blob: 7dc668cbbfb7b1520e80f1dbc3450d2e7cb4449f [file] [log] [blame]
// -*- mode:c++ -*-
// Copyright (c) 2010-2013,2016,2018-2019 ARM Limited
// All rights reserved
//
// The license below extends only to copyright in the software and shall
// not be construed as granting a license to any other intellectual
// property including but not limited to intellectual property relating
// to a hardware implementation of the functionality of the software
// licensed hereunder. You may use the software subject to the license
// terms below provided that you ensure that this notice is replicated
// unmodified and in its entirety in all distributions of the software,
// modified or unmodified, in source code or in binary form.
//
// 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.
output header {{
template <class Micro>
class VfpMacroRegRegOp : public VfpMacroOp
{
public:
VfpMacroRegRegOp(ExtMachInst _machInst, IntRegIndex _dest,
IntRegIndex _op1, bool _wide) :
VfpMacroOp("VfpMacroRegRegOp", _machInst, No_OpClass, _wide)
{
numMicroops = machInst.fpscrLen + 1;
assert(numMicroops > 1);
microOps = new StaticInstPtr[numMicroops];
for (unsigned i = 0; i < numMicroops; i++) {
VfpMicroMode mode = VfpMicroop;
if (i == 0)
mode = VfpFirstMicroop;
else if (i == numMicroops - 1)
mode = VfpLastMicroop;
microOps[i] = new Micro(_machInst, _dest, _op1, mode);
nextIdxs(_dest, _op1);
}
}
};
template <class VfpOp>
StaticInstPtr
decodeVfpRegRegOp(ExtMachInst machInst,
IntRegIndex dest, IntRegIndex op1, bool wide)
{
if (machInst.fpscrLen == 0 || VfpMacroOp::inScalarBank(dest)) {
return new VfpOp(machInst, dest, op1);
} else {
return new VfpMacroRegRegOp<VfpOp>(machInst, dest, op1, wide);
}
}
template <class Micro>
class VfpMacroRegImmOp : public VfpMacroOp
{
public:
VfpMacroRegImmOp(ExtMachInst _machInst, IntRegIndex _dest, uint64_t _imm,
bool _wide) :
VfpMacroOp("VfpMacroRegImmOp", _machInst, No_OpClass, _wide)
{
numMicroops = machInst.fpscrLen + 1;
microOps = new StaticInstPtr[numMicroops];
for (unsigned i = 0; i < numMicroops; i++) {
VfpMicroMode mode = VfpMicroop;
if (i == 0)
mode = VfpFirstMicroop;
else if (i == numMicroops - 1)
mode = VfpLastMicroop;
microOps[i] = new Micro(_machInst, _dest, _imm, mode);
nextIdxs(_dest);
}
}
};
template <class VfpOp>
StaticInstPtr
decodeVfpRegImmOp(ExtMachInst machInst,
IntRegIndex dest, uint64_t imm, bool wide)
{
if (machInst.fpscrLen == 0 || VfpMacroOp::inScalarBank(dest)) {
return new VfpOp(machInst, dest, imm);
} else {
return new VfpMacroRegImmOp<VfpOp>(machInst, dest, imm, wide);
}
}
template <class Micro>
class VfpMacroRegRegImmOp : public VfpMacroOp
{
public:
VfpMacroRegRegImmOp(ExtMachInst _machInst, IntRegIndex _dest,
IntRegIndex _op1, uint64_t _imm, bool _wide) :
VfpMacroOp("VfpMacroRegRegImmOp", _machInst, No_OpClass, _wide)
{
numMicroops = machInst.fpscrLen + 1;
microOps = new StaticInstPtr[numMicroops];
for (unsigned i = 0; i < numMicroops; i++) {
VfpMicroMode mode = VfpMicroop;
if (i == 0)
mode = VfpFirstMicroop;
else if (i == numMicroops - 1)
mode = VfpLastMicroop;
microOps[i] = new Micro(_machInst, _dest, _op1, _imm, mode);
nextIdxs(_dest, _op1);
}
}
};
template <class VfpOp>
StaticInstPtr
decodeVfpRegRegImmOp(ExtMachInst machInst, IntRegIndex dest,
IntRegIndex op1, uint64_t imm, bool wide)
{
if (machInst.fpscrLen == 0 || VfpMacroOp::inScalarBank(dest)) {
return new VfpOp(machInst, dest, op1, imm);
} else {
return new VfpMacroRegRegImmOp<VfpOp>(machInst, dest, op1, imm, wide);
}
}
template <class Micro>
class VfpMacroRegRegRegOp : public VfpMacroOp
{
public:
VfpMacroRegRegRegOp(ExtMachInst _machInst, IntRegIndex _dest,
IntRegIndex _op1, IntRegIndex _op2, bool _wide) :
VfpMacroOp("VfpMacroRegRegRegOp", _machInst, No_OpClass, _wide)
{
numMicroops = machInst.fpscrLen + 1;
microOps = new StaticInstPtr[numMicroops];
for (unsigned i = 0; i < numMicroops; i++) {
VfpMicroMode mode = VfpMicroop;
if (i == 0)
mode = VfpFirstMicroop;
else if (i == numMicroops - 1)
mode = VfpLastMicroop;
microOps[i] = new Micro(_machInst, _dest, _op1, _op2, mode);
nextIdxs(_dest, _op1, _op2);
}
}
};
template <class VfpOp>
StaticInstPtr
decodeVfpRegRegRegOp(ExtMachInst machInst, IntRegIndex dest,
IntRegIndex op1, IntRegIndex op2, bool wide)
{
if (machInst.fpscrLen == 0 || VfpMacroOp::inScalarBank(dest)) {
return new VfpOp(machInst, dest, op1, op2);
} else {
return new VfpMacroRegRegRegOp<VfpOp>(machInst, dest, op1, op2, wide);
}
}
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
vmsrCode = vmsrEnabledCheckCode + '''
MiscDest = Op1;
'''
vmsrIop = ArmInstObjParams("vmsr", "Vmsr", "FpRegRegImmOp",
{ "code": vmsrCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" },
["IsSerializeAfter","IsNonSpeculative"])
header_output += FpRegRegImmOpDeclare.subst(vmsrIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmsrIop);
exec_output += PredOpExecute.subst(vmsrIop);
vmsrFpscrCode = vmsrEnabledCheckCode + '''
Fpscr = Op1 & ~FpCondCodesMask;
FpCondCodes = Op1 & FpCondCodesMask;
'''
vmsrFpscrIop = ArmInstObjParams("vmsr", "VmsrFpscr", "FpRegRegOp",
{ "code": vmsrFpscrCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" },
["IsSerializeAfter","IsNonSpeculative",
"IsSquashAfter"])
header_output += FpRegRegOpDeclare.subst(vmsrFpscrIop);
decoder_output += FpRegRegOpConstructor.subst(vmsrFpscrIop);
exec_output += PredOpExecute.subst(vmsrFpscrIop);
vmrsCode = vmrsEnabledCheckCode + '''
CPSR cpsr = Cpsr;
if (!isSecure(xc->tcBase()) && (cpsr.mode != MODE_HYP)) {
HCR hcr = Hcr;
bool hypTrap = false;
switch(xc->tcBase()->flattenRegId(RegId(MiscRegClass, op1)).index()) {
case MISCREG_FPSID:
hypTrap = hcr.tid0;
break;
case MISCREG_MVFR0:
case MISCREG_MVFR1:
hypTrap = hcr.tid3;
break;
}
if (hypTrap) {
