[RISCV] Add vendor-defined XTHeadBa (address-generation) extension

[philipp.tomsich]

Summary

The vendor-defined XTHeadBa (predating the standard Zba extension)
extension adds an address-generation instruction (th.addsl) with
similar semantics as sh[123]add from Zba. It is supported by the C9xx
cores (e.g., found in the wild in the Allwinner D1) by Alibaba T-Head.

The current (as of this commit) public documentation for XTHeadBa is
available from:

https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf

Support for these instructions has already landed in GNU Binutils:

https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8254c3d2c94ae5458095ea6c25446ba89134b9da

[craig.topper]

Update docs/ReleaseNotes.rst too

llvm/test/CodeGen/RISCV/rv64xtheadba.ll

63

sh2add -> th.addsl

[philipp.tomsich]

• Update llvm/docs/ReleaseNotes.rst
• Fix comment in testcase: sh2add -> th.addsl

[philipp.tomsich]

• fixup .ll tests to not reference an unused check-label

[• pcwang-thead]

llvm/test/CodeGen/RISCV/rv64xtheadba.ll

8

Rename these tests from sh[1,2,3]add to th_addsl_[1,2,3]?

67

And this one too.

[philipp.tomsich]

• updated the test cases to have function names matching the instructions

[reames]

This generally looks reasonable to me, and I support adding this vendor extension once normal code review is complete.

llvm/lib/Target/RISCV/RISCVInstrInfoXTHead.td

16

Stray whitespace.

46

It looks like you're missing the decoder namerspace here.

llvm/test/MC/RISCV/rv64xtheadba-valid.s

10 ↗

(On Diff #494246)

Is 0 a valid encoding? If so, can you add a test covering that?

p.s. The choice to use uimm2 seems slightly odd here since it wastes one value for the shift. Was the 0 encoding possibly used for something else? If it was, maybe we need a different immediate type or an assert somewhere?

[philipp.tomsich]

llvm/test/MC/RISCV/rv64xtheadba-valid.s

10 ↗

(On Diff #494246)

Yes, 0 is a valid encoding.
Yes, it wastes encoding space and is non-sensical and should normally not be generated.

We'll add the test case. We already have the equivalent in the binutils tests (see https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=gas/testsuite/gas/riscv/x-thead-ba.s;h=5081c06e6aa85ec4ec0882780be30b25a80e11b1;hb=8254c3d2c94ae5458095ea6c25446ba89134b9da)

[reames]

llvm/test/MC/RISCV/rv64xtheadba-valid.s

10 ↗

(On Diff #494246)

Great, that's the ease answer implementation wise. :)

[philipp.tomsich]

• adds rv32 test cases
• adds test for th.addsl rd,rs1,rs2,0
• adds decoder namespace

[craig.topper]

Why didn't the DecoderNamespace require changes to RISCVDisassembler::getInstruction to lookup in the correct table?

[craig.topper]

In D143029#4103955, @craig.topper wrote:

Why didn't the DecoderNamespace require changes to RISCVDisassembler::getInstruction to lookup in the correct table?

I applied this patch locally and the tests fail. So I guess it does changes?

[philipp.tomsich]

• Add missing handling of the XTHeadBa decoder namespace in the RISCV Disassembler

[craig.topper]

LGTM

llvm/test/MC/RISCV/rv32xtheadba-valid.s

1 ↗

(On Diff #495023)

I think this "Bitmanip base" was copied from Zbb? Should we adjust it to include T-Head?

llvm/test/MC/RISCV/rv64xtheadba-invalid.s

1 ↗

(On Diff #495023)

Can we merge this test with rv32xtheadba-invalid.s?

llvm/test/MC/RISCV/rv64xtheadba-valid.s

1 ↗

(On Diff #495023)

Can we merge this with rv32xtheadba-valid.s?

[philipp.tomsich]

• make clang-format happy

[craig.topper]

llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp

479

I update this function so we don't have to keep repeating Size = 4. Please rebase.

[philipp.tomsich]

• rebased and updated RISCVDisassembler.cpp
• merged RV32 and RV64 assembler tests into a single file

[philipp.tomsich]

llvm/test/MC/RISCV/rv32xtheadba-valid.s

1 ↗

(On Diff #495023)

This can't be merged, as the mnemonics are different.
Updated the comment.

[craig.topper]

LGTM other than test comment.

llvm/test/MC/RISCV/XTHeadBa-valid.s

2

"base" is wrong here, isn't it? This is the addressing extension. Is it wrong in the Zba test too?

[philipp.tomsich]

llvm/test/MC/RISCV/XTHeadBa-valid.s

2

Yes, it is wrong in the Zba tests: I'll send up NFC patches for those.

