[RISCV] Add matching of codegen patterns to RISCV Bit Manipulation Zbbp asm instructions

[PaoloS]

Summary

This patch provides optimization of bit manipulation operations by enabling
the +experimental-b target feature.
It adds matching of single block patterns of instructions to specific bit-manip
instructions belonging to both the permutation and the base subsets of the
experimental B extension of RISC-V.
It adds also the correspondent codegen tests.

This patch is based on Clifford Wolf's proposal for the bit manipulation
extension of RISCV:
https://github.com/riscv/riscv-bitmanip/blob/master/bitmanip-0.92.pdf

[PaoloS]

Added missing pattern-matching for *w instructions.
Added tests.
Added both i32 and i64 code versions to both i32 and i64 test files.
Removed NOT labels from tests.

[lewis-revill]

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

870

I'm not quite sure given the tests here how these patterns are used? Is @llvm.fshl.i32 lowered to this pattern?

[PaoloS]

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

870

Precisely.

[PaoloS]

Updated the test:

• the tests have been updated from the top of all the sub-patches together so that they are exactly the same as they would be if updated with the whole final patch.
• labels specific to the sub-extension have been added alongside the generic RISCVIB label (that activates all the sub-extensions) so that we can see how differently the patterns are matched with the specific subextension or with all of them together.
• the tests will probably fail if run by checking out the commit of a subextension and if updated they'll change. These tests are designed to work with the final squashed patch.

Added the pattern-matching for rori and roriw.
Added the correspondent tests.

[lewis-revill]

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

663

A little confused by this comment here, do you mean there is no pattern matching rather than there is no encoding.

llvm/test/CodeGen/RISCV/rv32Zbbp.ll

433

Nitpick: See comment on first patch regarding these extra tests.

[lewis-revill]

See nitpicks, otherwise LGTM

[PaoloS]

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

663

Yeah, I mean there's no encoding at all. By consequence no pattern-matching

[PaoloS]

llvm/test/CodeGen/RISCV/rv32Zbbp.ll

433

I'm going to comment those.

[PaoloS]

Aligned complex pattern declaration.
Clarified comment about missing 'roli' instruction.
Added comment to inefficient tests.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::UDIVREM, XLenVT, Expand);
   setOperationAction(ISD::SMUL_LOHI, XLenVT, Expand);
   setOperationAction(ISD::UMUL_LOHI, XLenVT, Expand);
 
   setOperationAction(ISD::SHL_PARTS, XLenVT, Custom);
   setOperationAction(ISD::SRL_PARTS, XLenVT, Custom);
   setOperationAction(ISD::SRA_PARTS, XLenVT, Custom);
 
+  if (!(Subtarget.hasStdExtZbb() || Subtarget.hasStdExtZbp())) {
     setOperationAction(ISD::ROTL, XLenVT, Expand);
     setOperationAction(ISD::ROTR, XLenVT, Expand);
+  }
 
   if (!Subtarget.hasStdExtZbp())
     setOperationAction(ISD::BSWAP, XLenVT, Expand);
 
   if (!Subtarget.hasStdExtZbb()) {
     setOperationAction(ISD::CTTZ, XLenVT, Expand);
     setOperationAction(ISD::CTLZ, XLenVT, Expand);
     setOperationAction(ISD::CTPOP, XLenVT, Expand);

llvm/lib/Target/RISCV/RISCVInstrInfoB.td

 // Codegen patterns
 //===----------------------------------------------------------------------===//
 def SLOIPat  : ComplexPattern<XLenVT, 2, "SelectSLOI", [or]>;
 def SROIPat  : ComplexPattern<XLenVT, 2, "SelectSROI", [or]>;
 def SLOIWPat  : ComplexPattern<XLenVT, 2, "SelectSLOIW", [sext_inreg]>;
 def SROIWPat  : ComplexPattern<XLenVT, 2, "SelectSROIW", [or]>;
 def SLLIUWPat  : ComplexPattern<XLenVT, 2, "SelectSLLIUW", [and]>;
 
+let Predicates = [HasStdExtZbbOrZbp] in {
+def : Pat<(and GPR:$rs1, (not GPR:$rs2)), (ANDN GPR:$rs1, GPR:$rs2)>;
+def : Pat<(or  GPR:$rs1, (not GPR:$rs2)), (ORN  GPR:$rs1, GPR:$rs2)>;
+def : Pat<(xor GPR:$rs1, (not GPR:$rs2)), (XNOR GPR:$rs1, GPR:$rs2)>;
+} // Predicates = [HasStdExtZbbOrZbp]
+
 let Predicates = [HasStdExtZbb] in {
 def : Pat<(xor (shl (xor GPR:$rs1, -1), GPR:$rs2), -1),
           (SLO GPR:$rs1, GPR:$rs2)>;
 def : Pat<(xor (srl (xor GPR:$rs1, -1), GPR:$rs2), -1),
           (SRO GPR:$rs1, GPR:$rs2)>;
 def : Pat<(SLOIPat GPR:$rs1, uimmlog2xlen:$shamt),
           (SLOI GPR:$rs1, uimmlog2xlen:$shamt)>;
 def : Pat<(SROIPat GPR:$rs1, uimmlog2xlen:$shamt),
           (SROI GPR:$rs1, uimmlog2xlen:$shamt)>;
 } // Predicates = [HasStdExtZbb]
 
