[AArch64][SVE2] Asm: support TBL/TBX instructions

[c-rhodes]

Summary

A three sources variant of the TBL instruction is added to the existing
SVE instruction in SVE2. This is implemented with minor changes to the
existing TableGen class. TBX is a new instruction with its own
definition.

The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

lib/Target/AArch64/AArch64SVEInstrInfo.td

let Predicates = [HasSVE2] in {
   defm STNT1B_ZZR_S : sve2_mem_cstnt_vs<0b001, "stnt1b", Z_s, ZPR32>;
   defm STNT1H_ZZR_S : sve2_mem_cstnt_vs<0b011, "stnt1h", Z_s, ZPR32>;
   defm STNT1W_ZZR_S : sve2_mem_cstnt_vs<0b101, "stnt1w", Z_s, ZPR32>;
 
   defm STNT1B_ZZR_D : sve2_mem_cstnt_vs<0b000, "stnt1b", Z_d, ZPR64>;
   defm STNT1H_ZZR_D : sve2_mem_cstnt_vs<0b010, "stnt1h", Z_d, ZPR64>;
   defm STNT1W_ZZR_D : sve2_mem_cstnt_vs<0b100, "stnt1w", Z_d, ZPR64>;
   defm STNT1D_ZZR_D : sve2_mem_cstnt_vs<0b110, "stnt1d", Z_d, ZPR64>;
+
+  // SVE table lookup (three sources)
+  defm TBL_ZZZZ : sve2_int_perm_tbl<"tbl">;
+  defm TBX_ZZZ  : sve2_int_perm_tbx<"tbx">;
 }
 
 let Predicates = [HasSVE2AES] in {
   // SVE2 crypto destructive binary operations
   def AESE_ZZZ_B : sve2_crypto_des_bin_op<0b00, "aese", ZPR8>;
   def AESD_ZZZ_B : sve2_crypto_des_bin_op<0b01, "aesd", ZPR8>;
 
   // SVE2 crypto unary operations

lib/Target/AArch64/SVEInstrFormats.td

multiclass sve_int_perm_dup_i<string asm> {
   def : InstAlias<"mov $Zd, $Sn",
                   (!cast<Instruction>(NAME # _S) ZPR32:$Zd, FPR32asZPR:$Sn, 0), 2>;
   def : InstAlias<"mov $Zd, $Dn",
                   (!cast<Instruction>(NAME # _D) ZPR64:$Zd, FPR64asZPR:$Dn, 0), 2>;
   def : InstAlias<"mov $Zd, $Qn",
                   (!cast<Instruction>(NAME # _Q) ZPR128:$Zd, FPR128asZPR:$Qn, 0), 2>;
 }
 
-class sve_int_perm_tbl<bits<2> sz8_64, string asm, ZPRRegOp zprty,
-                       RegisterOperand VecList>
+class sve_int_perm_tbl<bits<2> sz8_64, bits<2> opc, string asm,
+                       ZPRRegOp zprty, RegisterOperand VecList>
 : I<(outs zprty:$Zd), (ins VecList:$Zn, zprty:$Zm),
   asm, "\t$Zd, $Zn, $Zm",
   "",
   []>, Sched<[]> {
   bits<5> Zd;
   bits<5> Zm;
   bits<5> Zn;
   let Inst{31-24} = 0b00000101;
   let Inst{23-22} = sz8_64;
   let Inst{21}    = 0b1;
   let Inst{20-16} = Zm;
-  let Inst{15-10} = 0b001100;
+  let Inst{15-13} = 0b001;
+  let Inst{12-11} = opc;
+  let Inst{10}    = 0b0;
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 }
 