That said: I admire your attention to detail.

[philipp.tomsich]

• fixes comment in XTHeadBa-valid.s ("bitmanip base" -> XTHeadBa)

llvm/docs/RISCVUsage.rst

 Vendor extensions are extensions which are not standardized by RISC-V International, and are instead defined by a hardware vendor.  The term vendor extension roughly parallels the definition of a `non-standard` extension from Section 1.3 of the Volume I: RISC-V Unprivileged ISA specification.  In particular, we expect to eventually accept both `custom` extensions and `non-conforming` extensions.
 
 Inclusion of a vendor extension will be considered on a case by case basis.  All proposals should be brought to the bi-weekly RISCV sync calls for discussion.  For a general idea of the factors likely to be considered, please see the `Clang documentation <https://clang.llvm.org/get_involved.html>`_.
 
 It is our intention to follow the naming conventions described in `riscv-non-isa/riscv-toolchain-conventions <https://github.com/riscv-non-isa/riscv-toolchain-conventions#conventions-for-vendor-extensions>`_.  Exceptions to this naming will need to be strongly motivated.
 
 The current vendor extensions supported are:
 
+``XTHeadBa``
+  LLVM implements `the THeadBa (address-generation) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`_  by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
+
 ``XTHeadVdot``
   LLVM implements `version 1.0.0 of the THeadV-family custom instructions specification <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.0/xthead-2022-12-04-2.2.0.pdf>`_ by T-HEAD of Alibaba.  All instructions are prefixed with `th.` as described in the specification, and the riscv-toolchain-convention document linked above.
 
 ``XVentanaCondOps``
   LLVM implements `version 1.0.0 of the VTx-family custom instructions specification <https://github.com/ventanamicro/ventana-custom-extensions/releases/download/v1.0.0/ventana-custom-extensions-v1.0.0.pdf>`_ by Ventana Micro Systems.  All instructions are prefixed with `vt.` as described in the specification, and the riscv-toolchain-convention document linked above.  These instructions are only available for riscv64 at this time.
 
 Specification Documents
 =======================
 For ratified specifications, please refer to the `official RISC-V International
 page <https://riscv.org/technical/specifications/>`_.  Make sure to check the
 `wiki for not yet integrated extensions
 <https://wiki.riscv.org/display/HOME/Recently+Ratified+Extensions>`_.
 

llvm/docs/ReleaseNotes.rst

 
 Changes to the RISC-V Backend
 -----------------------------
 
 * Assembler support for version 1.0.1 of the Zcb extension was added.
 * Zca, Zcf, and Zcd extensions were upgraded to version 1.0.1.
 * vsetvli intrinsics no longer have side effects. They may now be combined,
   moved, deleted, etc. by optimizations.
+* Adds support for the vendor-defined XTHeadBa (address-generation) extension.
 
 Changes to the WebAssembly Backend
 ----------------------------------
 
 * ...
 
 Changes to the Windows Target
 -----------------------------

llvm/lib/Support/RISCVISAInfo.cpp

static const RISCVSupportedExtension SupportedExtensions[] = {
     {"zicboz", RISCVExtensionVersion{1, 0}},
     {"zicbop", RISCVExtensionVersion{1, 0}},
 
     {"svnapot", RISCVExtensionVersion{1, 0}},
     {"svpbmt", RISCVExtensionVersion{1, 0}},
     {"svinval", RISCVExtensionVersion{1, 0}},
 
     // vendor-defined ('X') extensions
+    {"xtheadba", RISCVExtensionVersion{1, 0}},
     {"xtheadvdot", RISCVExtensionVersion{1, 0}},
     {"xventanacondops", RISCVExtensionVersion{1, 0}},
 };
 
 static const RISCVSupportedExtension SupportedExperimentalExtensions[] = {
     {"zihintntl", RISCVExtensionVersion{0, 2}},
 
     {"zca", RISCVExtensionVersion{1, 0}},

llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp

if ((Bytes[0] & 0x3) == 0x3) {
     }
     if (STI.getFeatureBits()[RISCV::FeatureVendorXVentanaCondOps]) {
       LLVM_DEBUG(dbgs() << "Trying Ventana custom opcode table:\n");
       Result = decodeInstruction(DecoderTableVentana32, MI, Insn, Address, this,
                                  STI);
       if (Result != MCDisassembler::Fail)
         return Result;
     }
+    if (STI.getFeatureBits()[RISCV::FeatureVendorXTHeadBa]) {
+      LLVM_DEBUG(dbgs() << "Trying XTHeadBa custom opcode table:\n");
+      Result = decodeInstruction(DecoderTableTHeadBa32, MI, Insn, Address, this,
+                                 STI);
+      if (Result != MCDisassembler::Fail)
+        return Result;
+    }
     if (STI.getFeatureBits()[RISCV::FeatureVendorXTHeadVdot]) {
       LLVM_DEBUG(dbgs() << "Trying XTHeadVdot custom opcode table:\n");
       Result =
           decodeInstruction(DecoderTableTHeadV32, MI, Insn, Address, this, STI);
       if (Result != MCDisassembler::Fail)
         return Result;
     }
 

