[AArch64] Allow non-standard INS/DUP encodings

[bsmith]

Summary

This patch allows the AArch64 backend to accept non-standard encodings of DUP (general) and INS (element).

The ARMv8 ARMARM states that for these instructions in A64 state, "Unspecified bits in "imm5" are ignored but should be set to zero by an assembler.", (imm4 for INS). This means that for both of these instructions the disassembler should accept any encoding with these ignored bits set to 1.

[t.p.northover]

Hi Bradley,

This looks good to me.

Tim.

[bsmith]

Thanks, committed as r234896.

lib/Target/AArch64/AArch64InstrFormats.td

def vi8lane : SIMDInsFromElement<".b", v16i8, i32, VectorIndexB> {
     let Inst{14-11} = idx2;
   }
   def vi16lane : SIMDInsFromElement<".h", v8i16, i32, VectorIndexH> {
     bits<3> idx;
     bits<3> idx2;
     let Inst{20-18} = idx;
     let Inst{17-16} = 0b10;
     let Inst{14-12} = idx2;
-    let Inst{11} = 0;
+    let Inst{11} = {?};
   }
   def vi32lane : SIMDInsFromElement<".s", v4i32, i32, VectorIndexS> {
     bits<2> idx;
     bits<2> idx2;
     let Inst{20-19} = idx;
     let Inst{18-16} = 0b100;
     let Inst{14-13} = idx2;
-    let Inst{12-11} = 0;
+    let Inst{12-11} = {?,?};
   }
   def vi64lane : SIMDInsFromElement<".d", v2i64, i64, VectorIndexD> {
     bits<1> idx;
     bits<1> idx2;
     let Inst{20} = idx;
     let Inst{19-16} = 0b1000;
     let Inst{14} = idx2;
-    let Inst{13-11} = 0;
+    let Inst{13-11} = {?,?,?};
   }
 
   // For all forms of the INS instruction, the "mov" mnemonic is the
   // preferred alias. Why they didn't just call the instruction "mov" in
   // the first place is a very good question indeed...
   def : SIMDInsMainMovAlias<".b", !cast<Instruction>(NAME#"vi8gpr"),
                          GPR32, VectorIndexB>;
   def : SIMDInsMainMovAlias<".h", !cast<Instruction>(NAME#"vi16gpr"),

lib/Target/AArch64/AArch64InstrInfo.td

def : Pat<(f32 (int_aarch64_neon_fminv (v2f32 V64:$Rn))),
           (FMINPv2i32p V64:$Rn)>;
 def : Pat<(f64 (int_aarch64_neon_fminv (v2f64 V128:$Rn))),
           (FMINPv2i64p V128:$Rn)>;
 
 //----------------------------------------------------------------------------
 // AdvSIMD INS/DUP instructions
 //----------------------------------------------------------------------------
 
-def DUPv8i8gpr  : SIMDDupFromMain<0, 0b00001, ".8b", v8i8, V64, GPR32>;
-def DUPv16i8gpr : SIMDDupFromMain<1, 0b00001, ".16b", v16i8, V128, GPR32>;
-def DUPv4i16gpr : SIMDDupFromMain<0, 0b00010, ".4h", v4i16, V64, GPR32>;
-def DUPv8i16gpr : SIMDDupFromMain<1, 0b00010, ".8h", v8i16, V128, GPR32>;
-def DUPv2i32gpr : SIMDDupFromMain<0, 0b00100, ".2s", v2i32, V64, GPR32>;
-def DUPv4i32gpr : SIMDDupFromMain<1, 0b00100, ".4s", v4i32, V128, GPR32>;
-def DUPv2i64gpr : SIMDDupFromMain<1, 0b01000, ".2d", v2i64, V128, GPR64>;
+def DUPv8i8gpr  : SIMDDupFromMain<0, {?,?,?,?,1}, ".8b", v8i8, V64, GPR32>;
+def DUPv16i8gpr : SIMDDupFromMain<1, {?,?,?,?,1}, ".16b", v16i8, V128, GPR32>;
+def DUPv4i16gpr : SIMDDupFromMain<0, {?,?,?,1,0}, ".4h", v4i16, V64, GPR32>;
+def DUPv8i16gpr : SIMDDupFromMain<1, {?,?,?,1,0}, ".8h", v8i16, V128, GPR32>;
+def DUPv2i32gpr : SIMDDupFromMain<0, {?,?,1,0,0}, ".2s", v2i32, V64, GPR32>;
+def DUPv4i32gpr : SIMDDupFromMain<1, {?,?,1,0,0}, ".4s", v4i32, V128, GPR32>;
+def DUPv2i64gpr : SIMDDupFromMain<1, {?,1,0,0,0}, ".2d", v2i64, V128, GPR64>;
 
 def DUPv2i64lane : SIMDDup64FromElement;
 def DUPv2i32lane : SIMDDup32FromElement<0, ".2s", v2i32, V64>;
 def DUPv4i32lane : SIMDDup32FromElement<1, ".4s", v4i32, V128>;
 def DUPv4i16lane : SIMDDup16FromElement<0, ".4h", v4i16, V64>;
 def DUPv8i16lane : SIMDDup16FromElement<1, ".8h", v8i16, V128>;
 def DUPv8i8lane  : SIMDDup8FromElement <0, ".8b", v8i8, V64>;
 def DUPv16i8lane : SIMDDup8FromElement <1, ".16b", v16i8, V128>;

test/MC/Disassembler/AArch64/arm64-advsimd.txt

 # CHECK: ins.h v2[1], v15[1]
 # CHECK: ins.b v2[1], v15[1]
 
 # CHECK: ins.d v2[1], v15[0]
 # CHECK: ins.s v2[3], v15[2]
 # CHECK: ins.h v2[7], v15[3]
 # CHECK: ins.b v2[10], v15[5]
 
+# INS/DUP (non-standard)
+0x60 0x0c 0x08 0x4e
+0x60 0x0c 0x0c 0x4e
+0x60 0x0c 0x0c 0x0e
+0x60 0x0c 0x0e 0x4e
+0x60 0x0c 0x0e 0x0e
+0x60 0x0c 0x0f 0x4e
+0x60 0x0c 0x0f 0x0e
+
+# CHECK: dup.2d  v0, x3
+# CHECK: dup.4s  v0, w3
+# CHECK: dup.2s  v0, w3
+# CHECK: dup.8h  v0, w3
+# CHECK: dup.4h  v0, w3
+# CHECK: dup.16b v0, w3
+# CHECK: dup.8b  v0, w3
+
+0xe2 0x75 0x18 0x6e
+0xe2 0x35 0x0c 0x6e
+0xe2 0x15 0x06 0x6e
+0xe2 0x0d 0x03 0x6e
+
+0xe2 0x05 0x18 0x6e
+0xe2 0x55 0x1c 0x6e
+0xe2 0x35 0x1e 0x6e
+0xe2 0x2d 0x15 0x6e
+
+# CHECK: ins.d v2[1], v15[1]
+# CHECK: ins.s v2[1], v15[1]
+# CHECK: ins.h v2[1], v15[1]
+# CHECK: ins.b v2[1], v15[1]
+
+# CHECK: ins.d v2[1], v15[0]
+# CHECK: ins.s v2[3], v15[2]
+# CHECK: ins.h v2[7], v15[3]
+# CHECK: ins.b v2[10], v15[5]
+
 0x00 0x1c 0x20 0x0e
 0x00 0x1c 0x20 0x4e
 
 # CHECK: and.8b  v0, v0, v0
 # CHECK: and.16b v0, v0, v0
 
 0x00 0x1c 0x60 0x0e