AArch64: legalize G_EXTRACT operations.

This is the dual problem to legalizing G_INSERTs so most of the code and
testing was cribbed from there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@306328 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 1d0d3dffa4c59cba76254764a759ed78e120c079..121823247e13fd2e1e1a11469148ce87503aac7b 100644 (file)
--- a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -158,7 +158,7 @@ LegalizerHelper::LegalizeResult LegalizerHelper::narrowScalar(MachineInstr &MI,
                                                               unsigned TypeIdx,
                                                               LLT NarrowTy) {
   // FIXME: Don't know how to handle secondary types yet.
-  if (TypeIdx != 0)
+  if (TypeIdx != 0 && MI.getOpcode() != TargetOpcode::G_EXTRACT)
     return UnableToLegalize;
 
   MIRBuilder.setInstr(MI);
@@ -193,6 +193,58 @@ LegalizerHelper::LegalizeResult LegalizerHelper::narrowScalar(MachineInstr &MI,
     MI.eraseFromParent();
     return Legalized;
   }
+  case TargetOpcode::G_EXTRACT: {
+    if (TypeIdx != 1)
+      return UnableToLegalize;
+
+    int64_t NarrowSize = NarrowTy.getSizeInBits();
+    int NumParts =
+        MRI.getType(MI.getOperand(1).getReg()).getSizeInBits() / NarrowSize;
+
+    SmallVector<unsigned, 2> SrcRegs, DstRegs;
+    SmallVector<uint64_t, 2> Indexes;
+    extractParts(MI.getOperand(1).getReg(), NarrowTy, NumParts, SrcRegs);
+
+    unsigned OpReg = MI.getOperand(0).getReg();
+    int64_t OpStart = MI.getOperand(2).getImm();
+    int64_t OpSize = MRI.getType(OpReg).getSizeInBits();
+    for (int i = 0; i < NumParts; ++i) {
+      unsigned SrcStart = i * NarrowSize;
+
+      if (SrcStart + NarrowSize <= OpStart || SrcStart >= OpStart + OpSize) {
+        // No part of the extract uses this subregister, ignore it.
+        continue;
+      } else if (SrcStart == OpStart && NarrowTy == MRI.getType(OpReg)) {
+        // The entire subregister is extracted, forward the value.
+        DstRegs.push_back(SrcRegs[i]);
+        continue;
+      }
+
+      // OpSegStart is where this destination segment would start in OpReg if it
+      // extended infinitely in both directions.
+      int64_t ExtractOffset, SegSize;
+      if (OpStart < SrcStart) {
+        ExtractOffset = 0;
+        SegSize = std::min(NarrowSize, OpStart + OpSize - SrcStart);
+      } else {
+        ExtractOffset = OpStart - SrcStart;
+        SegSize = std::min(SrcStart + NarrowSize - OpStart, OpSize);
+      }
+
+      unsigned SegReg = SrcRegs[i];
+      if (ExtractOffset != 0 || SegSize != NarrowSize) {
+        // A genuine extract is needed.
+        SegReg = MRI.createGenericVirtualRegister(LLT::scalar(SegSize));
+        MIRBuilder.buildExtract(SegReg, SrcRegs[i], ExtractOffset);
+      }
+
+      DstRegs.push_back(SegReg);
+    }
+
+    MIRBuilder.buildMerge(MI.getOperand(0).getReg(), DstRegs);
+    MI.eraseFromParent();
+    return Legalized;
+  }
   case TargetOpcode::G_INSERT: {
     if (TypeIdx != 0)
       return UnableToLegalize;

diff --git a/lib/CodeGen/GlobalISel/LegalizerInfo.cpp b/lib/CodeGen/GlobalISel/LegalizerInfo.cpp
index 595802f2228b9937317a05c57a71b2508bab8876..ef1ce7f87ebc287be46a09e17d8cd87d6e5a8a75 100644 (file)
--- a/lib/CodeGen/GlobalISel/LegalizerInfo.cpp
+++ b/lib/CodeGen/GlobalISel/LegalizerInfo.cpp
@@ -42,6 +42,7 @@ LegalizerInfo::LegalizerInfo() : TablesInitialized(false) {
 