llvm/lib/Target/RISCV/RISCVFeatures.td

 
 def FeatureVendorXVentanaCondOps
     : SubtargetFeature<"xventanacondops", "HasVendorXVentanaCondOps", "true",
                        "'XVentanaCondOps' (Ventana Conditional Ops)">;
 def HasVendorXVentanaCondOps : Predicate<"Subtarget->hasVendorXVentanaCondOps()">,
                                 AssemblerPredicate<(all_of FeatureVendorXVentanaCondOps),
                                 "'XVentanaCondOps' (Ventana Conditional Ops)">;
 
+def FeatureVendorXTHeadBa
+    : SubtargetFeature<"xtheadba", "HasVendorXTHeadBa", "true",
+                       "'xtheadba' (T-Head address calculation instructions)">;
+def HasVendorXTHeadBa : Predicate<"Subtarget->hasVendorXTHeadBa()">,
+                                  AssemblerPredicate<(all_of FeatureVendorXTHeadBa),
+                                  "'xtheadba' (T-Head address calculation instructions)">;
+
 def FeatureVendorXTHeadVdot
     : SubtargetFeature<"xtheadvdot", "HasVendorXTHeadVdot", "true",
                        "'xtheadvdot' (T-Head Vector Extensions for Dot)",
                        [FeatureStdExtV]>;
 def HasVendorXTHeadVdot : Predicate<"Subtarget->hasVendorXTHeadVdot()">,
                                     AssemblerPredicate<(all_of FeatureVendorXTHeadVdot),
                                     "'xtheadvdot' (T-Head Vector Extensions for Dot)">;
 

llvm/lib/Target/RISCV/RISCVInstrInfoXTHead.td

 //===-- RISCVInstrInfoXTHead.td ----------------------------*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file describes the vendor extensions defined by T-Head of Alibaba.
 //
 //===----------------------------------------------------------------------===//
 
 //===----------------------------------------------------------------------===//
 // Instruction class templates
 //===----------------------------------------------------------------------===//
 class THInstVdotVV<bits<6> funct6, RISCVVFormat opv, dag outs, dag ins,
                    string opcodestr, string argstr>
     : RVInstVV<funct6, opv, outs, ins, opcodestr, argstr> {
   let Inst{26} = 0;
   let Opcode = OPC_CUSTOM_0.Value;
   let DecoderNamespace = "THeadV";
 }
 
 class THInstVdotVX<bits<6> funct6, RISCVVFormat opv, dag outs, dag ins,

 
 // op vd, rs1, vs2, vm (reverse the order of rs1 and vs2)
 class THVdotALUrVX<bits<6> funct6, RISCVVFormat opv, string opcodestr>
     : THInstVdotVX<funct6, opv, (outs VR:$vd),
                    (ins GPR:$rs1, VR:$vs2, VMaskOp:$vm),
                    opcodestr, "$vd, $rs1, $vs2$vm">;
 } // hasSideEffects = 0, mayLoad = 0, mayStore = 0
 
+let Predicates = [HasVendorXTHeadBa], DecoderNamespace = "THeadBa",
+  hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
+class THShiftALU_rri<bits<3> funct3, string opcodestr>
+    : RVInstR<0, funct3, OPC_CUSTOM_0, (outs GPR:$rd),
+              (ins GPR:$rs1, GPR:$rs2, uimm2:$uimm2),
+        opcodestr, "$rd, $rs1, $rs2, $uimm2"> {
+  bits<2> uimm2;
+  let Inst{31-27} = 0;
+  let Inst{26-25} = uimm2;
+}
+
 //===----------------------------------------------------------------------===//
 // Combination of instruction classes.
 // Use these multiclasses to define instructions more easily.
 //===----------------------------------------------------------------------===//
 multiclass THVdotVMAQA_VX<string opcodestr, bits<6> funct6> {
   def _VX : THVdotALUrVX<funct6, OPMVX, opcodestr # ".vx">;
 }
 
 multiclass THVdotVMAQA<string opcodestr, bits<6> funct6> {
   def _VV   : THVdotALUrVV<funct6, OPMVX, opcodestr # ".vv">;
   defm ""   : THVdotVMAQA_VX<opcodestr, funct6>;
 }
 
