summaryrefslogtreecommitdiff
path: root/test/CodeGen/X86/atom-call-reg-indirect.ll
diff options
context:
space:
mode:
authorPreston Gurd <preston.gurd@intel.com>2013-03-27 19:14:02 +0000
committerPreston Gurd <preston.gurd@intel.com>2013-03-27 19:14:02 +0000
commit1edadea42f6f5c393b4fdb9d7ce1cf7eb9c24ab4 (patch)
tree0703e20d41246fa36a72779d0d1ba5b58b6ee2d7 /test/CodeGen/X86/atom-call-reg-indirect.ll
parente915047fed99221afb8c540d8a7e81038a6483f1 (diff)
downloadllvm-1edadea42f6f5c393b4fdb9d7ce1cf7eb9c24ab4.tar.gz
llvm-1edadea42f6f5c393b4fdb9d7ce1cf7eb9c24ab4.tar.bz2
llvm-1edadea42f6f5c393b4fdb9d7ce1cf7eb9c24ab4.tar.xz
For the current Atom processor, the fastest way to handle a call
indirect through a memory address is to load the memory address into a register and then call indirect through the register. This patch implements this improvement by modifying SelectionDAG to force a function address which is a memory reference to be loaded into a virtual register. Patch by Sriram Murali. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178171 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/atom-call-reg-indirect.ll')
-rw-r--r--test/CodeGen/X86/atom-call-reg-indirect.ll45
1 files changed, 45 insertions, 0 deletions
diff --git a/test/CodeGen/X86/atom-call-reg-indirect.ll b/test/CodeGen/X86/atom-call-reg-indirect.ll
new file mode 100644
index 0000000000..632781130d
--- /dev/null
+++ b/test/CodeGen/X86/atom-call-reg-indirect.ll
@@ -0,0 +1,45 @@
+; RUN: llc < %s -mcpu=atom -mtriple=i686-linux | FileCheck -check-prefix=ATOM32 %s
+; RUN: llc < %s -mcpu=core2 -mtriple=i686-linux | FileCheck -check-prefix=ATOM-NOT32 %s
+; RUN: llc < %s -mcpu=atom -mtriple=x86_64-linux | FileCheck -check-prefix=ATOM64 %s
+; RUN: llc < %s -mcpu=core2 -mtriple=x86_64-linux | FileCheck -check-prefix=ATOM-NOT64 %s
+
+
+; fn_ptr.ll
+%class.A = type { i32 (...)** }
+
+define i32 @test1() #0 {
+ ;ATOM: test1
+entry:
+ %call = tail call %class.A* @_Z3facv()
+ %0 = bitcast %class.A* %call to void (%class.A*)***
+ %vtable = load void (%class.A*)*** %0, align 8
+ %1 = load void (%class.A*)** %vtable, align 8
+ ;ATOM32: movl (%ecx), %ecx
+ ;ATOM32: calll *%ecx
+ ;ATOM-NOT32: calll *(%ecx)
+ ;ATOM64: movq (%rcx), %rcx
+ ;ATOM64: callq *%rcx
+ ;ATOM-NOT64: callq *(%rcx)
+ tail call void %1(%class.A* %call)
+ ret i32 0
+}
+
+declare %class.A* @_Z3facv() #1
+
+; virt_fn.ll
+@p = external global void (i32)**
+
+define i32 @test2() #0 {
+ ;ATOM: test2
+entry:
+ %0 = load void (i32)*** @p, align 8
+ %1 = load void (i32)** %0, align 8
+ ;ATOM32: movl (%eax), %eax
+ ;ATOM32: calll *%eax
+ ;ATOM-NOT: calll *(%eax)
+ ;ATOM64: movq (%rax), %rax
+ ;ATOM64: callq *%rax
+ ;ATOM-NOT64: callq *(%rax)
+ tail call void %1(i32 2)
+ ret i32 0
+}