   DefaultActions[TargetOpcode::G_BRCOND] = WidenScalar;
   DefaultActions[TargetOpcode::G_INSERT] = NarrowScalar;
+  DefaultActions[TargetOpcode::G_EXTRACT] = NarrowScalar;
   DefaultActions[TargetOpcode::G_FNEG] = Lower;
 }
 
@@ -75,8 +76,7 @@ LegalizerInfo::getAction(const InstrAspect &Aspect) const {
 
   // FIXME: the long-term plan calls for expansion in terms of load/store (if
   // they're not legal).
-  if (Aspect.Opcode == TargetOpcode::G_EXTRACT ||
-      Aspect.Opcode == TargetOpcode::G_MERGE_VALUES ||
+  if (Aspect.Opcode == TargetOpcode::G_MERGE_VALUES ||
       Aspect.Opcode == TargetOpcode::G_UNMERGE_VALUES)
     return std::make_pair(Legal, Aspect.Type);
 

diff --git a/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp b/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
index 3c70013ea296b227837afdf2d822c4ced21dfeb3..6fcc3907dd2805f8703ef56a663b0831c55ff5b2 100644 (file)
--- a/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
+++ b/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
@@ -482,6 +482,9 @@ MachineInstrBuilder MachineIRBuilder::buildMerge(unsigned Res,
          "input operands do not cover output register");
 #endif
 
+  if (Ops.size() == 1)
+    return buildCopy(Res, Ops[0]);
+
   MachineInstrBuilder MIB = buildInstr(TargetOpcode::G_MERGE_VALUES);
   MIB.addDef(Res);
   for (unsigned i = 0; i < Ops.size(); ++i)

diff --git a/lib/Target/AArch64/AArch64LegalizerInfo.cpp b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
index 01196817f3112a6bd1b370461beda35a266dd3a1..257797749494e45618f3291fd5f5abc2239a92e7 100644 (file)
--- a/lib/Target/AArch64/AArch64LegalizerInfo.cpp
+++ b/lib/Target/AArch64/AArch64LegalizerInfo.cpp
@@ -99,6 +99,12 @@ AArch64LegalizerInfo::AArch64LegalizerInfo() {
     // G_INSERT (It seems entirely reasonable that inputs shouldn't overlap).
   }
 
+  for (auto Ty : {s1, s8, s16, s32, s64, p0})
+    setAction({G_EXTRACT, Ty}, Legal);
+
+  for (auto Ty : {s32, s64})
+    setAction({G_EXTRACT, 1, Ty}, Legal);
+
   for (unsigned MemOp : {G_LOAD, G_STORE}) {
     for (auto Ty : {s8, s16, s32, s64, p0, v2s32})
       setAction({MemOp, Ty}, Legal);

diff --git a/lib/Target/X86/X86LegalizerInfo.cpp b/lib/Target/X86/X86LegalizerInfo.cpp
index a584eabcc1b285f1f220b75edce1f49024a95850..53215296ce654b62acff17e93a61310e25cd5a79 100644 (file)
--- a/lib/Target/X86/X86LegalizerInfo.cpp
+++ b/lib/Target/X86/X86LegalizerInfo.cpp
@@ -228,10 +228,14 @@ void X86LegalizerInfo::setLegalizerInfoAVX() {
     for (auto Ty : {v8s32, v4s64})
       setAction({MemOp, Ty}, Legal);
 
-  for (auto Ty : {v32s8, v16s16, v8s32, v4s64})
+  for (auto Ty : {v32s8, v16s16, v8s32, v4s64}) {
     setAction({G_INSERT, Ty}, Legal);
-  for (auto Ty : {v16s8, v8s16, v4s32, v2s64})
+    setAction({G_EXTRACT, 1, Ty}, Legal);
+  }
+  for (auto Ty : {v16s8, v8s16, v4s32, v2s64}) {
     setAction({G_INSERT, 1, Ty}, Legal);
+    setAction({G_EXTRACT, Ty}, Legal);
+  }
 }
 
 void X86LegalizerInfo::setLegalizerInfoAVX2() {
@@ -280,10 +284,14 @@ void X86LegalizerInfo::setLegalizerInfoAVX512() {
     for (auto Ty : {v16s32, v8s64})
       setAction({MemOp, Ty}, Legal);
 