 //===----------------------------------------------------------------------===//
 // Instructions
 //===----------------------------------------------------------------------===//
+let Predicates = [HasVendorXTHeadBa] in {
+def TH_ADDSL : THShiftALU_rri<0b001, "th.addsl">,
+               Sched<[WriteSHXADD, ReadSHXADD, ReadSHXADD]>;
+} // Predicates = [HasVendorXTHeadBa]
+
 let Predicates = [HasVendorXTHeadVdot],
     Constraints = "@earlyclobber $vd",
     RVVConstraint = WidenV in {
 defm THVdotVMAQA      : THVdotVMAQA<"th.vmaqa",     0b100000>;
 defm THVdotVMAQAU     : THVdotVMAQA<"th.vmaqau",    0b100010>;
 defm THVdotVMAQASU    : THVdotVMAQA<"th.vmaqasu",   0b100100>;
 defm THVdotVMAQAUS    : THVdotVMAQA_VX<"th.vmaqaus",0b100110>;
 }

 multiclass VPatTernaryVMAQA_VV_VX<string intrinsic, string instruction,
                                   list<VTypeInfoToWide> vtilist>
     : VPatTernaryVMAQA_VV<intrinsic, instruction, vtilist>,
       VPatTernaryVMAQA_VX<intrinsic, instruction, vtilist>;
 
 //===----------------------------------------------------------------------===//
 // Pseudo-instructions and codegen patterns
 //===----------------------------------------------------------------------===//
+let Predicates = [HasVendorXTHeadBa] in {
+def : Pat<(add GPR:$rs1, (shl GPR:$rs2, uimm2:$uimm2)),
+          (TH_ADDSL GPR:$rs1, GPR:$rs2, uimm2:$uimm2)>;
+
+// Reuse complex patterns from StdExtZba
+def : Pat<(add sh1add_op:$rs1, non_imm12:$rs2),
+          (TH_ADDSL GPR:$rs2, sh1add_op:$rs1, 1)>;
+def : Pat<(add sh2add_op:$rs1, non_imm12:$rs2),
+          (TH_ADDSL GPR:$rs2, sh2add_op:$rs1, 2)>;
+def : Pat<(add sh3add_op:$rs1, non_imm12:$rs2),
+          (TH_ADDSL GPR:$rs2, sh3add_op:$rs1, 3)>;
+
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 6)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 1), 1)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 10)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 2), 1)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 18)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 3), 1)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 12)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 1), 2)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 20)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 2), 2)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 36)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 3), 2)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 24)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 1), 3)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 40)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 2), 3)>;
+def : Pat<(add (mul_oneuse GPR:$rs1, (XLenVT 72)), GPR:$rs2),
+          (TH_ADDSL GPR:$rs2, (TH_ADDSL GPR:$rs1, GPR:$rs1, 3), 3)>;
+
+def : Pat<(add GPR:$r, CSImm12MulBy4:$i),
+          (TH_ADDSL GPR:$r, (ADDI X0, (SimmShiftRightBy2XForm CSImm12MulBy4:$i)), 2)>;
+def : Pat<(add GPR:$r, CSImm12MulBy8:$i),
+          (TH_ADDSL GPR:$r, (ADDI X0, (SimmShiftRightBy3XForm CSImm12MulBy8:$i)), 3)>;
+
+def : Pat<(mul GPR:$r, C3LeftShift:$i),
+          (SLLI (TH_ADDSL GPR:$r, GPR:$r, 1),
+                (TrailingZeros C3LeftShift:$i))>;
+def : Pat<(mul GPR:$r, C5LeftShift:$i),
+          (SLLI (TH_ADDSL GPR:$r, GPR:$r, 2),
+                (TrailingZeros C5LeftShift:$i))>;
+def : Pat<(mul GPR:$r, C9LeftShift:$i),
+          (SLLI (TH_ADDSL GPR:$r, GPR:$r, 3),
+                (TrailingZeros C9LeftShift:$i))>;
+
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 11)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 2), 1)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 19)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 3), 1)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 13)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 1), 2)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 21)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 2), 2)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 37)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 3), 2)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 25)),
+    (TH_ADDSL (TH_ADDSL GPR:$r, GPR:$r, 2), (TH_ADDSL GPR:$r, GPR:$r, 2), 2)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 41)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 2), 3)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 73)),
+    (TH_ADDSL GPR:$r, (TH_ADDSL GPR:$r, GPR:$r, 3), 3)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 27)),
+    (TH_ADDSL (TH_ADDSL GPR:$r, GPR:$r, 3), (TH_ADDSL GPR:$r, GPR:$r, 3), 1)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 45)),
+    (TH_ADDSL (TH_ADDSL GPR:$r, GPR:$r, 3), (TH_ADDSL GPR:$r, GPR:$r, 3), 2)>;
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 81)),
+    (TH_ADDSL (TH_ADDSL GPR:$r, GPR:$r, 3), (TH_ADDSL GPR:$r, GPR:$r, 3), 3)>;
+
+def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 200)),
+    (SLLI (TH_ADDSL (TH_ADDSL GPR:$r, GPR:$r, 2),
+           (TH_ADDSL GPR:$r, GPR:$r, 2), 2), 3)>;
+} // Predicates = [HasVendorXTHeadBa]
+
 defm PseudoTHVdotVMAQA      : VPseudoVMAQA_VV_VX;
 defm PseudoTHVdotVMAQAU     : VPseudoVMAQA_VV_VX;
 defm PseudoTHVdotVMAQASU    : VPseudoVMAQA_VV_VX;
 defm PseudoTHVdotVMAQAUS    : VPseudoVMAQA_VX;
 
