[CIR] Implement inline builtin functions #163911
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mmha
requested review from
andykaylor,
bcardosolopes,
lanza and
xlauko
as code owners
llvmbot
added
clang
Member
|
@llvm/pr-subscribers-clangir Author: Morris Hafner (mmha) ChangesThis patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch. Full diff: https://github.com/llvm/llvm-project/pull/163911.diff 4 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
index 4897c29b58a1f..319da852bef74 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
@@ -1674,7 +1674,25 @@ CIRGenCallee CIRGenFunction::emitDirectCallee(const GlobalDecl &gd) {
// name to make it clear it's not the actual builtin.
auto fn = cast<cir::FuncOp>(curFn);
if (fn.getName() != fdInlineName && onlyHasInlineBuiltinDeclaration(fd)) {
- cgm.errorNYI("Inline only builtin function calls");
+ cir::FuncOp clone =
+ mlir::cast_or_null<cir::FuncOp>(cgm.getGlobalValue(fdInlineName));
+
+ if (!clone) {
+ // Create a forward declaration - the body will be generated in
+ // generateCode when the function definition is processed
+ cir::FuncOp calleeFunc = emitFunctionDeclPointer(cgm, gd);
+ mlir::OpBuilder::InsertionGuard guard(builder);
+ builder.setInsertionPointToStart(cgm.getModule().getBody());
+
+ clone = builder.create<cir::FuncOp>(calleeFunc.getLoc(), fdInlineName,
+ calleeFunc.getFunctionType());
+ clone.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::InternalLinkage));
+ clone.setSymVisibility("private");
+ clone.setInlineKindAttr(cir::InlineAttr::get(
+ &cgm.getMLIRContext(), cir::InlineKind::AlwaysInline));
+ }
+ return CIRGenCallee::forDirect(clone, gd);
}
// Replaceable builtins provide their own implementation of a builtin. If we
diff --git a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
index ba36cbeeb620e..59dee9d542819 100644
--- a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
@@ -551,6 +551,50 @@ cir::FuncOp CIRGenFunction::generateCode(clang::GlobalDecl gd, cir::FuncOp fn,
const auto funcDecl = cast<FunctionDecl>(gd.getDecl());
curGD = gd;
+ if (funcDecl->isInlineBuiltinDeclaration()) {
+ // When generating code for a builtin with an inline declaration, use a
+ // mangled name to hold the actual body, while keeping an external
+ // declaration in case the function pointer is referenced somewhere.
+ std::string fdInlineName = fn.getName().str() + ".inline";
+ cir::FuncOp clone =
+ mlir::cast_or_null<cir::FuncOp>(cgm.getGlobalValue(fdInlineName));
+ if (!clone) {
+ mlir::OpBuilder::InsertionGuard guard(builder);
+ builder.setInsertionPoint(fn);
+ clone = builder.create<cir::FuncOp>(fn.getLoc(), fdInlineName,
+ fn.getFunctionType());
+ clone.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::InternalLinkage));
+ clone.setSymVisibility("private");
+ clone.setInlineKindAttr(cir::InlineAttr::get(
+ &cgm.getMLIRContext(), cir::InlineKind::AlwaysInline));
+ }
+ fn.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::ExternalLinkage));
+ fn.setSymVisibility("private");
+ fn = clone;
+ } else {
+ // Detect the unusual situation where an inline version is shadowed by a
+ // non-inline version. In that case we should pick the external one
+ // everywhere. That's GCC behavior too.