 multiclass sve_int_perm_tbl<string asm> {
-  def _B : sve_int_perm_tbl<0b00, asm, ZPR8, Z_b>;
-  def _H : sve_int_perm_tbl<0b01, asm, ZPR16, Z_h>;
-  def _S : sve_int_perm_tbl<0b10, asm, ZPR32, Z_s>;
-  def _D : sve_int_perm_tbl<0b11, asm, ZPR64, Z_d>;
+  def _B : sve_int_perm_tbl<0b00, 0b10, asm, ZPR8,  Z_b>;
+  def _H : sve_int_perm_tbl<0b01, 0b10, asm, ZPR16, Z_h>;
+  def _S : sve_int_perm_tbl<0b10, 0b10, asm, ZPR32, Z_s>;
+  def _D : sve_int_perm_tbl<0b11, 0b10, asm, ZPR64, Z_d>;
 
   def : InstAlias<asm # "\t$Zd, $Zn, $Zm",
                  (!cast<Instruction>(NAME # _B) ZPR8:$Zd, ZPR8:$Zn, ZPR8:$Zm), 0>;
   def : InstAlias<asm # "\t$Zd, $Zn, $Zm",
                  (!cast<Instruction>(NAME # _H) ZPR16:$Zd, ZPR16:$Zn, ZPR16:$Zm), 0>;
   def : InstAlias<asm # "\t$Zd, $Zn, $Zm",
                  (!cast<Instruction>(NAME # _S) ZPR32:$Zd, ZPR32:$Zn, ZPR32:$Zm), 0>;
   def : InstAlias<asm # "\t$Zd, $Zn, $Zm",
                  (!cast<Instruction>(NAME # _D) ZPR64:$Zd, ZPR64:$Zn, ZPR64:$Zm), 0>;
 }
 
+multiclass sve2_int_perm_tbl<string asm> {
+  def _B : sve_int_perm_tbl<0b00, 0b01, asm, ZPR8,  ZZ_b>;
+  def _H : sve_int_perm_tbl<0b01, 0b01, asm, ZPR16, ZZ_h>;
+  def _S : sve_int_perm_tbl<0b10, 0b01, asm, ZPR32, ZZ_s>;
+  def _D : sve_int_perm_tbl<0b11, 0b01, asm, ZPR64, ZZ_d>;
+}
+
+class sve2_int_perm_tbx<bits<2> sz8_64, string asm, ZPRRegOp zprty>
+: I<(outs zprty:$Zd), (ins zprty:$Zn, zprty:$Zm),
+  asm, "\t$Zd, $Zn, $Zm",
+  "",
+  []>, Sched<[]> {
+  bits<5> Zd;
+  bits<5> Zm;
+  bits<5> Zn;
+  let Inst{31-24} = 0b00000101;
+  let Inst{23-22} = sz8_64;
+  let Inst{21}    = 0b1;
+  let Inst{20-16} = Zm;
+  let Inst{15-10} = 0b001011;
+  let Inst{9-5}   = Zn;
+  let Inst{4-0}   = Zd;
+}
+
+multiclass sve2_int_perm_tbx<string asm> {
+  def _B : sve2_int_perm_tbx<0b00, asm, ZPR8>;
+  def _H : sve2_int_perm_tbx<0b01, asm, ZPR16>;
+  def _S : sve2_int_perm_tbx<0b10, asm, ZPR32>;
+  def _D : sve2_int_perm_tbx<0b11, asm, ZPR64>;
+}
+
 class sve_int_perm_reverse_z<bits<2> sz8_64, string asm, ZPRRegOp zprty>
 : I<(outs zprty:$Zd), (ins zprty:$Zn),
   asm, "\t$Zd, $Zn",
   "",
   []>, Sched<[]> {
   bits<5> Zd;
   bits<5> Zn;
   let Inst{31-24} = 0b00000101;