+let Predicates = [HasStdExtZbbOrZbp] in {
+def : Pat<(rotl GPR:$rs1, GPR:$rs2), (ROL GPR:$rs1, GPR:$rs2)>;
+def : Pat<(fshl GPR:$rs1, GPR:$rs1, GPR:$rs2), (ROL GPR:$rs1, GPR:$rs2)>;
+def : Pat<(rotr GPR:$rs1, GPR:$rs2), (ROR GPR:$rs1, GPR:$rs2)>;
+def : Pat<(fshr GPR:$rs1, GPR:$rs1, GPR:$rs2), (ROR GPR:$rs1, GPR:$rs2)>;
+} // Predicates = [HasStdExtZbbOrZbp]
+
 let Predicates = [HasStdExtZbp, IsRV32] in {
 def : Pat<(or (or (and (srl GPR:$rs1, (i32 1)), (i32 0x55555555)), GPR:$rs1),
               (and (shl GPR:$rs1, (i32 1)), (i32 0xAAAAAAAA))),
           (GORCI GPR:$rs1, (i32 1))>;
 def : Pat<(or (or (and (srl GPR:$rs1, (i32 2)), (i32 0x33333333)), GPR:$rs1),
               (and (shl GPR:$rs1, (i32 2)), (i32 0xCCCCCCCC))),
           (GORCI GPR:$rs1, (i32 2))>;
 def : Pat<(or (or (and (srl GPR:$rs1, (i32 4)), (i32 0x0F0F0F0F)), GPR:$rs1),

 def : Pat<(umin GPR:$rs1, GPR:$rs2), (MINU GPR:$rs1, GPR:$rs2)>;
 def : Pat<(riscv_selectcc GPR:$rs1, GPR:$rs2, (XLenVT 12), GPR:$rs1, GPR:$rs2),
           (MINU  GPR:$rs1, GPR:$rs2)>;
 def : Pat<(umax GPR:$rs1, GPR:$rs2), (MAXU GPR:$rs1, GPR:$rs2)>;
 def : Pat<(riscv_selectcc GPR:$rs2, GPR:$rs1, (XLenVT 12), GPR:$rs1, GPR:$rs2),
           (MAXU  GPR:$rs1, GPR:$rs2)>;
 } // Predicates = [HasStdExtZbb]
 
+let Predicates = [HasStdExtZbbOrZbp, IsRV32] in
+def : Pat<(or (and GPR:$rs1, 0x0000FFFF), (shl GPR:$rs2, (i32 16))),
+          (PACK GPR:$rs1, GPR:$rs2)>;
+let Predicates = [HasStdExtZbbOrZbp, IsRV64] in
+def : Pat<(or (and GPR:$rs1, 0x00000000FFFFFFFF), (shl GPR:$rs2, (i64 32))),
+          (PACK GPR:$rs1, GPR:$rs2)>;
+let Predicates = [HasStdExtZbbOrZbp, IsRV32] in
+def : Pat<(or (and GPR:$rs2, 0xFFFF0000), (srl GPR:$rs1, (i32 16))),
+          (PACKU GPR:$rs1, GPR:$rs2)>;
+let Predicates = [HasStdExtZbbOrZbp, IsRV64] in
+def : Pat<(or (and GPR:$rs2, 0xFFFFFFFF00000000), (srl GPR:$rs1, (i64 32))),
+          (PACKU GPR:$rs1, GPR:$rs2)>;
+let Predicates = [HasStdExtZbbOrZbp] in
+def : Pat<(or (and (shl GPR:$rs2, (XLenVT 8)), 0xFF00),
+              (and GPR:$rs1, 0x00FF)),
+          (PACKH GPR:$rs1, GPR:$rs2)>;
+
 let Predicates = [HasStdExtZbp, IsRV32] in {
 def : Pat<(or (or (and (shl GPR:$rs1, (i32 8)), (i32 0x00FF0000)),
                   (and GPR:$rs1, (i32 0xFF0000FF))),
               (and (srl GPR:$rs1, (i32 8)), (i32 0x0000FF00))),
           (SHFLI GPR:$rs1, (i32 8))>;
 def : Pat<(or (or (and (shl GPR:$rs1, (i32 4)), (i32 0x0F000F00)),
                   (and GPR:$rs1, (i32 0xF00FF00F))),
               (and (srl GPR:$rs1, (i32 4)), (i32 0x00F000F0))),

