8ec4d8047367cb3ea68569d
2006-02-16 07:22:51 +01:00
/// @retval Full-system mode: NoFault if FP is enabled, FenFault
/// if not. Non-full-system mode: always returns NoFault.
arch: teach ISA parser how to split code across files This patch encompasses several interrelated and interdependent changes to the ISA generation step. The end goal is to reduce the size of the generated compilation units for instruction execution and decoding so that batch compilation can proceed with all CPUs active without exhausting physical memory. The ISA parser (src/arch/isa_parser.py) has been improved so that it can accept 'split [output_type];' directives at the top level of the grammar and 'split(output_type)' python calls within 'exec {{ ... }}' blocks. This has the effect of "splitting" the files into smaller compilation units. I use air-quotes around "splitting" because the files themselves are not split, but preprocessing directives are inserted to have the same effect. Architecturally, the ISA parser has had some changes in how it works. In general, it emits code sooner. It doesn't generate per-CPU files, and instead defers to the C preprocessor to create the duplicate copies for each CPU type. Likewise there are more files emitted and the C preprocessor does more substitution that used to be done by the ISA parser. Finally, the build system (SCons) needs to be able to cope with a dynamic list of source files coming out of the ISA parser. The changes to the SCons{cript,truct} files support this. In broad strokes, the targets requested on the command line are hidden from SCons until all the build dependencies are determined, otherwise it would try, realize it can't reach the goal, and terminate in failure. Since build steps (i.e. running the ISA parser) must be taken to determine the file list, several new build stages have been inserted at the very start of the build. First, the build dependencies from the ISA parser will be emitted to arch/$ISA/generated/inc.d, which is then read by a new SCons builder to finalize the dependencies. (Once inc.d exists, the ISA parser will not need to be run to complete this step.) Once the dependencies are known, the 'Environments' are made by the makeEnv() function. This function used to be called before the build began but now happens during the build. It is easy to see that this step is quite slow; this is a known issue and it's important to realize that it was already slow, but there was no obvious cause to attribute it to since nothing was displayed to the terminal. Since new steps that used to be performed serially are now in a potentially-parallel build phase, the pathname handling in the SCons scripts has been tightened up to deal with chdir() race conditions. In general, pathnames are computed earlier and more likely to be stored, passed around, and processed as absolute paths rather than relative paths. In the end, some of these issues had to be fixed by inserting serializing dependencies in the build. Minor note: For the null ISA, we just provide a dummy inc.d so SCons is never compelled to try to generate it. While it seems slightly wrong to have anything in src/arch/*/generated (i.e. a non-generated 'generated' file), it's by far the simplest solution.
2014-05-10 00:58:47 +02:00
inline Fault checkFpEnableFault(CPU_EXEC_CONTEXT *xc)
{
2010-12-20 22:24:40 +01:00
Fault fault = NoFault; // dummy... this ipr access should not fault
if (FullSystem && !ICSR_FPE(xc->readMiscReg(IPR_ICSR))) {
fault = std::make_shared<FloatEnableFault>();
}
return fault;
}
}};
output header {{
/**
* Base class for general floating-point instructions. Includes
* support for various Alpha rounding and trapping modes. Only FP
* instructions that require this support are derived from this
* class; the rest derive directly from AlphaStaticInst.
*/
class AlphaFP : public AlphaStaticInst
{
public:
/// Alpha FP rounding modes.
enum RoundingMode {
2010-12-20 22:24:40 +01:00
Chopped = 0, ///< round toward zero
Minus_Infinity = 1, ///< round toward minus infinity
2010-12-20 22:24:40 +01:00
Normal = 2, ///< round to nearest (default)
Dynamic = 3, ///< use FPCR setting (in instruction)
Plus_Infinity = 3 ///< round to plus inifinity (in FPCR)
};
/// Alpha FP trapping modes.
