Add calling convention for C++ TLS on Darwin

[manmanren]

Summary

This patch adds a new target-independent calling convention for C++ TLS access functions.

The target-specific implementation for X86-64 is defined as following:

• Arguments are passed as for the default C calling convention
• The same applies for the return value(s)
• the callee preserves all GPRs - except RAX and RDI

It attempts to make the code in the caller as little intrusive as possible. The access function makes C-style TLS function calls in the entry and exit block, C-style TLS functions save a lot more registers than normal calls. The added calling convention ties into the existing implementation of the C-style TLS functions, so we can't simply use existing calling conventions such as preserve_mostcc.

Cheers,
Manman

[qcolombet]

Hi Manman,

I think you want to add the LLVM list as a subscriber.

Cheers,
-Quentin

[manmanren]

ping

Cheers,
Manman

[silvas]

Small question: how Darwin-specific is this? Should it be darwin_cxx_tlscc or do you expect other platforms to use it?

[manmanren]

In D14780#299119, @silvas wrote:

Small question: how Darwin-specific is this? Should it be darwin_cxx_tlscc or do you expect other platforms to use it?

This is currently specific to Darwin. I don't see another example of putting the platform in the name of a calling convention. But I can be convinced :]

[lhames]

Perhaps a less generic name would be good, since not all platforms will use cxx_tlscc for C++ TLS. Maybe 'darwin_cxx_tlscc' as Sean suggested, or 'cxx_fast_tlscc'?

Otherwise this looks good to me. :)

docs/LangRef.rst

410–411

I think this should be "...attempts to make the code as non-intrusive as possible." or "This calling convention aims to minimize overhead in the caller by preserving as many registers as possible".

[manmanren]

In D14780#301744, @lhames wrote:

Perhaps a less generic name would be good, since not all platforms will use cxx_tlscc for C++ TLS. Maybe 'darwin_cxx_tlscc' as Sean suggested, or 'cxx_fast_tlscc'?

Thanks for reviewing! I changed the name to cxx_fast_tlscc as suggested and committed r254737.

Cheer,
Manman

Otherwise this looks good to me. :)

[qcolombet]

This landed in r254737.
Closing.

docs/BitCodeFormat.rst

 * *callingconv*: The calling convention number:
   * ``ccc``: code 0
   * ``fastcc``: code 8
   * ``coldcc``: code 9
   * ``webkit_jscc``: code 12
   * ``anyregcc``: code 13
   * ``preserve_mostcc``: code 14
   * ``preserve_allcc``: code 15
+  * ``cxx_tlscc``: code 17
   * ``x86_stdcallcc``: code 64
   * ``x86_fastcallcc``: code 65
   * ``arm_apcscc``: code 66
   * ``arm_aapcscc``: code 67
   * ``arm_aapcs_vfpcc``: code 68
 
 * isproto*: Non-zero if this entry represents a declaration rather than a
   definition

docs/LangRef.rst

- On X86-64 the callee preserves all general purpose registers, except for
       all floating-point registers (XMMs/YMMs).
 
     The idea behind this convention is to support calls to runtime functions
     that don't need to call out to any other functions.
 
     This calling convention, like the `PreserveMost` calling convention, will be
     used by a future version of the ObjectiveC runtime and should be considered
     experimental at this time.
+"``cxx_tlscc``" - The `CXX_TLS` calling convention for access functions
+    This calling convention attempts to make the code in the caller as little
+    intrusive as possible. This calling convention behaves identical to the `C`
+    calling convention on how arguments and return values are passed, but it
+    uses a different set of caller/callee-saved registers. Given that C-style
+    TLS on Darwin has its own special CSRs, we can't use the existing
+    `PreserveMost`.
+
+    - On X86-64 the callee preserves all general purpose registers, except for
+      RDI and RAX.
 "``cc <n>``" - Numbered convention
     Any calling convention may be specified by number, allowing
     target-specific calling conventions to be used. Target specific
     calling conventions start at 64.
 
 More calling conventions can be added/defined on an as-needed basis, to
 support Pascal conventions or any other well-known target-independent
 convention.

include/llvm/IR/CallingConv.h

enum {
 
     // PreserveAll - Calling convention for runtime calls that preserves
     // (almost) all registers.
     PreserveAll = 15,
 
     // Swift - Calling convention for Swift.
     Swift = 16,
 
+    // C++ TLS - Calling convention for access functions.
+    CXX_TLS = 17,
+
     // Target - This is the start of the target-specific calling conventions,
     // e.g. fastcall and thiscall on X86.
     FirstTargetCC = 64,
 
     /// X86_StdCall - stdcall is the calling conventions mostly used by the
     /// Win32 API. It is basically the same as the C convention with the
     /// difference in that the callee is responsible for popping the arguments
     /// from the stack.