 def : Pat<(xor (riscv_srlw (xor GPR:$rs1, -1), GPR:$rs2), -1),
           (SROW GPR:$rs1, GPR:$rs2)>;
 def : Pat<(SLOIWPat GPR:$rs1, uimmlog2xlen:$shamt),
           (SLOIW GPR:$rs1, uimmlog2xlen:$shamt)>;
 def : Pat<(SROIWPat GPR:$rs1, uimmlog2xlen:$shamt),
           (SROIW GPR:$rs1, uimmlog2xlen:$shamt)>;
 } // Predicates = [HasStdExtZbb, IsRV64]
 
+let Predicates = [HasStdExtZbbOrZbp, IsRV64] in {
+def : Pat<(or (riscv_sllw (assertsexti32 GPR:$rs1), (assertsexti32 GPR:$rs2)),
+              (riscv_srlw (assertsexti32 GPR:$rs1),
+                          (sub (i64 0), (assertsexti32 GPR:$rs2)))),
+          (ROLW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(or (riscv_sllw (assertsexti32 GPR:$rs1),
+                          (sub (i64 0), (assertsexti32 GPR:$rs2))),
+              (riscv_srlw (assertsexti32 GPR:$rs1), (assertsexti32 GPR:$rs2))),
+          (RORW GPR:$rs1, GPR:$rs2)>;
+} // Predicates = [HasStdExtZbbOrZbp, IsRV64]
+
 let Predicates = [HasStdExtZbp, IsRV64] in {
 def : Pat<(sext_inreg (or (or (and (srl GPR:$rs1, (i64 1)), (i64 0x55555555)),
                               GPR:$rs1),
                           (and (shl GPR:$rs1, (i64 1)), (i64 0xAAAAAAAA))),
                       i32),
           (GORCIW GPR:$rs1, (i64 1))>;
 def : Pat<(sext_inreg (or (or (and (srl GPR:$rs1, (i64 2)), (i64 0x33333333)),
                               GPR:$rs1),

 
 let Predicates = [HasStdExtZbb, IsRV64] in {
 def : Pat<(add (ctlz (and GPR:$rs1, (i64 0xFFFFFFFF))), (i64 -32)),
           (CLZW GPR:$rs1)>;
 // We don't pattern-match CTZW here as it has the same pattern and result as
 // RV64 CTZ
 def : Pat<(ctpop (and GPR:$rs1, (i64 0xFFFFFFFF))), (PCNTW GPR:$rs1)>;
 } // Predicates = [HasStdExtZbb, IsRV64]
+
+let Predicates = [HasStdExtZbbOrZbp, IsRV64] in {
+def : Pat<(sext_inreg (or (shl (assertsexti32 GPR:$rs2), (i64 16)),
+                          (and (assertsexti32 GPR:$rs1), 0x000000000000FFFF)),
+                      i32),
+          (PACKW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(or (and (assertsexti32 GPR:$rs2), 0xFFFFFFFFFFFF0000),
+              (srl (and (assertsexti32 GPR:$rs1), 0x00000000FFFF0000),
+                   (i64 16))),
+          (PACKUW GPR:$rs1, GPR:$rs2)>;
+} // Predicates = [HasStdExtZbbOrZbp, IsRV64]