/// For instructions that produce integer results, the
/// "Underflow Enable" modes really mean "Overflow Enable", and
/// the assembly modifier is V rather than U.
enum TrappingMode {
/// default: nothing enabled
2010-12-20 22:24:40 +01:00
Imprecise = 0, ///< no modifier
/// underflow/overflow traps enabled, inexact disabled
2010-12-20 22:24:40 +01:00
Underflow_Imprecise = 1, ///< /U or /V
Underflow_Precise = 5, ///< /SU or /SV
/// underflow/overflow and inexact traps enabled
Underflow_Inexact_Precise = 7 ///< /SUI or /SVI
};
protected:
/// Map Alpha rounding mode to C99 constants from <fenv.h>.
static const int alphaToC99RoundingMode[];
/// Map enum RoundingMode values to disassembly suffixes.
static const char *roundingModeSuffix[];
/// Map enum TrappingMode values to FP disassembly suffixes.
static const char *fpTrappingModeSuffix[];
/// Map enum TrappingMode values to integer disassembly suffixes.
static const char *intTrappingModeSuffix[];
/// This instruction's rounding mode.
RoundingMode roundingMode;
/// This instruction's trapping mode.
TrappingMode trappingMode;
/// Have we warned about this instruction's unsupported
/// rounding mode (if applicable)?
mutable bool warnedOnRounding;
/// Have we warned about this instruction's unsupported
/// trapping mode (if applicable)?
mutable bool warnedOnTrapping;
/// Constructor
Changes to support automatic renaming of the shadow registers at decode time. This requires using an ExtMachInst (uint64_t) instead of the normal MachInst; the ExtMachInst is packed with extra decode context information. In the case of Alpha, the PAL mode is included. The shadow registers are folded into the normal integer registers to ease renaming indexing. Include the removed Opcdec class of instructions for faulting when a pal mode only instruction is decoded in non-pal mode. arch/alpha/ev5.cc: Changes to automatically map the shadow registers if the instruction is in PAL mode. arch/alpha/isa/branch.isa: arch/alpha/isa/decoder.isa: arch/alpha/isa/fp.isa: arch/alpha/isa/int.isa: arch/alpha/isa/mem.isa: arch/alpha/isa/pal.isa: arch/alpha/isa/unimp.isa: Changes for automatically using the shadow registers. Now instructions must decode based on an ExtMachInst, which is a MachInst with any decode context information concatenated onto the higher order bits. arch/alpha/isa/main.isa: Changes for automatically using the shadow registers. Now instructions must decode based on an ExtMachInst, which is a MachInst with any decode context information concatenated onto the higher order bits. The decoder (for Alpha) uses the 32nd bit in order to determine if the machine is in PAL mode. If it is, then it refers to the reg_redir table to determine the true index of the register it is using. Also include the opcdec instruction definition. arch/alpha/isa_traits.hh: Define ExtMachInst type that is used by the static inst in order to decode the instruction, given the context of being in pal mode or not. Redefine the number of Int registers, splitting it into NumIntArchRegs (32) and NumIntRegs (32 + 8 shadow registers). Change the dependence tags to reflect the integer registers include the 8 shadow registers. Define function to make an ExtMachInst. Currently it is somewhat specific to Alpha; in the future it must be decided to make this more generic and possibly slower, or leave it specific to each architecture and ifdef it within the CPU. arch/isa_parser.py: Have static insts decode on the ExtMachInst. base/remote_gdb.cc: Support the automatic remapping of shadow registers. Remote GDB must now look at the PC being read in order to tell if it should use the normal register indices or the shadow register indices. cpu/o3/regfile.hh: Comment out the pal registers; they are now a part of the integer registers. cpu/simple/cpu.cc: Create an ExtMachInst to decode on, based on the normal MachInst and the PC of the instructoin. cpu/static_inst.hh: Change from MachInst to ExtMachInst to support shadow register renaming. --HG-- extra : convert_revision : 1d23eabf735e297068e1917445a6348e9f8c88d5
2006-03-03 21:28:25 +01:00
AlphaFP(const char *mnem, ExtMachInst _machInst, OpClass __opClass)
: AlphaStaticInst(mnem, _machInst, __opClass),
roundingMode((enum RoundingMode)FP_ROUNDMODE),
trappingMode((enum TrappingMode)FP_TRAPMODE),
warnedOnRounding(false),
warnedOnTrapping(false)
{
}
int getC99RoundingMode(uint64_t fpcr_val) const;
// This differs from the AlphaStaticInst version only in
// printing suffixes for non-default rounding & trapping modes.