return std::make_shared<HypervisorTrap>(machInst, imm,
EC_TRAPPED_CP10_MRC_VMRS);
}
}
Dest = MiscOp1;
'''
vmrsIop = ArmInstObjParams("vmrs", "Vmrs", "FpRegRegImmOp",
{ "code": vmrsCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" },
["IsSerializeBefore"])
header_output += FpRegRegImmOpDeclare.subst(vmrsIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmrsIop);
exec_output += PredOpExecute.subst(vmrsIop);
vmrsFpscrIop = ArmInstObjParams("vmrs", "VmrsFpscr", "FpRegRegOp",
{ "code": vmrsEnabledCheckCode + \
"Dest = Fpscr | FpCondCodes;",
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" },
["IsSerializeBefore"])
header_output += FpRegRegOpDeclare.subst(vmrsFpscrIop);
decoder_output += FpRegRegOpConstructor.subst(vmrsFpscrIop);
exec_output += PredOpExecute.subst(vmrsFpscrIop);
vmrsApsrFpscrCode = vfpEnabledCheckCode + '''
FPSCR fpscr = FpCondCodes;
CondCodesNZ = (fpscr.n << 1) | fpscr.z;
CondCodesC = fpscr.c;
CondCodesV = fpscr.v;
'''
vmrsApsrFpscrIop = ArmInstObjParams("vmrs", "VmrsApsrFpscr", "PredOp",
{ "code": vmrsApsrFpscrCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" })
header_output += BasicDeclare.subst(vmrsApsrFpscrIop);
decoder_output += BasicConstructor.subst(vmrsApsrFpscrIop);
exec_output += PredOpExecute.subst(vmrsApsrFpscrIop);
vmovImmSCode = vfpEnabledCheckCode + '''
FpDest_uw = bits(imm, 31, 0);
'''
vmovImmSIop = ArmInstObjParams("vmov", "VmovImmS", "FpRegImmOp",
{ "code": vmovImmSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegImmOpDeclare.subst(vmovImmSIop);
decoder_output += FpRegImmOpConstructor.subst(vmovImmSIop);
exec_output += PredOpExecute.subst(vmovImmSIop);
vmovImmDCode = vfpEnabledCheckCode + '''
FpDestP0_uw = bits(imm, 31, 0);
FpDestP1_uw = bits(imm, 63, 32);
'''
vmovImmDIop = ArmInstObjParams("vmov", "VmovImmD", "FpRegImmOp",
{ "code": vmovImmDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegImmOpDeclare.subst(vmovImmDIop);
decoder_output += FpRegImmOpConstructor.subst(vmovImmDIop);
exec_output += PredOpExecute.subst(vmovImmDIop);
vmovImmQCode = vfpEnabledCheckCode + '''
FpDestP0_uw = bits(imm, 31, 0);
FpDestP1_uw = bits(imm, 63, 32);
FpDestP2_uw = bits(imm, 31, 0);
FpDestP3_uw = bits(imm, 63, 32);
'''
vmovImmQIop = ArmInstObjParams("vmov", "VmovImmQ", "FpRegImmOp",
{ "code": vmovImmQCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegImmOpDeclare.subst(vmovImmQIop);
decoder_output += FpRegImmOpConstructor.subst(vmovImmQIop);
exec_output += PredOpExecute.subst(vmovImmQIop);
vmovRegSCode = vfpEnabledCheckCode + '''
FpDest_uw = FpOp1_uw;
'''
vmovRegSIop = ArmInstObjParams("vmov", "VmovRegS", "FpRegRegOp",
{ "code": vmovRegSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegOpDeclare.subst(vmovRegSIop);
decoder_output += FpRegRegOpConstructor.subst(vmovRegSIop);
exec_output += PredOpExecute.subst(vmovRegSIop);
vmovRegDCode = vfpEnabledCheckCode + '''
FpDestP0_uw = FpOp1P0_uw;
FpDestP1_uw = FpOp1P1_uw;
'''
vmovRegDIop = ArmInstObjParams("vmov", "VmovRegD", "FpRegRegOp",
{ "code": vmovRegDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegOpDeclare.subst(vmovRegDIop);
decoder_output += FpRegRegOpConstructor.subst(vmovRegDIop);
exec_output += PredOpExecute.subst(vmovRegDIop);
vmovRegQCode = vfpEnabledCheckCode + '''
FpDestP0_uw = FpOp1P0_uw;
FpDestP1_uw = FpOp1P1_uw;
FpDestP2_uw = FpOp1P2_uw;
FpDestP3_uw = FpOp1P3_uw;
'''
vmovRegQIop = ArmInstObjParams("vmov", "VmovRegQ", "FpRegRegOp",
{ "code": vmovRegQCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegOpDeclare.subst(vmovRegQIop);
decoder_output += FpRegRegOpConstructor.subst(vmovRegQIop);
exec_output += PredOpExecute.subst(vmovRegQIop);
vmovCoreRegBCode = simdEnabledCheckCode + '''
FpDest_uw = insertBits(FpDest_uw, imm * 8 + 7, imm * 8, Op1_ub);
'''
vmovCoreRegBIop = ArmInstObjParams("vmov", "VmovCoreRegB", "FpRegRegImmOp",
{ "code": vmovCoreRegBCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovCoreRegBIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovCoreRegBIop);
exec_output += PredOpExecute.subst(vmovCoreRegBIop);
vmovCoreRegHCode = simdEnabledCheckCode + '''
FpDest_uw = insertBits(FpDest_uw, imm * 16 + 15, imm * 16, Op1_uh);
'''
vmovCoreRegHIop = ArmInstObjParams("vmov", "VmovCoreRegH", "FpRegRegImmOp",
{ "code": vmovCoreRegHCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovCoreRegHIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovCoreRegHIop);
exec_output += PredOpExecute.subst(vmovCoreRegHIop);
vmovCoreRegWCode = vfpEnabledCheckCode + '''
FpDest_uw = Op1_uw;
'''
vmovCoreRegWIop = ArmInstObjParams("vmov", "VmovCoreRegW", "FpRegRegOp",
{ "code": vmovCoreRegWCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegOpDeclare.subst(vmovCoreRegWIop);
decoder_output += FpRegRegOpConstructor.subst(vmovCoreRegWIop);
exec_output += PredOpExecute.subst(vmovCoreRegWIop);
vmovRegCoreUBCode = vfpEnabledCheckCode + '''
assert(imm < 4);
Dest = bits(FpOp1_uw, imm * 8 + 7, imm * 8);
'''
vmovRegCoreUBIop = ArmInstObjParams(
"vmov", "VmovRegCoreUB", "FpRegRegImmOp",
{ "code": vmovRegCoreUBCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovRegCoreUBIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovRegCoreUBIop);
exec_output += PredOpExecute.subst(vmovRegCoreUBIop);
vmovRegCoreUHCode = vfpEnabledCheckCode + '''
assert(imm < 2);
Dest = bits(FpOp1_uw, imm * 16 + 15, imm * 16);
'''
vmovRegCoreUHIop = ArmInstObjParams(
"vmov", "VmovRegCoreUH", "FpRegRegImmOp",
{ "code": vmovRegCoreUHCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovRegCoreUHIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovRegCoreUHIop);
exec_output += PredOpExecute.subst(vmovRegCoreUHIop);
vmovRegCoreSBCode = vfpEnabledCheckCode + '''
assert(imm < 4);
