Implement the local-dynamic TLS model for x86 (PR3985)

This implements codegen support for accesses to thread-local variables
using the local-dynamic model, and adds a clean-up pass so that the base
address for the TLS block can be re-used between local-dynamic access on
an execution path.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@157818 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 79 insertions, 0 deletions

diff --git a/test/CodeGen/X86/tls-local-dynamic.ll b/test/CodeGen/X86/tls-local-dynamic.ll
new file mode 100644
index 0000000000..c5fd16bbec
--- /dev/null
+++ b/test/CodeGen/X86/tls-local-dynamic.ll
@@ -0,0 +1,59 @@
+; RUN: llc < %s -march=x86-64 -mtriple=x86_64-linux-gnu -relocation-model=pic | FileCheck  %s
+
+@x = internal thread_local global i32 0, align 4
+@y = internal thread_local global i32 0, align 4
+
+; get_x and get_y are here to prevent x and y to be optimized away as 0
+
+define i32* @get_x() {
+entry:
+  ret i32* @x
+; FIXME: This function uses a single thread-local variable,
+; so we might want to fall back to general-dynamic here.
+; CHECK:       get_x:
+; CHECK:       leaq x@TLSLD(%rip), %rdi
+; CHECK-NEXT:  callq __tls_get_addr@PLT
+; CHECK:       x@DTPOFF
+}
+
+define i32* @get_y() {
+entry:
+  ret i32* @y
+}
+
+define i32 @f(i32 %i) {
+entry:
+  %cmp = icmp eq i32 %i, 1
+  br i1 %cmp, label %return, label %if.else
+; This bb does not access TLS, so should not call __tls_get_addr.
+; CHECK:       f:
+; CHECK-NOT:   __tls_get_addr
+; CHECK:       je
+
+
+if.else:
+  %0 = load i32* @x, align 4
+  %cmp1 = icmp eq i32 %i, 2
+  br i1 %cmp1, label %if.then2, label %return
+; Now we call __tls_get_addr.
+; CHECK:       # %if.else
+; CHECK:       leaq x@TLSLD(%rip), %rdi
+; CHECK-NEXT:  callq __tls_get_addr@PLT
+; CHECK:       x@DTPOFF
+
+
+if.then2:
+  %1 = load i32* @y, align 4
+  %add = add nsw i32 %1, %0
+  br label %return
+; This accesses TLS, but is dominated by the previous block,
+; so should not have to call __tls_get_addr again.
+; CHECK:       # %if.then2
+; CHECK-NOT:   __tls_get_addr
+; CHECK:       y@DTPOFF
+
+
+return:
+  %retval.0 = phi i32 [ %add, %if.then2 ], [ 5, %entry ], [ %0, %if.else ]
+  ret i32 %retval.0
+}
diff --git a/test/CodeGen/X86/tls-pic.ll b/test/CodeGen/X86/tls-pic.ll
index b83416d4b3..c3e8216725 100644
--- a/test/CodeGen/X86/tls-pic.ll
+++ b/test/CodeGen/X86/tls-pic.ll
@@ -2,6 +2,8 @@
 ; RUN: llc < %s -march=x86-64 -mtriple=x86_64-linux-gnu -relocation-model=pic | FileCheck -check-prefix=X64 %s
 
 @i = thread_local global i32 15
+@j = internal thread_local global i32 42
+@k = internal thread_local global i32 42
 
 define i32 @f1() {
 entry:
@@ -64,4 +66,22 @@ entry:
 ; X64:   callq __tls_get_addr@PLT
 
 
+define i32 @f5() nounwind {
+entry:
+	%0 = load i32* @j, align 4
+	%1 = load i32* @k, align 4
+	%add = add nsw i32 %0, %1
+	ret i32 %add
+}
 
+; X32:    f5:
+; X32:      leal {{[jk]}}@TLSLDM
+; X32-NEXT: calll ___tls_get_addr@PLT
+; X32-NEXT: movl {{[jk]}}@DTPOFF(%eax)
+; X32-NEXT: addl {{[jk]}}@DTPOFF(%eax)
+
+; X64:    f5:
+; X64:      leaq {{[jk]}}@TLSLD(%rip), %rdi
+; X64-NEXT: callq	__tls_get_addr@PLT
+; X64-NEXT: movl {{[jk]}}@DTPOFF(%rax)
+; X64-NEXT: addl {{[jk]}}@DTPOFF(%rax)