-  for (auto Ty : {v64s8, v32s16, v16s32, v8s64})
+  for (auto Ty : {v64s8, v32s16, v16s32, v8s64}) {
     setAction({G_INSERT, Ty}, Legal);
-  for (auto Ty : {v32s8, v16s16, v8s32, v4s64, v16s8, v8s16, v4s32, v2s64})
+    setAction({G_EXTRACT, 1, Ty}, Legal);
+  }
+  for (auto Ty : {v32s8, v16s16, v8s32, v4s64, v16s8, v8s16, v4s32, v2s64}) {
     setAction({G_INSERT, 1, Ty}, Legal);
+    setAction({G_EXTRACT, Ty}, Legal);
+  }
 
   /************ VLX *******************/
   if (!Subtarget.hasVLX())

diff --git a/test/CodeGen/AArch64/GlobalISel/legalize-combines.mir b/test/CodeGen/AArch64/GlobalISel/legalize-combines.mir
index e3e0175d39ac59c678477339a1c5338e90778e28..fbacc28d7434edbba673879426e682c21e9692b4 100644 (file)
--- a/test/CodeGen/AArch64/GlobalISel/legalize-combines.mir
+++ b/test/CodeGen/AArch64/GlobalISel/legalize-combines.mir
@@ -57,11 +57,11 @@ body: |
     %0:_(s64) = COPY %x0
 
     ; CHECK-LABEL: name: test_combines_4
-    ; CHECK: %2(<2 x s32>) = G_EXTRACT %1(s128), 0
-    ; CHECK: %3(<2 x s32>) = G_ADD %2, %2
+    ; CHECK: %2(s64) = COPY %0(s64)
+    ; CHECK: %3(s64) = G_ADD %2, %2
     %1:_(s128) = G_MERGE_VALUES %0, %0
-    %2:_(<2 x s32>) = G_EXTRACT %1, 0
-    %3:_(<2 x s32>) = G_ADD %2, %2
+    %2:_(s64) = G_EXTRACT %1, 0
+    %3:_(s64) = G_ADD %2, %2
 ...
 
 ---

diff --git a/test/CodeGen/AArch64/GlobalISel/legalize-exceptions.ll b/test/CodeGen/AArch64/GlobalISel/legalize-exceptions.ll
index 23e7d5163e5a51cc4c11950ef34edce7997e86a7..d2452b86170f8947841c56e244a864b847da6599 100644 (file)
--- a/test/CodeGen/AArch64/GlobalISel/legalize-exceptions.ll
+++ b/test/CodeGen/AArch64/GlobalISel/legalize-exceptions.ll
@@ -22,12 +22,11 @@ declare void @_Unwind_Resume(i8*)
 ; CHECK:     [[SEL:%[0-9]+]](s32) = G_PTRTOINT [[SEL_PTR]]
 ; CHECK:     [[STRUCT_SEL:%[0-9]+]](s64) = G_INSERT {{%[0-9]+}}, [[SEL]](s32), 0
 
-; CHECK:     [[STRUCT:%[0-9]+]](s128) = G_MERGE_VALUES [[STRUCT_PTR]](s64), [[STRUCT_SEL]]
-
-; CHECK:     [[PTR:%[0-9]+]](p0) = G_EXTRACT [[STRUCT]](s128), 0
+; CHECK:     [[PTR:%[0-9]+]](p0) = COPY [[STRUCT_PTR]](s64)
 ; CHECK:     G_STORE [[PTR]](p0), {{%[0-9]+}}(p0)
 
-; CHECK:     [[SEL:%[0-9]+]](s32) = G_EXTRACT [[STRUCT]](s128), 64
+; CHECK:     [[SEL_TMP:%[0-9]+]](s32) = G_EXTRACT [[STRUCT_SEL]](s64), 0
+; CHECK:     [[SEL:%[0-9]+]](s32) = COPY [[SEL_TMP]]
 ; CHECK:     G_STORE [[SEL]](s32), {{%[0-9]+}}(p0)
 