Dest = sext<8>(bits(FpOp1_uw, imm * 8 + 7, imm * 8));
'''
vmovRegCoreSBIop = ArmInstObjParams(
"vmov", "VmovRegCoreSB", "FpRegRegImmOp",
{ "code": vmovRegCoreSBCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovRegCoreSBIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovRegCoreSBIop);
exec_output += PredOpExecute.subst(vmovRegCoreSBIop);
vmovRegCoreSHCode = vfpEnabledCheckCode + '''
assert(imm < 2);
Dest = sext<16>(bits(FpOp1_uw, imm * 16 + 15, imm * 16));
'''
vmovRegCoreSHIop = ArmInstObjParams(
"vmov", "VmovRegCoreSH", "FpRegRegImmOp",
{ "code": vmovRegCoreSHCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vmovRegCoreSHIop);
decoder_output += FpRegRegImmOpConstructor.subst(vmovRegCoreSHIop);
exec_output += PredOpExecute.subst(vmovRegCoreSHIop);
vmovRegCoreWCode = vfpEnabledCheckCode + '''
Dest = FpOp1_uw;
'''
vmovRegCoreWIop = ArmInstObjParams("vmov", "VmovRegCoreW", "FpRegRegOp",
{ "code": vmovRegCoreWCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegOpDeclare.subst(vmovRegCoreWIop);
decoder_output += FpRegRegOpConstructor.subst(vmovRegCoreWIop);
exec_output += PredOpExecute.subst(vmovRegCoreWIop);
vmov2Reg2CoreCode = vfpEnabledCheckCode + '''
FpDestP0_uw = Op1_uw;
FpDestP1_uw = Op2_uw;
'''
vmov2Reg2CoreIop = ArmInstObjParams(
"vmov", "Vmov2Reg2Core", "FpRegRegRegOp",
{ "code": vmov2Reg2CoreCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmov2Reg2CoreIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmov2Reg2CoreIop);
exec_output += PredOpExecute.subst(vmov2Reg2CoreIop);
vmov2Core2RegCode = vfpEnabledCheckCode + '''
Dest_uw = FpOp2P0_uw;
Op1_uw = FpOp2P1_uw;
'''
vmov2Core2RegIop = ArmInstObjParams(
"vmov", "Vmov2Core2Reg", "FpRegRegRegOp",
{ "code": vmov2Core2RegCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMiscOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmov2Core2RegIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmov2Core2RegIop);
exec_output += PredOpExecute.subst(vmov2Core2RegIop);
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
singleSimpleCode = vfpEnabledCheckCode + '''
GEM5_VAR_USED FPSCR fpscr = (FPSCR) FpscrExc;
FpDest = %(op)s;
'''
singleCode = singleSimpleCode + '''
FpscrExc = fpscr;
'''
singleTernOp = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
float cOp1 = FpOp1;
float cOp2 = FpOp2;
float cOp3 = FpDestP0;
FpDestP0 = ternaryOp(fpscr, %(palam)s, %(op)s,
fpscr.fz, fpscr.dn, fpscr.rMode);
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
singleBinOp = "binaryOp(fpscr, FpOp1, FpOp2," + \
"%(func)s, fpscr.fz, fpscr.dn, fpscr.rMode)"
singleUnaryOp = "unaryOp(fpscr, FpOp1, %(func)s, fpscr.fz, fpscr.rMode)"
doubleCode = vfpEnabledCheckCode + '''
GEM5_VAR_USED FPSCR fpscr = (FPSCR) FpscrExc;
double dest = %(op)s;
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
doubleTernOp = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
double cOp2 = dbl(FpOp2P0_uw, FpOp2P1_uw);
double cOp3 = dbl(FpDestP0_uw, FpDestP1_uw);
double cDest = ternaryOp(fpscr, %(palam)s, %(op)s,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
doubleBinOp = '''
binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
%(func)s, fpscr.fz, fpscr.dn, fpscr.rMode);
'''
doubleUnaryOp = '''
unaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw), %(func)s,
fpscr.fz, fpscr.rMode)
'''
def buildTernaryFpOp(Name, base, opClass, singleOp, doubleOp, paramStr):
global header_output, decoder_output, exec_output
code = singleTernOp % { "op": singleOp, "palam": paramStr }
sIop = ArmInstObjParams(Name.lower() + "s", Name + "S", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
code = doubleTernOp % { "op": doubleOp, "palam": paramStr }
dIop = ArmInstObjParams(Name.lower() + "d", Name + "D", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
declareTempl = eval(base + "Declare");
constructorTempl = eval(base + "Constructor");
for iop in sIop, dIop:
header_output += declareTempl.subst(iop)
decoder_output += constructorTempl.subst(iop)
exec_output += PredOpExecute.subst(iop)
buildTernaryFpOp("Vfma", "FpRegRegRegOp", "SimdFloatMultAccOp",
"fpMulAdd<float>", "fpMulAdd<double>", " cOp1, cOp2, cOp3" )
buildTernaryFpOp("Vfms", "FpRegRegRegOp", "SimdFloatMultAccOp",
"fpMulAdd<float>", "fpMulAdd<double>", "-cOp1, cOp2, cOp3" )
buildTernaryFpOp("Vfnma", "FpRegRegRegOp", "SimdFloatMultAccOp",
"fpMulAdd<float>", "fpMulAdd<double>", "-cOp1, cOp2, -cOp3" )
buildTernaryFpOp("Vfnms", "FpRegRegRegOp", "SimdFloatMultAccOp",
"fpMulAdd<float>", "fpMulAdd<double>", " cOp1, cOp2, -cOp3" )
def buildBinFpOp(name, Name, base, opClass, singleOp, doubleOp):
global header_output, decoder_output, exec_output
code = singleCode % { "op": singleBinOp }
code = code % { "func": singleOp }
sIop = ArmInstObjParams(name + "s", Name + "S", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
code = doubleCode % { "op": doubleBinOp }
code = code % { "func": doubleOp }
dIop = ArmInstObjParams(name + "d", Name + "D", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
declareTempl = eval(base + "Declare");
constructorTempl = eval(base + "Constructor");
for iop in sIop, dIop:
header_output += declareTempl.subst(iop)
decoder_output += constructorTempl.subst(iop)
exec_output += PredOpExecute.subst(iop)
buildBinFpOp("vadd", "Vadd", "FpRegRegRegOp", "SimdFloatAddOp", "fpAddS",
"fpAddD")
buildBinFpOp("vsub", "Vsub", "FpRegRegRegOp", "SimdFloatAddOp", "fpSubS",
"fpSubD")
buildBinFpOp("vdiv", "Vdiv", "FpRegRegRegOp", "SimdFloatDivOp", "fpDivS",
"fpDivD")
buildBinFpOp("vmul", "Vmul", "FpRegRegRegOp", "SimdFloatMultOp", "fpMulS",
"fpMulD")
def buildBinOp(name, base, opClass, op):
'''
Create backported aarch64 instructions that use fplib.
Because they are backported, these instructions are unconditional.