+ for (const FunctionDecl *pd = funcDecl->getPreviousDecl(); pd;
+ pd = pd->getPreviousDecl()) {
+ if (LLVM_UNLIKELY(pd->isInlineBuiltinDeclaration())) {
+ std::string inlineName = funcDecl->getName().str() + ".inline";
+ if (auto inlineFn = mlir::cast_or_null<cir::FuncOp>(
+ cgm.getGlobalValue(inlineName))) {
+ // Replace all uses of the .inline function with the regular function
+ if (inlineFn
+ .replaceAllSymbolUses(fn.getSymNameAttr(), cgm.getModule())
+ .failed())
+ llvm_unreachable("Failed to replace inline builtin symbol uses");
+ inlineFn.erase();
+ }
+ break;
+ }
+ }
+ }
+
SourceLocation loc = funcDecl->getLocation();
Stmt *body = funcDecl->getBody();
SourceRange bodyRange =
diff --git a/clang/lib/CIR/CodeGen/CIRGenModule.cpp b/clang/lib/CIR/CodeGen/CIRGenModule.cpp
index 127f763e11204..74d39b73fff8f 100644
--- a/clang/lib/CIR/CodeGen/CIRGenModule.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenModule.cpp
@@ -1917,6 +1917,17 @@ void CIRGenModule::setFunctionAttributes(GlobalDecl globalDecl,
const Decl *decl = globalDecl.getDecl();
func.setGlobalVisibilityAttr(getGlobalVisibilityAttrFromDecl(decl));
}
+
+ // If we plan on emitting this inline builtin, we can't treat it as a builtin.
+ const auto *fd = cast<FunctionDecl>(globalDecl.getDecl());
+ if (fd->isInlineBuiltinDeclaration()) {
+ const FunctionDecl *fdBody;
+ bool hasBody = fd->hasBody(fdBody);
+ (void)hasBody;
+ assert(hasBody && "Inline builtin declarations should always have an "
+ "available body!");
+ assert(!cir::MissingFeatures::attributeNoBuiltin());
+ }
}
void CIRGenModule::setCIRFunctionAttributesForDefinition(
diff --git a/clang/test/CIR/CodeGen/builtin_inline.c b/clang/test/CIR/CodeGen/builtin_inline.c
new file mode 100644
index 0000000000000..83a3ba6e53f4b
--- /dev/null
+++ b/clang/test/CIR/CodeGen/builtin_inline.c
@@ -0,0 +1,91 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s -check-prefix=CIR
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm -disable-llvm-passes %s -o %t-cir.ll
+// RUN: FileCheck --input-file=%t-cir.ll %s -check-prefix=LLVM
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -emit-llvm -disable-llvm-passes %s -o %t.ll
+// RUN: FileCheck --input-file=%t.ll %s -check-prefix=OGCG
+
+typedef unsigned long size_t;
+
+// Normal inline builtin declaration
+// When a builtin is redefined with extern inline + always_inline attributes,
+// the compiler creates a .inline version to avoid conflicts with the builtin
+
+extern inline __attribute__((always_inline)) __attribute__((gnu_inline))
+void *memcpy(void *a, const void *b, size_t c) {
+ return __builtin_memcpy(a, b, c);
+}
+
+void *test_inline_builtin_memcpy(void *a, const void *b, size_t c) {
+ return memcpy(a, b, c);
+}
+
+// CIR: cir.func internal private{{.*}}@memcpy.inline({{.*}}) -> !cir.ptr<!void> inline(always)
+
+// CIR-LABEL: @test_inline_builtin_memcpy(
+// CIR: cir.call @memcpy.inline(
+// CIR: }
+
+// LLVM: define internal ptr @memcpy.inline(ptr{{.*}}, ptr{{.*}}, i64{{.*}}) #{{[0-9]+}}
+
+// LLVM-LABEL: @test_inline_builtin_memcpy(
+// LLVM: call ptr @memcpy.inline(
+
+// OGCG-LABEL: @test_inline_builtin_memcpy(
+// OGCG: call ptr @memcpy.inline(
+
+// OGCG: define internal ptr @memcpy.inline(ptr{{.*}} %a, ptr{{.*}} %b, i64{{.*}} %c) #{{[0-9]+}}
+
+// Shadowing case
+// When a non-inline function definition shadows an inline builtin declaration,
+// the .inline version should be replaced with the regular function and removed.