llvm/test/CodeGen/RISCV/rv32Zbbp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32I
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-b -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32IB
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-zbb -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32IB
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-zbp -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV32IB
+
+define i32 @andn_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: andn_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    not a1, a1
+; RV32I-NEXT:    and a0, a1, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: andn_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    andn a0, a0, a1
+; RV32IB-NEXT:    ret
+  %neg = xor i32 %b, -1
+  %and = and i32 %neg, %a
+  ret i32 %and
+}
+
+define i64 @andn_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: andn_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    not a3, a3
+; RV32I-NEXT:    not a2, a2
+; RV32I-NEXT:    and a0, a2, a0
+; RV32I-NEXT:    and a1, a3, a1
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: andn_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    andn a0, a0, a2
+; RV32IB-NEXT:    andn a1, a1, a3
+; RV32IB-NEXT:    ret
+  %neg = xor i64 %b, -1
+  %and = and i64 %neg, %a
+  ret i64 %and
+}
+
+define i32 @orn_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: orn_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    not a1, a1
+; RV32I-NEXT:    or a0, a1, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: orn_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    orn a0, a0, a1
+; RV32IB-NEXT:    ret
+  %neg = xor i32 %b, -1
+  %or = or i32 %neg, %a
+  ret i32 %or
+}
+
+define i64 @orn_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: orn_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    not a3, a3
+; RV32I-NEXT:    not a2, a2
+; RV32I-NEXT:    or a0, a2, a0
+; RV32I-NEXT:    or a1, a3, a1
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: orn_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    orn a0, a0, a2
+; RV32IB-NEXT:    orn a1, a1, a3
+; RV32IB-NEXT:    ret
+  %neg = xor i64 %b, -1
+  %or = or i64 %neg, %a
+  ret i64 %or
+}
+
+define i32 @xnor_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: xnor_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    xor a0, a0, a1
+; RV32I-NEXT:    not a0, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: xnor_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    xnor a0, a0, a1
+; RV32IB-NEXT:    ret
+  %neg = xor i32 %a, -1
+  %xor = xor i32 %neg, %b
+  ret i32 %xor
+}
+
+define i64 @xnor_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: xnor_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    xor a1, a1, a3
+; RV32I-NEXT:    xor a0, a0, a2
+; RV32I-NEXT:    not a0, a0
+; RV32I-NEXT:    not a1, a1
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: xnor_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    xnor a0, a0, a2
+; RV32IB-NEXT:    xnor a1, a1, a3
+; RV32IB-NEXT:    ret
+  %neg = xor i64 %a, -1
+  %xor = xor i64 %neg, %b
+  ret i64 %xor
+}
+
+declare i32 @llvm.fshl.i32(i32, i32, i32)
+
+define i32 @rol_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: rol_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    sll a2, a0, a1