'''
global header_output, decoder_output, exec_output
inst_datas = [
(
"s",
'''
FpDest_uw = fplib%(op)s<>(FpOp1_uw, FpOp2_uw, fpscr);
'''
),
(
"d",
'''
uint64_t op1 = ((uint64_t)FpOp1P0_uw |
((uint64_t)FpOp1P1_uw << 32));
uint64_t op2 = ((uint64_t)FpOp2P0_uw |
((uint64_t)FpOp2P1_uw << 32));
uint64_t dest = fplib%(op)s<>(op1, op2, fpscr);
FpDestP0_uw = dest;
FpDestP1_uw = dest >> 32;
'''
)
]
Name = name[0].upper() + name[1:]
declareTempl = eval(base + "Declare");
constructorTempl = eval(base + "Constructor");
for size_suffix, code in inst_datas:
code = (
'''
FPSCR fpscr = (FPSCR)FpscrExc;
''' +
code +
'''
FpscrExc = fpscr;
'''
)
iop = ArmInstObjParams(
name + size_suffix,
Name + size_suffix.upper(),
base,
{
"code": code % {"op": op},
"op_class": opClass
},
[]
)
header_output += declareTempl.subst(iop)
decoder_output += constructorTempl.subst(iop)
exec_output += BasicExecute.subst(iop)
ops = [
("vminnm", "FpRegRegRegOp", "SimdFloatCmpOp", "MinNum"),
("vmaxnm", "FpRegRegRegOp", "SimdFloatCmpOp", "MaxNum"),
]
for op in ops:
buildBinOp(*op)
def buildUnaryFpOp(name, Name, base, opClass, singleOp, doubleOp = None):
if doubleOp is None:
doubleOp = singleOp
global header_output, decoder_output, exec_output
code = singleCode % { "op": singleUnaryOp }
code = code % { "func": singleOp }
sIop = ArmInstObjParams(name + "s", Name + "S", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
code = doubleCode % { "op": doubleUnaryOp }
code = code % { "func": doubleOp }
dIop = ArmInstObjParams(name + "d", Name + "D", base,
{ "code": code,
"predicate_test": predicateTest,
"op_class": opClass }, [])
declareTempl = eval(base + "Declare");
constructorTempl = eval(base + "Constructor");
for iop in sIop, dIop:
header_output += declareTempl.subst(iop)
decoder_output += constructorTempl.subst(iop)
exec_output += PredOpExecute.subst(iop)
buildUnaryFpOp("vsqrt", "Vsqrt", "FpRegRegOp", "SimdFloatSqrtOp", "sqrtf",
"sqrt")
def buildSimpleUnaryFpOp(name, Name, base, opClass, singleOp,
doubleOp = None):
if doubleOp is None:
doubleOp = singleOp
global header_output, decoder_output, exec_output
sIop = ArmInstObjParams(name + "s", Name + "S", base,
{ "code": singleSimpleCode % { "op": singleOp },
"predicate_test": predicateTest,
"op_class": opClass }, [])
dIop = ArmInstObjParams(name + "d", Name + "D", base,
{ "code": doubleCode % { "op": doubleOp },
"predicate_test": predicateTest,
"op_class": opClass }, [])
declareTempl = eval(base + "Declare");
constructorTempl = eval(base + "Constructor");
for iop in sIop, dIop:
header_output += declareTempl.subst(iop)
decoder_output += constructorTempl.subst(iop)
exec_output += PredOpExecute.subst(iop)
buildSimpleUnaryFpOp("vneg", "Vneg", "FpRegRegOp", "SimdFloatMiscOp",
"-FpOp1", "-dbl(FpOp1P0_uw, FpOp1P1_uw)")
buildSimpleUnaryFpOp("vabs", "Vabs", "FpRegRegOp", "SimdFloatMiscOp",
"fabsf(FpOp1)", "fabs(dbl(FpOp1P0_uw, FpOp1P1_uw))")
buildSimpleUnaryFpOp("vrintp", "VRIntP", "FpRegRegOp", "SimdFloatMiscOp",
"fplibRoundInt<uint32_t>(FpOp1, FPRounding_POSINF, false, fpscr)",
"fplibRoundInt<uint64_t>(dbl(FpOp1P0_uw, FpOp1P1_uw), " \
"FPRounding_POSINF, false, fpscr)"
)
buildSimpleUnaryFpOp("vrintm", "VRIntM", "FpRegRegOp", "SimdFloatMiscOp",
"fplibRoundInt<uint32_t>(FpOp1, FPRounding_NEGINF, false, fpscr)",
"fplibRoundInt<uint64_t>(dbl(FpOp1P0_uw, FpOp1P1_uw), " \
"FPRounding_NEGINF, false, fpscr)"
)
buildSimpleUnaryFpOp("vrinta", "VRIntA", "FpRegRegOp", "SimdFloatMiscOp",
"fplibRoundInt<uint32_t>(FpOp1, FPRounding_TIEAWAY, false, fpscr)",
"fplibRoundInt<uint64_t>(dbl(FpOp1P0_uw, FpOp1P1_uw), " \
"FPRounding_TIEAWAY, false, fpscr)"
)
buildSimpleUnaryFpOp("vrintn", "VRIntN", "FpRegRegOp", "SimdFloatMiscOp",
"fplibRoundInt<uint32_t>(FpOp1, FPRounding_TIEEVEN, false, fpscr)",
"fplibRoundInt<uint64_t>(dbl(FpOp1P0_uw, FpOp1P1_uw), " \
"FPRounding_TIEEVEN, false, fpscr)"
)
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
vmlaSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
float mid = binaryOp(fpscr, FpOp1, FpOp2,
fpMulS, fpscr.fz, fpscr.dn, fpscr.rMode);
FpDest = binaryOp(fpscr, FpDest, mid, fpAddS,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpscrExc = fpscr;
'''
vmlaSIop = ArmInstObjParams("vmlas", "VmlaS", "FpRegRegRegOp",
{ "code": vmlaSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmlaSIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmlaSIop);
exec_output += PredOpExecute.subst(vmlaSIop);
vmlaDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double mid = binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
fpMulD, fpscr.fz, fpscr.dn, fpscr.rMode);
double dest = binaryOp(fpscr, dbl(FpDestP0_uw, FpDestP1_uw),
mid, fpAddD, fpscr.fz,
fpscr.dn, fpscr.rMode);
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
vmlaDIop = ArmInstObjParams("vmlad", "VmlaD", "FpRegRegRegOp",
{ "code": vmlaDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmlaDIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmlaDIop);
exec_output += PredOpExecute.subst(vmlaDIop);
vmlsSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
float mid = binaryOp(fpscr, FpOp1, FpOp2,
fpMulS, fpscr.fz, fpscr.dn, fpscr.rMode);
FpDest = binaryOp(fpscr, FpDest, -mid, fpAddS,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpscrExc = fpscr;
'''
vmlsSIop = ArmInstObjParams("vmlss", "VmlsS", "FpRegRegRegOp",
{ "code": vmlsSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmlsSIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmlsSIop);
exec_output += PredOpExecute.subst(vmlsSIop);
vmlsDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double mid = binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
fpMulD, fpscr.fz, fpscr.dn, fpscr.rMode);
double dest = binaryOp(fpscr, dbl(FpDestP0_uw, FpDestP1_uw),
-mid, fpAddD, fpscr.fz,
fpscr.dn, fpscr.rMode);
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
vmlsDIop = ArmInstObjParams("vmlsd", "VmlsD", "FpRegRegRegOp",
{ "code": vmlsDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vmlsDIop);
decoder_output += FpRegRegRegOpConstructor.subst(vmlsDIop);
exec_output += PredOpExecute.subst(vmlsDIop);
vnmlaSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
float mid = binaryOp(fpscr, FpOp1, FpOp2,
fpMulS, fpscr.fz, fpscr.dn, fpscr.rMode);
FpDest = binaryOp(fpscr, -FpDest, -mid, fpAddS,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpscrExc = fpscr;
'''
vnmlaSIop = ArmInstObjParams("vnmlas", "VnmlaS", "FpRegRegRegOp",
{ "code": vnmlaSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmlaSIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmlaSIop);
exec_output += PredOpExecute.subst(vnmlaSIop);
vnmlaDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double mid = binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
fpMulD, fpscr.fz, fpscr.dn, fpscr.rMode);
double dest = binaryOp(fpscr, -dbl(FpDestP0_uw, FpDestP1_uw),
-mid, fpAddD, fpscr.fz,
fpscr.dn, fpscr.rMode);
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
vnmlaDIop = ArmInstObjParams("vnmlad", "VnmlaD", "FpRegRegRegOp",
{ "code": vnmlaDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmlaDIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmlaDIop);
exec_output += PredOpExecute.subst(vnmlaDIop);
vnmlsSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
float mid = binaryOp(fpscr, FpOp1, FpOp2,
fpMulS, fpscr.fz, fpscr.dn, fpscr.rMode);
FpDest = binaryOp(fpscr, -FpDest, mid, fpAddS,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpscrExc = fpscr;
'''
vnmlsSIop = ArmInstObjParams("vnmlss", "VnmlsS", "FpRegRegRegOp",
{ "code": vnmlsSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmlsSIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmlsSIop);
exec_output += PredOpExecute.subst(vnmlsSIop);
vnmlsDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double mid = binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
fpMulD, fpscr.fz, fpscr.dn, fpscr.rMode);
double dest = binaryOp(fpscr, -dbl(FpDestP0_uw, FpDestP1_uw),
mid, fpAddD, fpscr.fz,
fpscr.dn, fpscr.rMode);
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
vnmlsDIop = ArmInstObjParams("vnmlsd", "VnmlsD", "FpRegRegRegOp",
{ "code": vnmlsDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultAccOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmlsDIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmlsDIop);
exec_output += PredOpExecute.subst(vnmlsDIop);
vnmulSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
FpDest = -binaryOp(fpscr, FpOp1, FpOp2, fpMulS,
fpscr.fz, fpscr.dn, fpscr.rMode);
FpscrExc = fpscr;
'''
vnmulSIop = ArmInstObjParams("vnmuls", "VnmulS", "FpRegRegRegOp",
{ "code": vnmulSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmulSIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmulSIop);
exec_output += PredOpExecute.subst(vnmulSIop);
vnmulDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double dest = -binaryOp(fpscr, dbl(FpOp1P0_uw, FpOp1P1_uw),
dbl(FpOp2P0_uw, FpOp2P1_uw),
fpMulD, fpscr.fz, fpscr.dn,
fpscr.rMode);
FpDestP0_uw = dblLow(dest);
FpDestP1_uw = dblHi(dest);
FpscrExc = fpscr;
'''
vnmulDIop = ArmInstObjParams("vnmuld", "VnmulD", "FpRegRegRegOp",
{ "code": vnmulDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatMultOp" }, [])
header_output += FpRegRegRegOpDeclare.subst(vnmulDIop);
decoder_output += FpRegRegRegOpConstructor.subst(vnmulDIop);
exec_output += PredOpExecute.subst(vnmulDIop);
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
vcvtUIntFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_uw) : "m" (FpOp1_uw));
FpDest = FpOp1_uw;
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtUIntFpSIop = ArmInstObjParams("vcvt", "VcvtUIntFpS", "FpRegRegOp",
{ "code": vcvtUIntFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtUIntFpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtUIntFpSIop);
exec_output += PredOpExecute.subst(vcvtUIntFpSIop);
vcvtUIntFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1P0_uw) : "m" (FpOp1P0_uw));
double cDest = (uint64_t)FpOp1P0_uw;
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtUIntFpDIop = ArmInstObjParams("vcvt", "VcvtUIntFpD", "FpRegRegOp",
{ "code": vcvtUIntFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtUIntFpDIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtUIntFpDIop);
exec_output += PredOpExecute.subst(vcvtUIntFpDIop);
vcvtSIntFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_sw) : "m" (FpOp1_sw));
FpDest = FpOp1_sw;
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtSIntFpSIop = ArmInstObjParams("vcvt", "VcvtSIntFpS", "FpRegRegOp",
{ "code": vcvtSIntFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtSIntFpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtSIntFpSIop);
exec_output += PredOpExecute.subst(vcvtSIntFpSIop);
vcvtSIntFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1P0_sw) : "m" (FpOp1P0_sw));
double cDest = FpOp1P0_sw;
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtSIntFpDIop = ArmInstObjParams("vcvt", "VcvtSIntFpD", "FpRegRegOp",
{ "code": vcvtSIntFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtSIntFpDIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtSIntFpDIop);
exec_output += PredOpExecute.subst(vcvtSIntFpDIop);
vjcvtSFixedFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
uint64_t cOp1 = ((uint64_t)FpOp1P0_uw | ((uint64_t)FpOp1P1_uw << 32));
uint8_t nz;
FpDest_uw = fplibFPToFixedJS(cOp1, fpscr, false, nz);
finishVfp(fpscr, state, fpscr.fz);
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vjcvtSFixedFpDIop = ArmInstObjParams(
"vjcvt", "VjcvtSFixedFpD", "FpRegRegOp",
{ "code": vjcvtSFixedFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vjcvtSFixedFpDIop);
decoder_output += FpRegRegOpConstructor.subst(vjcvtSFixedFpDIop);
exec_output += PredOpExecute.subst(vjcvtSFixedFpDIop);
vcvtFpUIntSRCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
vfpFlushToZero(fpscr, FpOp1);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_uw = vfpFpToFixed<float>(FpOp1, false, 32, 0, false);
__asm__ __volatile__("" :: "m" (FpDest_uw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpUIntSRIop = ArmInstObjParams("vcvt", "VcvtFpUIntSR", "FpRegRegOp",
{ "code": vcvtFpUIntSRCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpUIntSRIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpUIntSRIop);
exec_output += PredOpExecute.subst(vcvtFpUIntSRIop);
vcvtFpUIntDRCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t result = vfpFpToFixed<double>(cOp1, false, 32, 0, false);
__asm__ __volatile__("" :: "m" (result));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = result;
FpscrExc = fpscr;
'''
vcvtFpUIntDRIop = ArmInstObjParams("vcvtr", "VcvtFpUIntDR", "FpRegRegOp",
{ "code": vcvtFpUIntDRCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpUIntDRIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpUIntDRIop);
exec_output += PredOpExecute.subst(vcvtFpUIntDRIop);
vcvtFpSIntSRCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
vfpFlushToZero(fpscr, FpOp1);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_sw = vfpFpToFixed<float>(FpOp1, true, 32, 0, false);
__asm__ __volatile__("" :: "m" (FpDest_sw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpSIntSRIop = ArmInstObjParams("vcvtr", "VcvtFpSIntSR", "FpRegRegOp",
{ "code": vcvtFpSIntSRCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpSIntSRIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpSIntSRIop);
exec_output += PredOpExecute.subst(vcvtFpSIntSRIop);
vcvtFpSIntDRCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
int64_t result = vfpFpToFixed<double>(cOp1, true, 32, 0, false);
__asm__ __volatile__("" :: "m" (result));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = result;
FpscrExc = fpscr;
'''
vcvtFpSIntDRIop = ArmInstObjParams("vcvtr", "VcvtFpSIntDR", "FpRegRegOp",
{ "code": vcvtFpSIntDRCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpSIntDRIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpSIntDRIop);
exec_output += PredOpExecute.subst(vcvtFpSIntDRIop);
round_mode_suffix_to_mode = {
'': 'VfpRoundZero',
'a': 'VfpRoundAway',
'm': 'VfpRoundDown',
'n': 'VfpRoundNearest',
'p': 'VfpRoundUpward',
}
def buildVcvt(code, className, roundModeSuffix):
global header_output, decoder_output, exec_output, \
vfpEnabledCheckCode, round_mode_suffix_to_mode
full_code = vfpEnabledCheckCode + code.format(
round_mode=round_mode_suffix_to_mode[roundModeSuffix],
)
iop = ArmInstObjParams(
"vcvt{}".format(roundModeSuffix),
className.format(roundModeSuffix),
