[AVX512] Enabling bit logic lowering

Added lowering tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224132 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 0f0bd81689e09728dc75e61c3857ea39bd7928fb..306260924ff52a4876bf554c6482d9f27588455c 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1570,6 +1570,13 @@ void X86TargetLowering::resetOperationActions() {
     setOperationAction(ISD::SETCC,              MVT::v4i1, Custom);
     setOperationAction(ISD::SETCC,              MVT::v2i1, Custom);
     setOperationAction(ISD::INSERT_SUBVECTOR,   MVT::v8i1, Legal);
+
+    setOperationAction(ISD::AND,                MVT::v8i32, Legal);
+    setOperationAction(ISD::OR,                 MVT::v8i32, Legal);
+    setOperationAction(ISD::XOR,                MVT::v8i32, Legal);
+    setOperationAction(ISD::AND,                MVT::v4i32, Legal);
+    setOperationAction(ISD::OR,                 MVT::v4i32, Legal);
+    setOperationAction(ISD::XOR,                MVT::v4i32, Legal);
   }
 
   // SIGN_EXTEND_INREGs are evaluated by the extend type. Handle the expansion

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index b31f521f8cd4d2ae16bce7dfd54981181907a43c..9fa8ce25b8769389d625e7b7e6a39ad1750f02c5 100644 (file)
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -2922,6 +2922,7 @@ let isCodeGenOnly = 1 in {
 ///
 multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
                                    SDNode OpNode> {
+  let Predicates = [HasAVX, NoVLX] in {
   defm V#NAME#PSY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedSingle,
         !strconcat(OpcodeStr, "ps"), f256mem,
         [(set VR256:$dst, (v4i64 (OpNode VR256:$src1, VR256:$src2)))],
@@ -2952,6 +2953,7 @@ multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
        [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
                                  (loadv2i64 addr:$src2)))], 0>,
                                                  PD, VEX_4V;
+  }
 
   let Constraints = "$src1 = $dst" in {
     defm PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,

diff --git a/test/CodeGen/X86/avx512-logic.ll b/test/CodeGen/X86/avx512-logic.ll
new file mode 100644 (file)
index 0000000..bee4f52
--- /dev/null
+++ b/test/CodeGen/X86/avx512-logic.ll
@@ -0,0 +1,101 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+
+; CHECK-LABEL: vpandd
+; CHECK: vpandd %zmm
+; CHECK: ret
+define <16 x i32> @vpandd(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1,
+                            i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = and <16 x i32> %a2, %b
+  ret <16 x i32> %x
+}
+
+; CHECK-LABEL: vpord
+; CHECK: vpord %zmm
+; CHECK: ret
+define <16 x i32> @vpord(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1,
+                            i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = or <16 x i32> %a2, %b
+  ret <16 x i32> %x
+}
+
+; CHECK-LABEL: vpxord
+; CHECK: vpxord %zmm
+; CHECK: ret
+define <16 x i32> @vpxord(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1,
+                            i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = xor <16 x i32> %a2, %b
+  ret <16 x i32> %x
+}
+
+; CHECK-LABEL: vpandq
+; CHECK: vpandq %zmm
+; CHECK: ret
+define <8 x i64> @vpandq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  %x = and <8 x i64> %a2, %b
+  ret <8 x i64> %x
+}
+
+; CHECK-LABEL: vporq
+; CHECK: vporq %zmm
+; CHECK: ret
+define <8 x i64> @vporq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  %x = or <8 x i64> %a2, %b
+  ret <8 x i64> %x
+}
+
+; CHECK-LABEL: vpxorq
+; CHECK: vpxorq %zmm
+; CHECK: ret
+define <8 x i64> @vpxorq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  %x = xor <8 x i64> %a2, %b
+  ret <8 x i64> %x
+}
+
+
+; CHECK-LABEL: orq_broadcast
+; CHECK: vporq LCP{{.*}}(%rip){1to8}, %zmm0, %zmm0
+; CHECK: ret
+define <8 x i64> @orq_broadcast(<8 x i64> %a) nounwind {
+  %b = or <8 x i64> %a, <i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2>
+  ret <8 x i64> %b
+}
+
+; CHECK-LABEL: andd512fold
+; CHECK: vpandd (%
+; CHECK: ret
+define <16 x i32> @andd512fold(<16 x i32> %y, <16 x i32>* %x) {
+entry:
+  %a = load <16 x i32>* %x, align 4
+  %b = and <16 x i32> %y, %a
+  ret <16 x i32> %b
+}
+
+; CHECK-LABEL: andqbrst
+; CHECK: vpandq  (%rdi){1to8}, %zmm
+; CHECK: ret
+define <8 x i64> @andqbrst(<8 x i64> %p1, i64* %ap) {
+entry:
+  %a = load i64* %ap, align 8
+  %b = insertelement <8 x i64> undef, i64 %a, i32 0
+  %c = shufflevector <8 x i64> %b, <8 x i64> undef, <8 x i32> zeroinitializer
+  %d = and <8 x i64> %p1, %c
+  ret <8 x i64>%d
+}