+; RV32I-NEXT:    neg a1, a1
+; RV32I-NEXT:    srl a0, a0, a1
+; RV32I-NEXT:    or a0, a2, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: rol_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    rol a0, a0, a1
+; RV32IB-NEXT:    ret
+  %or = tail call i32 @llvm.fshl.i32(i32 %a, i32 %a, i32 %b)
+  ret i32 %or
+}
+
+declare i64 @llvm.fshl.i64(i64, i64, i64)
+
+define i64 @rol_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: rol_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    andi a3, a2, 63
+; RV32I-NEXT:    addi t1, a3, -32
+; RV32I-NEXT:    addi a6, zero, 31
+; RV32I-NEXT:    bltz t1, .LBB7_2
+; RV32I-NEXT:  # %bb.1:
+; RV32I-NEXT:    sll a7, a0, t1
+; RV32I-NEXT:    j .LBB7_3
+; RV32I-NEXT:  .LBB7_2:
+; RV32I-NEXT:    sll a4, a1, a2
+; RV32I-NEXT:    sub a3, a6, a3
+; RV32I-NEXT:    srli a5, a0, 1
+; RV32I-NEXT:    srl a3, a5, a3
+; RV32I-NEXT:    or a7, a4, a3
+; RV32I-NEXT:  .LBB7_3:
+; RV32I-NEXT:    neg a4, a2
+; RV32I-NEXT:    andi a5, a4, 63
+; RV32I-NEXT:    addi a3, a5, -32
+; RV32I-NEXT:    bltz a3, .LBB7_7
+; RV32I-NEXT:  # %bb.4:
+; RV32I-NEXT:    mv t0, zero
+; RV32I-NEXT:    bgez a3, .LBB7_8
+; RV32I-NEXT:  .LBB7_5:
+; RV32I-NEXT:    srl a3, a0, a4
+; RV32I-NEXT:    sub a4, a6, a5
+; RV32I-NEXT:    slli a1, a1, 1
+; RV32I-NEXT:    sll a1, a1, a4
+; RV32I-NEXT:    or a4, a3, a1
+; RV32I-NEXT:    or a1, a7, t0
+; RV32I-NEXT:    bgez t1, .LBB7_9
+; RV32I-NEXT:  .LBB7_6:
+; RV32I-NEXT:    sll a0, a0, a2
+; RV32I-NEXT:    or a0, a0, a4
+; RV32I-NEXT:    ret
+; RV32I-NEXT:  .LBB7_7:
+; RV32I-NEXT:    srl t0, a1, a4
+; RV32I-NEXT:    bltz a3, .LBB7_5
+; RV32I-NEXT:  .LBB7_8:
+; RV32I-NEXT:    srl a4, a1, a3
+; RV32I-NEXT:    or a1, a7, t0
+; RV32I-NEXT:    bltz t1, .LBB7_6
+; RV32I-NEXT:  .LBB7_9:
+; RV32I-NEXT:    or a0, zero, a4
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: rol_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    andi a3, a2, 63
+; RV32IB-NEXT:    addi t1, a3, -32
+; RV32IB-NEXT:    addi a6, zero, 31
+; RV32IB-NEXT:    bltz t1, .LBB7_2
+; RV32IB-NEXT:  # %bb.1:
+; RV32IB-NEXT:    sll a7, a0, t1
+; RV32IB-NEXT:    j .LBB7_3
+; RV32IB-NEXT:  .LBB7_2:
+; RV32IB-NEXT:    sll a4, a1, a2
+; RV32IB-NEXT:    sub a3, a6, a3
+; RV32IB-NEXT:    srli a5, a0, 1
+; RV32IB-NEXT:    srl a3, a5, a3
+; RV32IB-NEXT:    or a7, a4, a3
+; RV32IB-NEXT:  .LBB7_3:
+; RV32IB-NEXT:    neg a4, a2
+; RV32IB-NEXT:    andi a5, a4, 63
+; RV32IB-NEXT:    addi a3, a5, -32
+; RV32IB-NEXT:    bltz a3, .LBB7_7
+; RV32IB-NEXT:  # %bb.4:
+; RV32IB-NEXT:    mv t0, zero
+; RV32IB-NEXT:    bgez a3, .LBB7_8
+; RV32IB-NEXT:  .LBB7_5:
+; RV32IB-NEXT:    srl a3, a0, a4
+; RV32IB-NEXT:    sub a4, a6, a5
+; RV32IB-NEXT:    slli a1, a1, 1
+; RV32IB-NEXT:    sll a1, a1, a4
+; RV32IB-NEXT:    or a4, a3, a1
+; RV32IB-NEXT:    or a1, a7, t0
+; RV32IB-NEXT:    bgez t1, .LBB7_9
+; RV32IB-NEXT:  .LBB7_6:
+; RV32IB-NEXT:    sll a0, a0, a2
+; RV32IB-NEXT:    or a0, a0, a4
+; RV32IB-NEXT:    ret
+; RV32IB-NEXT:  .LBB7_7:
+; RV32IB-NEXT:    srl t0, a1, a4
+; RV32IB-NEXT:    bltz a3, .LBB7_5
+; RV32IB-NEXT:  .LBB7_8:
+; RV32IB-NEXT:    srl a4, a1, a3
+; RV32IB-NEXT:    or a1, a7, t0
+; RV32IB-NEXT:    bltz t1, .LBB7_6
+; RV32IB-NEXT:  .LBB7_9:
+; RV32IB-NEXT:    or a0, zero, a4
+; RV32IB-NEXT:    ret
+  %or = tail call i64 @llvm.fshl.i64(i64 %a, i64 %a, i64 %b)
+  ret i64 %or
+}
+
