Skip to content
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

[Driver,CodeGen] Support -mtls-dialect= #79256

Merged
merged 5 commits into from
Jan 26, 2024

Conversation

MaskRay
Copy link
Member

@MaskRay MaskRay commented Jan 24, 2024

GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth [email protected]

Created using spr 1.3.4
@llvmbot llvmbot added clang Clang issues not falling into any other category clang:driver 'clang' and 'clang++' user-facing binaries. Not 'clang-cl' clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:codegen labels Jan 24, 2024
@llvmbot
Copy link
Collaborator

llvmbot commented Jan 24, 2024

@llvm/pr-subscribers-clang-driver
@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-clang

Author: Fangrui Song (MaskRay)

Changes

GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

  • x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
  • RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see #66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

Co-authored-by: Paul Kirth <[email protected]>


Full diff: https://github.com/llvm/llvm-project/pull/79256.diff

6 Files Affected:

  • (modified) clang/include/clang/Basic/CodeGenOptions.def (+3)
  • (modified) clang/include/clang/Driver/Options.td (+5)
  • (modified) clang/lib/CodeGen/BackendUtil.cpp (+1)
  • (modified) clang/lib/Driver/ToolChains/Clang.cpp (+23)
  • (added) clang/test/CodeGen/RISCV/tls-dialect.c (+13)
  • (added) clang/test/Driver/tls-dialect.c (+19)
diff --git a/clang/include/clang/Basic/CodeGenOptions.def b/clang/include/clang/Basic/CodeGenOptions.def
index 2f2e45d5cf63dfa..7c0bfe328496147 100644
--- a/clang/include/clang/Basic/CodeGenOptions.def
+++ b/clang/include/clang/Basic/CodeGenOptions.def
@@ -369,6 +369,9 @@ ENUM_CODEGENOPT(VecLib, llvm::driver::VectorLibrary, 3, llvm::driver::VectorLibr
 /// The default TLS model to use.
 ENUM_CODEGENOPT(DefaultTLSModel, TLSModel, 2, GeneralDynamicTLSModel)
 
+/// Whether to enable TLSDESC. AArch64 enables TLSDESC regardless of this value.
+CODEGENOPT(EnableTLSDESC, 1, 0)
+
 /// Bit size of immediate TLS offsets (0 == use the default).
 VALUE_CODEGENOPT(TLSSize, 8, 0)
 
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 7f4fa33748facaf..773bc1dcda01d5c 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -4419,6 +4419,8 @@ def mtls_size_EQ : Joined<["-"], "mtls-size=">, Group<m_Group>,
   HelpText<"Specify bit size of immediate TLS offsets (AArch64 ELF only): "
            "12 (for 4KB) | 24 (for 16MB, default) | 32 (for 4GB) | 48 (for 256TB, needs -mcmodel=large)">,
   MarshallingInfoInt<CodeGenOpts<"TLSSize">>;
+def mtls_dialect_EQ : Joined<["-"], "mtls-dialect=">, Group<m_Group>,
+  Flags<[TargetSpecific]>, HelpText<"Which thread-local storage dialect to use for dynamic accesses of TLS variables">;
 def mimplicit_it_EQ : Joined<["-"], "mimplicit-it=">, Group<m_Group>;
 def mdefault_build_attributes : Joined<["-"], "mdefault-build-attributes">, Group<m_Group>;
 def mno_default_build_attributes : Joined<["-"], "mno-default-build-attributes">, Group<m_Group>;
@@ -7066,6 +7068,9 @@ def fexperimental_assignment_tracking_EQ : Joined<["-"], "fexperimental-assignme
   Values<"disabled,enabled,forced">, NormalizedValues<["Disabled","Enabled","Forced"]>,
   MarshallingInfoEnum<CodeGenOpts<"AssignmentTrackingMode">, "Enabled">;
 
+def enable_tlsdesc : Flag<["-"], "enable-tlsdesc">,
+  MarshallingInfoFlag<CodeGenOpts<"EnableTLSDESC">>;
+
 } // let Visibility = [CC1Option]
 