std::string
generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};
}};
output decoder {{
int
AlphaFP::getC99RoundingMode(uint64_t fpcr_val) const
{
if (roundingMode == Dynamic) {
return alphaToC99RoundingMode[bits(fpcr_val, 59, 58)];
}
else {
return alphaToC99RoundingMode[roundingMode];
}
}
std::string
AlphaFP::generateDisassembly(Addr pc, const SymbolTable *symtab) const
{
std::string mnem_str(mnemonic);
#ifndef SS_COMPATIBLE_DISASSEMBLY
std::string suffix("");
suffix += ((_destRegIdx[0] >= FP_Reg_Base)
? fpTrappingModeSuffix[trappingMode]
: intTrappingModeSuffix[trappingMode]);
suffix += roundingModeSuffix[roundingMode];
if (suffix != "") {
mnem_str = csprintf("%s/%s", mnemonic, suffix);
}
#endif
std::stringstream ss;
ccprintf(ss, "%-10s ", mnem_str.c_str());
// just print the first two source regs... if there's
// a third one, it's a read-modify-write dest (Rc),
// e.g. for CMOVxx
if (_numSrcRegs > 0) {
printReg(ss, _srcRegIdx[0]);
}
if (_numSrcRegs > 1) {
ss << ",";
printReg(ss, _srcRegIdx[1]);
}
// just print the first dest... if there's a second one,
// it's generally implicit
if (_numDestRegs > 0) {
if (_numSrcRegs > 0)
ss << ",";
printReg(ss, _destRegIdx[0]);
}
return ss.str();
}
const int AlphaFP::alphaToC99RoundingMode[] = {
2010-12-20 22:24:40 +01:00
M5_FE_TOWARDZERO, // Chopped
M5_FE_DOWNWARD, // Minus_Infinity
M5_FE_TONEAREST, // Normal
M5_FE_UPWARD // Dynamic in inst, Plus_Infinity in FPCR
};
const char *AlphaFP::roundingModeSuffix[] = { "c", "m", "", "d" };
// mark invalid trapping modes, but don't fail on them, because
// you could decode anything on a misspeculated path
const char *AlphaFP::fpTrappingModeSuffix[] =
{ "", "u", "INVTM2", "INVTM3", "INVTM4", "su", "INVTM6", "sui" };
const char *AlphaFP::intTrappingModeSuffix[] =
{ "", "v", "INVTM2", "INVTM3", "INVTM4", "sv", "INVTM6", "svi" };
}};
// FP instruction class execute method template. Handles non-standard
// rounding modes.
