Skip to content

Add ARM64 encodings for groups IF_SVE_CE,CF #98409

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 11 commits into from
Feb 29, 2024
Merged

Add ARM64 encodings for groups IF_SVE_CE,CF #98409

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
376b4d9
Add ARM64 encodings for groups IF_SVE_CE,CF
snickolls-arm Feb 12, 2024
f2cfb1e
Move left-shifted integers into explicit types
snickolls-arm Feb 14, 2024
fc39e86
Merge branch 'main' into github-IF_SVE_CE,CF
snickolls-arm Feb 21, 2024
028d5c6
Address review comments
snickolls-arm Feb 12, 2024
9f8fc26
Fix formatting errors
snickolls-arm Feb 21, 2024
e2299de
Merge branch 'main' into github-IF_SVE_CE,CF
snickolls-arm Feb 23, 2024
1e43376
Fix a typo with register ID and add a scalable option for moving to and
snickolls-arm Feb 12, 2024
8fd7a28
Add shortcut for pmov with zero index
snickolls-arm Feb 12, 2024
d1f8d6d
Fix formatting
snickolls-arm Feb 27, 2024
54b0b75
Merge branch 'main' into github-IF_SVE_CE,CF
snickolls-arm Feb 28, 2024
e88f014
Merge remote-tracking branch 'origin/main' into github-IF_SVE_CE,CF
kunalspathak Feb 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 44 additions & 0 deletions src/coreclr/jit/codegenarm64test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4725,6 +4725,50 @@ void CodeGen::genArm64EmitterUnitTestsSve()
theEmitter->emitIns_R_R_R(INS_sve_lsr, EA_SCALABLE, REG_V0, REG_P0, REG_V0, INS_OPTS_SCALABLE_S,
INS_SCALABLE_OPTS_WIDE); // LSR <Zdn>.<T>, <Pg>/M, <Zdn>.<T>, <Zm>.D

// IF_SVE_CE_2A
theEmitter->emitIns_R_R(INS_sve_pmov, EA_SCALABLE, REG_P2, REG_V12, INS_OPTS_NONE,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.B, <Zn>

// IF_SVE_CE_2B
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_P15, REG_V7, 7, INS_OPTS_SCALABLE_D,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.D, <Zn>[<imm>]
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_P7, REG_V16, 0, INS_OPTS_SCALABLE_D,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.D, <Zn>[<imm>]

// IF_SVE_CE_2C
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_P0, REG_V31, 1, INS_OPTS_SCALABLE_H,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.H, <Zn>[<imm>]
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V1, REG_P1, 0, INS_OPTS_SCALABLE_H,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.H, <Zn>[<imm>]

// IF_SVE_CE_2D
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_P3, REG_V9, 3, INS_OPTS_SCALABLE_S,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.S, <Zn>[<imm>]
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_P10, REG_V4, 0, INS_OPTS_SCALABLE_S,
INS_SCALABLE_OPTS_TO_PREDICATE); // PMOV <Pd>.S, <Zn>[<imm>]

// IF_SVE_CF_2A
theEmitter->emitIns_R_R(INS_sve_pmov, EA_SCALABLE, REG_V11, REG_P12, INS_OPTS_NONE,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>, <Pn>.B

// IF_SVE_CF_2B
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V6, REG_P8, 7, INS_OPTS_SCALABLE_D,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.D
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V9, REG_P7, 0, INS_OPTS_SCALABLE_D,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.D

// IF_SVE_CF_2C
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V8, REG_P4, 1, INS_OPTS_SCALABLE_H,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.H
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V5, REG_P9, 0, INS_OPTS_SCALABLE_H,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.H

// IF_SVE_CF_2D
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V14, REG_P2, 3, INS_OPTS_SCALABLE_S,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.S
theEmitter->emitIns_R_R_I(INS_sve_pmov, EA_SCALABLE, REG_V3, REG_P15, 0, INS_OPTS_SCALABLE_S,
INS_SCALABLE_OPTS_TO_VECTOR); // PMOV <Zd>[<imm>], <Pn>.S

// IF_SVE_CJ_2A
theEmitter->emitIns_R_R(INS_sve_rev, EA_SCALABLE, REG_P1, REG_P2,
INS_OPTS_SCALABLE_B); // REV <Pd>.<T>, <Pn>.<T>
Expand Down
249 changes: 249 additions & 0 deletions src/coreclr/jit/emitarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1121,6 +1121,52 @@ void emitter::emitInsSanityCheck(instrDesc* id)
assert(isValidUimm4From1(emitGetInsSC(id)));
break;

case IF_SVE_CE_2A: // ................ ......nnnnn.DDDD -- SVE move predicate from vector
assert(isPredicateRegister(id->idReg1())); // DDDD
assert(isVectorRegister(id->idReg2())); // nnnnn
break;

case IF_SVE_CE_2B: // .........i...ii. ......nnnnn.DDDD -- SVE move predicate from vector
assert(isPredicateRegister(id->idReg1())); // DDDD
assert(isVectorRegister(id->idReg2())); // nnnnn
assert(isValidUimm<3>(emitGetInsSC(id)));
break;

case IF_SVE_CE_2C: // ..............i. ......nnnnn.DDDD -- SVE move predicate from vector
assert(isPredicateRegister(id->idReg1())); // DDDD
assert(isVectorRegister(id->idReg2())); // nnnnn
assert(isValidUimm<1>(emitGetInsSC(id))); // i
break;

case IF_SVE_CE_2D: // .............ii. ......nnnnn.DDDD -- SVE move predicate from vector
assert(isPredicateRegister(id->idReg1())); // DDDD
assert(isVectorRegister(id->idReg2())); // nnnnn
assert(isValidUimm<3>(emitGetInsSC(id))); // ii
break;

case IF_SVE_CF_2A: // ................ .......NNNNddddd -- SVE move predicate into vector
assert(isVectorRegister(id->idReg1())); // ddddd
assert(isPredicateRegister(id->idReg2())); // NNNN
break;

case IF_SVE_CF_2B: // .........i...ii. .......NNNNddddd -- SVE move predicate into vector
assert(isVectorRegister(id->idReg1())); // ddddd
assert(isPredicateRegister(id->idReg2())); // NNNN
assert(isValidUimm<3>(emitGetInsSC(id)));
break;

case IF_SVE_CF_2C: // ..............i. .......NNNNddddd -- SVE move predicate into vector
assert(isVectorRegister(id->idReg1())); // ddddd
assert(isPredicateRegister(id->idReg2())); // NNNN
assert(isValidUimm<1>(emitGetInsSC(id))); // i
break;

case IF_SVE_CF_2D: // .............ii. .......NNNNddddd -- SVE move predicate into vector
assert(isVectorRegister(id->idReg1())); // ddddd
assert(isPredicateRegister(id->idReg2())); // NNNN
assert(isValidUimm<2>(emitGetInsSC(id))); // ii
break;

case IF_SVE_CI_3A: // ........xx..MMMM .......NNNN.DDDD -- SVE permute predicate elements
elemsize = id->idOpSize();
assert(insOptsScalableStandard(id->idInsOpt()));
Expand Down Expand Up @@ -8629,6 +8675,25 @@ void emitter::emitIns_R_R(instruction ins,
}
break;

case INS_sve_pmov:
if (sopt == INS_SCALABLE_OPTS_TO_PREDICATE)
Copy link
Contributor

@a74nh a74nh Feb 23, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you add a check so that INS_OPTS_SCALABLE_B is required for opt (and all the other places IF_SVE_CE_2A and IF_SVE_CF_2A are checked)

This ensures the code calling it doesn't call with a different sized scalable. Eg: We want theEmitter->emitIns_R_R(INS_sve_pmov, EA_SCALABLE, REG_P2, REG_V12, INS_OPTS_SCALABLE_H, INS_SCALABLE_OPTS_TO_PREDICATE); to fail