 //===----------------------------------------------------------------------===//
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
index ec203f6f28bc173..7877e20d77f7724 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -401,6 +401,7 @@ static bool initTargetOptions(DiagnosticsEngine &Diags,
   Options.UniqueBasicBlockSectionNames =
       CodeGenOpts.UniqueBasicBlockSectionNames;
   Options.TLSSize = CodeGenOpts.TLSSize;
+  Options.EnableTLSDESC = CodeGenOpts.EnableTLSDESC;
   Options.EmulatedTLS = CodeGenOpts.EmulatedTLS;
   Options.DebuggerTuning = CodeGenOpts.getDebuggerTuning();
   Options.EmitStackSizeSection = CodeGenOpts.StackSizeSection;
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 5dc614e11aab599..93fd579eb92ba50 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -5822,6 +5822,29 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
     Args.AddLastArg(CmdArgs, options::OPT_mtls_size_EQ);
   }
 
+  if (Arg *A = Args.getLastArg(options::OPT_mtls_dialect_EQ)) {
+    StringRef V = A->getValue();
+    bool SupportedArgument = false, EnableTLSDESC = false;
+    bool Unsupported = !Triple.isOSBinFormatELF();
+    if (Triple.isRISCV()) {
+      SupportedArgument = V == "desc" || V == "trad";
+      EnableTLSDESC = V == "desc";
+    } else if (Triple.isX86()) {
+      SupportedArgument = V == "gnu";
+    } else {
+      Unsupported = true;
+    }
+    if (Unsupported) {
+      D.Diag(diag::err_drv_unsupported_opt_for_target)
+          << A->getSpelling() << TripleStr;
+    } else if (!SupportedArgument) {
+      D.Diag(diag::err_drv_unsupported_option_argument_for_target)
+          << A->getSpelling() << V << TripleStr;
+    } else if (EnableTLSDESC) {
+      CmdArgs.push_back("-enable-tlsdesc");
+    }
+  }
+
   // Add the target cpu
   std::string CPU = getCPUName(D, Args, Triple, /*FromAs*/ false);
   if (!CPU.empty()) {
diff --git a/clang/test/CodeGen/RISCV/tls-dialect.c b/clang/test/CodeGen/RISCV/tls-dialect.c
new file mode 100644
index 000000000000000..5a1bfa539321269
--- /dev/null
+++ b/clang/test/CodeGen/RISCV/tls-dialect.c
@@ -0,0 +1,13 @@
+/// cc1 -enable-tlsdesc (due to -mtls-dialect=desc) enables TLSDESC.
+// RUN: %clang_cc1 -triple riscv64 -S -mrelocation-model pic -pic-level 1 -enable-tlsdesc %s -o - | FileCheck %s --check-prefix=DESC
+// RUN: %clang_cc1 -triple riscv64 -S -mrelocation-model pic -pic-level 1 %s -o - | FileCheck %s --check-prefix=NODESC
+
+__thread int x;
+
+// DESC:       %tlsdesc_hi
+// DESC-NOT:   %tls_gd_pcrel_hi
+// NODESC:     %tls_gd_pcrel_hi
+// NODESC-NOT: %tlsdesc_hi
+int use() {
+  return x;
+}
diff --git a/clang/test/Driver/tls-dialect.c b/clang/test/Driver/tls-dialect.c
new file mode 100644
index 000000000000000..d90200998e3b06b
--- /dev/null
+++ b/clang/test/Driver/tls-dialect.c
@@ -0,0 +1,19 @@
+// RUN: %clang -### --target=riscv64-freebsd -mtls-dialect=desc %s 2>&1 | FileCheck --check-prefix=DESC %s
+// RUN: %clang -### --target=riscv64-linux -mtls-dialect=trad %s 2>&1 | FileCheck --check-prefix=NODESC %s
+// RUN: %clang -### --target=riscv64-linux %s 2>&1 | FileCheck --check-prefix=NODESC %s
+// RUN: %clang -### --target=x86_64-linux -mtls-dialect=gnu %s 2>&1 | FileCheck --check-prefix=NODESC %s
+
+/// Unsupported target
+/// GCC supports -mtls-dialect= for AArch64, but we just unsupport it for AArch64 as it is very rarely used.
+// RUN: not %clang --target=aarch64-linux -mtls-dialect=desc %s 2>&1 | FileCheck --check-prefix=UNSUPPORTED-TARGET %s
+// RUN: not %clang --target=x86_64-apple-macos -mtls-dialect=desc %s 2>&1 | FileCheck -check-prefix=UNSUPPORTED-TARGET %s
+
+/// Unsupported argument
+// RUN: not %clang -### --target=riscv64-linux -mtls-dialect=gnu2 %s 2>&1 | FileCheck --check-prefix=UNSUPPORTED-ARG %s
+// RUN: not %clang -### --target=x86_64-linux -mtls-dialect=gnu2 %s 2>&1 | FileCheck --check-prefix=UNSUPPORTED-ARG %s
+
+// DESC: "-cc1" {{.*}}"-enable-tlsdesc"
+// NODESC-NOT: "-enable-tlsdesc"
+
+// UNSUPPORTED-TARGET: error: unsupported option '-mtls-dialect=' for target
+// UNSUPPORTED-ARG: error: unsupported argument 'gnu2' to option '-mtls-dialect=' for target