+
+extern inline __attribute__((always_inline)) __attribute__((gnu_inline))
+void *memmove(void *a, const void *b, size_t c) {
+ return __builtin_memmove(a, b, c);
+}
+
+void *memmove(void *a, const void *b, size_t c) {
+ char *dst = (char *)a;
+ const char *src = (const char *)b;
+ if (dst < src) {
+ for (size_t i = 0; i < c; i++) {
+ dst[i] = src[i];
+ }
+ } else {
+ for (size_t i = c; i > 0; i--) {
+ dst[i-1] = src[i-1];
+ }
+ }
+ return a;
+}
+
+void *test_shadowed_memmove(void *a, const void *b, size_t c) {
+ return memmove(a, b, c);
+}
+
+// CIR: cir.func{{.*}}@memmove({{.*}}) -> !cir.ptr<!void>{{.*}}{
+// CIR-NOT: @memmove.inline
+
+// CIR-LABEL: @test_shadowed_memmove(
+// CIR: cir.call @memmove(
+// CIR-NOT: @memmove.inline
+// CIR: }
+
+// LLVM: define dso_local ptr @memmove(ptr{{.*}}, ptr{{.*}}, i64{{.*}}) #{{[0-9]+}}
+// LLVM-NOT: @memmove.inline
+
+// LLVM-LABEL: @test_shadowed_memmove(
+// TODO - this deviation from OGCG is expected until we implement the nobuiltin
+// attribute. See CIRGenFunction::emitDirectCallee
+// LLVM: call ptr @memmove(
+// LLVM-NOT: @memmove.inline
+// LLVM: }
+
+// OGCG: define dso_local ptr @memmove(ptr{{.*}} %a, ptr{{.*}} %b, i64{{.*}} %c) #{{[0-9]+}}
+// OGCG-NOT: @memmove.inline
+
+// OGCG-LABEL: @test_shadowed_memmove(
+// OGCG: call void @llvm.memmove.p0.p0.i64(
+// OGCG-NOT: @memmove.inline
+// OGCG: }
|
Member
|
@llvm/pr-subscribers-clang Author: Morris Hafner (mmha) ChangesThis patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch. Full diff: https://github.com/llvm/llvm-project/pull/163911.diff 4 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
index 4897c29b58a1f..319da852bef74 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
@@ -1674,7 +1674,25 @@ CIRGenCallee CIRGenFunction::emitDirectCallee(const GlobalDecl &gd) {
// name to make it clear it's not the actual builtin.
auto fn = cast<cir::FuncOp>(curFn);
if (fn.getName() != fdInlineName && onlyHasInlineBuiltinDeclaration(fd)) {
- cgm.errorNYI("Inline only builtin function calls");
+ cir::FuncOp clone =
+ mlir::cast_or_null<cir::FuncOp>(cgm.getGlobalValue(fdInlineName));
+
+ if (!clone) {
+ // Create a forward declaration - the body will be generated in
+ // generateCode when the function definition is processed
+ cir::FuncOp calleeFunc = emitFunctionDeclPointer(cgm, gd);
+ mlir::OpBuilder::InsertionGuard guard(builder);
+ builder.setInsertionPointToStart(cgm.getModule().getBody());
+
+ clone = builder.create<cir::FuncOp>(calleeFunc.getLoc(), fdInlineName,
+ calleeFunc.getFunctionType());
+ clone.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::InternalLinkage));
+ clone.setSymVisibility("private");
+ clone.setInlineKindAttr(cir::InlineAttr::get(
+ &cgm.getMLIRContext(), cir::InlineKind::AlwaysInline));
+ }
+ return CIRGenCallee::forDirect(clone, gd);
}
// Replaceable builtins provide their own implementation of a builtin. If we
diff --git a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
index ba36cbeeb620e..59dee9d542819 100644
--- a/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenFunction.cpp
@@ -551,6 +551,50 @@ cir::FuncOp CIRGenFunction::generateCode(clang::GlobalDecl gd, cir::FuncOp fn,
const auto funcDecl = cast<FunctionDecl>(gd.getDecl());
curGD = gd;
+ if (funcDecl->isInlineBuiltinDeclaration()) {
+ // When generating code for a builtin with an inline declaration, use a
+ // mangled name to hold the actual body, while keeping an external
+ // declaration in case the function pointer is referenced somewhere.
