[CIR] Change default assumption about allowing builtins #144004
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
The code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs.
llvmbot
added
clang
Member
|
@llvm/pr-subscribers-clang Author: Andy Kaylor (andykaylor) ChangesThe code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs. Full diff: https://github.com/llvm/llvm-project/pull/144004.diff 2 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
index 5d04faf443b8d..b3b23eb41bfa3 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
@@ -1052,7 +1052,8 @@ CIRGenCallee CIRGenFunction::emitDirectCallee(const GlobalDecl &gd) {
bool isPredefinedLibFunction =
cgm.getASTContext().BuiltinInfo.isPredefinedLibFunction(builtinID);
- bool hasAttributeNoBuiltin = false;
+ // Assume nobuiltins everywhere until we actually read the attributes.
+ bool hasAttributeNoBuiltin = true;
assert(!cir::MissingFeatures::attributeNoBuiltin());
// When directing calling an inline builtin, call it through it's mangled
diff --git a/clang/test/CIR/CodeGen/libc.c b/clang/test/CIR/CodeGen/libc.c
new file mode 100644
index 0000000000000..f65fe92cd36a0
--- /dev/null
+++ b/clang/test/CIR/CodeGen/libc.c
@@ -0,0 +1,55 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s
+
+// Note: In the final implementation, we will want these to generate
+// CIR-specific libc operations. This test is just a placeholder
+// to make sure we can compile these to normal function calls
+// until the special handling is implemented.
+
+void *memcpy(void *, const void *, unsigned long);
+void testMemcpy(void *dst, const void *src, unsigned long size) {
+ memcpy(dst, src, size);
+ // CHECK: cir.call @memcpy
+}
+
+void *memmove(void *, const void *, unsigned long);
+void testMemmove(void *src, const void *dst, unsigned long size) {
+ memmove(dst, src, size);
+ // CHECK: cir.call @memmove
+}
+
+void *memset(void *, int, unsigned long);
+void testMemset(void *dst, int val, unsigned long size) {
+ memset(dst, val, size);
+ // CHECK: cir.call @memset
+}
+
+double fabs(double);
+double testFabs(double x) {
+ return fabs(x);
+ // CHECK: cir.call @fabs
+}
+
+float fabsf(float);
+float testFabsf(float x) {
+ return fabsf(x);
+ // CHECK: cir.call @fabsf
+}
+
+int abs(int);
+int testAbs(int x) {
+ return abs(x);
+ // CHECK: cir.call @abs
+}
+
+long labs(long);
+long testLabs(long x) {
+ return labs(x);
+ // CHECK: cir.call @labs
+}
+
+long long llabs(long long);
+long long testLlabs(long long x) {
+ return llabs(x);
+ // CHECK: cir.call @llabs
+}
|
Member
|
@llvm/pr-subscribers-clangir Author: Andy Kaylor (andykaylor) ChangesThe code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs. Full diff: https://github.com/llvm/llvm-project/pull/144004.diff 2 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
index 5d04faf443b8d..b3b23eb41bfa3 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
@@ -1052,7 +1052,8 @@ CIRGenCallee CIRGenFunction::emitDirectCallee(const GlobalDecl &gd) {
bool isPredefinedLibFunction =
cgm.getASTContext().BuiltinInfo.isPredefinedLibFunction(builtinID);
- bool hasAttributeNoBuiltin = false;
+ // Assume nobuiltins everywhere until we actually read the attributes.
+ bool hasAttributeNoBuiltin = true;
assert(!cir::MissingFeatures::attributeNoBuiltin());
// When directing calling an inline builtin, call it through it's mangled
diff --git a/clang/test/CIR/CodeGen/libc.c b/clang/test/CIR/CodeGen/libc.c
new file mode 100644
index 0000000000000..f65fe92cd36a0
--- /dev/null
+++ b/clang/test/CIR/CodeGen/libc.c
@@ -0,0 +1,55 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s
+
+// Note: In the final implementation, we will want these to generate
+// CIR-specific libc operations. This test is just a placeholder
+// to make sure we can compile these to normal function calls
+// until the special handling is implemented.
+
+void *memcpy(void *, const void *, unsigned long);
+void testMemcpy(void *dst, const void *src, unsigned long size) {
+ memcpy(dst, src, size);
+ // CHECK: cir.call @memcpy
+}
+
+void *memmove(void *, const void *, unsigned long);
+void testMemmove(void *src, const void *dst, unsigned long size) {
+ memmove(dst, src, size);
+ // CHECK: cir.call @memmove
+}
+
+void *memset(void *, int, unsigned long);
+void testMemset(void *dst, int val, unsigned long size) {
+ memset(dst, val, size);
+ // CHECK: cir.call @memset
+}
+
+double fabs(double);
+double testFabs(double x) {
+ return fabs(x);
+ // CHECK: cir.call @fabs
+}
+
+float fabsf(float);
+float testFabsf(float x) {
+ return fabsf(x);
+ // CHECK: cir.call @fabsf
+}
+
+int abs(int);
+int testAbs(int x) {
+ return abs(x);
+ // CHECK: cir.call @abs
+}
+
+long labs(long);
+long testLabs(long x) {
+ return labs(x);
+ // CHECK: cir.call @labs
+}
+
+long long llabs(long long);
+long long testLlabs(long long x) {
+ return llabs(x);
+ // CHECK: cir.call @llabs
+}
|
erichkeane
approved these changes
Jun 13, 2025
Collaborator
erichkeane
left a comment
erichkeane
left a comment
There was a problem hiding this comment.
Seems like a reasonable assumption for now. THOUGH, I wonder if we'd be better off at one point instead to figure out which ones we DO know how to handle, and use the 'nobuiltin' as a fallback instead? That way we can exercise the builtin-code as it gets implemented.
tomtor
pushed a commit
to tomtor/llvm-project
that referenced
this pull request
Jun 14, 2025
The code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs.
akuhlens
pushed a commit
to akuhlens/llvm-project
that referenced
this pull request
Jun 24, 2025
The code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
The code to read the "nobuiltins" attributes hasn't been implemented yet, but we were defaulting to the assumption that use of builtins is allowed for function calls that we recognize as standard C library calls and have builtin equivalents of. This change reverses that assumption so that when such calls are encountered, we just emit the call. This is a better default assumption, and since our builtin handling for these functions isn't implemented yet, it also allows us to compile more programs.