diff --git a/test/CodeGen/X86/avx512vl-logic.ll b/test/CodeGen/X86/avx512vl-logic.ll
new file mode 100644 (file)
index 0000000..02cb8f9
--- /dev/null
+++ b/test/CodeGen/X86/avx512vl-logic.ll
@@ -0,0 +1,137 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512vl | FileCheck %s
+
+; 256-bit
+
+; CHECK-LABEL: vpandd256
+; CHECK: vpandd %ymm
+; CHECK: ret
+define <8 x i32> @vpandd256(<8 x i32> %a, <8 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = and <8 x i32> %a2, %b
+  ret <8 x i32> %x
+}
+
+; CHECK-LABEL: vpord256
+; CHECK: vpord %ymm
+; CHECK: ret
+define <8 x i32> @vpord256(<8 x i32> %a, <8 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = or <8 x i32> %a2, %b
+  ret <8 x i32> %x
+}
+
+; CHECK-LABEL: vpxord256
+; CHECK: vpxord %ymm
+; CHECK: ret
+define <8 x i32> @vpxord256(<8 x i32> %a, <8 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <8 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %x = xor <8 x i32> %a2, %b
+  ret <8 x i32> %x
+}
+
+; CHECK-LABEL: vpandq256
+; CHECK: vpandq %ymm
+; CHECK: ret
+define <4 x i64> @vpandq256(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i64> %a, <i64 1, i64 1, i64 1, i64 1>
+  %x = and <4 x i64> %a2, %b
+  ret <4 x i64> %x
+}
+
+; CHECK-LABEL: vporq256
+; CHECK: vporq %ymm
+; CHECK: ret
+define <4 x i64> @vporq256(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i64> %a, <i64 1, i64 1, i64 1, i64 1>
+  %x = or <4 x i64> %a2, %b
+  ret <4 x i64> %x
+}
+
+; CHECK-LABEL: vpxorq256
+; CHECK: vpxorq %ymm
+; CHECK: ret
+define <4 x i64> @vpxorq256(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i64> %a, <i64 1, i64 1, i64 1, i64 1>
+  %x = xor <4 x i64> %a2, %b
+  ret <4 x i64> %x
+}
+
+; 128-bit
+
+; CHECK-LABEL: vpandd128
+; CHECK: vpandd %xmm
+; CHECK: ret
+define <4 x i32> @vpandd128(<4 x i32> %a, <4 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i32> %a, <i32 1, i32 1, i32 1, i32 1>
+  %x = and <4 x i32> %a2, %b
+  ret <4 x i32> %x
+}
+
+; CHECK-LABEL: vpord128
+; CHECK: vpord %xmm
+; CHECK: ret
+define <4 x i32> @vpord128(<4 x i32> %a, <4 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i32> %a, <i32 1, i32 1, i32 1, i32 1>
+  %x = or <4 x i32> %a2, %b
+  ret <4 x i32> %x
+}
+
+; CHECK-LABEL: vpxord128
+; CHECK: vpxord %xmm
+; CHECK: ret
+define <4 x i32> @vpxord128(<4 x i32> %a, <4 x i32> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <4 x i32> %a, <i32 1, i32 1, i32 1, i32 1>
+  %x = xor <4 x i32> %a2, %b
+  ret <4 x i32> %x
+}
+
+; CHECK-LABEL: vpandq128
+; CHECK: vpandq %xmm
+; CHECK: ret
+define <2 x i64> @vpandq128(<2 x i64> %a, <2 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <2 x i64> %a, <i64 1, i64 1>
+  %x = and <2 x i64> %a2, %b
+  ret <2 x i64> %x
+}
+
+; CHECK-LABEL: vporq128
+; CHECK: vporq %xmm
+; CHECK: ret
+define <2 x i64> @vporq128(<2 x i64> %a, <2 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <2 x i64> %a, <i64 1, i64 1>
+  %x = or <2 x i64> %a2, %b
+  ret <2 x i64> %x
+}
+
+; CHECK-LABEL: vpxorq128
+; CHECK: vpxorq %xmm
+; CHECK: ret
+define <2 x i64> @vpxorq128(<2 x i64> %a, <2 x i64> %b) nounwind uwtable readnone ssp {
+entry:
+  ; Force the execution domain with an add.
+  %a2 = add <2 x i64> %a, <i64 1, i64 1>
+  %x = xor <2 x i64> %a2, %b
+  ret <2 x i64> %x
+}