-
Notifications
You must be signed in to change notification settings - Fork 12.1k
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.
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
[X86] Change how we treat functions with explicit sections as small/large #88172
Conversation
…arge Following llvm#78348, we should treat functions with an explicit section as small, unless the section name is (or has the prefix) ".ltext". Clang emits global initializers into a ".text.startup" section on Linux. If we mix small/medium code model object files with large code model object files, we'll end up mixing sections with and without the large section flag. Reland of llvm#87838 with a check for non-ELF platforms in TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests fail.
@llvm/pr-subscribers-backend-x86 Author: Arthur Eubanks (aeubanks) ChangesFollowing #78348, we should treat functions with an explicit section as small, unless the section name is (or has the prefix) ".ltext". Clang emits global initializers into a ".text.startup" section on Linux. If we mix small/medium code model object files with large code model object files, we'll end up mixing sections with and without the large section flag. Reland of #87838 with a check for non-ELF platforms in TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests fail. Full diff: https://github.com/llvm/llvm-project/pull/88172.diff 2 Files Affected:
diff --git a/llvm/lib/Target/TargetMachine.cpp b/llvm/lib/Target/TargetMachine.cpp
index a7fe329b064ee16..4ae6a468df21f36 100644
--- a/llvm/lib/Target/TargetMachine.cpp
+++ b/llvm/lib/Target/TargetMachine.cpp
@@ -43,6 +43,9 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
if (getTargetTriple().getArch() != Triple::x86_64)
return false;
+ if (!getTargetTriple().isOSBinFormatELF())
+ return getCodeModel() == CodeModel::Large;
+
auto *GO = GVal->getAliaseeObject();
// Be conservative if we can't find an underlying GlobalObject.
@@ -51,9 +54,20 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
auto *GV = dyn_cast<GlobalVariable>(GO);
+ auto IsPrefix = [](StringRef Name, StringRef Prefix) {
+ return Name.consume_front(Prefix) && (Name.empty() || Name[0] == '.');
+ };
+
// Functions/GlobalIFuncs are only large under the large code model.
- if (!GV)
+ if (!GV) {
+ // Handle explicit sections as we do for GlobalVariables with an explicit
+ // section, see comments below.
+ if (GO->hasSection()) {
+ StringRef Name = GO->getSection();
+ return IsPrefix(Name, ".ltext");
+ }
return getCodeModel() == CodeModel::Large;
+ }
if (GV->isThreadLocal())
return false;
@@ -73,11 +87,8 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
// data sections. The code model attribute overrides this above.
if (GV->hasSection()) {
StringRef Name = GV->getSection();
- auto IsPrefix = [&](StringRef Prefix) {
- StringRef S = Name;
- return S.consume_front(Prefix) && (S.empty() || S[0] == '.');
- };
- return IsPrefix(".lbss") || IsPrefix(".ldata") || IsPrefix(".lrodata");
+ return IsPrefix(Name, ".lbss") || IsPrefix(Name, ".ldata") ||
+ IsPrefix(Name, ".lrodata");
}
// Respect large data threshold for medium and large code models.