"FpRegRegOp",
{ "code": full_code,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" },
[]
)
header_output += FpRegRegOpDeclare.subst(iop);
decoder_output += FpRegRegOpConstructor.subst(iop);
exec_output += PredOpExecute.subst(iop);
code = '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
gem5::RoundingMode old_rm = gem5::getFpRound();
gem5::setFpRound(gem5::RoundingMode::TowardZero);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_uw = vfpFpToFixed<float>(
FpOp1, false, 32, 0, true, {round_mode});
__asm__ __volatile__("" :: "m" (FpDest_uw));
gem5::setFpRound(old_rm);
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
for round_mode_suffix in round_mode_suffix_to_mode:
buildVcvt(code, "Vcvt{}FpUIntS", round_mode_suffix)
code = '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
gem5::RoundingMode old_rm = gem5::getFpRound();
gem5::setFpRound(gem5::RoundingMode::TowardZero);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t result = vfpFpToFixed<double>(
cOp1, false, 32, 0, true, {round_mode});
__asm__ __volatile__("" :: "m" (result));
gem5::setFpRound(old_rm);
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = result;
FpscrExc = fpscr;
'''
for round_mode_suffix in round_mode_suffix_to_mode:
buildVcvt(code, "Vcvt{}FpUIntD", round_mode_suffix)
code = '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
gem5::RoundingMode old_rm = gem5::getFpRound();
gem5::setFpRound(gem5::RoundingMode::TowardZero);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_sw = vfpFpToFixed<float>(
FpOp1, true, 32, 0, true, {round_mode});
__asm__ __volatile__("" :: "m" (FpDest_sw));
gem5::setFpRound(old_rm);
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
for round_mode_suffix in round_mode_suffix_to_mode:
buildVcvt(code, "Vcvt{}FpSIntS", round_mode_suffix)
code = '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
gem5::RoundingMode old_rm = gem5::getFpRound();
gem5::setFpRound(gem5::RoundingMode::TowardZero);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
int64_t result = vfpFpToFixed<double>(
cOp1, true, 32, 0, true, {round_mode});
__asm__ __volatile__("" :: "m" (result));
gem5::setFpRound(old_rm);
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = result;
FpscrExc = fpscr;
'''
for round_mode_suffix in round_mode_suffix_to_mode:
buildVcvt(code, "Vcvt{}FpSIntD", round_mode_suffix)
vcvtFpSFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
double cDest = fixFpSFpDDest(FpscrExc, FpOp1);
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtFpSFpDIop = ArmInstObjParams("vcvt", "VcvtFpSFpD", "FpRegRegOp",
{ "code": vcvtFpSFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpSFpDIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpSFpDIop);
exec_output += PredOpExecute.subst(vcvtFpSFpDIop);
vcvtFpDFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
FpDest = fixFpDFpSDest(FpscrExc, cOp1);
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpDFpSIop = ArmInstObjParams("vcvt", "VcvtFpDFpS", "FpRegRegOp",
{ "code": vcvtFpDFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpDFpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpDFpSIop);
exec_output += PredOpExecute.subst(vcvtFpDFpSIop);
vcvtFpHTFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest = vcvtFpHFpS(fpscr, fpscr.dn, fpscr.ahp,
bits(fpToBits(FpOp1), 31, 16));
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpHTFpSIop = ArmInstObjParams("vcvtt", "VcvtFpHTFpS", "FpRegRegOp",
{ "code": vcvtFpHTFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpHTFpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpHTFpSIop);
exec_output += PredOpExecute.subst(vcvtFpHTFpSIop);
vcvtFpHBFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest = vcvtFpHFpS(fpscr, fpscr.dn, fpscr.ahp,
bits(fpToBits(FpOp1), 15, 0));
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpHBFpSIop = ArmInstObjParams("vcvtb", "VcvtFpHBFpS", "FpRegRegOp",
{ "code": vcvtFpHBFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpHBFpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpHBFpSIop);
exec_output += PredOpExecute.subst(vcvtFpHBFpSIop);
vcvtFpSFpHTCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1), "=m" (FpDest_uw)
: "m" (FpOp1), "m" (FpDest_uw));
FpDest_uw = insertBits(FpDest_uw, 31, 16,,
vcvtFpSFpH(fpscr, fpscr.fz, fpscr.dn,
fpscr.rMode, fpscr.ahp, FpOp1));
__asm__ __volatile__("" :: "m" (FpDest_uw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpSFpHTIop = ArmInstObjParams("vcvtt", "VcvtFpSFpHT", "FpRegRegOp",
{ "code": vcvtFpHTFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpSFpHTIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpSFpHTIop);
exec_output += PredOpExecute.subst(vcvtFpSFpHTIop);
vcvtFpSFpHBCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1), "=m" (FpDest_uw)
: "m" (FpOp1), "m" (FpDest_uw));
FpDest_uw = insertBits(FpDest_uw, 15, 0,
vcvtFpSFpH(fpscr, fpscr.fz, fpscr.dn,
fpscr.rMode, fpscr.ahp, FpOp1));
__asm__ __volatile__("" :: "m" (FpDest_uw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpSFpHBIop = ArmInstObjParams("vcvtb", "VcvtFpSFpHB", "FpRegRegOp",
{ "code": vcvtFpSFpHBCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcvtFpSFpHBIop);
decoder_output += FpRegRegOpConstructor.subst(vcvtFpSFpHBIop);
exec_output += PredOpExecute.subst(vcvtFpSFpHBIop);
vcmpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpDest, FpOp1);
if (FpDest == FpOp1) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (FpDest < FpOp1) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (FpDest > FpOp1) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
const uint32_t qnan = 0x7fc00000;
const bool nan1 = std::isnan(FpDest);
const bool signal1 = nan1 && ((fpToBits(FpDest) & qnan) != qnan);
const bool nan2 = std::isnan(FpOp1);
const bool signal2 = nan2 && ((fpToBits(FpOp1) & qnan) != qnan);
if (signal1 || signal2)
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpSIop = ArmInstObjParams("vcmps", "VcmpS", "FpRegRegOp",
{ "code": vcmpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcmpSIop);
decoder_output += FpRegRegOpConstructor.subst(vcmpSIop);
exec_output += PredOpExecute.subst(vcmpSIop);
vcmpDCode = vfpEnabledCheckCode + '''
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
double cDest = dbl(FpDestP0_uw, FpDestP1_uw);
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, cDest, cOp1);
if (cDest == cOp1) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (cDest < cOp1) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (cDest > cOp1) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
const uint64_t qnan = 0x7ff8000000000000ULL;
const bool nan1 = std::isnan(cDest);
const bool signal1 = nan1 && ((fpToBits(cDest) & qnan) != qnan);
const bool nan2 = std::isnan(cOp1);
const bool signal2 = nan2 && ((fpToBits(cOp1) & qnan) != qnan);
if (signal1 || signal2)
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpDIop = ArmInstObjParams("vcmpd", "VcmpD", "FpRegRegOp",
{ "code": vcmpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcmpDIop);
decoder_output += FpRegRegOpConstructor.subst(vcmpDIop);
exec_output += PredOpExecute.subst(vcmpDIop);
vcmpZeroSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpDest);
// This only handles imm == 0 for now.