lib/AsmParser/LLLexer.cpp

#define KEYWORD(STR)                                                           \
   KEYWORD(x86_64_win64cc);
   KEYWORD(webkit_jscc);
   KEYWORD(anyregcc);
   KEYWORD(preserve_mostcc);
   KEYWORD(preserve_allcc);
   KEYWORD(ghccc);
   KEYWORD(hhvmcc);
   KEYWORD(hhvm_ccc);
+  KEYWORD(cxx_tlscc);
 
   KEYWORD(cc);
   KEYWORD(c);
 
   KEYWORD(attributes);
 
   KEYWORD(alwaysinline);
   KEYWORD(argmemonly);

lib/AsmParser/LLParser.cpp

 ///   ::= 'x86_64_win64cc'
 ///   ::= 'webkit_jscc'
 ///   ::= 'anyregcc'
 ///   ::= 'preserve_mostcc'
 ///   ::= 'preserve_allcc'
 ///   ::= 'ghccc'
 ///   ::= 'hhvmcc'
 ///   ::= 'hhvm_ccc'
+///   ::= 'cxx_tlscc'
 ///   ::= 'cc' UINT
 ///
 bool LLParser::ParseOptionalCallingConv(unsigned &CC) {
   switch (Lex.getKind()) {
   default:                       CC = CallingConv::C; return false;
   case lltok::kw_ccc:            CC = CallingConv::C; break;
   case lltok::kw_fastcc:         CC = CallingConv::Fast; break;
   case lltok::kw_coldcc:         CC = CallingConv::Cold; break;
bool LLParser::ParseOptionalCallingConv(unsigned &CC) {
   case lltok::kw_x86_64_win64cc: CC = CallingConv::X86_64_Win64; break;
   case lltok::kw_webkit_jscc:    CC = CallingConv::WebKit_JS; break;
   case lltok::kw_anyregcc:       CC = CallingConv::AnyReg; break;
   case lltok::kw_preserve_mostcc:CC = CallingConv::PreserveMost; break;
   case lltok::kw_preserve_allcc: CC = CallingConv::PreserveAll; break;
   case lltok::kw_ghccc:          CC = CallingConv::GHC; break;
   case lltok::kw_hhvmcc:         CC = CallingConv::HHVM; break;
   case lltok::kw_hhvm_ccc:       CC = CallingConv::HHVM_C; break;
+  case lltok::kw_cxx_tlscc:      CC = CallingConv::CXX_TLS; break;
   case lltok::kw_cc: {
       Lex.Lex();
       return ParseUInt32(CC);
     }
   }
 
   Lex.Lex();
   return false;

lib/AsmParser/LLToken.h

enum Kind {
     kw_msp430_intrcc,
     kw_ptx_kernel, kw_ptx_device,
     kw_spir_kernel, kw_spir_func,
     kw_x86_64_sysvcc, kw_x86_64_win64cc,
     kw_webkit_jscc, kw_anyregcc,
     kw_preserve_mostcc, kw_preserve_allcc,
     kw_ghccc,
     kw_hhvmcc, kw_hhvm_ccc,
+    kw_cxx_tlscc,
 
     // Attributes:
     kw_attributes,
     kw_alwaysinline,
     kw_argmemonly,
     kw_sanitize_address,
     kw_builtin,
     kw_byval,

lib/IR/AsmWriter.cpp

static void PrintCallingConv(unsigned cc, raw_ostream &Out) {
   switch (cc) {
   default:                         Out << "cc" << cc; break;
   case CallingConv::Fast:          Out << "fastcc"; break;
   case CallingConv::Cold:          Out << "coldcc"; break;
   case CallingConv::WebKit_JS:     Out << "webkit_jscc"; break;
   case CallingConv::AnyReg:        Out << "anyregcc"; break;
   case CallingConv::PreserveMost:  Out << "preserve_mostcc"; break;
   case CallingConv::PreserveAll:   Out << "preserve_allcc"; break;
+  case CallingConv::CXX_TLS:       Out << "cxx_tlscc"; break;
   case CallingConv::GHC:           Out << "ghccc"; break;
   case CallingConv::X86_StdCall:   Out << "x86_stdcallcc"; break;
   case CallingConv::X86_FastCall:  Out << "x86_fastcallcc"; break;
   case CallingConv::X86_ThisCall:  Out << "x86_thiscallcc"; break;
   case CallingConv::X86_VectorCall:Out << "x86_vectorcallcc"; break;
   case CallingConv::Intel_OCL_BI:  Out << "intel_ocl_bicc"; break;
   case CallingConv::ARM_APCS:      Out << "arm_apcscc"; break;
   case CallingConv::ARM_AAPCS:     Out << "arm_aapcscc"; break;