 let Predicates = [HasVendorXTHeadVdot] in {
 defm : VPatTernaryVMAQA_VV_VX<"int_riscv_th_vmaqa",  "PseudoTHVdotVMAQA",  AllQuadWidenableInt8NoVLMulVectors>;
 defm : VPatTernaryVMAQA_VV_VX<"int_riscv_th_vmaqau", "PseudoTHVdotVMAQAU", AllQuadWidenableInt8NoVLMulVectors>;
 defm : VPatTernaryVMAQA_VV_VX<"int_riscv_th_vmaqasu","PseudoTHVdotVMAQASU",AllQuadWidenableInt8NoVLMulVectors>;
 defm : VPatTernaryVMAQA_VX<"int_riscv_th_vmaqaus",   "PseudoTHVdotVMAQAUS",AllQuadWidenableInt8NoVLMulVectors>;
 }

llvm/test/CodeGen/RISCV/attributes.ll

 ; RUN: llc -mtriple=riscv64 -mattr=+zbkb,+zbkc,+zbkx,+zksed,+zksh %s -o - | FileCheck --check-prefix=RV64COMBINEINTOZKS %s
 ; RUN: llc -mtriple=riscv64 -mattr=+zicbom %s -o - | FileCheck --check-prefix=RV64ZICBOM %s
 ; RUN: llc -mtriple=riscv64 -mattr=+zicboz %s -o - | FileCheck --check-prefix=RV64ZICBOZ %s
 ; RUN: llc -mtriple=riscv64 -mattr=+zicbop %s -o - | FileCheck --check-prefix=RV64ZICBOP %s
 ; RUN: llc -mtriple=riscv64 -mattr=+svnapot %s -o - | FileCheck --check-prefix=RV64SVNAPOT %s
 ; RUN: llc -mtriple=riscv64 -mattr=+svpbmt %s -o - | FileCheck --check-prefix=RV64SVPBMT %s
 ; RUN: llc -mtriple=riscv64 -mattr=+svinval %s -o - | FileCheck --check-prefix=RV64SVINVAL %s
 ; RUN: llc -mtriple=riscv64 -mattr=+xventanacondops %s -o - | FileCheck --check-prefix=RV64XVENTANACONDOPS %s
+; RUN: llc -mtriple=riscv64 -mattr=+xtheadba %s -o - | FileCheck --check-prefix=RV64XTHEADBA %s
 ; RUN: llc -mtriple=riscv64 -mattr=+xtheadvdot %s -o - | FileCheck --check-prefix=RV64XTHEADVDOT %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-zawrs %s -o - | FileCheck --check-prefix=RV64ZAWRS %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-ztso %s -o - | FileCheck --check-prefix=RV64ZTSO %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-zca %s -o - | FileCheck --check-prefix=RV64ZCA %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-zcb %s -o - | FileCheck --check-prefix=RV64ZCB %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-zcd %s -o - | FileCheck --check-prefix=RV64ZCD %s
 
 ; RV32M: .attribute 5, "rv32i2p0_m2p0"