Copy link
Contributor

@a74nh a74nh Feb 23, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually, what would be even more useful is if opt is INS_OPTS_SCALABLE_H/S/D then it should just do:
return emitIns_R_R_I(INS_sve_pmov, attr, reg1, reg2, 0, opt, sopt, INS_SCALABLE_OPTS_TO_PREDICATE);

Why? Because it's a useful shortcut. Most of the time the code calling all of this will be wanting to do a pmov with immediate 0, and it's a pain to have to write:
if (size == B) then emitIns_R_R(pmov,....) else emitIns_R_R_I(pmov,....,0);

{
assert(isPredicateRegister(reg1));
assert(isVectorRegister(reg2));
fmt = IF_SVE_CE_2A;
}
else if (sopt == INS_SCALABLE_OPTS_TO_VECTOR)
{
assert(isVectorRegister(reg1));
assert(isPredicateRegister(reg2));
fmt = IF_SVE_CF_2A;
}
else
{
assert(!"invalid instruction");
}
break;

case INS_sve_movs:
{
assert(opt == INS_OPTS_SCALABLE_B);
Expand Down Expand Up @@ -9646,6 +9711,57 @@ void emitter::emitIns_R_R_I(instruction ins,
fmt = IF_SVE_AM_2A;
break;

case INS_sve_pmov:
if (sopt == INS_SCALABLE_OPTS_TO_PREDICATE)
{
assert(isPredicateRegister(reg1));
assert(isVectorRegister(reg2));
switch (opt)
{
case INS_OPTS_SCALABLE_D:
assert(isValidUimm<3>(imm));
fmt = IF_SVE_CE_2B;
break;
case INS_OPTS_SCALABLE_S:
assert(isValidUimm<2>(imm));
fmt = IF_SVE_CE_2D;
break;
case INS_OPTS_SCALABLE_H:
assert(isValidUimm<1>(imm));
fmt = IF_SVE_CE_2C;
break;
default:
unreached();
}
}
else if (sopt == INS_SCALABLE_OPTS_TO_VECTOR)
{
assert(isVectorRegister(reg1));
assert(isPredicateRegister(reg2));
switch (opt)
{
case INS_OPTS_SCALABLE_D:
assert(isValidUimm<3>(imm));
fmt = IF_SVE_CF_2B;
break;
case INS_OPTS_SCALABLE_S:
assert(isValidUimm<2>(imm));
fmt = IF_SVE_CF_2D;
break;
case INS_OPTS_SCALABLE_H:
assert(isValidUimm<1>(imm));
fmt = IF_SVE_CF_2C;
break;
default:
unreached();
}
}
else
{
unreached();
}
break;

case INS_sve_sqrshrn:
case INS_sve_sqrshrun:
case INS_sve_uqrshrn:
Expand Down Expand Up @@ -18889,6 +19005,10 @@ void emitter::emitIns_Call(EmitCallType callType,

