Add support of RDSEED defined in AVX2 extension

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178314 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/IR/IntrinsicsX86.td

@@ -2550,14 +2550,19 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 }
 
 //===----------------------------------------------------------------------===//
-// RDRAND intrinsics. Return a random value and whether it is valid.
+// RDRAND intrinsics - Return a random value and whether it is valid.
+// RDSEED intrinsics - Return a NIST SP800-90B & C compliant random value and
+// whether it is valid.
 
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   // These are declared side-effecting so they don't get eliminated by CSE or
   // LICM.
   def int_x86_rdrand_16 : Intrinsic<[llvm_i16_ty, llvm_i32_ty], [], []>;
   def int_x86_rdrand_32 : Intrinsic<[llvm_i32_ty, llvm_i32_ty], [], []>;
   def int_x86_rdrand_64 : Intrinsic<[llvm_i64_ty, llvm_i32_ty], [], []>;
+  def int_x86_rdseed_16 : Intrinsic<[llvm_i16_ty, llvm_i32_ty], [], []>;
+  def int_x86_rdseed_32 : Intrinsic<[llvm_i32_ty, llvm_i32_ty], [], []>;
+  def int_x86_rdseed_64 : Intrinsic<[llvm_i64_ty, llvm_i32_ty], [], []>;
 }
 
 //===----------------------------------------------------------------------===//

lib/Target/X86/X86.td

@@ -126,6 +126,8 @@ def FeatureADX     : SubtargetFeature<"adx", "HasADX", "true",
                                       "Support ADX instructions">;
 def FeaturePRFCHW  : SubtargetFeature<"prfchw", "HasPRFCHW", "true",
                                       "Support PRFCHW instructions">;
+def FeatureRDSEED  : SubtargetFeature<"rdseed", "HasRDSEED", "true",
+                                      "Support RDSEED instruction">;
 def FeatureLeaForSP : SubtargetFeature<"lea-sp", "UseLeaForSP", "true",
                                      "Use LEA for adjusting the stack pointer">;
 def FeatureSlowDivide : SubtargetFeature<"idiv-to-divb",

lib/Target/X86/X86ISelLowering.cpp

@@ -10914,16 +10914,23 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, SelectionDAG &DAG) {
   switch (IntNo) {
   default: return SDValue();    // Don't custom lower most intrinsics.
 
-  // RDRAND intrinsics.
+  // RDRAND/RDSEED intrinsics.
   case Intrinsic::x86_rdrand_16:
   case Intrinsic::x86_rdrand_32:
-  case Intrinsic::x86_rdrand_64: {
+  case Intrinsic::x86_rdrand_64:
+  case Intrinsic::x86_rdseed_16:
+  case Intrinsic::x86_rdseed_32:
+  case Intrinsic::x86_rdseed_64: {
+    unsigned Opcode = (IntNo == Intrinsic::x86_rdseed_16 ||
+                       IntNo == Intrinsic::x86_rdseed_32 ||
+                       IntNo == Intrinsic::x86_rdseed_64) ? X86ISD::RDSEED :
+                                                            X86ISD::RDRAND;
     // Emit the node with the right value type.
     SDVTList VTs = DAG.getVTList(Op->getValueType(0), MVT::Glue, MVT::Other);
-    SDValue Result = DAG.getNode(X86ISD::RDRAND, dl, VTs, Op.getOperand(0));
+    SDValue Result = DAG.getNode(Opcode, dl, VTs, Op.getOperand(0));
 
-    // If the value returned by RDRAND was valid (CF=1), return 1. Otherwise
-    // return the value from Rand, which is always 0, casted to i32.
+    // If the value returned by RDRAND/RDSEED was valid (CF=1), return 1.
+    // Otherwise return the value from Rand, which is always 0, casted to i32.
     SDValue Ops[] = { DAG.getZExtOrTrunc(Result, dl, Op->getValueType(1)),
                       DAG.getConstant(1, Op->getValueType(1)),
                       DAG.getConstant(X86::COND_B, MVT::i32),
@@ -12781,6 +12788,7 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::WIN_FTOL:           return "X86ISD::WIN_FTOL";
   case X86ISD::SAHF:               return "X86ISD::SAHF";
   case X86ISD::RDRAND:             return "X86ISD::RDRAND";
+  case X86ISD::RDSEED:             return "X86ISD::RDSEED";
   case X86ISD::FMADD:              return "X86ISD::FMADD";
   case X86ISD::FMSUB:              return "X86ISD::FMSUB";
   case X86ISD::FNMADD:             return "X86ISD::FNMADD";
@@ -15847,8 +15855,10 @@ static SDValue checkBoolTestSetCCCombine(SDValue Cmp, X86::CondCode &CC) {
       if (Op.getOpcode() == ISD::ZERO_EXTEND ||
           Op.getOpcode() == ISD::TRUNCATE)
         Op = Op.getOperand(0);
-      // A special case for rdrand, where 0 is set if false cond is found.
-      if (Op.getOpcode() != X86ISD::RDRAND || Op.getResNo() != 0)
+      // A special case for rdrand/rdseed, where 0 is set if false cond is
+      // found.
+      if ((Op.getOpcode() != X86ISD::RDRAND &&
+           Op.getOpcode() != X86ISD::RDSEED) || Op.getResNo() != 0)
         return SDValue();
     }
     // Quit if false value is not the constant 0 or 1.