+ std::string fdInlineName = fn.getName().str() + ".inline";
+ cir::FuncOp clone =
+ mlir::cast_or_null<cir::FuncOp>(cgm.getGlobalValue(fdInlineName));
+ if (!clone) {
+ mlir::OpBuilder::InsertionGuard guard(builder);
+ builder.setInsertionPoint(fn);
+ clone = builder.create<cir::FuncOp>(fn.getLoc(), fdInlineName,
+ fn.getFunctionType());
+ clone.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::InternalLinkage));
+ clone.setSymVisibility("private");
+ clone.setInlineKindAttr(cir::InlineAttr::get(
+ &cgm.getMLIRContext(), cir::InlineKind::AlwaysInline));
+ }
+ fn.setLinkageAttr(cir::GlobalLinkageKindAttr::get(
+ &cgm.getMLIRContext(), cir::GlobalLinkageKind::ExternalLinkage));
+ fn.setSymVisibility("private");
+ fn = clone;
+ } else {
+ // Detect the unusual situation where an inline version is shadowed by a
+ // non-inline version. In that case we should pick the external one
+ // everywhere. That's GCC behavior too.
+ for (const FunctionDecl *pd = funcDecl->getPreviousDecl(); pd;
+ pd = pd->getPreviousDecl()) {
+ if (LLVM_UNLIKELY(pd->isInlineBuiltinDeclaration())) {
+ std::string inlineName = funcDecl->getName().str() + ".inline";
+ if (auto inlineFn = mlir::cast_or_null<cir::FuncOp>(
+ cgm.getGlobalValue(inlineName))) {
+ // Replace all uses of the .inline function with the regular function
+ if (inlineFn
+ .replaceAllSymbolUses(fn.getSymNameAttr(), cgm.getModule())
+ .failed())
+ llvm_unreachable("Failed to replace inline builtin symbol uses");
+ inlineFn.erase();
+ }
+ break;
+ }
+ }
+ }
+
SourceLocation loc = funcDecl->getLocation();
Stmt *body = funcDecl->getBody();
SourceRange bodyRange =
diff --git a/clang/lib/CIR/CodeGen/CIRGenModule.cpp b/clang/lib/CIR/CodeGen/CIRGenModule.cpp
index 127f763e11204..74d39b73fff8f 100644
--- a/clang/lib/CIR/CodeGen/CIRGenModule.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenModule.cpp
@@ -1917,6 +1917,17 @@ void CIRGenModule::setFunctionAttributes(GlobalDecl globalDecl,
const Decl *decl = globalDecl.getDecl();
func.setGlobalVisibilityAttr(getGlobalVisibilityAttrFromDecl(decl));
}
+
+ // If we plan on emitting this inline builtin, we can't treat it as a builtin.
+ const auto *fd = cast<FunctionDecl>(globalDecl.getDecl());
+ if (fd->isInlineBuiltinDeclaration()) {
+ const FunctionDecl *fdBody;
+ bool hasBody = fd->hasBody(fdBody);
+ (void)hasBody;
+ assert(hasBody && "Inline builtin declarations should always have an "
+ "available body!");
+ assert(!cir::MissingFeatures::attributeNoBuiltin());
+ }
}
void CIRGenModule::setCIRFunctionAttributesForDefinition(
diff --git a/clang/test/CIR/CodeGen/builtin_inline.c b/clang/test/CIR/CodeGen/builtin_inline.c
new file mode 100644
index 0000000000000..83a3ba6e53f4b
--- /dev/null
+++ b/clang/test/CIR/CodeGen/builtin_inline.c
@@ -0,0 +1,91 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s -check-prefix=CIR
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm -disable-llvm-passes %s -o %t-cir.ll
+// RUN: FileCheck --input-file=%t-cir.ll %s -check-prefix=LLVM
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -emit-llvm -disable-llvm-passes %s -o %t.ll
+// RUN: FileCheck --input-file=%t.ll %s -check-prefix=OGCG
+
+typedef unsigned long size_t;
+
+// Normal inline builtin declaration
+// When a builtin is redefined with extern inline + always_inline attributes,