assert(imm == 0);
if (FpDest == imm) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (FpDest < imm) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (FpDest > imm) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
const uint32_t qnan = 0x7fc00000;
const bool nan = std::isnan(FpDest);
const bool signal = nan && ((fpToBits(FpDest) & qnan) != qnan);
if (signal)
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpZeroSIop = ArmInstObjParams("vcmpZeros", "VcmpZeroS", "FpRegImmOp",
{ "code": vcmpZeroSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegImmOpDeclare.subst(vcmpZeroSIop);
decoder_output += FpRegImmOpConstructor.subst(vcmpZeroSIop);
exec_output += PredOpExecute.subst(vcmpZeroSIop);
vcmpZeroDCode = vfpEnabledCheckCode + '''
// This only handles imm == 0 for now.
assert(imm == 0);
double cDest = dbl(FpDestP0_uw, FpDestP1_uw);
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, cDest);
if (cDest == imm) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (cDest < imm) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (cDest > imm) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
const uint64_t qnan = 0x7ff8000000000000ULL;
const bool nan = std::isnan(cDest);
const bool signal = nan && ((fpToBits(cDest) & qnan) != qnan);
if (signal)
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpZeroDIop = ArmInstObjParams("vcmpZerod", "VcmpZeroD", "FpRegImmOp",
{ "code": vcmpZeroDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegImmOpDeclare.subst(vcmpZeroDIop);
decoder_output += FpRegImmOpConstructor.subst(vcmpZeroDIop);
exec_output += PredOpExecute.subst(vcmpZeroDIop);
vcmpeSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpDest, FpOp1);
if (FpDest == FpOp1) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (FpDest < FpOp1) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (FpDest > FpOp1) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpeSIop = ArmInstObjParams("vcmpes", "VcmpeS", "FpRegRegOp",
{ "code": vcmpeSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcmpeSIop);
decoder_output += FpRegRegOpConstructor.subst(vcmpeSIop);
exec_output += PredOpExecute.subst(vcmpeSIop);
vcmpeDCode = vfpEnabledCheckCode + '''
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
double cDest = dbl(FpDestP0_uw, FpDestP1_uw);
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, cDest, cOp1);
if (cDest == cOp1) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (cDest < cOp1) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (cDest > cOp1) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpeDIop = ArmInstObjParams("vcmped", "VcmpeD", "FpRegRegOp",
{ "code": vcmpeDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegRegOpDeclare.subst(vcmpeDIop);
decoder_output += FpRegRegOpConstructor.subst(vcmpeDIop);
exec_output += PredOpExecute.subst(vcmpeDIop);
vcmpeZeroSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpDest);
if (FpDest == imm) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (FpDest < imm) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (FpDest > imm) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpeZeroSIop = ArmInstObjParams("vcmpeZeros", "VcmpeZeroS", "FpRegImmOp",
{ "code": vcmpeZeroSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegImmOpDeclare.subst(vcmpeZeroSIop);
decoder_output += FpRegImmOpConstructor.subst(vcmpeZeroSIop);
exec_output += PredOpExecute.subst(vcmpeZeroSIop);
vcmpeZeroDCode = vfpEnabledCheckCode + '''
double cDest = dbl(FpDestP0_uw, FpDestP1_uw);
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, cDest);
if (cDest == imm) {
fpscr.n = 0; fpscr.z = 1; fpscr.c = 1; fpscr.v = 0;
} else if (cDest < imm) {
fpscr.n = 1; fpscr.z = 0; fpscr.c = 0; fpscr.v = 0;
} else if (cDest > imm) {
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 0;
} else {
fpscr.ioc = 1;
fpscr.n = 0; fpscr.z = 0; fpscr.c = 1; fpscr.v = 1;
}
FpCondCodes = fpscr & FpCondCodesMask;
FpscrExc = fpscr;
'''
vcmpeZeroDIop = ArmInstObjParams("vcmpeZerod", "VcmpeZeroD", "FpRegImmOp",
{ "code": vcmpeZeroDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCmpOp" }, [])
header_output += FpRegImmOpDeclare.subst(vcmpeZeroDIop);
decoder_output += FpRegImmOpConstructor.subst(vcmpeZeroDIop);
exec_output += PredOpExecute.subst(vcmpeZeroDIop);
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
vselSCode = vfpEnabledCheckCode + '''
if (testPredicate(CondCodesNZ, CondCodesC, CondCodesV, cond)) {
FpDest = FpOp1;
} else {
FpDest = FpOp2;
} '''
vselSIop = ArmInstObjParams("vsels", "VselS", "FpRegRegRegCondOp",
{ "code" : vselSCode,
"predicate_test" : predicateTest,
"op_class" : "SimdFloatCmpOp" }, [] )
header_output += FpRegRegRegCondOpDeclare.subst(vselSIop);
decoder_output += FpRegRegRegCondOpConstructor.subst(vselSIop);
exec_output += PredOpExecute.subst(vselSIop);
vselDCode = vfpEnabledCheckCode + '''
if (testPredicate(CondCodesNZ, CondCodesC, CondCodesV, cond)) {
FpDestP0_uw = FpOp1P0_uw;
FpDestP1_uw = FpOp1P1_uw;
} else {
FpDestP0_uw = FpOp2P0_uw;
FpDestP1_uw = FpOp2P1_uw;
} '''
vselDIop = ArmInstObjParams("vseld", "VselD", "FpRegRegRegCondOp",
{ "code" : vselDCode,
"predicate_test" : predicateTest,
"op_class" : "SimdFloatCmpOp" }, [] )
header_output += FpRegRegRegCondOpDeclare.subst(vselDIop);
decoder_output += FpRegRegRegCondOpConstructor.subst(vselDIop);
exec_output += PredOpExecute.subst(vselDIop);
}};
let {{
header_output = ""
decoder_output = ""
exec_output = ""
vcvtFpSFixedSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_sw = vfpFpToFixed<float>(FpOp1, true, 32, imm);
__asm__ __volatile__("" :: "m" (FpDest_sw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpSFixedSIop = ArmInstObjParams(
"vcvt", "VcvtFpSFixedS", "FpRegRegImmOp",
{ "code": vcvtFpSFixedSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpSFixedSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpSFixedSIop);
exec_output += PredOpExecute.subst(vcvtFpSFixedSIop);
vcvtFpSFixedDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t mid = vfpFpToFixed<double>(cOp1, true, 32, imm);
__asm__ __volatile__("" :: "m" (mid));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = mid;
FpDestP1_uw = mid >> 32;
FpscrExc = fpscr;
'''
vcvtFpSFixedDIop = ArmInstObjParams(
"vcvt", "VcvtFpSFixedD", "FpRegRegImmOp",
{ "code": vcvtFpSFixedDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpSFixedDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpSFixedDIop);
exec_output += PredOpExecute.subst(vcvtFpSFixedDIop);
vcvtFpUFixedSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_uw = vfpFpToFixed<float>(FpOp1, false, 32, imm);
__asm__ __volatile__("" :: "m" (FpDest_uw));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpUFixedSIop = ArmInstObjParams(
"vcvt", "VcvtFpUFixedS", "FpRegRegImmOp",
{ "code": vcvtFpUFixedSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpUFixedSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpUFixedSIop);
exec_output += PredOpExecute.subst(vcvtFpUFixedSIop);
vcvtFpUFixedDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t mid = vfpFpToFixed<double>(cOp1, false, 32, imm);
__asm__ __volatile__("" :: "m" (mid));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = mid;
FpDestP1_uw = mid >> 32;
FpscrExc = fpscr;
'''
vcvtFpUFixedDIop = ArmInstObjParams(
"vcvt", "VcvtFpUFixedD", "FpRegRegImmOp",