test/MC/AArch64/SVE2/tbl-diagnostics.s

+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2  2>&1 < %s| FileCheck %s
+
+tbl z0.b, { z1.b, z2.b }, z3.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: tbl z0.b, { z1.b, z2.b }, z3.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Invalid vector list.
+
+tbl z0.d, { }, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: vector register expected
+// CHECK-NEXT: tbl z0.d, { }, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+tbl z0.d, { z1.d, z2.d, z3.d }, z4.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
+// CHECK-NEXT: tbl z0.d, { z1.d, z2.d, z3.d }, z4.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+tbl z0.d, { z1.d, z2.b }, z3.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: mismatched register size suffix
+// CHECK-NEXT: tbl z0.d, { z1.d, z2.b }, z3.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+tbl z0.d, { z1.d, z21.d }, z3.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: registers must be sequential
+// CHECK-NEXT: tbl z0.d, { z1.d, z21.d }, z3.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+tbl z0.d, { v0.2d, v1.2d }, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
+// CHECK-NEXT: tbl z0.d, { v0.2d, v1.2d }, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Negative tests for instructions that are incompatible with movprfx
+
+movprfx z31.d, p0/z, z6.d
+tbl  z31.d, { z30.d, z31.d }, z31.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: tbl  z31.d, { z30.d, z31.d }, z31.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+movprfx z31, z6
+tbl  z31.d, { z30.d, z31.d }, z31.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: tbl  z31.d, { z30.d, z31.d }, z31.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:

test/MC/AArch64/SVE2/tbl.s

+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve2 - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d - | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+tbl  z28.b, { z29.b, z30.b }, z31.b
+// CHECK-INST: tbl  z28.b, { z29.b, z30.b }, z31.b
+// CHECK-ENCODING: [0xbc,0x2b,0x3f,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: bc 2b 3f 05 <unknown>
+
+tbl  z28.h, { z29.h, z30.h }, z31.h
+// CHECK-INST: tbl  z28.h, { z29.h, z30.h }, z31.h
+// CHECK-ENCODING: [0xbc,0x2b,0x7f,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: bc 2b 7f 05 <unknown>
+
+tbl  z28.s, { z29.s, z30.s }, z31.s
+// CHECK-INST: tbl  z28.s, { z29.s, z30.s }, z31.s
+// CHECK-ENCODING: [0xbc,0x2b,0xbf,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: bc 2b bf 05 <unknown>
+
+tbl  z28.d, { z29.d, z30.d }, z31.d
+// CHECK-INST: tbl  z28.d, { z29.d, z30.d }, z31.d
+// CHECK-ENCODING: [0xbc,0x2b,0xff,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: bc 2b ff 05 <unknown>

test/MC/AArch64/SVE2/tbx-diagnostics.s

+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2  2>&1 < %s| FileCheck %s
+
+tbx z0.b, z1.b, z2.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: tbx z0.b, z1.b, z2.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Negative tests for instructions that are incompatible with movprfx
+
+movprfx z31.d, p0/z, z6.d
+tbx  z31.d, z31.d, z31.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: tbx  z31.d, z31.d, z31.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+movprfx z31, z6
+tbx  z31.d, z31.d, z31.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: tbx  z31.d, z31.d, z31.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:

test/MC/AArch64/SVE2/tbx.s

+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve2 - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d - | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+tbx  z31.b, z31.b, z31.b
+// CHECK-INST: tbx  z31.b, z31.b, z31.b
+// CHECK-ENCODING: [0xff,0x2f,0x3f,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: ff 2f 3f 05 <unknown>
+
+tbx  z31.h, z31.h, z31.h
+// CHECK-INST: tbx  z31.h, z31.h, z31.h
+// CHECK-ENCODING: [0xff,0x2f,0x7f,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: ff 2f 7f 05 <unknown>
+
+tbx  z31.s, z31.s, z31.s
+// CHECK-INST: tbx  z31.s, z31.s, z31.s
+// CHECK-ENCODING: [0xff,0x2f,0xbf,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: ff 2f bf 05 <unknown>
+
+tbx  z31.d, z31.d, z31.d
+// CHECK-INST: tbx  z31.d, z31.d, z31.d
+// CHECK-ENCODING: [0xff,0x2f,0xff,0x05]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: ff 2f ff 05 <unknown>