Skip to content

[Clang][SME] Refactor checkArmStreamingBuiltin. #145941

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 1 commit into from
Jul 8, 2025
+47 −39
Merged

[Clang][SME] Refactor checkArmStreamingBuiltin. #145941

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
36 changes: 18 additions & 18 deletions clang/include/clang/Basic/arm_sve.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -264,22 +264,22 @@ let SVETargetGuard = "sve,bf16", SMETargetGuard = "sme,bf16" in {
def SVLD1RQ_BF : SInst<"svld1rq[_{2}]", "dPc", "b", MergeNone, "aarch64_sve_ld1rq", [VerifyRuntimeMode]>;
}

multiclass StructLoad<string name, string proto, string i> {
def : SInst<name, proto, "csilUcUsUiUlhfdm", MergeNone, i, [IsStructLoad, VerifyRuntimeMode]>;
multiclass StructLoad<string name, string proto, string i, list<FlagType> f = []> {
def : SInst<name, proto, "csilUcUsUiUlhfdm", MergeNone, i, !listconcat(f, [IsStructLoad])>;
let SVETargetGuard = "sve,bf16", SMETargetGuard = "sme,bf16" in {
def: SInst<name, proto, "b", MergeNone, i, [IsStructLoad, VerifyRuntimeMode]>;
def: SInst<name, proto, "b", MergeNone, i, !listconcat(f, [IsStructLoad])>;
}
}

// Load N-element structure into N vectors (scalar base)
defm SVLD2 : StructLoad<"svld2[_{2}]", "2Pc", "aarch64_sve_ld2_sret">;
defm SVLD3 : StructLoad<"svld3[_{2}]", "3Pc", "aarch64_sve_ld3_sret">;
defm SVLD4 : StructLoad<"svld4[_{2}]", "4Pc", "aarch64_sve_ld4_sret">;
defm SVLD2 : StructLoad<"svld2[_{2}]", "2Pc", "aarch64_sve_ld2_sret", [VerifyRuntimeMode]>;
Copy link
Collaborator

Choose a reason for hiding this comment

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

Why do these need the runtime mode verified, and not the ld2q ? (those are valid with FEAT_SME2p1, although currently their target guard says SMETargetGuard = InvalidMode)

I guess that is what you meant with:

[1] The change uncovered an issue whereby a couple of builtins where tagged with VerifyRuntimeMode but did not include both streaming and non-streaming guards. Some of those builtins are available in SME2p1 but to keep the PR mostly NFC I've ensured they are only available in non-streaming mode, thus matching the existing implementation.

?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Exactly. I created #147362 to add the necessary guards to enable the sme2p1 builtins.

defm SVLD3 : StructLoad<"svld3[_{2}]", "3Pc", "aarch64_sve_ld3_sret", [VerifyRuntimeMode]>;
defm SVLD4 : StructLoad<"svld4[_{2}]", "4Pc", "aarch64_sve_ld4_sret", [VerifyRuntimeMode]>;
Copy link
Contributor

Choose a reason for hiding this comment

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

I've probably missed them, but I can't find any streaming or non-streaming guards for these instructions which are passing the VerifyRuntimeMode flag?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

SVEEmitter automatically prefixes SVETargetGuard with "sve," and SMETargetGuard with "sme," (see FormatGuard lambda within Intrinsic::Intrinsic()`.