+declare i32 @llvm.fshr.i32(i32, i32, i32)
+
+define i32 @ror_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: ror_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    srl a2, a0, a1
+; RV32I-NEXT:    neg a1, a1
+; RV32I-NEXT:    sll a0, a0, a1
+; RV32I-NEXT:    or a0, a0, a2
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: ror_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    ror a0, a0, a1
+; RV32IB-NEXT:    ret
+  %or = tail call i32 @llvm.fshr.i32(i32 %a, i32 %a, i32 %b)
+  ret i32 %or
+}
+
+declare i64 @llvm.fshr.i64(i64, i64, i64)
+
+define i64 @ror_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: ror_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    andi a3, a2, 63
+; RV32I-NEXT:    addi t1, a3, -32
+; RV32I-NEXT:    addi a6, zero, 31
+; RV32I-NEXT:    bltz t1, .LBB9_2
+; RV32I-NEXT:  # %bb.1:
+; RV32I-NEXT:    srl a7, a1, t1
+; RV32I-NEXT:    j .LBB9_3
+; RV32I-NEXT:  .LBB9_2:
+; RV32I-NEXT:    srl a4, a0, a2
+; RV32I-NEXT:    sub a3, a6, a3
+; RV32I-NEXT:    slli a5, a1, 1
+; RV32I-NEXT:    sll a3, a5, a3
+; RV32I-NEXT:    or a7, a4, a3
+; RV32I-NEXT:  .LBB9_3:
+; RV32I-NEXT:    neg a4, a2
+; RV32I-NEXT:    andi a5, a4, 63
+; RV32I-NEXT:    addi a3, a5, -32
+; RV32I-NEXT:    bltz a3, .LBB9_7
+; RV32I-NEXT:  # %bb.4:
+; RV32I-NEXT:    mv t0, zero
+; RV32I-NEXT:    bgez a3, .LBB9_8
+; RV32I-NEXT:  .LBB9_5:
+; RV32I-NEXT:    sll a3, a1, a4
+; RV32I-NEXT:    sub a4, a6, a5
+; RV32I-NEXT:    srli a0, a0, 1
+; RV32I-NEXT:    srl a0, a0, a4
+; RV32I-NEXT:    or a4, a3, a0
+; RV32I-NEXT:    or a0, t0, a7
+; RV32I-NEXT:    bgez t1, .LBB9_9
+; RV32I-NEXT:  .LBB9_6:
+; RV32I-NEXT:    srl a1, a1, a2
+; RV32I-NEXT:    or a1, a4, a1
+; RV32I-NEXT:    ret
+; RV32I-NEXT:  .LBB9_7:
+; RV32I-NEXT:    sll t0, a0, a4
+; RV32I-NEXT:    bltz a3, .LBB9_5
+; RV32I-NEXT:  .LBB9_8:
+; RV32I-NEXT:    sll a4, a0, a3
+; RV32I-NEXT:    or a0, t0, a7
+; RV32I-NEXT:    bltz t1, .LBB9_6
+; RV32I-NEXT:  .LBB9_9:
+; RV32I-NEXT:    or a1, a4, zero
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: ror_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    andi a3, a2, 63
+; RV32IB-NEXT:    addi t1, a3, -32
+; RV32IB-NEXT:    addi a6, zero, 31
+; RV32IB-NEXT:    bltz t1, .LBB9_2
+; RV32IB-NEXT:  # %bb.1:
+; RV32IB-NEXT:    srl a7, a1, t1
+; RV32IB-NEXT:    j .LBB9_3
+; RV32IB-NEXT:  .LBB9_2:
+; RV32IB-NEXT:    srl a4, a0, a2
+; RV32IB-NEXT:    sub a3, a6, a3
+; RV32IB-NEXT:    slli a5, a1, 1
+; RV32IB-NEXT:    sll a3, a5, a3
+; RV32IB-NEXT:    or a7, a4, a3
+; RV32IB-NEXT:  .LBB9_3:
+; RV32IB-NEXT:    neg a4, a2
+; RV32IB-NEXT:    andi a5, a4, 63
+; RV32IB-NEXT:    addi a3, a5, -32
+; RV32IB-NEXT:    bltz a3, .LBB9_7
+; RV32IB-NEXT:  # %bb.4:
+; RV32IB-NEXT:    mv t0, zero
+; RV32IB-NEXT:    bgez a3, .LBB9_8
+; RV32IB-NEXT:  .LBB9_5:
+; RV32IB-NEXT:    sll a3, a1, a4
+; RV32IB-NEXT:    sub a4, a6, a5
+; RV32IB-NEXT:    srli a0, a0, 1
+; RV32IB-NEXT:    srl a0, a0, a4
+; RV32IB-NEXT:    or a4, a3, a0
+; RV32IB-NEXT:    or a0, t0, a7
+; RV32IB-NEXT:    bgez t1, .LBB9_9
+; RV32IB-NEXT:  .LBB9_6:
+; RV32IB-NEXT:    srl a1, a1, a2
+; RV32IB-NEXT:    or a1, a4, a1
+; RV32IB-NEXT:    ret
+; RV32IB-NEXT:  .LBB9_7:
+; RV32IB-NEXT:    sll t0, a0, a4
+; RV32IB-NEXT:    bltz a3, .LBB9_5
+; RV32IB-NEXT:  .LBB9_8:
+; RV32IB-NEXT:    sll a4, a0, a3
+; RV32IB-NEXT:    or a0, t0, a7