def template FloatingPointExecute {{
arch: teach ISA parser how to split code across files This patch encompasses several interrelated and interdependent changes to the ISA generation step. The end goal is to reduce the size of the generated compilation units for instruction execution and decoding so that batch compilation can proceed with all CPUs active without exhausting physical memory. The ISA parser (src/arch/isa_parser.py) has been improved so that it can accept 'split [output_type];' directives at the top level of the grammar and 'split(output_type)' python calls within 'exec {{ ... }}' blocks. This has the effect of "splitting" the files into smaller compilation units. I use air-quotes around "splitting" because the files themselves are not split, but preprocessing directives are inserted to have the same effect. Architecturally, the ISA parser has had some changes in how it works. In general, it emits code sooner. It doesn't generate per-CPU files, and instead defers to the C preprocessor to create the duplicate copies for each CPU type. Likewise there are more files emitted and the C preprocessor does more substitution that used to be done by the ISA parser. Finally, the build system (SCons) needs to be able to cope with a dynamic list of source files coming out of the ISA parser. The changes to the SCons{cript,truct} files support this. In broad strokes, the targets requested on the command line are hidden from SCons until all the build dependencies are determined, otherwise it would try, realize it can't reach the goal, and terminate in failure. Since build steps (i.e. running the ISA parser) must be taken to determine the file list, several new build stages have been inserted at the very start of the build. First, the build dependencies from the ISA parser will be emitted to arch/$ISA/generated/inc.d, which is then read by a new SCons builder to finalize the dependencies. (Once inc.d exists, the ISA parser will not need to be run to complete this step.) Once the dependencies are known, the 'Environments' are made by the makeEnv() function. This function used to be called before the build began but now happens during the build. It is easy to see that this step is quite slow; this is a known issue and it's important to realize that it was already slow, but there was no obvious cause to attribute it to since nothing was displayed to the terminal. Since new steps that used to be performed serially are now in a potentially-parallel build phase, the pathname handling in the SCons scripts has been tightened up to deal with chdir() race conditions. In general, pathnames are computed earlier and more likely to be stored, passed around, and processed as absolute paths rather than relative paths. In the end, some of these issues had to be fixed by inserting serializing dependencies in the build. Minor note: For the null ISA, we just provide a dummy inc.d so SCons is never compelled to try to generate it. While it seems slightly wrong to have anything in src/arch/*/generated (i.e. a non-generated 'generated' file), it's by far the simplest solution.
2014-05-10 00:58:47 +02:00
Fault %(class_name)s::execute(CPU_EXEC_CONTEXT *xc,
Trace::InstRecord *traceData) const
{
if (trappingMode != Imprecise && !warnedOnTrapping) {
warn("%s: non-standard trapping mode not supported",
generateDisassembly(0, NULL));
warnedOnTrapping = true;
}
Fault fault = NoFault;
%(fp_enable_check)s;
%(op_decl)s;
%(op_rd)s;
#if USE_FENV
if (roundingMode == Normal) {
%(code)s;
} else {
m5_fesetround(getC99RoundingMode(
xc->readMiscReg(MISCREG_FPCR)));
%(code)s;
m5_fesetround(M5_FE_TONEAREST);
}
#else
if (roundingMode != Normal && !warnedOnRounding) {
warn("%s: non-standard rounding mode not supported",
generateDisassembly(0, NULL));
warnedOnRounding = true;
}
%(code)s;
#endif