lib/Target/X86/X86ISelLowering.h

@@ -356,6 +356,10 @@ namespace llvm {
       // RDRAND - Get a random integer and indicate whether it is valid in CF.
       RDRAND,
 
+      // RDSEED - Get a NIST SP800-90B & C compliant random integer and
+      // indicate whether it is valid in CF.
+      RDSEED,
+
       // PCMP*STRI
       PCMPISTRI,
       PCMPESTRI,

lib/Target/X86/X86InstrInfo.td

@@ -142,6 +142,9 @@ def X86sahf    : SDNode<"X86ISD::SAHF",     SDTX86sahf>;
 def X86rdrand  : SDNode<"X86ISD::RDRAND",   SDTX86rdrand,
                         [SDNPHasChain, SDNPSideEffect]>;
 
+def X86rdseed  : SDNode<"X86ISD::RDSEED",   SDTX86rdrand,
+                        [SDNPHasChain, SDNPSideEffect]>;
+
 def X86cas : SDNode<"X86ISD::LCMPXCHG_DAG", SDTX86cas,
                         [SDNPHasChain, SDNPInGlue, SDNPOutGlue, SDNPMayStore,
                          SDNPMayLoad, SDNPMemOperand]>;
@@ -607,6 +610,7 @@ def HasHLE       : Predicate<"Subtarget->hasHLE()">;
 def HasTSX       : Predicate<"Subtarget->hasRTM() || Subtarget->hasHLE()">;
 def HasADX       : Predicate<"Subtarget->hasADX()">;
 def HasPRFCHW    : Predicate<"Subtarget->hasPRFCHW()">;
+def HasRDSEED    : Predicate<"Subtarget->hasRDSEED()">;
 def HasPrefetchW : Predicate<"Subtarget->has3DNow() || Subtarget->hasPRFCHW()">;
 def FPStackf32   : Predicate<"!Subtarget->hasSSE1()">;
 def FPStackf64   : Predicate<"!Subtarget->hasSSE2()">;
@@ -1633,6 +1637,21 @@ let Predicates = [HasRDRAND], Defs = [EFLAGS] in {
                      [(set GR64:$dst, EFLAGS, (X86rdrand))]>, TB;
 }
 
+//===----------------------------------------------------------------------===//
+// RDSEED Instruction
+//
+let Predicates = [HasRDSEED], Defs = [EFLAGS] in {
+  def RDSEED16r : I<0xC7, MRM7r, (outs GR16:$dst), (ins),
+                    "rdseed{w}\t$dst",
+                    [(set GR16:$dst, EFLAGS, (X86rdseed))]>, OpSize, TB;
+  def RDSEED32r : I<0xC7, MRM7r, (outs GR32:$dst), (ins),
+                    "rdseed{l}\t$dst",
+                    [(set GR32:$dst, EFLAGS, (X86rdseed))]>, TB;
+  def RDSEED64r : RI<0xC7, MRM7r, (outs GR64:$dst), (ins),
+                     "rdseed{q}\t$dst",
+                     [(set GR64:$dst, EFLAGS, (X86rdseed))]>, TB;
+}
+
 //===----------------------------------------------------------------------===//
 // LZCNT Instruction
 //

lib/Target/X86/X86Subtarget.cpp

@@ -334,6 +334,10 @@ void X86Subtarget::AutoDetectSubtargetFeatures() {
         HasADX = true;
         ToggleFeature(X86::FeatureADX);
       }
+      if (IsIntel && ((EBX >> 18) & 0x1)) {
+        HasRDSEED = true;
+        ToggleFeature(X86::FeatureRDSEED);
+      }
     }
   }
 }
@@ -454,6 +458,7 @@ void X86Subtarget::initializeEnvironment() {
   HasHLE = false;
   HasADX = false;
   HasPRFCHW = false;
+  HasRDSEED = false;
   IsBTMemSlow = false;
   IsUAMemFast = false;
   HasVectorUAMem = false;

lib/Target/X86/X86Subtarget.h

@@ -130,6 +130,9 @@ class X86Subtarget : public X86GenSubtargetInfo {
   /// HasPRFCHW - Processor has PRFCHW instructions.
   bool HasPRFCHW;
 