lib/Target/X86/X86RegisterInfo.cpp

if (HasAVX)
       return CSR_64_AllRegs_AVX_SaveList;
     return CSR_64_AllRegs_SaveList;
   case CallingConv::PreserveMost:
     return CSR_64_RT_MostRegs_SaveList;
   case CallingConv::PreserveAll:
     if (HasAVX)
       return CSR_64_RT_AllRegs_AVX_SaveList;
     return CSR_64_RT_AllRegs_SaveList;
+  case CallingConv::CXX_TLS:
+    if (Is64Bit)
+      return CSR_64_TLS_Darwin_SaveList;
+    break;
   case CallingConv::Intel_OCL_BI: {
     if (HasAVX512 && IsWin64)
       return CSR_Win64_Intel_OCL_BI_AVX512_SaveList;
     if (HasAVX512 && Is64Bit)
       return CSR_64_Intel_OCL_BI_AVX512_SaveList;
     if (HasAVX && IsWin64)
       return CSR_Win64_Intel_OCL_BI_AVX_SaveList;
     if (HasAVX && Is64Bit)
if (HasAVX)
       return CSR_64_AllRegs_AVX_RegMask;
     return CSR_64_AllRegs_RegMask;
   case CallingConv::PreserveMost:
     return CSR_64_RT_MostRegs_RegMask;
   case CallingConv::PreserveAll:
     if (HasAVX)
       return CSR_64_RT_AllRegs_AVX_RegMask;
     return CSR_64_RT_AllRegs_RegMask;
+  case CallingConv::CXX_TLS:
+    if (Is64Bit)
+      return CSR_64_TLS_Darwin_RegMask;
+    break;
   case CallingConv::Intel_OCL_BI: {
     if (HasAVX512 && IsWin64)
       return CSR_Win64_Intel_OCL_BI_AVX512_RegMask;
     if (HasAVX512 && Is64Bit)
       return CSR_64_Intel_OCL_BI_AVX512_RegMask;
     if (HasAVX && IsWin64)
       return CSR_Win64_Intel_OCL_BI_AVX_RegMask;
     if (HasAVX && Is64Bit)

test/CodeGen/X86/cxx_tlscc64.ll

+; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -enable-shrink-wrap=true | FileCheck --check-prefix=SHRINK %s
+%struct.S = type { i8 }
+
+@sg = internal thread_local global %struct.S zeroinitializer, align 1
+@__dso_handle = external global i8
+@__tls_guard = internal thread_local unnamed_addr global i1 false
+
+declare void @_ZN1SC1Ev(%struct.S*)
+declare void @_ZN1SD1Ev(%struct.S*)
+declare i32 @_tlv_atexit(void (i8*)*, i8*, i8*)
+
+; Every GPR should be saved - except rdi, rax, and rsp
+; CHECK-LABEL: _ZTW2sg
+; CHECK: pushq %r11
+; CHECK: pushq %r10
+; CHECK: pushq %r9
+; CHECK: pushq %r8
+; CHECK: pushq %rsi
+; CHECK: pushq %rdx
+; CHECK: pushq %rcx
+; CHECK: pushq %rbx
+; CHECK: callq
+; CHECK: jne
+; CHECK: callq
+; CHECK: tlv_atexit
+; CHECK: callq
+; CHECK: popq %rbx
+; CHECK: popq %rcx
+; CHECK: popq %rdx
+; CHECK: popq %rsi
+; CHECK: popq %r8
+; CHECK: popq %r9
+; CHECK: popq %r10
+; CHECK: popq %r11
+; SHRINK-LABEL: _ZTW2sg
+; SHRINK: callq
+; SHRINK: jne
+; SHRINK: pushq %r11
+; SHRINK: pushq %r10
+; SHRINK: pushq %r9
+; SHRINK: pushq %r8
+; SHRINK: pushq %rsi
+; SHRINK: pushq %rdx
+; SHRINK: pushq %rcx
+; SHRINK: pushq %rbx
+; SHRINK: callq
+; SHRINK: tlv_atexit
+; SHRINK: popq %rbx
+; SHRINK: popq %rcx
+; SHRINK: popq %rdx
+; SHRINK: popq %rsi
+; SHRINK: popq %r8
+; SHRINK: popq %r9
+; SHRINK: popq %r10
+; SHRINK: popq %r11
+; SHRINK: LBB{{.*}}:
+; SHRINK: callq
+define cxx_tlscc nonnull %struct.S* @_ZTW2sg() {
+  %.b.i = load i1, i1* @__tls_guard, align 1
+  br i1 %.b.i, label %__tls_init.exit, label %init.i
+
+init.i:
+  store i1 true, i1* @__tls_guard, align 1
+  tail call void @_ZN1SC1Ev(%struct.S* nonnull @sg) #2
+  %1 = tail call i32 @_tlv_atexit(void (i8*)* nonnull bitcast (void (%struct.S*)* @_ZN1SD1Ev to void (i8*)*), i8* nonnull getelementptr inbounds (%struct.S, %struct.S* @sg, i64 0, i32 0), i8* nonnull @__dso_handle) #2
+  br label %__tls_init.exit
+
+__tls_init.exit:
+  ret %struct.S* @sg
+}