 ; RV64ZICBOM: .attribute 5, "rv64i2p0_zicbom1p0"
 ; RV64ZICBOZ: .attribute 5, "rv64i2p0_zicboz1p0"
 ; RV64ZAWRS: .attribute 5, "rv64i2p0_zawrs1p0"
 ; RV64ZICBOP: .attribute 5, "rv64i2p0_zicbop1p0"
 ; RV64SVNAPOT: .attribute 5, "rv64i2p0_svnapot1p0"
 ; RV64SVPBMT: .attribute 5, "rv64i2p0_svpbmt1p0"
 ; RV64SVINVAL: .attribute 5, "rv64i2p0_svinval1p0"
 ; RV64XVENTANACONDOPS: .attribute 5, "rv64i2p0_xventanacondops1p0"
+; RV64XTHEADBA: .attribute 5, "rv64i2p0_xtheadba1p0"
 ; RV64XTHEADVDOT: .attribute 5, "rv64i2p0_f2p0_d2p0_v1p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0_xtheadvdot1p0"
 ; RV64ZTSO: .attribute 5, "rv64i2p0_ztso0p1"
 ; RV64ZCA: .attribute 5, "rv64i2p0_zca1p0"
 ; RV64ZCB: .attribute 5, "rv64i2p0_zca1p0_zcb1p0"
 ; RV64ZCD: .attribute 5, "rv64i2p0_zcd1p0"
 
 define i32 @addi(i32 %a) {
   %1 = add i32 %a, 1
   ret i32 %1
 }

llvm/test/CodeGen/RISCV/rv32xtheadba.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --extra_scrub
+; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefixes=RV32I
+; RUN: llc -mtriple=riscv32 -mattr=+m,+xtheadba -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefixes=RV32XTHEADBA
+
+define signext i16 @th_addsl_1(i64 %0, ptr %1) {
+; RV32I-LABEL: th_addsl_1:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a0, a0, 1
+; RV32I-NEXT:    add a0, a2, a0
+; RV32I-NEXT:    lh a0, 0(a0)
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: th_addsl_1:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a2, a0, 1
+; RV32XTHEADBA-NEXT:    lh a0, 0(a0)
+; RV32XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i16, ptr %1, i64 %0
+  %4 = load i16, ptr %3
+  ret i16 %4
+}
+
+define signext i32 @th_addsl_2(i64 %0, ptr %1) {
+; RV32I-LABEL: th_addsl_2:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a0, a0, 2
+; RV32I-NEXT:    add a0, a2, a0
+; RV32I-NEXT:    lw a0, 0(a0)
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: th_addsl_2:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a2, a0, 2
+; RV32XTHEADBA-NEXT:    lw a0, 0(a0)
+; RV32XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i32, ptr %1, i64 %0
+  %4 = load i32, ptr %3
+  ret i32 %4
+}
+
+define i64 @th_addsl_3(i64 %0, ptr %1) {
+; RV32I-LABEL: th_addsl_3:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a0, a0, 3
+; RV32I-NEXT:    add a2, a2, a0
+; RV32I-NEXT:    lw a0, 0(a2)
+; RV32I-NEXT:    lw a1, 4(a2)
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: th_addsl_3:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a1, a2, a0, 3
+; RV32XTHEADBA-NEXT:    lw a0, 0(a1)
+; RV32XTHEADBA-NEXT:    lw a1, 4(a1)
+; RV32XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i64, ptr %1, i64 %0
+  %4 = load i64, ptr %3
+  ret i64 %4
+}
+
+; Type legalization inserts a sext_inreg after the first add. That add will be
+; selected as th.addsl which does not sign extend. SimplifyDemandedBits is unable
+; to remove the sext_inreg because it has multiple uses. The ashr will use the
+; sext_inreg to become sraiw. This leaves the sext_inreg only used by the shl.
+; If the shl is selected as sllw, we don't need the sext_inreg.
+define i64 @th_addsl_2_extra_sext(i32 %x, i32 %y, i32 %z) {
+; RV32I-LABEL: th_addsl_2_extra_sext:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a0, a0, 2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    sll a1, a2, a0
+; RV32I-NEXT:    srai a2, a0, 2
+; RV32I-NEXT:    mul a0, a1, a2
+; RV32I-NEXT:    mulh a1, a1, a2
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: th_addsl_2_extra_sext:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV32XTHEADBA-NEXT:    sll a1, a2, a0
+; RV32XTHEADBA-NEXT:    srai a2, a0, 2
+; RV32XTHEADBA-NEXT:    mul a0, a1, a2
+; RV32XTHEADBA-NEXT:    mulh a1, a1, a2
+; RV32XTHEADBA-NEXT:    ret
+  %a = shl i32 %x, 2
+  %b = add i32 %a, %y
+  %c = shl i32 %z, %b
+  %d = ashr i32 %b, 2
+  %e = sext i32 %c to i64
+  %f = sext i32 %d to i64
+  %g = mul i64 %e, %f
+  ret i64 %g
+}
+
+define i32 @addmul6(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul6:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 6
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul6:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 6
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul10(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul10:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 10
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul10:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 10
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul12(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul12:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 12
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul12:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 12
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul18(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul18:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 18
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul18:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 18
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul20(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul20:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 20
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul20:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 20
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul24(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul24:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 24
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul24:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 24
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul36(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul36:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 36
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul36:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 36
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul40(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul40:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 40
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul40:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 40
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @addmul72(i32 %a, i32 %b) {
+; RV32I-LABEL: addmul72:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a2, 72
+; RV32I-NEXT:    mul a0, a0, a2
+; RV32I-NEXT:    add a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: addmul72:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV32XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 72
+  %d = add i32 %c, %b
+  ret i32 %d
+}
+
+define i32 @mul96(i32 %a) {
+; RV32I-LABEL: mul96:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 96
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: mul96:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV32XTHEADBA-NEXT:    slli a0, a0, 5
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 96
+  ret i32 %c
+}
+
+define i32 @mul160(i32 %a) {
+; RV32I-LABEL: mul160:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 160
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: mul160:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    slli a0, a0, 5
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 160
+  ret i32 %c
+}
+
+define i32 @mul200(i32 %a) {
+; RV32I-LABEL: mul200:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 200
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: mul200:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV32XTHEADBA-NEXT:    slli a0, a0, 3
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 200
+  ret i32 %c
+}
+
+define i32 @mul288(i32 %a) {
+; RV32I-LABEL: mul288:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 288
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32XTHEADBA-LABEL: mul288:
+; RV32XTHEADBA:       # %bb.0:
+; RV32XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV32XTHEADBA-NEXT:    slli a0, a0, 5
+; RV32XTHEADBA-NEXT:    ret
+  %c = mul i32 %a, 288
+  ret i32 %c
+}
+