Changed the fault enum into a class, and fixed everything up to work with it. Next, the faults need to be pulled out of all the other code so that they are only used to communicate between the CPU and the ISA. SConscript: The new faults.cc file in sim allocates the system wide faults. When these faults are generated through a function interface in the ISA, this file may go away. arch/alpha/alpha_memory.cc: Changed Fault to Fault * and took the underscores out of fault names. arch/alpha/alpha_memory.hh: Changed Fault to Fault *. Also, added an include for the alpha faults. arch/alpha/ev5.cc: Changed the fault_addr array into a fault_addr function. Once all of the faults can be expected to have the same type, fault_addr can go away completely and the info it provided will come from the fault itself. Also, Fault was changed to Fault *, and underscores were taken out of fault names. arch/alpha/isa/decoder.isa: Changed Fault to Fault * and took the underscores out fault names. arch/alpha/isa/fp.isa: Changed Fault to Fault *, and took the underscores out of fault names. arch/alpha/isa/main.isa: Changed Fault to Fault *, removed underscores from fault names, and made an include of the alpha faults show up in all the generated files. arch/alpha/isa/mem.isa: Changed Fault to Fault * and removed underscores from fault names. arch/alpha/isa/unimp.isa: arch/alpha/isa/unknown.isa: cpu/exec_context.hh: cpu/ozone/cpu.hh: cpu/simple/cpu.cc: dev/alpha_console.cc: dev/ide_ctrl.cc: dev/isa_fake.cc: dev/pciconfigall.cc: dev/pcidev.cc: dev/pcidev.hh: dev/tsunami_cchip.cc: dev/tsunami_io.cc: dev/tsunami_pchip.cc: Changed Fault to Fault *, and removed underscores from fault names. arch/alpha/isa_traits.hh: Changed the include of arch/alpha/faults.hh to sim/faults.hh, since the alpha faults weren't needed. cpu/base_dyn_inst.cc: Changed Fault to Fault *, and removed underscores from fault names. This file probably shouldn't use the Unimplemented Opcode fault. cpu/base_dyn_inst.hh: Changed Fault to Fault * and took the underscores out of the fault names. cpu/exec_context.cc: cpu/o3/alpha_dyn_inst.hh: cpu/o3/alpha_dyn_inst_impl.hh: cpu/o3/fetch.hh: dev/alpha_console.hh: dev/baddev.hh: dev/ide_ctrl.hh: dev/isa_fake.hh: dev/ns_gige.hh: dev/pciconfigall.hh: dev/sinic.hh: dev/tsunami_cchip.hh: dev/tsunami_io.hh: dev/tsunami_pchip.hh: dev/uart.hh: dev/uart8250.hh: Changed Fault to Fault *. cpu/o3/alpha_cpu.hh: Changed Fault to Fault *, removed underscores from fault names. cpu/o3/alpha_cpu_impl.hh: Changed Fault to Fault *, removed underscores from fault names, and changed the fault_addr array to the fault_addr function. Once all faults are from the ISA, this function will probably go away. cpu/o3/commit_impl.hh: cpu/o3/fetch_impl.hh: dev/baddev.cc: Changed Fault to Fault *, and removed underscores from the fault names. cpu/o3/regfile.hh: Added an include for the alpha specific faults which will hopefully go away once the ipr stuff is moved, changed Fault to Fault *, and removed the underscores from fault names. cpu/simple/cpu.hh: Changed Fault to Fault * dev/ns_gige.cc: Changed Fault to Fault *, and removdd underscores from fault names. dev/sinic.cc: Changed Fault to Fault *, and removed the underscores from fault names. dev/uart8250.cc: Chanted Fault to Fault *, and removed underscores from fault names. kern/kernel_stats.cc: Removed underscores from fault names, and from NumFaults. kern/kernel_stats.hh: Changed the predeclaration of Fault from an enum to a class, and changd the "fault" function to work with the classes instead of the enum. Once there are no system wide faults anymore, this code will simplify back to something like it was originally. sim/faults.cc: This allocates the system wide faults. sim/faults.hh: This declares the system wide faults. sim/syscall_emul.cc: sim/syscall_emul.hh: Removed the underscores from fault names. --HG-- rename : arch/alpha/faults.cc => sim/faults.cc rename : arch/alpha/faults.hh => sim/faults.hh extra : convert_revision : 253d39258237333ae8ec4d8047367cb3ea68569d
2006-02-16 07:22:51 +01:00
if (fault == NoFault) {
%(op_wb)s;
}
return fault;
}
}};
// FP instruction class execute method template where no dynamic
// rounding mode control is needed. Like BasicExecute, but includes
// check & warning for non-standard trapping mode.