case IF_SVE_CZ_4A_A:
case IF_SVE_CZ_4A_L:
case IF_SVE_CE_2A:
case IF_SVE_CE_2B:
case IF_SVE_CE_2C:
case IF_SVE_CE_2D:
case IF_SVE_CF_2A:
case IF_SVE_CF_2B:
case IF_SVE_CF_2C:
Expand Down Expand Up @@ -23047,6 +23167,68 @@ BYTE* emitter::emitOutput_InstrSve(BYTE* dst, instrDesc* id)
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CE_2A: // ................ ......nnnnn.DDDD -- SVE move predicate from vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_P_3_to_0(id->idReg1()); // DDDD
code |= insEncodeReg_V_9_to_5(id->idReg2()); // nnnnn
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CE_2B: // .........i...ii. ......nnnnn.DDDD -- SVE move predicate from vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_P_3_to_0(id->idReg1()); // DDDD
code |= insEncodeReg_V_9_to_5(id->idReg2()); // nnnnn
code |= insEncodeSplitUimm<22, 22, 18, 17>(emitGetInsSC(id)); // i...ii
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CE_2C: // ..............i. ......nnnnn.DDDD -- SVE move predicate from vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_P_3_to_0(id->idReg1()); // DDDD
code |= insEncodeReg_V_9_to_5(id->idReg2()); // nnnnn
code |= insEncodeUimm<17, 17>(emitGetInsSC(id)); // i
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CE_2D: // .............ii. ......nnnnn.DDDD -- SVE move predicate from vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_P_3_to_0(id->idReg1()); // DDDD
code |= insEncodeReg_V_9_to_5(id->idReg2()); // nnnnn
code |= insEncodeUimm<18, 17>(emitGetInsSC(id)); // ii
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CF_2A: // ................ .......NNNNddddd -- SVE move predicate into vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_V_4_to_0(id->idReg1()); // ddddd
code |= insEncodeReg_P_8_to_5(id->idReg2()); // NNNN
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CF_2B: // .........i...ii. .......NNNNddddd -- SVE move predicate into vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_V_4_to_0(id->idReg1()); // ddddd
code |= insEncodeReg_P_8_to_5(id->idReg2()); // NNNN
code |= insEncodeSplitUimm<22, 22, 18, 17>(emitGetInsSC(id)); // i...ii
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CF_2C: // ..............i. .......NNNNddddd -- SVE move predicate into vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_V_4_to_0(id->idReg1()); // ddddd
code |= insEncodeReg_P_8_to_5(id->idReg2()); // NNNN
code |= insEncodeUimm<17, 17>(emitGetInsSC(id)); // i
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CF_2D: // .............ii. .......NNNNddddd -- SVE move predicate into vector
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_V_4_to_0(id->idReg1()); // ddddd
code |= insEncodeReg_P_8_to_5(id->idReg2()); // NNNN
code |= insEncodeUimm<18, 17>(emitGetInsSC(id)); // ii
dst += emitOutput_Instr(dst, code);
break;

case IF_SVE_CI_3A: // ........xx..MMMM .......NNNN.DDDD -- SVE permute predicate elements
code = emitInsCodeSve(ins, fmt);
code |= insEncodeReg_P_3_to_0(id->idReg1()); // DDDD
Expand Down Expand Up @@ -24840,6 +25022,18 @@ void emitter::emitDispReg(regNumber reg, emitAttr attr, bool addComma)
emitDispComma();
}

//------------------------------------------------------------------------
// emitDispSveReg: Display a scalable vector register name
//
void emitter::emitDispSveReg(regNumber reg, bool addComma)
{
assert(isVectorRegister(reg));
printf(emitSveRegName(reg));

if (addComma)
emitDispComma();
}

//------------------------------------------------------------------------
// emitDispSveReg: Display a scalable vector register name with an arrangement suffix
//
Expand All @@ -24858,6 +25052,16 @@ void emitter::emitDispSveReg(regNumber reg, insOpts opt, bool addComma)
emitDispComma();
}

//------------------------------------------------------------------------
// emitDispSveRegIndex: Display a scalable vector register with indexed element
//
void emitter::emitDispSveRegIndex(regNumber reg, ssize_t index, bool addComma)
{
assert(isVectorRegister(reg));
printf(emitSveRegName(reg));
emitDispElementIndex(index, addComma);
}

//------------------------------------------------------------------------
// emitDispVectorReg: Display a SIMD vector register name with an arrangement suffix
//
Expand Down Expand Up @@ -26980,6 +27184,39 @@ void emitter::emitDispInsHelp(
emitDispSveReg(id->idReg3(), id->idInsOpt(), false); // mmmmm
break;