llvm/test/CodeGen/RISCV/rv64xtheadba.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --extra_scrub
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefixes=RV64I
+; RUN: llc -mtriple=riscv64 -mattr=+m,+xtheadba -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefixes=RV64XTHEADBA
+
+define signext i16 @th_addsl_1(i64 %0, ptr %1) {
+; RV64I-LABEL: th_addsl_1:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 1
+; RV64I-NEXT:    add a0, a1, a0
+; RV64I-NEXT:    lh a0, 0(a0)
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: th_addsl_1:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV64XTHEADBA-NEXT:    lh a0, 0(a0)
+; RV64XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i16, ptr %1, i64 %0
+  %4 = load i16, ptr %3
+  ret i16 %4
+}
+
+define signext i32 @th_addsl_2(i64 %0, ptr %1) {
+; RV64I-LABEL: th_addsl_2:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 2
+; RV64I-NEXT:    add a0, a1, a0
+; RV64I-NEXT:    lw a0, 0(a0)
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: th_addsl_2:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV64XTHEADBA-NEXT:    lw a0, 0(a0)
+; RV64XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i32, ptr %1, i64 %0
+  %4 = load i32, ptr %3
+  ret i32 %4
+}
+
+define i64 @th_addsl_3(i64 %0, ptr %1) {
+; RV64I-LABEL: th_addsl_3:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 3
+; RV64I-NEXT:    add a0, a1, a0
+; RV64I-NEXT:    ld a0, 0(a0)
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: th_addsl_3:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV64XTHEADBA-NEXT:    ld a0, 0(a0)
+; RV64XTHEADBA-NEXT:    ret
+  %3 = getelementptr inbounds i64, ptr %1, i64 %0
+  %4 = load i64, ptr %3
+  ret i64 %4
+}
+
+; Type legalization inserts a sext_inreg after the first add. That add will be
+; selected as th.addsl which does not sign extend. SimplifyDemandedBits is unable
+; to remove the sext_inreg because it has multiple uses. The ashr will use the
+; sext_inreg to become sraiw. This leaves the sext_inreg only used by the shl.
+; If the shl is selected as sllw, we don't need the sext_inreg.
+define i64 @th_addsl_2_extra_sext(i32 %x, i32 %y, i32 %z) {
+; RV64I-LABEL: th_addsl_2_extra_sext:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    sllw a1, a2, a0
+; RV64I-NEXT:    sraiw a0, a0, 2
+; RV64I-NEXT:    mul a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: th_addsl_2_extra_sext:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV64XTHEADBA-NEXT:    sllw a1, a2, a0
+; RV64XTHEADBA-NEXT:    sraiw a0, a0, 2
+; RV64XTHEADBA-NEXT:    mul a0, a1, a0
+; RV64XTHEADBA-NEXT:    ret
+  %a = shl i32 %x, 2
+  %b = add i32 %a, %y
+  %c = shl i32 %z, %b
+  %d = ashr i32 %b, 2
+  %e = sext i32 %c to i64
+  %f = sext i32 %d to i64
+  %g = mul i64 %e, %f
+  ret i64 %g
+}
+
+define i64 @addmul6(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul6:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 6
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul6:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 6
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul10(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul10:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 10
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul10:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 10
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul12(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul12:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 12
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul12:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 12
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul18(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul18:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 18
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul18:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 1
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 18
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul20(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul20:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 20
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul20:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 20
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul24(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul24:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 24
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul24:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 24
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul36(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul36:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 36
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul36:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 2
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 36
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul40(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul40:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 40
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul40:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 40
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul72(i64 %a, i64 %b) {
+; RV64I-LABEL: addmul72:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 72
+; RV64I-NEXT:    mul a0, a0, a2
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: addmul72:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV64XTHEADBA-NEXT:    th.addsl a0, a1, a0, 3
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 72
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @mul96(i64 %a) {
+; RV64I-LABEL: mul96:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 96
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: mul96:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 1
+; RV64XTHEADBA-NEXT:    slli a0, a0, 5
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 96
+  ret i64 %c
+}
+
+define i64 @mul160(i64 %a) {
+; RV64I-LABEL: mul160:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 160
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: mul160:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    slli a0, a0, 5
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 160
+  ret i64 %c
+}
+
+define i64 @mul200(i64 %a) {
+; RV64I-LABEL: mul200:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 200
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: mul200:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 2
+; RV64XTHEADBA-NEXT:    slli a0, a0, 3
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 200
+  ret i64 %c
+}
+
+define i64 @mul288(i64 %a) {
+; RV64I-LABEL: mul288:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 288
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: mul288:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a0, a0, a0, 3
+; RV64XTHEADBA-NEXT:    slli a0, a0, 5
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 288
+  ret i64 %c
+}
+

llvm/test/MC/RISCV/XTHeadBa-invalid.s

+# RUN: not llvm-mc -triple riscv32 -mattr=+xtheadba < %s 2>&1 | FileCheck %s
+# RUN: not llvm-mc -triple riscv64 -mattr=+xtheadba < %s 2>&1 | FileCheck %s
+
+# Too few operands
+th.addsl t0, t1 # CHECK: :[[@LINE]]:1: error: too few operands for instruction
+# Too few operands
+th.addsl t0, t1, t2 # CHECK: :[[@LINE]]:1: error: too few operands for instruction
+# Immediate operand out of range
+th.addsl t0, t1, t2, 4 # CHECK: :[[@LINE]]:22: error: immediate must be an integer in the range [0, 3]
+# Immediate operand out of range
+th.addsl t0, t1, t2, -1 # CHECK: :[[@LINE]]:22: error: immediate must be an integer in the range [0, 3]

llvm/test/MC/RISCV/XTHeadBa-valid.s

+# With XTHeadBa (address generation) extension:
+# RUN: llvm-mc %s -triple=riscv32 -mattr=+xtheadba -show-encoding \
+# RUN:     | FileCheck -check-prefixes=CHECK-ASM,CHECK-ASM-AND-OBJ %s
+# RUN: llvm-mc %s -triple=riscv64 -mattr=+xtheadba -show-encoding \
+# RUN:     | FileCheck -check-prefixes=CHECK-ASM,CHECK-ASM-AND-OBJ %s
+# RUN: llvm-mc -filetype=obj -triple=riscv32 -mattr=+xtheadba < %s \
+# RUN:     | llvm-objdump --mattr=+xtheadba -d -r - \
+# RUN:     | FileCheck --check-prefix=CHECK-ASM-AND-OBJ %s
+# RUN: llvm-mc -filetype=obj -triple=riscv64 -mattr=+xtheadba < %s \
+# RUN:     | llvm-objdump --mattr=+xtheadba -d -r - \
+# RUN:     | FileCheck --check-prefix=CHECK-ASM-AND-OBJ %s
+
+# CHECK-ASM-AND-OBJ: th.addsl t0, t1, t2, 0
+# CHECK-ASM: encoding: [0x8b,0x12,0x73,0x00]
+th.addsl t0, t1, t2, 0
+# CHECK-ASM-AND-OBJ: th.addsl t0, t1, t2, 1
+# CHECK-ASM: encoding: [0x8b,0x12,0x73,0x02]
+th.addsl t0, t1, t2, 1
+# CHECK-ASM-AND-OBJ: th.addsl t0, t1, t2, 2
+# CHECK-ASM: encoding: [0x8b,0x12,0x73,0x04]
+th.addsl t0, t1, t2, 2
+# CHECK-ASM-AND-OBJ: th.addsl t0, t1, t2, 3
+# CHECK-ASM: encoding: [0x8b,0x12,0x73,0x06]
+th.addsl t0, t1, t2, 3