+// the compiler creates a .inline version to avoid conflicts with the builtin
+
+extern inline __attribute__((always_inline)) __attribute__((gnu_inline))
+void *memcpy(void *a, const void *b, size_t c) {
+ return __builtin_memcpy(a, b, c);
+}
+
+void *test_inline_builtin_memcpy(void *a, const void *b, size_t c) {
+ return memcpy(a, b, c);
+}
+
+// CIR: cir.func internal private{{.*}}@memcpy.inline({{.*}}) -> !cir.ptr<!void> inline(always)
+
+// CIR-LABEL: @test_inline_builtin_memcpy(
+// CIR: cir.call @memcpy.inline(
+// CIR: }
+
+// LLVM: define internal ptr @memcpy.inline(ptr{{.*}}, ptr{{.*}}, i64{{.*}}) #{{[0-9]+}}
+
+// LLVM-LABEL: @test_inline_builtin_memcpy(
+// LLVM: call ptr @memcpy.inline(
+
+// OGCG-LABEL: @test_inline_builtin_memcpy(
+// OGCG: call ptr @memcpy.inline(
+
+// OGCG: define internal ptr @memcpy.inline(ptr{{.*}} %a, ptr{{.*}} %b, i64{{.*}} %c) #{{[0-9]+}}
+
+// Shadowing case
+// When a non-inline function definition shadows an inline builtin declaration,
+// the .inline version should be replaced with the regular function and removed.
+
+extern inline __attribute__((always_inline)) __attribute__((gnu_inline))
+void *memmove(void *a, const void *b, size_t c) {
+ return __builtin_memmove(a, b, c);
+}
+
+void *memmove(void *a, const void *b, size_t c) {
+ char *dst = (char *)a;
+ const char *src = (const char *)b;
+ if (dst < src) {
+ for (size_t i = 0; i < c; i++) {
+ dst[i] = src[i];
+ }
+ } else {
+ for (size_t i = c; i > 0; i--) {
+ dst[i-1] = src[i-1];
+ }
+ }
+ return a;
+}
+
+void *test_shadowed_memmove(void *a, const void *b, size_t c) {
+ return memmove(a, b, c);
+}
+
+// CIR: cir.func{{.*}}@memmove({{.*}}) -> !cir.ptr<!void>{{.*}}{
+// CIR-NOT: @memmove.inline
+
+// CIR-LABEL: @test_shadowed_memmove(
+// CIR: cir.call @memmove(
+// CIR-NOT: @memmove.inline
+// CIR: }
+
+// LLVM: define dso_local ptr @memmove(ptr{{.*}}, ptr{{.*}}, i64{{.*}}) #{{[0-9]+}}
+// LLVM-NOT: @memmove.inline
+
+// LLVM-LABEL: @test_shadowed_memmove(
+// TODO - this deviation from OGCG is expected until we implement the nobuiltin
+// attribute. See CIRGenFunction::emitDirectCallee
+// LLVM: call ptr @memmove(
+// LLVM-NOT: @memmove.inline
+// LLVM: }
+
+// OGCG: define dso_local ptr @memmove(ptr{{.*}} %a, ptr{{.*}} %b, i64{{.*}} %c) #{{[0-9]+}}
+// OGCG-NOT: @memmove.inline
+
+// OGCG-LABEL: @test_shadowed_memmove(
+// OGCG: call void @llvm.memmove.p0.p0.i64(
+// OGCG-NOT: @memmove.inline
+// OGCG: }
|
xlauko
reviewed
Oct 17, 2025
xlauko
reviewed
Oct 17, 2025
andykaylor
reviewed
Oct 18, 2025
This patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch.
mmha
force-pushed
the
cir-inline-builtins
branch
from
fc2e164 to
ac95da4
Compare
Lukacma
pushed a commit
to Lukacma/llvm-project
that referenced
this pull request
Oct 29, 2025
This patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch.
aokblast
pushed a commit
to aokblast/llvm-project
that referenced
this pull request
Oct 30, 2025
This patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch implements the handling of inline builtin functions in CIR. There is a known limitation in CIR where direct calls to shadowed inline builtin functions are generated instead of the intrinsic. This is expected to be fixed by the introduction of the nobuiltin attribute in a future patch.