case IF_SVE_CE_2A: // ................ ......nnnnn.DDDD -- SVE move predicate from vector
emitDispPredicateReg(id->idReg1(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_B, true); // DDDD
emitDispSveReg(id->idReg2(), false); // nnnnn
break;
case IF_SVE_CE_2B: // .........i...ii. ......nnnnn.DDDD -- SVE move predicate from vector
emitDispPredicateReg(id->idReg1(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_D, true); // DDDD
emitDispSveRegIndex(id->idReg2(), emitGetInsSC(id), false); // nnnnn
break;
case IF_SVE_CE_2C: // ..............i. ......nnnnn.DDDD -- SVE move predicate from vector
emitDispPredicateReg(id->idReg1(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_H, true); // DDDD
emitDispSveRegIndex(id->idReg2(), emitGetInsSC(id), false); // nnnnn
break;
case IF_SVE_CE_2D: // .............ii. ......nnnnn.DDDD -- SVE move predicate from vector
emitDispPredicateReg(id->idReg1(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_S, true); // DDDD
emitDispSveRegIndex(id->idReg2(), emitGetInsSC(id), false); // nnnnn
break;
case IF_SVE_CF_2A: // ................ .......NNNNddddd -- SVE move predicate into vector
emitDispSveReg(id->idReg1(), true); // ddddd
emitDispPredicateReg(id->idReg2(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_B, false); // NNNN
break;
case IF_SVE_CF_2B: // .........i...ii. .......NNNNddddd -- SVE move predicate into vector
emitDispSveRegIndex(id->idReg1(), emitGetInsSC(id), true); // ddddd
emitDispPredicateReg(id->idReg2(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_D, false); // NNNN
break;
case IF_SVE_CF_2C: // ..............i. .......NNNNddddd -- SVE move predicate into vector
emitDispSveRegIndex(id->idReg1(), emitGetInsSC(id), true); // ddddd
emitDispPredicateReg(id->idReg2(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_H, false); // NNNN
break;
case IF_SVE_CF_2D: // .............ii. .......NNNNddddd -- SVE move predicate into vector
emitDispSveRegIndex(id->idReg1(), emitGetInsSC(id), true); // ddddd
emitDispPredicateReg(id->idReg2(), insGetPredicateType(fmt), INS_OPTS_SCALABLE_S, false); // NNNN
break;

// <Pd>.<T>, <Pn>.<T>, <Pm>.<T>
case IF_SVE_CI_3A: // ........xx..MMMM .......NNNN.DDDD -- SVE permute predicate elements
emitDispPredicateReg(id->idReg1(), insGetPredicateType(fmt, 1), id->idInsOpt(), true); // DDDD
Expand Down Expand Up @@ -30608,6 +30845,18 @@ emitter::insExecutionCharacteristics emitter::getInsExecutionCharacteristics(ins
result.insLatency = PERFSCORE_LATENCY_2C;
break;

case IF_SVE_CE_2A: // ................ ......nnnnn.DDDD -- SVE move predicate from vector
case IF_SVE_CE_2B: // .........i...ii. ......nnnnn.DDDD -- SVE move predicate from vector
case IF_SVE_CE_2C: // ..............i. ......nnnnn.DDDD -- SVE move predicate from vector
case IF_SVE_CE_2D: // .............ii. ......nnnnn.DDDD -- SVE move predicate from vector
case IF_SVE_CF_2A: // ................ .......NNNNddddd -- SVE move predicate into vector
case IF_SVE_CF_2B: // .........i...ii. .......NNNNddddd -- SVE move predicate into vector
case IF_SVE_CF_2C: // ..............i. .......NNNNddddd -- SVE move predicate into vector
case IF_SVE_CF_2D: // .............ii. .......NNNNddddd -- SVE move predicate into vector
result.insThroughput = PERFSCORE_THROUGHPUT_140C; // @ToDo currently undocumented
result.insLatency = PERFSCORE_LATENCY_140C;
break;

case IF_SVE_CI_3A: // ........xx..MMMM .......NNNN.DDDD -- SVE permute predicate elements
case IF_SVE_CJ_2A: // ........xx...... .......NNNN.DDDD -- SVE reverse predicate elements
case IF_SVE_CK_2A: // ................ .......NNNN.DDDD -- SVE unpack predicate elements
Expand Down
Loading