@efriedma-quic
Copy link
Collaborator

What's the expected interaction here with LTO? Modifying TargetOptions has no effect if we're generating a bitcode file.

@MaskRay
Copy link
Member Author

MaskRay commented Jan 25, 2024

What's the expected interaction here with LTO? Modifying TargetOptions has no effect if we're generating a bitcode file.

Thanks for reminding me this. There seems no motivation to mix trad/desc (so modules flag metadata or function attribute would be overkill), so I think we can do it the -ffunction-sections/-femulated-tls way by passing an option to the linker...

Created using spr 1.3.4
Copy link

github-actions bot commented Jan 25, 2024

✅ With the latest revision this PR passed the C/C++ code formatter.

Created using spr 1.3.4
Copy link
Contributor

@ilovepi ilovepi left a comment

Choose a reason for hiding this comment

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

LGTM

Created using spr 1.3.4
@MaskRay MaskRay merged commit 36b4a9c into main Jan 26, 2024
3 of 4 checks passed
@MaskRay MaskRay deleted the users/MaskRay/spr/drivercodegen-support-mtls-dialect branch January 26, 2024 17:25
llvmbot pushed a commit to llvmbot/llvm-project that referenced this pull request Jan 26, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
llvmbot pushed a commit to llvmbot/llvm-project that referenced this pull request Jan 26, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
tstellar pushed a commit that referenced this pull request Jan 27, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see #66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
tstellar pushed a commit to tstellar/llvm-project that referenced this pull request Feb 14, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
tstellar pushed a commit to tstellar/llvm-project that referenced this pull request Feb 14, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
tstellar pushed a commit to tstellar/llvm-project that referenced this pull request Feb 14, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
tstellar pushed a commit to tstellar/llvm-project that referenced this pull request Feb 14, 2024
GCC supports -mtls-dialect= for several architectures to select TLSDESC.
This patch supports the following values

* x86: "gnu". "gnu2" (TLSDESC) is not supported yet.
* RISC-V: "trad" (general dynamic), "desc" (TLSDESC, see llvm#66915)

AArch64 toolchains seem to support TLSDESC from the beginning, and the
general dynamic model has poor support. Nobody seems to use the option
-mtls-dialect= at all, so we don't bother with it.
There also seems very little interest in AArch32's TLSDESC support.

TLSDESC does not change IR, but affects object file generation. Without
a backend option the option is a no-op for in-process ThinLTO.

There seems no motivation to have fine-grained control mixing trad/desc
for TLS, so we just pass -mllvm, and don't bother with a modules flag
metadata or function attribute.

Co-authored-by: Paul Kirth <[email protected]>
(cherry picked from commit 36b4a9c)
@pointhex pointhex mentioned this pull request May 7, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
clang:codegen clang:driver 'clang' and 'clang++' user-facing binaries. Not 'clang-cl' clang:frontend Language frontend issues, e.g. anything involving "Sema" clang Clang issues not falling into any other category
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants