AVX-512: Added masked SHIFT commands, more encoding tests

llvm-svn: 189005

2 files changed, 44 insertions, 23 deletions

diff --git a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
index 7173d51e6a2..0c9fd91e310 100644
--- a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
+++ b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
@@ -868,11 +868,12 @@ void X86MCCodeEmitter::EmitVEXOpcodePrefix(uint64_t TSFlags, unsigned &CurByte,
   case X86II::MRM6r: case X86II::MRM7r:
     // MRM0r-MRM7r instructions forms:
     //  dst(VEX_4V), src(ModR/M), imm8
-    VEX_4V = getVEXRegisterEncoding(MI, CurOp);
-    if (HasEVEX && X86II::is32ExtendedReg(MI.getOperand(CurOp).getReg()))
-        EVEX_V2 = 0x0;
-    CurOp++;
-    
+    if (HasVEX_4V) {
+      VEX_4V = getVEXRegisterEncoding(MI, CurOp);
+      if (HasEVEX && X86II::is32ExtendedReg(MI.getOperand(CurOp).getReg()))
+          EVEX_V2 = 0x0;
+      CurOp++;
+    }    
     if (HasEVEX_K)
       EVEX_aaa = getWriteMaskRegisterEncoding(MI, CurOp++);
 
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index c3fb8019cc9..0cb946ddb8c 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -1720,78 +1720,98 @@ defm VPTESTMQZ  : avx512_vptest<0x27, "vptestmq", VK8, VR512,  f512mem, memopv8i
 // AVX-512  Shift instructions
 //===----------------------------------------------------------------------===//
 multiclass avx512_shift_rmi<bits<8> opc, Format ImmFormR, Format ImmFormM,
-                         string OpcodeStr,
-                         SDNode OpNode, RegisterClass RC, ValueType vt,
-                         X86MemOperand x86memop, PatFrag mem_frag> {
+                         string OpcodeStr, SDNode OpNode, RegisterClass RC,
+                         ValueType vt, X86MemOperand x86memop, PatFrag mem_frag,
+                         RegisterClass KRC> {
   def ri : AVX512BIi8<opc, ImmFormR, (outs RC:$dst),
        (ins RC:$src1, i32i8imm:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
        [(set RC:$dst, (vt (OpNode RC:$src1, (i32 imm:$src2))))],
         SSE_INTSHIFT_ITINS_P.rr>, EVEX_4V;
+  def rik : AVX512BIi8<opc, ImmFormR, (outs RC:$dst),
+       (ins KRC:$mask, RC:$src1, i32i8imm:$src2),
+           !strconcat(OpcodeStr,
+                "\t{$src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2}"),
+       [], SSE_INTSHIFT_ITINS_P.rr>, EVEX_4V, EVEX_K;
   def mi: AVX512BIi8<opc, ImmFormM, (outs RC:$dst),
        (ins x86memop:$src1, i32i8imm:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
        [(set RC:$dst, (OpNode (mem_frag addr:$src1),
                      (i32 imm:$src2)))], SSE_INTSHIFT_ITINS_P.rm>, EVEX_4V;
+  def mik: AVX512BIi8<opc, ImmFormM, (outs RC:$dst),
+       (ins KRC:$mask, x86memop:$src1, i32i8imm:$src2),
+           !strconcat(OpcodeStr,
+                "\t{$src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2}"),
+       [], SSE_INTSHIFT_ITINS_P.rm>, EVEX_4V, EVEX_K;
 }
 
 multiclass avx512_shift_rrm<bits<8> opc, string OpcodeStr, SDNode OpNode,
                           RegisterClass RC, ValueType vt, ValueType SrcVT,
-                          PatFrag bc_frag> {
+                          PatFrag bc_frag, RegisterClass KRC> {
   // src2 is always 128-bit
   def rr : AVX512BI<opc, MRMSrcReg, (outs RC:$dst),
        (ins RC:$src1, VR128X:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
        [(set RC:$dst, (vt (OpNode RC:$src1, (SrcVT VR128X:$src2))))],
         SSE_INTSHIFT_ITINS_P.rr>, EVEX_4V;
+  def rrk : AVX512BI<opc, MRMSrcReg, (outs RC:$dst),
+       (ins KRC:$mask, RC:$src1, VR128X:$src2),
+           !strconcat(OpcodeStr,
+                "\t{$src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2}"),
+       [], SSE_INTSHIFT_ITINS_P.rr>, EVEX_4V, EVEX_K;
   def rm : AVX512BI<opc, MRMSrcMem, (outs RC:$dst),
        (ins RC:$src1, i128mem:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
        [(set RC:$dst, (vt (OpNode RC:$src1,
                        (bc_frag (memopv2i64 addr:$src2)))))],
                         SSE_INTSHIFT_ITINS_P.rm>, EVEX_4V;
+  def rmk : AVX512BI<opc, MRMSrcMem, (outs RC:$dst),
+       (ins KRC:$mask, RC:$src1, i128mem:$src2),
+           !strconcat(OpcodeStr,
+                "\t{$src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2}"),
+       [], SSE_INTSHIFT_ITINS_P.rm>, EVEX_4V, EVEX_K;
 }
 
 defm VPSRLDZ : avx512_shift_rmi<0x72, MRM2r, MRM2m, "vpsrld", X86vsrli,
-                           VR512, v16i32, i512mem, memopv16i32>, EVEX_V512,
-                           EVEX_CD8<32, CD8VF>;
+                           VR512, v16i32, i512mem, memopv16i32, VK16WM>,
+                           EVEX_V512, EVEX_CD8<32, CD8VF>;
 defm VPSRLDZ : avx512_shift_rrm<0xD2, "vpsrld", X86vsrl,
-                           VR512, v16i32, v4i32, bc_v4i32>, EVEX_V512,
+                           VR512, v16i32, v4i32, bc_v4i32, VK16WM>, EVEX_V512,
                            EVEX_CD8<32, CD8VQ>;
                            
 defm VPSRLQZ : avx512_shift_rmi<0x73, MRM2r, MRM2m, "vpsrlq", X86vsrli,
-                           VR512, v8i64, i512mem, memopv8i64>, EVEX_V512,
+                           VR512, v8i64, i512mem, memopv8i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VF>, VEX_W;
 defm VPSRLQZ : avx512_shift_rrm<0xD3, "vpsrlq", X86vsrl,
-                           VR512, v8i64, v2i64, bc_v2i64>, EVEX_V512,
+                           VR512, v8i64, v2i64, bc_v2i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VQ>, VEX_W;
 
 defm VPSLLDZ : avx512_shift_rmi<0x72, MRM6r, MRM6m, "vpslld", X86vshli,
-                           VR512, v16i32, i512mem, memopv16i32>, EVEX_V512,
+                           VR512, v16i32, i512mem, memopv16i32, VK16WM>, EVEX_V512,
                            EVEX_CD8<32, CD8VF>;
 defm VPSLLDZ : avx512_shift_rrm<0xF2, "vpslld", X86vshl,
-                           VR512, v16i32, v4i32, bc_v4i32>, EVEX_V512,
+                           VR512, v16i32, v4i32, bc_v4i32, VK16WM>, EVEX_V512,
                            EVEX_CD8<32, CD8VQ>;
                            
 defm VPSLLQZ : avx512_shift_rmi<0x73, MRM6r, MRM6m, "vpsllq", X86vshli,
-                           VR512, v8i64, i512mem, memopv8i64>, EVEX_V512,
+                           VR512, v8i64, i512mem, memopv8i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VF>, VEX_W;
 defm VPSLLQZ : avx512_shift_rrm<0xF3, "vpsllq", X86vshl,
-                           VR512, v8i64, v2i64, bc_v2i64>, EVEX_V512,
+                           VR512, v8i64, v2i64, bc_v2i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VQ>, VEX_W;
 
 defm VPSRADZ : avx512_shift_rmi<0x72, MRM4r, MRM4m, "vpsrad", X86vsrai,
-                           VR512, v16i32, i512mem, memopv16i32>, EVEX_V512,
-                           EVEX_CD8<32, CD8VF>;
+                           VR512, v16i32, i512mem, memopv16i32, VK16WM>,
+                           EVEX_V512, EVEX_CD8<32, CD8VF>;
 defm VPSRADZ : avx512_shift_rrm<0xE2, "vpsrad", X86vsra,
-                           VR512, v16i32, v4i32, bc_v4i32>, EVEX_V512,
+                           VR512, v16i32, v4i32, bc_v4i32, VK16WM>, EVEX_V512,
                            EVEX_CD8<32, CD8VQ>;
                            
 defm VPSRAQZ : avx512_shift_rmi<0x72, MRM4r, MRM4m, "vpsraq", X86vsrai,
-                           VR512, v8i64, i512mem, memopv8i64>, EVEX_V512,
+                           VR512, v8i64, i512mem, memopv8i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VF>, VEX_W;
 defm VPSRAQZ : avx512_shift_rrm<0xE2, "vpsraq", X86vsra,
-                           VR512, v8i64, v2i64, bc_v2i64>, EVEX_V512,
+                           VR512, v8i64, v2i64, bc_v2i64, VK8WM>, EVEX_V512,
                            EVEX_CD8<64, CD8VQ>, VEX_W;
 
 //===-------------------------------------------------------------------===//