+  /// HasRDSEED - Processor has RDSEED instructions.
+  bool HasRDSEED;
+
   /// IsBTMemSlow - True if BT (bit test) of memory instructions are slow.
   bool IsBTMemSlow;
 
@@ -266,6 +269,7 @@ class X86Subtarget : public X86GenSubtargetInfo {
   bool hasHLE() const { return HasHLE; }
   bool hasADX() const { return HasADX; }
   bool hasPRFCHW() const { return HasPRFCHW; }
+  bool hasRDSEED() const { return HasRDSEED; }
   bool isBTMemSlow() const { return IsBTMemSlow; }
   bool isUnalignedMemAccessFast() const { return IsUAMemFast; }
   bool hasVectorUAMem() const { return HasVectorUAMem; }

test/CodeGen/X86/bool-simplify.ll

@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86-64 -mattr=+sse41,-avx,+rdrand | FileCheck %s
+; RUN: llc < %s -march=x86-64 -mattr=+sse41,-avx,+rdrand,+rdseed | FileCheck %s
 
 define i32 @foo(<2 x i64> %c, i32 %a, i32 %b) {
   %t1 = call i32 @llvm.x86.sse41.ptestz(<2 x i64> %c, <2 x i64> %c)
@@ -84,7 +84,55 @@ define i64 @rnd64(i64 %arg) nounwind uwtable {
 ; CHECK: ret
 }
 
+define i16 @seed16(i16 %arg) nounwind uwtable {
+  %1 = tail call { i16, i32 } @llvm.x86.rdseed.16() nounwind
+  %2 = extractvalue { i16, i32 } %1, 0
+  %3 = extractvalue { i16, i32 } %1, 1
+  %4 = icmp eq i32 %3, 0
+  %5 = select i1 %4, i16 0, i16 %arg
+  %6 = add i16 %5, %2
+  ret i16 %6
+; CHECK: seed16
+; CHECK: rdseed
+; CHECK: cmov
+; CHECK-NOT: cmov
+; CHECK: ret
+}
+
+define i32 @seed32(i32 %arg) nounwind uwtable {
+  %1 = tail call { i32, i32 } @llvm.x86.rdseed.32() nounwind
+  %2 = extractvalue { i32, i32 } %1, 0
+  %3 = extractvalue { i32, i32 } %1, 1
+  %4 = icmp eq i32 %3, 0
+  %5 = select i1 %4, i32 0, i32 %arg
+  %6 = add i32 %5, %2
+  ret i32 %6
+; CHECK: seed32
+; CHECK: rdseed
+; CHECK: cmov
+; CHECK-NOT: cmov
+; CHECK: ret
+}
+
+define i64 @seed64(i64 %arg) nounwind uwtable {
+  %1 = tail call { i64, i32 } @llvm.x86.rdseed.64() nounwind
+  %2 = extractvalue { i64, i32 } %1, 0
+  %3 = extractvalue { i64, i32 } %1, 1
+  %4 = icmp eq i32 %3, 0
+  %5 = select i1 %4, i64 0, i64 %arg
+  %6 = add i64 %5, %2
+  ret i64 %6
+; CHECK: seed64
+; CHECK: rdseed
+; CHECK: cmov
+; CHECK-NOT: cmov
+; CHECK: ret
+}
+
 declare i32 @llvm.x86.sse41.ptestz(<2 x i64>, <2 x i64>) nounwind readnone
 declare { i16, i32 } @llvm.x86.rdrand.16() nounwind
 declare { i32, i32 } @llvm.x86.rdrand.32() nounwind
 declare { i64, i32 } @llvm.x86.rdrand.64() nounwind
+declare { i16, i32 } @llvm.x86.rdseed.16() nounwind
+declare { i32, i32 } @llvm.x86.rdseed.32() nounwind
+declare { i64, i32 } @llvm.x86.rdseed.64() nounwind