{ "code": vcvtFpUFixedDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpUFixedDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpUFixedDIop);
exec_output += PredOpExecute.subst(vcvtFpUFixedDIop);
vcvtSFixedFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_sw) : "m" (FpOp1_sw));
FpDest = vfpSFixedToFpS(fpscr.fz, fpscr.dn, FpOp1_sw, 32, imm);
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtSFixedFpSIop = ArmInstObjParams(
"vcvt", "VcvtSFixedFpS", "FpRegRegImmOp",
{ "code": vcvtSFixedFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtSFixedFpSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtSFixedFpSIop);
exec_output += PredOpExecute.subst(vcvtSFixedFpSIop);
vcvtSFixedFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
uint64_t mid = ((uint64_t)FpOp1P0_uw | ((uint64_t)FpOp1P1_uw << 32));
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (mid) : "m" (mid));
double cDest = vfpSFixedToFpD(fpscr.fz, fpscr.dn, mid, 32, imm);
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtSFixedFpDIop = ArmInstObjParams(
"vcvt", "VcvtSFixedFpD", "FpRegRegImmOp",
{ "code": vcvtSFixedFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtSFixedFpDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtSFixedFpDIop);
exec_output += PredOpExecute.subst(vcvtSFixedFpDIop);
vcvtUFixedFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_uw) : "m" (FpOp1_uw));
FpDest = vfpUFixedToFpS(fpscr.fz, fpscr.dn, FpOp1_uw, 32, imm);
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtUFixedFpSIop = ArmInstObjParams(
"vcvt", "VcvtUFixedFpS", "FpRegRegImmOp",
{ "code": vcvtUFixedFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtUFixedFpSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtUFixedFpSIop);
exec_output += PredOpExecute.subst(vcvtUFixedFpSIop);
vcvtUFixedFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
uint64_t mid = ((uint64_t)FpOp1P0_uw | ((uint64_t)FpOp1P1_uw << 32));
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (mid) : "m" (mid));
double cDest = vfpUFixedToFpD(fpscr.fz, fpscr.dn, mid, 32, imm);
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtUFixedFpDIop = ArmInstObjParams(
"vcvt", "VcvtUFixedFpD", "FpRegRegImmOp",
{ "code": vcvtUFixedFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtUFixedFpDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtUFixedFpDIop);
exec_output += PredOpExecute.subst(vcvtUFixedFpDIop);
vcvtFpSHFixedSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_sh = vfpFpToFixed<float>(FpOp1, true, 16, imm);
__asm__ __volatile__("" :: "m" (FpDest_sh));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpSHFixedSIop = ArmInstObjParams("vcvt", "VcvtFpSHFixedS",
"FpRegRegImmOp",
{ "code": vcvtFpSHFixedSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpSHFixedSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpSHFixedSIop);
exec_output += PredOpExecute.subst(vcvtFpSHFixedSIop);
vcvtFpSHFixedDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t result = vfpFpToFixed<double>(cOp1, true, 16, imm);
__asm__ __volatile__("" :: "m" (result));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = result;
FpDestP1_uw = result >> 32;
FpscrExc = fpscr;
'''
vcvtFpSHFixedDIop = ArmInstObjParams("vcvt", "VcvtFpSHFixedD",
"FpRegRegImmOp",
{ "code": vcvtFpSHFixedDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpSHFixedDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpSHFixedDIop);
exec_output += PredOpExecute.subst(vcvtFpSHFixedDIop);
vcvtFpUHFixedSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
vfpFlushToZero(fpscr, FpOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1) : "m" (FpOp1));
FpDest_uh = vfpFpToFixed<float>(FpOp1, false, 16, imm);
__asm__ __volatile__("" :: "m" (FpDest_uh));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtFpUHFixedSIop = ArmInstObjParams("vcvt", "VcvtFpUHFixedS",
"FpRegRegImmOp",
{ "code": vcvtFpUHFixedSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpUHFixedSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpUHFixedSIop);
exec_output += PredOpExecute.subst(vcvtFpUHFixedSIop);
vcvtFpUHFixedDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
double cOp1 = dbl(FpOp1P0_uw, FpOp1P1_uw);
vfpFlushToZero(fpscr, cOp1);
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (cOp1) : "m" (cOp1));
uint64_t mid = vfpFpToFixed<double>(cOp1, false, 16, imm);
__asm__ __volatile__("" :: "m" (mid));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = mid;
FpDestP1_uw = mid >> 32;
FpscrExc = fpscr;
'''
vcvtFpUHFixedDIop = ArmInstObjParams("vcvt", "VcvtFpUHFixedD",
"FpRegRegImmOp",
{ "code": vcvtFpUHFixedDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtFpUHFixedDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtFpUHFixedDIop);
exec_output += PredOpExecute.subst(vcvtFpUHFixedDIop);
vcvtSHFixedFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_sh) : "m" (FpOp1_sh));
FpDest = vfpSFixedToFpS(fpscr.fz, fpscr.dn, FpOp1_sh, 16, imm);
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtSHFixedFpSIop = ArmInstObjParams("vcvt", "VcvtSHFixedFpS",
"FpRegRegImmOp",
{ "code": vcvtSHFixedFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtSHFixedFpSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtSHFixedFpSIop);
exec_output += PredOpExecute.subst(vcvtSHFixedFpSIop);
vcvtSHFixedFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
uint64_t mid = ((uint64_t)FpOp1P0_uw | ((uint64_t)FpOp1P1_uw << 32));
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (mid) : "m" (mid));
double cDest = vfpSFixedToFpD(fpscr.fz, fpscr.dn, mid, 16, imm);
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtSHFixedFpDIop = ArmInstObjParams("vcvt", "VcvtSHFixedFpD",
"FpRegRegImmOp",
{ "code": vcvtSHFixedFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtSHFixedFpDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtSHFixedFpDIop);
exec_output += PredOpExecute.subst(vcvtSHFixedFpDIop);
vcvtUHFixedFpSCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (FpOp1_uh) : "m" (FpOp1_uh));
FpDest = vfpUFixedToFpS(fpscr.fz, fpscr.dn, FpOp1_uh, 16, imm);
__asm__ __volatile__("" :: "m" (FpDest));
finishVfp(fpscr, state, fpscr.fz);
FpscrExc = fpscr;
'''
vcvtUHFixedFpSIop = ArmInstObjParams("vcvt", "VcvtUHFixedFpS",
"FpRegRegImmOp",
{ "code": vcvtUHFixedFpSCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtUHFixedFpSIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtUHFixedFpSIop);
exec_output += PredOpExecute.subst(vcvtUHFixedFpSIop);
vcvtUHFixedFpDCode = vfpEnabledCheckCode + '''
FPSCR fpscr = (FPSCR) FpscrExc;
uint64_t mid = ((uint64_t)FpOp1P0_uw | ((uint64_t)FpOp1P1_uw << 32));
VfpSavedState state = prepFpState(fpscr.rMode);
__asm__ __volatile__("" : "=m" (mid) : "m" (mid));
double cDest = vfpUFixedToFpD(fpscr.fz, fpscr.dn, mid, 16, imm);
__asm__ __volatile__("" :: "m" (cDest));
finishVfp(fpscr, state, fpscr.fz);
FpDestP0_uw = dblLow(cDest);
FpDestP1_uw = dblHi(cDest);
FpscrExc = fpscr;
'''
vcvtUHFixedFpDIop = ArmInstObjParams("vcvt", "VcvtUHFixedFpD",
"FpRegRegImmOp",
{ "code": vcvtUHFixedFpDCode,
"predicate_test": predicateTest,
"op_class": "SimdFloatCvtOp" }, [])
header_output += FpRegRegImmOpDeclare.subst(vcvtUHFixedFpDIop);
decoder_output += FpRegRegImmOpConstructor.subst(vcvtUHFixedFpDIop);
exec_output += PredOpExecute.subst(vcvtUHFixedFpDIop);
}};