// Load N-element structure into N vectors (scalar base, VL displacement)
defm SVLD2_VNUM : StructLoad<"svld2_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2_sret">;
defm SVLD3_VNUM : StructLoad<"svld3_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3_sret">;
defm SVLD4_VNUM : StructLoad<"svld4_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4_sret">;
defm SVLD2_VNUM : StructLoad<"svld2_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2_sret", [VerifyRuntimeMode]>;
defm SVLD3_VNUM : StructLoad<"svld3_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3_sret", [VerifyRuntimeMode]>;
defm SVLD4_VNUM : StructLoad<"svld4_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4_sret", [VerifyRuntimeMode]>;

// Load one octoword and replicate (scalar base)
let SVETargetGuard = "sve,f64mm", SMETargetGuard = InvalidMode in {
Expand Down Expand Up @@ -434,21 +434,21 @@ def SVST1H_SCATTER_INDEX_S : MInst<"svst1h_scatter[_{2}base]_index[_{d}]", "v
def SVST1W_SCATTER_INDEX_S : MInst<"svst1w_scatter[_{2}base]_index[_{d}]", "vPuld", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_scalar_offset">;
} // let SVETargetGuard = "sve"

multiclass StructStore<string name, string proto, string i> {
def : SInst<name, proto, "csilUcUsUiUlhfdm", MergeNone, i, [IsStructStore, VerifyRuntimeMode]>;
multiclass StructStore<string name, string proto, string i, list<FlagType> f = []> {
def : SInst<name, proto, "csilUcUsUiUlhfdm", MergeNone, i, !listconcat(f, [IsStructStore])>;
let SVETargetGuard = "sve,bf16", SMETargetGuard = "sme,bf16" in {
def: SInst<name, proto, "b", MergeNone, i, [IsStructStore, VerifyRuntimeMode]>;
def: SInst<name, proto, "b", MergeNone, i, !listconcat(f, [IsStructStore])>;
}
}
// Store N vectors into N-element structure (scalar base)
defm SVST2 : StructStore<"svst2[_{d}]", "vPp2", "aarch64_sve_st2">;
defm SVST3 : StructStore<"svst3[_{d}]", "vPp3", "aarch64_sve_st3">;
defm SVST4 : StructStore<"svst4[_{d}]", "vPp4", "aarch64_sve_st4">;
defm SVST2 : StructStore<"svst2[_{d}]", "vPp2", "aarch64_sve_st2", [VerifyRuntimeMode]>;
defm SVST3 : StructStore<"svst3[_{d}]", "vPp3", "aarch64_sve_st3", [VerifyRuntimeMode]>;
defm SVST4 : StructStore<"svst4[_{d}]", "vPp4", "aarch64_sve_st4", [VerifyRuntimeMode]>;

// Store N vectors into N-element structure (scalar base, VL displacement)
defm SVST2_VNUM : StructStore<"svst2_vnum[_{d}]", "vPpl2", "aarch64_sve_st2">;
defm SVST3_VNUM : StructStore<"svst3_vnum[_{d}]", "vPpl3", "aarch64_sve_st3">;
defm SVST4_VNUM : StructStore<"svst4_vnum[_{d}]", "vPpl4", "aarch64_sve_st4">;
defm SVST2_VNUM : StructStore<"svst2_vnum[_{d}]", "vPpl2", "aarch64_sve_st2", [VerifyRuntimeMode]>;
defm SVST3_VNUM : StructStore<"svst3_vnum[_{d}]", "vPpl3", "aarch64_sve_st3", [VerifyRuntimeMode]>;
defm SVST4_VNUM : StructStore<"svst4_vnum[_{d}]", "vPpl4", "aarch64_sve_st4", [VerifyRuntimeMode]>;

// Store one vector, with no truncation, non-temporal (scalar base)
def SVSTNT1 : MInst<"svstnt1[_{d}]", "vPpd", "csilUcUsUiUlhfdm", [IsStore, VerifyRuntimeMode], MemEltTyDefault, "aarch64_sve_stnt1">;
Expand Down
47 changes: 26 additions & 21 deletions clang/lib/Sema/SemaARM.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -569,34 +569,39 @@ static bool checkArmStreamingBuiltin(Sema &S, CallExpr *TheCall,
// * When compiling for SVE only, the caller must be in non-streaming mode.
// * When compiling for both SVE and SME, the caller can be in either mode.
if (BuiltinType == SemaARM::VerifyRuntimeMode) {
llvm::StringMap<bool> CallerFeatureMapWithoutSVE;
S.Context.getFunctionFeatureMap(CallerFeatureMapWithoutSVE, FD);
CallerFeatureMapWithoutSVE["sve"] = false;
llvm::StringMap<bool> CallerFeatures;
S.Context.getFunctionFeatureMap(CallerFeatures, FD);