diff --git a/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll b/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
index 016c9a4d7b8390e..66a6fd376754275 100644
--- a/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
+++ b/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
@@ -13,9 +13,20 @@
; RUN: llvm-readelf -S %t | FileCheck %s --check-prefix=LARGE-DS
; SMALL: .text {{.*}} AX {{.*}}
+; SMALL: .ltext {{.*}} AXl {{.*}}
+; SMALL: .ltext.2 {{.*}} AXl {{.*}}
+; SMALL: .foo {{.*}} AX {{.*}}
; SMALL-DS: .text.func {{.*}} AX {{.*}}
+; SMALL-DS: .ltext {{.*}} AXl {{.*}}
+; SMALL-DS: .ltext.2 {{.*}} AXl {{.*}}
+; SMALL-DS: .foo {{.*}} AX {{.*}}
; LARGE: .ltext {{.*}} AXl {{.*}}
+; LARGE: .ltext.2 {{.*}} AXl {{.*}}
+; LARGE: .foo {{.*}} AX {{.*}}
; LARGE-DS: .ltext.func {{.*}} AXl {{.*}}
+; LARGE-DS: .ltext {{.*}} AXl {{.*}}
+; LARGE-DS: .ltext.2 {{.*}} AXl {{.*}}
+; LARGE-DS: .foo {{.*}} AX {{.*}}
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64--linux"
@@ -23,3 +34,15 @@ target triple = "x86_64--linux"
define void @func() {
ret void
}
+
+define void @ltext() section ".ltext" {
+ ret void
+}
+
+define void @ltext2() section ".ltext.2" {
+ ret void
+}
+
+define void @foo() section ".foo" {
+ ret void
+}
|
ping |
@@ -43,6 +43,9 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const { | |||
if (getTargetTriple().getArch() != Triple::x86_64) | |||
return false; | |||
|
|||
if (!getTargetTriple().isOSBinFormatELF()) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add a comment about what's going on here. Essentially, all the logic below is ELF-specific. The large code model is supported with other object file formats mainly for the benefit of JIT compilation, and for that we fall back to the global code model option.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
done
This matches other types of relocations, e.g. to constant pool. And makes things more consistent with PIC large code model. Some users of the large code model may not place small data in the lower 2GB of the address space (e.g. ClangBuiltLinux/linux#2016), so just unconditionally use 64-bit relocations in the large code model. So now functions in a section not marked large will use 64-bit relocations to reference everything when using the large code model. This also fixes some lldb tests broken by #88172 (https://lab.llvm.org/buildbot/#/builders/68/builds/72458).
…89101) This matches other types of relocations, e.g. to constant pool. And makes things more consistent with PIC large code model. Some users of the large code model may not place small data in the lower 2GB of the address space (e.g. ClangBuiltLinux/linux#2016), so just unconditionally use 64-bit relocations in the large code model. So now functions in a section not marked large will use 64-bit relocations to reference everything when using the large code model. This also fixes some lldb tests broken by llvm#88172 (https://lab.llvm.org/buildbot/#/builders/68/builds/72458). (cherry picked from commit 6cea7c4)
…89101) This matches other types of relocations, e.g. to constant pool. And makes things more consistent with PIC large code model. Some users of the large code model may not place small data in the lower 2GB of the address space (e.g. ClangBuiltLinux/linux#2016), so just unconditionally use 64-bit relocations in the large code model. So now functions in a section not marked large will use 64-bit relocations to reference everything when using the large code model. This also fixes some lldb tests broken by llvm#88172 (https://lab.llvm.org/buildbot/#/builders/68/builds/72458). (cherry picked from commit 6cea7c4)
…arge (llvm#88172) Following llvm#78348, we should treat functions with an explicit section as small, unless the section name is (or has the prefix) ".ltext". Clang emits global initializers into a ".text.startup" section on Linux. If we mix small/medium code model object files with large code model object files, we'll end up mixing sections with and without the large section flag. Reland of llvm#87838 with a check for non-ELF platforms in TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests fail.
…arge (llvm#88172) Following llvm#78348, we should treat functions with an explicit section as small, unless the section name is (or has the prefix) ".ltext". Clang emits global initializers into a ".text.startup" section on Linux. If we mix small/medium code model object files with large code model object files, we'll end up mixing sections with and without the large section flag. Reland of llvm#87838 with a check for non-ELF platforms in TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests fail.
Following #78348, we should treat functions with an explicit section as small, unless the section name is (or has the prefix) ".ltext".
Clang emits global initializers into a ".text.startup" section on Linux. If we mix small/medium code model object files with large code model object files, we'll end up mixing sections with and without the large section flag.
Reland of #87838 with a check for non-ELF platforms in TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests fail.