test/CodeGen/X86/rdseed.ll

@@ -0,0 +1,48 @@
+; RUN: llc < %s -march=x86-64 -mcpu=core-avx-i -mattr=+rdseed | FileCheck %s
+
+declare {i16, i32} @llvm.x86.rdseed.16()
+declare {i32, i32} @llvm.x86.rdseed.32()
+declare {i64, i32} @llvm.x86.rdseed.64()
+
+define i32 @_rdseed16_step(i16* %random_val) {
+  %call = call {i16, i32} @llvm.x86.rdseed.16()
+  %randval = extractvalue {i16, i32} %call, 0
+  store i16 %randval, i16* %random_val
+  %isvalid = extractvalue {i16, i32} %call, 1
+  ret i32 %isvalid
+; CHECK: _rdseed16_step:
+; CHECK: rdseedw	%ax
+; CHECK: movw	%ax, (%r[[A0:di|cx]])
+; CHECK: movzwl	%ax, %ecx
+; CHECK: movl	$1, %eax
+; CHECK: cmovael	%ecx, %eax
+; CHECK: ret
+}
+
+define i32 @_rdseed32_step(i32* %random_val) {
+  %call = call {i32, i32} @llvm.x86.rdseed.32()
+  %randval = extractvalue {i32, i32} %call, 0
+  store i32 %randval, i32* %random_val
+  %isvalid = extractvalue {i32, i32} %call, 1
+  ret i32 %isvalid
+; CHECK: _rdseed32_step:
+; CHECK: rdseedl	%e[[T0:[a-z]+]]
+; CHECK: movl	%e[[T0]], (%r[[A0]])
+; CHECK: movl	$1, %eax
+; CHECK: cmovael	%e[[T0]], %eax
+; CHECK: ret
+}
+
+define i32 @_rdseed64_step(i64* %random_val) {
+  %call = call {i64, i32} @llvm.x86.rdseed.64()
+  %randval = extractvalue {i64, i32} %call, 0
+  store i64 %randval, i64* %random_val
+  %isvalid = extractvalue {i64, i32} %call, 1
+  ret i32 %isvalid
+; CHECK: _rdseed64_step:
+; CHECK: rdseedq	%r[[T1:[a-z]+]]
+; CHECK: movq	%r[[T1]], (%r[[A0]])
+; CHECK: movl	$1, %eax
+; CHECK: cmovael	%e[[T1]], %eax
+; CHECK: ret
+}

test/MC/X86/x86_64-rand-encoding.s

@@ -0,0 +1,49 @@
+// RUN: llvm-mc -triple x86_64-unknown-unknown --show-encoding %s | FileCheck %s
+
+// CHECK: rdrandw %ax
+// CHECK: encoding: [0x66,0x0f,0xc7,0xf0]
+          rdrand %ax
+
+// CHECK: rdrandl %eax
+// CHECK: encoding: [0x0f,0xc7,0xf0]
+          rdrand %eax
+
+// CHECK: rdrandq %rax
+// CHECK: encoding: [0x48,0x0f,0xc7,0xf0]
+          rdrand %rax
+
+// CHECK: rdrandw %r11w
+// CHECK: encoding: [0x66,0x41,0x0f,0xc7,0xf3]
+          rdrand %r11w
+
+// CHECK: rdrandl %r11d
+// CHECK: encoding: [0x41,0x0f,0xc7,0xf3]
+          rdrand %r11d
+
+// CHECK: rdrandq %r11
+// CHECK: encoding: [0x49,0x0f,0xc7,0xf3]
+          rdrand %r11
+
+// CHECK: rdseedw %ax
+// CHECK: encoding: [0x66,0x0f,0xc7,0xf8]
+          rdseed %ax
+
+// CHECK: rdseedl %eax
+// CHECK: encoding: [0x0f,0xc7,0xf8]
+          rdseed %eax
+
+// CHECK: rdseedq %rax
+// CHECK: encoding: [0x48,0x0f,0xc7,0xf8]
+          rdseed %rax
+
+// CHECK: rdseedw %r11w
+// CHECK: encoding: [0x66,0x41,0x0f,0xc7,0xfb]
+          rdseed %r11w
+
+// CHECK: rdseedl %r11d
+// CHECK: encoding: [0x41,0x0f,0xc7,0xfb]
+          rdseed %r11d
+
+// CHECK: rdseedq %r11
+// CHECK: encoding: [0x49,0x0f,0xc7,0xfb]
+          rdseed %r11