// Avoid emitting diagnostics for a function that can never compile.
if (FnType == SemaARM::ArmStreaming && !CallerFeatureMapWithoutSVE["sme"])
if (FnType == SemaARM::ArmStreaming && !CallerFeatures["sme"])
return false;

llvm::StringMap<bool> CallerFeatureMapWithoutSME;
S.Context.getFunctionFeatureMap(CallerFeatureMapWithoutSME, FD);
CallerFeatureMapWithoutSME["sme"] = false;
const auto FindTopLevelPipe = [](const char *S) {
unsigned Depth = 0;
unsigned I = 0, E = strlen(S);
for (; I < E; ++I) {
if (S[I] == '|' && Depth == 0)
break;
if (S[I] == '(')
++Depth;
else if (S[I] == ')')
--Depth;
}
return I;
};

const char *RequiredFeatures =
S.Context.BuiltinInfo.getRequiredFeatures(BuiltinID);
unsigned PipeIdx = FindTopLevelPipe(RequiredFeatures);
assert(PipeIdx != 0 && PipeIdx != strlen(RequiredFeatures) &&
"Expected feature string of the form 'SVE-EXPR|SME-EXPR'");
StringRef NonStreamingBuiltinGuard = StringRef(RequiredFeatures, PipeIdx);
StringRef StreamingBuiltinGuard = StringRef(RequiredFeatures + PipeIdx + 1);

// We know the builtin requires either some combination of SVE flags, or
// some combination of SME flags, but we need to figure out which part
// of the required features is satisfied by the target features.
//
// For a builtin with target guard 'sve2p1|sme2', if we compile with
// '+sve2p1,+sme', then we know that it satisfies the 'sve2p1' part if we
// evaluate the features for '+sve2p1,+sme,+nosme'.
//
// Similarly, if we compile with '+sve2,+sme2', then we know it satisfies
// the 'sme2' part if we evaluate the features for '+sve2,+sme2,+nosve'.
StringRef BuiltinTargetGuards(
S.Context.BuiltinInfo.getRequiredFeatures(BuiltinID));
bool SatisfiesSVE = Builtin::evaluateRequiredTargetFeatures(
BuiltinTargetGuards, CallerFeatureMapWithoutSME);
NonStreamingBuiltinGuard, CallerFeatures);
bool SatisfiesSME = Builtin::evaluateRequiredTargetFeatures(
BuiltinTargetGuards, CallerFeatureMapWithoutSVE);
StreamingBuiltinGuard, CallerFeatures);

if ((SatisfiesSVE && SatisfiesSME) ||
(SatisfiesSVE && FnType == SemaARM::ArmStreamingCompatible))
Expand Down
3 changes: 3 additions & 0 deletions clang/utils/TableGen/SveEmitter.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1901,6 +1901,9 @@ void SVEEmitter::createStreamingAttrs(raw_ostream &OS, ACLEKind Kind) {
if (!Def->isFlagSet(VerifyRuntimeMode) && !Def->getSVEGuard().empty() &&
!Def->getSMEGuard().empty())
report_fatal_error("Missing VerifyRuntimeMode flag");
if (Def->isFlagSet(VerifyRuntimeMode) &&
(Def->getSVEGuard().empty() || Def->getSMEGuard().empty()))
report_fatal_error("VerifyRuntimeMode requires SVE and SME guards");

if (Def->isFlagSet(IsStreamingFlag))
StreamingMap["ArmStreaming"].insert(Def->getMangledName());
Expand Down
Loading