+; RV32IB-NEXT:    bltz t1, .LBB9_6
+; RV32IB-NEXT:  .LBB9_9:
+; RV32IB-NEXT:    or a1, a4, zero
+; RV32IB-NEXT:    ret
+  %or = tail call i64 @llvm.fshr.i64(i64 %a, i64 %a, i64 %b)
+  ret i64 %or
+}
+
+define i32 @pack_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: pack_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    lui a2, 16
+; RV32I-NEXT:    addi a2, a2, -1
+; RV32I-NEXT:    and a0, a0, a2
+; RV32I-NEXT:    slli a1, a1, 16
+; RV32I-NEXT:    or a0, a1, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: pack_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    pack a0, a0, a1
+; RV32IB-NEXT:    ret
+  %shl = and i32 %a, 65535
+  %shl1 = shl i32 %b, 16
+  %or = or i32 %shl1, %shl
+  ret i32 %or
+}
+
+define i64 @pack_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: pack_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    mv a1, a2
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: pack_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    mv a1, a2
+; RV32IB-NEXT:    ret
+  %shl = and i64 %a, 4294967295
+  %shl1 = shl i64 %b, 32
+  %or = or i64 %shl1, %shl
+  ret i64 %or
+}
+
+define i32 @packu_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: packu_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    srli a0, a0, 16
+; RV32I-NEXT:    lui a2, 1048560
+; RV32I-NEXT:    and a1, a1, a2
+; RV32I-NEXT:    or a0, a1, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: packu_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    packu a0, a0, a1
+; RV32IB-NEXT:    ret
+  %shr = lshr i32 %a, 16
+  %shr1 = and i32 %b, -65536
+  %or = or i32 %shr1, %shr
+  ret i32 %or
+}
+
+define i64 @packu_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: packu_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    mv a0, a1
+; RV32I-NEXT:    mv a1, a3
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: packu_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    mv a0, a1
+; RV32IB-NEXT:    mv a1, a3
+; RV32IB-NEXT:    ret
+  %shr = lshr i64 %a, 32
+  %shr1 = and i64 %b, -4294967296
+  %or = or i64 %shr1, %shr
+  ret i64 %or
+}
+
+define i32 @packh_i32(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: packh_i32:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    andi a0, a0, 255
+; RV32I-NEXT:    slli a1, a1, 8
+; RV32I-NEXT:    lui a2, 16
+; RV32I-NEXT:    addi a2, a2, -256
+; RV32I-NEXT:    and a1, a1, a2
+; RV32I-NEXT:    or a0, a1, a0
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: packh_i32:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    packh a0, a0, a1
+; RV32IB-NEXT:    ret
+  %and = and i32 %a, 255
+  %and1 = shl i32 %b, 8
+  %shl = and i32 %and1, 65280
+  %or = or i32 %shl, %and
+  ret i32 %or
+}
+
+define i64 @packh_i64(i64 %a, i64 %b) nounwind {
+; RV32I-LABEL: packh_i64:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    andi a0, a0, 255
+; RV32I-NEXT:    slli a1, a2, 8
+; RV32I-NEXT:    lui a2, 16
+; RV32I-NEXT:    addi a2, a2, -256
+; RV32I-NEXT:    and a1, a1, a2
+; RV32I-NEXT:    or a0, a1, a0
+; RV32I-NEXT:    mv a1, zero
+; RV32I-NEXT:    ret
+;
+; RV32IB-LABEL: packh_i64:
+; RV32IB:       # %bb.0:
+; RV32IB-NEXT:    packh a0, a0, a2
+; RV32IB-NEXT:    mv a1, zero
+; RV32IB-NEXT:    ret
+  %and = and i64 %a, 255
+  %and1 = shl i64 %b, 8
+  %shl = and i64 %and1, 65280
+  %or = or i64 %shl, %and
+  ret i64 %or
+}

