From 8be316bf232d03f387df9093fb71fa1d6d392a88 Mon Sep 17 00:00:00 2001
From: Nemanja Ivanovic <nemanja.i.ibm@gmail.com>
Date: Sun, 5 Jul 2015 06:03:51 +0000
Subject: [PATCH] Add missing builtins to the PPC back end for ABI compliance
 (vol. 2)

This patch corresponds to review:
http://reviews.llvm.org/D10874

Back end portion of the second round of additions to altivec.h.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241398 91177308-0d34-0410-b5e6-96231b3b80d8
---
 include/llvm/IR/IntrinsicsPowerPC.td         |  6 ++++
 lib/Target/PowerPC/PPCInstrVSX.td            |  6 ++++
 test/CodeGen/PowerPC/vsx-elementary-arith.ll | 31 ++++++++++++++++++++
 3 files changed, 43 insertions(+)

diff --git a/include/llvm/IR/IntrinsicsPowerPC.td b/include/llvm/IR/IntrinsicsPowerPC.td
index 05adc5a757b..630f89a0bdd 100644
--- a/include/llvm/IR/IntrinsicsPowerPC.td
+++ b/include/llvm/IR/IntrinsicsPowerPC.td
@@ -694,6 +694,12 @@ def int_ppc_vsx_xvrspip :
 def int_ppc_vsx_xvrdpip :
       Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty], [IntrNoMem]>;
 
+// Vector rsqrte
+def int_ppc_vsx_xvrsqrtesp : GCCBuiltin<"__builtin_vsx_xvrsqrtesp">,
+      Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
+def int_ppc_vsx_xvrsqrtedp : GCCBuiltin<"__builtin_vsx_xvrsqrtedp">,
+      Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty], [IntrNoMem]>;
+
 // Vector compare
 def int_ppc_vsx_xvcmpeqdp :
       PowerPC_VSX_Intrinsic<"xvcmpeqdp", [llvm_v2i64_ty],
diff --git a/lib/Target/PowerPC/PPCInstrVSX.td b/lib/Target/PowerPC/PPCInstrVSX.td
index 43ba4994fde..3cb67deacbf 100644
--- a/lib/Target/PowerPC/PPCInstrVSX.td
+++ b/lib/Target/PowerPC/PPCInstrVSX.td
@@ -989,6 +989,12 @@ def : Pat<(int_ppc_vsx_xvdivsp v4f32:$A, v4f32:$B),
 def : Pat<(int_ppc_vsx_xvdivdp v2f64:$A, v2f64:$B),
           (XVDIVDP $A, $B)>;
 
+// Recip. square root estimate
+def : Pat<(int_ppc_vsx_xvrsqrtesp v4f32:$A),
+          (XVRSQRTESP $A)>;
+def : Pat<(int_ppc_vsx_xvrsqrtedp v2f64:$A),
+          (XVRSQRTEDP $A)>;
+
 } // AddedComplexity
 } // HasVSX
 
diff --git a/test/CodeGen/PowerPC/vsx-elementary-arith.ll b/test/CodeGen/PowerPC/vsx-elementary-arith.ll
index d8f76bb989e..5416f667aef 100644
--- a/test/CodeGen/PowerPC/vsx-elementary-arith.ll
+++ b/test/CodeGen/PowerPC/vsx-elementary-arith.ll
@@ -116,5 +116,36 @@ entry:
 ; CHECK: xssqrtdp {{[0-9]+}}
 }
 
+; Vector forms
+; Function Attrs: nounwind
+define <4 x float> @emit_xvrsqrtesp() {
+entry:
+; CHECK-LABEL: @emit_xvrsqrtesp
+  %vf = alloca <4 x float>, align 16
+  %vfr = alloca <4 x float>, align 16
+  %0 = load <4 x float>, <4 x float>* %vf, align 16
+  %call = call <4 x float> @llvm.ppc.vsx.xvrsqrtesp(<4 x float> %0)
+; CHECK: xvrsqrtesp {{[0-9]+}}, {{[0-9]+}}
+  ret <4 x float> %call
+}
+
+; Function Attrs: nounwind
+define <2 x double> @emit_xvrsqrtedp() {
+entry:
+; CHECK-LABEL: @emit_xvrsqrtedp
+  %vd = alloca <2 x double>, align 16
+  %vdr = alloca <2 x double>, align 16
+  %0 = load <2 x double>, <2 x double>* %vd, align 16
+  %call = call <2 x double> @llvm.ppc.vsx.xvrsqrtedp(<2 x double> %0)
+  ret <2 x double> %call
+; CHECK: xvrsqrtedp {{[0-9]+}}, {{[0-9]+}}
+}
+
 ; Function Attrs: nounwind
 declare double @sqrt(double)
+
+; Function Attrs: nounwind readnone
+declare <4 x float> @llvm.ppc.vsx.xvrsqrtesp(<4 x float>)
+
+; Function Attrs: nounwind readnone
+declare <2 x double> @llvm.ppc.vsx.xvrsqrtedp(<2 x double>)