 define void @bar() personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {

diff --git a/test/CodeGen/AArch64/GlobalISel/legalize-extracts.mir b/test/CodeGen/AArch64/GlobalISel/legalize-extracts.mir
new file mode 100644 (file)
index 0000000..dc6b59b
--- /dev/null
+++ b/test/CodeGen/AArch64/GlobalISel/legalize-extracts.mir
@@ -0,0 +1,85 @@
+# RUN: llc -mtriple=aarch64-linux-gnu -O0 -run-pass=legalizer -global-isel %s -o - | FileCheck %s
+
+---
+name:            test_extracts_1
+body: |
+  bb.0:
+    liveins: %w0
+
+    ; Low part of extraction takes entirity of the low register entirely, so
+    ; value stored is forwarded directly from first load.
+
+    ; CHECK-LABEL: name: test_extracts_1
+    ; CHECK: [[LO:%[0-9]+]](s64) = G_LOAD
+    ; CHECK: {{%[0-9]+}}(s64) = G_LOAD
+    ; CHECK: [[VAL:%[0-9]+]](s64) = COPY [[LO]]
+    ; CHECK: G_STORE [[VAL]]
+    %0:_(s64) = COPY %x0
+    %1:_(s32) = COPY %w1
+    %2:_(p0) = COPY %x2
+    %3:_(s128) = G_LOAD %2(p0) :: (load 16)
+    %4:_(s64) = G_EXTRACT %3(s128), 0
+    G_STORE %4(s64), %2(p0) :: (store 8)
+    RET_ReallyLR
+...
+
+---
+name:            test_extracts_2
+body: |
+  bb.0:
+    liveins: %w0
+
+      ; Low extraction wipes takes whole low register. High extraction is real.
+    ; CHECK-LABEL: name: test_extracts_2
+    ; CHECK: [[LO_TMP:%[0-9]+]](s64) = G_LOAD
+    ; CHECK: [[HI:%[0-9]+]](s64) = G_LOAD
+    ; CHECK: [[LO:%[0-9]+]](s64) = COPY [[LO_TMP]]
+    ; CHECK: [[NEWHI_TMP:%[0-9]+]](s32) = G_EXTRACT [[HI]](s64), 0
+    ; CHECK: [[NEWHI:%[0-9]+]](s32) = COPY [[NEWHI_TMP]]
+    ; CHECK: G_STORE [[LO]]
+    ; CHECK: G_STORE [[NEWHI]]
+    %0:_(s64) = COPY %x0
+    %1:_(s32) = COPY %w1
+    %2:_(p0) = COPY %x2
+    %3:_(s128) = G_LOAD %2(p0) :: (load 16)
+    %4:_(s64) = G_EXTRACT %3(s128), 0
+    %5:_(s32) = G_EXTRACT %3(s128), 64
+    G_STORE %4(s64), %2(p0) :: (store 8)
+    G_STORE %5(s32), %2(p0) :: (store 4)
+    RET_ReallyLR
+...
+
+---
+name:            test_extracts_3
+body: |
+  bb.0:
+    liveins: %x0, %x1, %x2
+
+
+    ; CHECK-LABEL: name: test_extracts_3
+    ; CHECK: [[LO:%[0-9]+]](s32) = G_EXTRACT %0(s64), 32
+    ; CHECK: [[HI:%[0-9]+]](s32) = G_EXTRACT %1(s64), 0
+    ; CHECK: %3(s64) = G_MERGE_VALUES [[LO]](s32), [[HI]](s32)
+    %0:_(s64) = COPY %x0
+    %1:_(s64) = COPY %x1
+    %2:_(s128) = G_MERGE_VALUES %0, %1
+    %3:_(s64) = G_EXTRACT %2, 32
+    RET_ReallyLR
+...
+
+---
+name:            test_extracts_4
+body: |
+  bb.0:
+    liveins: %x0, %x1, %x2
+
+
+    ; CHECK-LABEL: name: test_extracts_4
+    ; CHECK: [[LO_TMP:%[0-9]+]](s32) = G_EXTRACT %0(s64), 32
+    ; CHECK: %3(s32) = COPY [[LO_TMP]]
+    %0:_(s64) = COPY %x0
+    %1:_(s64) = COPY %x1
+    %2:_(s128) = G_MERGE_VALUES %0, %1
+    %3:_(s32) = G_EXTRACT %2, 32
+    RET_ReallyLR
+...