llvm/test/CodeGen/RISCV/rv64Zbbp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64I
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-b -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IB
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-zbb -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IB
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-zbp -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IB
+
+define signext i32 @andn_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: andn_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    not a1, a1
+; RV64I-NEXT:    and a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: andn_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    andn a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i32 %b, -1
+  %and = and i32 %neg, %a
+  ret i32 %and
+}
+
+define i64 @andn_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: andn_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    not a1, a1
+; RV64I-NEXT:    and a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: andn_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    andn a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i64 %b, -1
+  %and = and i64 %neg, %a
+  ret i64 %and
+}
+
+define signext i32 @orn_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: orn_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    not a1, a1
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: orn_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    orn a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i32 %b, -1
+  %or = or i32 %neg, %a
+  ret i32 %or
+}
+
+define i64 @orn_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: orn_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    not a1, a1
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: orn_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    orn a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i64 %b, -1
+  %or = or i64 %neg, %a
+  ret i64 %or
+}
+
+define signext i32 @xnor_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: xnor_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    xor a0, a0, a1
+; RV64I-NEXT:    not a0, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: xnor_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    xnor a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i32 %a, -1
+  %xor = xor i32 %neg, %b
+  ret i32 %xor
+}
+
+define i64 @xnor_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: xnor_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    xor a0, a0, a1
+; RV64I-NEXT:    not a0, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: xnor_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    xnor a0, a0, a1
+; RV64IB-NEXT:    ret
+  %neg = xor i64 %a, -1
+  %xor = xor i64 %neg, %b
+  ret i64 %xor
+}
+
+declare i32 @llvm.fshl.i32(i32, i32, i32)
+
+define signext i32 @rol_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: rol_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    sllw a2, a0, a1
+; RV64I-NEXT:    neg a1, a1
+; RV64I-NEXT:    srlw a0, a0, a1
+; RV64I-NEXT:    or a0, a2, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: rol_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    rolw a0, a0, a1
+; RV64IB-NEXT:    ret
+  %1 = tail call i32 @llvm.fshl.i32(i32 %a, i32 %a, i32 %b)
+  ret i32 %1
+}
+
+declare i64 @llvm.fshl.i64(i64, i64, i64)
+
+define i64 @rol_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: rol_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    sll a2, a0, a1
+; RV64I-NEXT:    neg a1, a1
+; RV64I-NEXT:    srl a0, a0, a1
+; RV64I-NEXT:    or a0, a2, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: rol_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    rol a0, a0, a1
+; RV64IB-NEXT:    ret
+  %or = tail call i64 @llvm.fshl.i64(i64 %a, i64 %a, i64 %b)
+  ret i64 %or
+}
+
+declare i32 @llvm.fshr.i32(i32, i32, i32)
+
+define signext i32 @ror_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: ror_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srlw a2, a0, a1
+; RV64I-NEXT:    neg a1, a1
+; RV64I-NEXT:    sllw a0, a0, a1
+; RV64I-NEXT:    or a0, a0, a2
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: ror_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    rorw a0, a0, a1
+; RV64IB-NEXT:    ret
+  %1 = tail call i32 @llvm.fshr.i32(i32 %a, i32 %a, i32 %b)
+  ret i32 %1
+}
+
+declare i64 @llvm.fshr.i64(i64, i64, i64)
+
+define i64 @ror_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: ror_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srl a2, a0, a1
+; RV64I-NEXT:    neg a1, a1
+; RV64I-NEXT:    sll a0, a0, a1
+; RV64I-NEXT:    or a0, a0, a2
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: ror_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    ror a0, a0, a1
+; RV64IB-NEXT:    ret
+  %or = tail call i64 @llvm.fshr.i64(i64 %a, i64 %a, i64 %b)
+  ret i64 %or
+}
+
+define signext i32 @pack_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: pack_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    lui a2, 16
+; RV64I-NEXT:    addiw a2, a2, -1
+; RV64I-NEXT:    and a0, a0, a2
+; RV64I-NEXT:    slli a1, a1, 16
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: pack_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    packw a0, a0, a1
+; RV64IB-NEXT:    ret
+  %shl = and i32 %a, 65535
+  %shl1 = shl i32 %b, 16
+  %or = or i32 %shl1, %shl
+  ret i32 %or
+}
+
+define i64 @pack_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: pack_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 32
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    slli a1, a1, 32
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: pack_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    pack a0, a0, a1
+; RV64IB-NEXT:    ret
+  %shl = and i64 %a, 4294967295
+  %shl1 = shl i64 %b, 32
+  %or = or i64 %shl1, %shl
+  ret i64 %or
+}
+
+define signext i32 @packu_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: packu_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srliw a0, a0, 16
+; RV64I-NEXT:    lui a2, 1048560
+; RV64I-NEXT:    and a1, a1, a2
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: packu_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    packuw a0, a0, a1
+; RV64IB-NEXT:    ret
+  %shr = lshr i32 %a, 16
+  %shr1 = and i32 %b, -65536
+  %or = or i32 %shr1, %shr
+  ret i32 %or
+}
+
+define i64 @packu_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: packu_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    addi a2, zero, -1
+; RV64I-NEXT:    slli a2, a2, 32
+; RV64I-NEXT:    and a1, a1, a2
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: packu_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    packu a0, a0, a1
+; RV64IB-NEXT:    ret
+  %shr = lshr i64 %a, 32
+  %shr1 = and i64 %b, -4294967296
+  %or = or i64 %shr1, %shr
+  ret i64 %or
+}
+
+define signext i32 @packh_i32(i32 signext %a, i32 signext %b) nounwind {
+; RV64I-LABEL: packh_i32:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    andi a0, a0, 255
+; RV64I-NEXT:    slli a1, a1, 8
+; RV64I-NEXT:    lui a2, 16
+; RV64I-NEXT:    addiw a2, a2, -256
+; RV64I-NEXT:    and a1, a1, a2
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: packh_i32:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    packh a0, a0, a1
+; RV64IB-NEXT:    ret
+  %and = and i32 %a, 255
+  %and1 = shl i32 %b, 8
+  %shl = and i32 %and1, 65280
+  %or = or i32 %shl, %and
+  ret i32 %or
+}
+
+define i64 @packh_i64(i64 %a, i64 %b) nounwind {
+; RV64I-LABEL: packh_i64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    andi a0, a0, 255
+; RV64I-NEXT:    slli a1, a1, 8
+; RV64I-NEXT:    lui a2, 16
+; RV64I-NEXT:    addiw a2, a2, -256
+; RV64I-NEXT:    and a1, a1, a2
+; RV64I-NEXT:    or a0, a1, a0
+; RV64I-NEXT:    ret
+;
+; RV64IB-LABEL: packh_i64:
+; RV64IB:       # %bb.0:
+; RV64IB-NEXT:    packh a0, a0, a1
+; RV64IB-NEXT:    ret
+  %and = and i64 %a, 255
+  %and1 = shl i64 %b, 8
+  %shl = and i64 %and1, 65280
+  %or = or i64 %shl, %and
+  ret i64 %or
+}