[llvm] r221983 - [PowerPC] Add VSX builtins for vec_div

[Bill Schmidt]

Author: wschmidt
Date: Fri Nov 14 06:10:40 2014
New Revision: 221983

URL: http://llvm.org/viewvc/llvm-project?rev=221983&view=rev
Log:
[PowerPC] Add VSX builtins for vec_div

This patch adds builtin support for xvdivdp and xvdivsp, along with a
test case.  Straightforward stuff.

There's a companion patch for Clang.

Added:
    llvm/trunk/test/CodeGen/PowerPC/vsx-div.ll
Modified:
    llvm/trunk/include/llvm/IR/IntrinsicsPowerPC.td
    llvm/trunk/lib/Target/PowerPC/PPCInstrVSX.td

Modified: llvm/trunk/include/llvm/IR/IntrinsicsPowerPC.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/IR/IntrinsicsPowerPC.td?rev=221983&r1=221982&r2=221983&view=diff
==============================================================================
--- llvm/trunk/include/llvm/IR/IntrinsicsPowerPC.td (original)
+++ llvm/trunk/include/llvm/IR/IntrinsicsPowerPC.td Fri Nov 14 06:10:40 2014
@@ -528,14 +528,17 @@ def int_ppc_vsx_stxvw4x :
 def int_ppc_vsx_stxvd2x :
       Intrinsic<[], [llvm_v2f64_ty, llvm_ptr_ty], [IntrReadWriteArgMem]>;
 
-// Vector maximum.
+// Vector and scalar maximum.
 def int_ppc_vsx_xvmaxdp : PowerPC_VSX_Vec_DDD_Intrinsic<"xvmaxdp">;
 def int_ppc_vsx_xvmaxsp : PowerPC_VSX_Vec_FFF_Intrinsic<"xvmaxsp">;
 def int_ppc_vsx_xsmaxdp : PowerPC_VSX_Sca_DDD_Intrinsic<"xsmaxdp">;
 
-// Vector minimum.
+// Vector and scalar minimum.
 def int_ppc_vsx_xvmindp : PowerPC_VSX_Vec_DDD_Intrinsic<"xvmindp">;
 def int_ppc_vsx_xvminsp : PowerPC_VSX_Vec_FFF_Intrinsic<"xvminsp">;
 def int_ppc_vsx_xsmindp : PowerPC_VSX_Sca_DDD_Intrinsic<"xsmindp">;
 
+// Vector divide.
+def int_ppc_vsx_xvdivdp : PowerPC_VSX_Vec_DDD_Intrinsic<"xvdivdp">;
+def int_ppc_vsx_xvdivsp : PowerPC_VSX_Vec_FFF_Intrinsic<"xvdivsp">;
 }

Modified: llvm/trunk/lib/Target/PowerPC/PPCInstrVSX.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCInstrVSX.td?rev=221983&r1=221982&r2=221983&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCInstrVSX.td (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCInstrVSX.td Fri Nov 14 06:10:40 2014
@@ -887,6 +887,12 @@ def : Pat<(f64 (selectcc i1:$lhs, i1:$rh
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETNE)),
           (SELECT_VSFRC (CRXOR $lhs, $rhs), $tval, $fval)>;
 
+// Divides.
+def : Pat<(int_ppc_vsx_xvdivsp v4f32:$A, v4f32:$B),
+          (XVDIVSP $A, $B)>;
+def : Pat<(int_ppc_vsx_xvdivdp v2f64:$A, v2f64:$B),
+          (XVDIVDP $A, $B)>;
+
 } // AddedComplexity
 } // HasVSX
 

Added: llvm/trunk/test/CodeGen/PowerPC/vsx-div.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/vsx-div.ll?rev=221983&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/vsx-div.ll (added)
+++ llvm/trunk/test/CodeGen/PowerPC/vsx-div.ll Fri Nov 14 06:10:40 2014
@@ -0,0 +1,29 @@
+; RUN: llc -mcpu=pwr7 -mattr=+vsx -O1 -mtriple=powerpc64-unknown-linux-gnu < %s | FileCheck %s
+
+ at vf = global <4 x float> <float -1.500000e+00, float 2.500000e+00, float -3.500000e+00, float 4.500000e+00>, align 16
+ at vd = global <2 x double> <double 3.500000e+00, double -7.500000e+00>, align 16
+ at vf_res = common global <4 x float> zeroinitializer, align 16
+ at vd_res = common global <2 x double> zeroinitializer, align 16
+
+define void @test1() {
+entry:
+  %0 = load <4 x float>* @vf, align 16
+  %1 = tail call <4 x float> @llvm.ppc.vsx.xvdivsp(<4 x float> %0, <4 x float> %0)
+  store <4 x float> %1, <4 x float>* @vf_res, align 16
+  ret void
+}
+; CHECK-LABEL: @test1
+; CHECK: xvdivsp
+
+define void @test2() {
+entry:
+  %0 = load <2 x double>* @vd, align 16
+  %1 = tail call <2 x double> @llvm.ppc.vsx.xvdivdp(<2 x double> %0, <2 x double> %0)
+  store <2 x double> %1, <2 x double>* @vd_res, align 16
+  ret void
+}
+; CHECK-LABEL: @test2
+; CHECK: xvdivdp
+
+declare <2 x double> @llvm.ppc.vsx.xvdivdp(<2 x double>, <2 x double>)
+declare <4 x float> @llvm.ppc.vsx.xvdivsp(<4 x float>, <4 x float>)