[AArch64] Assembler support for the ARMv8.2a dot product instructions

Dot product is an optional ARMv8.2a extension, see also the public architecture
specification here:
https://developer.arm.com/products/architecture/a-profile/exploration-tools.
This patch adds AArch64 assembler support for these dot product instructions.

Differential Revision: https://reviews.llvm.org/D36515

llvm-svn: 310480
diff --git a/llvm/lib/Target/AArch64/AArch64.td b/llvm/lib/Target/AArch64/AArch64.td
index 436bf11..0c64e83 100644
--- a/llvm/lib/Target/AArch64/AArch64.td
+++ b/llvm/lib/Target/AArch64/AArch64.td
@@ -122,6 +122,10 @@
     "use-reciprocal-square-root", "UseRSqrt", "true",
     "Use the reciprocal square root approximation">;
 
+def FeatureDotProd : SubtargetFeature<
+    "dotprod", "HasDotProd", "true",
+    "Enable dot product support">;
+
 def FeatureNoNegativeImmediates : SubtargetFeature<"no-neg-immediates",
                                         "NegativeImmediates", "false",
                                         "Convert immediates and instructions "
diff --git a/llvm/lib/Target/AArch64/AArch64InstrFormats.td b/llvm/lib/Target/AArch64/AArch64InstrFormats.td
index 077edeb..0e50999 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrFormats.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrFormats.td
@@ -4374,6 +4374,12 @@
   let Inst{4-0}   = Rd;
 }
 
+class BaseSIMDThreeSameVectorDot<bit Q, bit U, string asm, string kind1,
+                                 string kind2> :
+        BaseSIMDThreeSameVector<Q, U, 0b100, 0b10010, V128, asm, kind1, [] > {
+  let AsmString = !strconcat(asm, "{\t$Rd" # kind1 # ", $Rn" # kind2 # ", $Rm" # kind2 # "}");
+}
+
 // All operand sizes distinguished in the encoding.
 multiclass SIMDThreeSameVector<bit U, bits<5> opc, string asm,
                                SDPatternOperator OpNode> {
@@ -6801,6 +6807,16 @@
   let Inst{4-0}   = Rd;
 }
 
+// ARMv8.2 Index Dot product instructions
+class BaseSIMDThreeSameVectorDotIndex<bit Q, bit U, string asm, string dst_kind,
+                                      string lhs_kind, string rhs_kind> :
+        BaseSIMDIndexedTied<Q, U, 0b0, 0b10, 0b1110, V128, V128, V128, VectorIndexS,
+                            asm, "", dst_kind, lhs_kind, rhs_kind, []> {
+  bits<2> idx;
+  let Inst{21}    = idx{0};  // L
+  let Inst{11}    = idx{1};  // H
+}
+
 multiclass SIMDFPIndexed<bit U, bits<4> opc, string asm,
                          SDPatternOperator OpNode> {
   let Predicates = [HasNEON, HasFullFP16] in {
@@ -9596,6 +9612,7 @@
 
 //----------------------------------------------------------------------------
 // Allow the size specifier tokens to be upper case, not just lower.
+def : TokenAlias<".4B", ".4b">;  // Add dot product
 def : TokenAlias<".8B", ".8b">;
 def : TokenAlias<".4H", ".4h">;
 def : TokenAlias<".2S", ".2s">;
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index 5049a39..e7f6eca 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -24,6 +24,8 @@
                                  AssemblerPredicate<"FeatureNEON", "neon">;
 def HasCrypto        : Predicate<"Subtarget->hasCrypto()">,
                                  AssemblerPredicate<"FeatureCrypto", "crypto">;
+def HasDotProd       : Predicate<"Subtarget->hasDotProd()">,
+                                 AssemblerPredicate<"FeatureDotProd", "dotprod">;
 def HasCRC           : Predicate<"Subtarget->hasCRC()">,
                                  AssemblerPredicate<"FeatureCRC", "crc">;
 def HasLSE           : Predicate<"Subtarget->hasLSE()">,
@@ -432,6 +434,18 @@
                        [(int_aarch64_isb (i32 imm32_0_15:$CRm))]>;
 }
 
+// ARMv8.2 Dot Product
+let Predicates = [HasDotProd] in {
+def UDOT2S    : BaseSIMDThreeSameVectorDot<0, 1, "udot", ".2s", ".8b">;
+def SDOT2S    : BaseSIMDThreeSameVectorDot<0, 0, "sdot", ".2s", ".8b">;
+def UDOT4S    : BaseSIMDThreeSameVectorDot<1, 1, "udot", ".4s", ".16b">;
+def SDOT4S    : BaseSIMDThreeSameVectorDot<1, 0, "sdot", ".4s", ".16b">;
+def UDOTIDX2S : BaseSIMDThreeSameVectorDotIndex<0, 1, "udot", ".2s", ".8b", ".4b">;
+def SDOTIDX2S : BaseSIMDThreeSameVectorDotIndex<0, 0, "sdot", ".2s", ".8b", ".4b">;
+def UDOTIDX4S : BaseSIMDThreeSameVectorDotIndex<1, 1, "udot", ".4s", ".16b", ".4b">;
+def SDOTIDX4S : BaseSIMDThreeSameVectorDotIndex<1, 0, "sdot", ".4s", ".16b", ".4b">;
+}
+
 def : InstAlias<"clrex", (CLREX 0xf)>;
 def : InstAlias<"isb", (ISB 0xf)>;
 
diff --git a/llvm/lib/Target/AArch64/AArch64Subtarget.h b/llvm/lib/Target/AArch64/AArch64Subtarget.h
index 5a1f45e..4232fab 100644
--- a/llvm/lib/Target/AArch64/AArch64Subtarget.h
+++ b/llvm/lib/Target/AArch64/AArch64Subtarget.h
@@ -62,6 +62,7 @@
   bool HasFPARMv8 = false;
   bool HasNEON = false;
   bool HasCrypto = false;
+  bool HasDotProd = false;
   bool HasCRC = false;
   bool HasLSE = false;
   bool HasRAS = false;
@@ -201,6 +202,7 @@
   bool hasFPARMv8() const { return HasFPARMv8; }
   bool hasNEON() const { return HasNEON; }
   bool hasCrypto() const { return HasCrypto; }
+  bool hasDotProd() const { return HasDotProd; }
   bool hasCRC() const { return HasCRC; }
   bool hasLSE() const { return HasLSE; }
   bool hasRAS() const { return HasRAS; }
diff --git a/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp b/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
index 4173765..03caaf0 100644
--- a/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
+++ b/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
@@ -1810,6 +1810,8 @@
       .Case(".d", true)
       // Needed for fp16 scalar pairwise reductions
       .Case(".2h", true)
+      // another special case for the ARMv8.2a dot product operand
+      .Case(".4b", true)
       .Default(false);
 }