Add a DAGCombine to turn (ctpop x) u< 2 into (x & x-1) == 0.

This shaves off 4 popcounts from the hacked 186.crafty source.

This is enabled even when a native popcount instruction is available. The
combined code is one operation longer but it should be faster nevertheless.

llvm-svn: 123621

1 files changed, 31 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/ctpop-combine.ll b/llvm/test/CodeGen/X86/ctpop-combine.ll
new file mode 100644
index 00000000000..b87637f9424
--- /dev/null
+++ b/llvm/test/CodeGen/X86/ctpop-combine.ll
@@ -0,0 +1,31 @@
+; RUN: llc -march=x86-64 < %s | FileCheck %s
+
+declare i64 @llvm.ctpop.i64(i64) nounwind readnone
+
+define i32 @test1(i64 %x) nounwind readnone {
+  %count = tail call i64 @llvm.ctpop.i64(i64 %x)
+  %cast = trunc i64 %count to i32
+  %cmp = icmp ugt i32 %cast, 1
+  %conv = zext i1 %cmp to i32
+  ret i32 %conv
+; CHECK: test1:
+; CHECK: leaq -1(%rdi)
+; CHECK-NEXT: testq
+; CHECK-NEXT: setne
+; CHECK: ret
+}
+
+
+define i32 @test2(i64 %x) nounwind readnone {
+  %count = tail call i64 @llvm.ctpop.i64(i64 %x)
+  %cast = trunc i64 %count to i32
+  %cmp = icmp ult i32 %cast, 2
+  %conv = zext i1 %cmp to i32
+  ret i32 %conv
+; CHECK: test2:
+; CHECK: leaq -1(%rdi)
+; CHECK-NEXT: testq
+; CHECK-NEXT: sete
+; CHECK: ret
+}
+