def template FPFixedRoundingExecute {{
arch: teach ISA parser how to split code across files This patch encompasses several interrelated and interdependent changes to the ISA generation step. The end goal is to reduce the size of the generated compilation units for instruction execution and decoding so that batch compilation can proceed with all CPUs active without exhausting physical memory. The ISA parser (src/arch/isa_parser.py) has been improved so that it can accept 'split [output_type];' directives at the top level of the grammar and 'split(output_type)' python calls within 'exec {{ ... }}' blocks. This has the effect of "splitting" the files into smaller compilation units. I use air-quotes around "splitting" because the files themselves are not split, but preprocessing directives are inserted to have the same effect. Architecturally, the ISA parser has had some changes in how it works. In general, it emits code sooner. It doesn't generate per-CPU files, and instead defers to the C preprocessor to create the duplicate copies for each CPU type. Likewise there are more files emitted and the C preprocessor does more substitution that used to be done by the ISA parser. Finally, the build system (SCons) needs to be able to cope with a dynamic list of source files coming out of the ISA parser. The changes to the SCons{cript,truct} files support this. In broad strokes, the targets requested on the command line are hidden from SCons until all the build dependencies are determined, otherwise it would try, realize it can't reach the goal, and terminate in failure. Since build steps (i.e. running the ISA parser) must be taken to determine the file list, several new build stages have been inserted at the very start of the build. First, the build dependencies from the ISA parser will be emitted to arch/$ISA/generated/inc.d, which is then read by a new SCons builder to finalize the dependencies. (Once inc.d exists, the ISA parser will not need to be run to complete this step.) Once the dependencies are known, the 'Environments' are made by the makeEnv() function. This function used to be called before the build began but now happens during the build. It is easy to see that this step is quite slow; this is a known issue and it's important to realize that it was already slow, but there was no obvious cause to attribute it to since nothing was displayed to the terminal. Since new steps that used to be performed serially are now in a potentially-parallel build phase, the pathname handling in the SCons scripts has been tightened up to deal with chdir() race conditions. In general, pathnames are computed earlier and more likely to be stored, passed around, and processed as absolute paths rather than relative paths. In the end, some of these issues had to be fixed by inserting serializing dependencies in the build. Minor note: For the null ISA, we just provide a dummy inc.d so SCons is never compelled to try to generate it. While it seems slightly wrong to have anything in src/arch/*/generated (i.e. a non-generated 'generated' file), it's by far the simplest solution.
2014-05-10 00:58:47 +02:00
Fault %(class_name)s::execute(CPU_EXEC_CONTEXT *xc,
Trace::InstRecord *traceData) const
{
if (trappingMode != Imprecise && !warnedOnTrapping) {
warn("%s: non-standard trapping mode not supported",
generateDisassembly(0, NULL));
warnedOnTrapping = true;
}
Fault fault = NoFault;
%(fp_enable_check)s;
%(op_decl)s;
%(op_rd)s;
%(code)s;
Changed the fault enum into a class, and fixed everything up to work with it. Next, the faults need to be pulled out of all the other code so that they are only used to communicate between the CPU and the ISA. SConscript: The new faults.cc file in sim allocates the system wide faults. When these faults are generated through a function interface in the ISA, this file may go away. arch/alpha/alpha_memory.cc: Changed Fault to Fault * and took the underscores out of fault names. arch/alpha/alpha_memory.hh: Changed Fault to Fault *. Also, added an include for the alpha faults. arch/alpha/ev5.cc: Changed the fault_addr array into a fault_addr function. Once all of the faults can be expected to have the same type, fault_addr can go away completely and the info it provided will come from the fault itself. Also, Fault was changed to Fault *, and underscores were taken out of fault names. arch/alpha/isa/decoder.isa: Changed Fault to Fault * and took the underscores out fault names. arch/alpha/isa/fp.isa: Changed Fault to Fault *, and took the underscores out of fault names. arch/alpha/isa/main.isa: Changed Fault to Fault *, removed underscores from fault names, and made an include of the alpha faults show up in all the generated files. arch/alpha/isa/mem.isa: Changed Fault to Fault * and removed underscores from fault names. arch/alpha/isa/unimp.isa: arch/alpha/isa/unknown.isa: cpu/exec_context.hh: cpu/ozone/cpu.hh: cpu/simple/cpu.cc: dev/alpha_console.cc: dev/ide_ctrl.cc: dev/isa_fake.cc: dev/pciconfigall.cc: dev/pcidev.cc: dev/pcidev.hh: dev/tsunami_cchip.cc: dev/tsunami_io.cc: dev/tsunami_pchip.cc: Changed Fault to Fault *, and removed underscores from fault names. arch/alpha/isa_traits.hh: Changed the include of arch/alpha/faults.hh to sim/faults.hh, since the alpha faults weren't needed. cpu/base_dyn_inst.cc: Changed Fault to Fault *, and removed underscores from fault names. This file probably shouldn't use the Unimplemented Opcode fault. cpu/base_dyn_inst.hh: Changed Fault to Fault * and took the underscores out of the fault names. cpu/exec_context.cc: cpu/o3/alpha_dyn_inst.hh: cpu/o3/alpha_dyn_inst_impl.hh: cpu/o3/fetch.hh: dev/alpha_console.hh: dev/baddev.hh: dev/ide_ctrl.hh: dev/isa_fake.hh: dev/ns_gige.hh: dev/pciconfigall.hh: dev/sinic.hh: dev/tsunami_cchip.hh: dev/tsunami_io.hh: dev/tsunami_pchip.hh: dev/uart.hh: dev/uart8250.hh: Changed Fault to Fault *. cpu/o3/alpha_cpu.hh: Changed Fault to Fault *, removed underscores from fault names. cpu/o3/alpha_cpu_impl.hh: Changed Fault to Fault *, removed underscores from fault names, and changed the fault_addr array to the fault_addr function. Once all faults are from the ISA, this function will probably go away. cpu/o3/commit_impl.hh: cpu/o3/fetch_impl.hh: dev/baddev.cc: Changed Fault to Fault *, and removed underscores from the fault names. cpu/o3/regfile.hh: Added an include for the alpha specific faults which will hopefully go away once the ipr stuff is moved, changed Fault to Fault *, and removed the underscores from fault names. cpu/simple/cpu.hh: Changed Fault to Fault * dev/ns_gige.cc: Changed Fault to Fault *, and removdd underscores from fault names. dev/sinic.cc: Changed Fault to Fault *, and removed the underscores from fault names. dev/uart8250.cc: Chanted Fault to Fault *, and removed underscores from fault names. kern/kernel_stats.cc: Removed underscores from fault names, and from NumFaults. kern/kernel_stats.hh: Changed the predeclaration of Fault from an enum to a class, and changd the "fault" function to work with the classes instead of the enum. Once there are no system wide faults anymore, this code will simplify back to something like it was originally. sim/faults.cc: This allocates the system wide faults. sim/faults.hh: This declares the system wide faults. sim/syscall_emul.cc: sim/syscall_emul.hh: Removed the underscores from fault names. --HG-- rename : arch/alpha/faults.cc => sim/faults.cc rename : arch/alpha/faults.hh => sim/faults.hh extra : convert_revision : 253d39258237333ae8ec4d8047367cb3ea68569d
2006-02-16 07:22:51 +01:00
if (fault == NoFault) {
%(op_wb)s;
}
return fault;
}
}};
def template FloatingPointDecode {{
{
AlphaStaticInst *i = new %(class_name)s(machInst);
if (FC == 31) {
i = makeNop(i);
}
return i;
}
}};
// General format for floating-point operate instructions:
// - Checks trapping and rounding mode flags. Trapping modes
// currently unimplemented (will fail).
// - Generates NOP if FC == 31.
def format FloatingPointOperate(code, *opt_args) {{
iop = InstObjParams(name, Name, 'AlphaFP', code, opt_args)
decode_block = FloatingPointDecode.subst(iop)
header_output = BasicDeclare.subst(iop)
decoder_output = BasicConstructor.subst(iop)
exec_output = FloatingPointExecute.subst(iop)
}};
// Special format for cvttq where rounding mode is pre-decoded
def format FPFixedRounding(code, class_suffix, *opt_args) {{
Name += class_suffix
iop = InstObjParams(name, Name, 'AlphaFP', code, opt_args)
decode_block = FloatingPointDecode.subst(iop)
header_output = BasicDeclare.subst(iop)
decoder_output = BasicConstructor.subst(iop)
exec_output = FPFixedRoundingExecute.subst(iop)
}};