[RISCV] Introduce codegen patterns for RV64M-only instructions

[asb]

Summary

This builds on D52977 to add RV64M support.

[shiva0217]

Probably need to add

def : Pat<(sext_inreg (sdiv GPR:$rs1, GPR:$rs2), i32),
          (DIVW GPR:$rs1, GPR:$rs2)>;
def : Pat<(sext_inreg (udiv GPR:$rs1, GPR:$rs2), i32),
          (DIVUW GPR:$rs1, GPR:$rs2)>;

Because DAG combine may create these patterns.

A simple example to observe the pattern:

int bar (int a, int b)
{
  return a / b;
}

[asb]

Thanks Shiva. I'm updating the patch series in a way that adds much more exhaustive testing for the *W patterns and fleshes out patterns for cases like this. I mused on the RV64I patch review thread whether there was something smarter to do here - maybe some canonicalisation. But regardless of that, it's helpful to have more exhaustive tests.

[asb]

Updated to:

• Add missing remw/remuw patterns
• Add exhaustive testing of the *W patterns introduced inrv64m
• Add and make use of sexti32 PatFrags

[asb]

Patch refresh, no functional changes.

[asb]

Marking as "planned changes". At least the following patterns are problematic:

def : Pat<(udiv (zexti32 GPR:$rs1), (zexti32 GPR:$rs2)),
          (DIVUW GPR:$rs1, GPR:$rs2)>;
def : Pat<(urem (zexti32 GPR:$rs1), (zexti32 GPR:$rs2)),
          (REMUW GPR:$rs1, GPR:$rs2)>;

DIVUW and REMUW will sign-extend the result so this pattern is only safe if we know that the upper 32-bits of the result aren't used.

[asb]

Updated the patch to address the correctness issues with the udiv and urem patterns. Like with D56264, a dag combine converts an ANY_EXTEND to a SIGN_EXTEND when it operates on an instruction where a 32-bit *W variant could be selected. This is advantageous as it can avoid unnecessary masking/sign-extension of the input operands.

See also the related discussion on llvm-dev http://lists.llvm.org/pipermail/llvm-dev/2018-December/128497.html

[efriedma]

lib/Target/RISCV/RISCVInstrInfoM.td

57 ↗

(On Diff #180031)

Are you sure (sdiv (sexti32 GPR:$rs1), (sexti32 GPR:$rs2)) is sufficient? I think it returns the wrong result for something like INT_MIN/-1 (which should be a positive 64-bit value).

The pattern where the result is sign-extended is fine, I think. And the corresponding "REMW" pattern is also fine.

[brucehoult]

lib/Target/RISCV/RISCVInstrInfoM.td

57 ↗

(On Diff #180031)

I'm not sure exactly what you're saying here.

A 32 bit INT_MIN/-1 which is 0x8000_0000/0xffffffff is defined to give INT_MIN 0x8000_0000 (which a DIVW will). On a 64 bit processor the end result will be 0xffff_ffff_8000_0000.

A 32 bit INT_MIN converted to 64 bit will be 0xffff_ffff_8000_0000, which when divided by a 64 bit -1 will give 0x0000_0001_0000_0000, which is incorrect for a 32 bit calculation.

Arguably, INT_MIN/-1 could have been defined to be 0 (the LSBs of the correct result), but it wasn't.

[brucehoult]

lib/Target/RISCV/RISCVInstrInfoM.td

57 ↗

(On Diff #180031)

Oops ... that was wrong. Doing 0xffff_ffff_8000_0000 / 0xffff_ffff_ffff_ffff in 64 bit will give 0x0000_0000_8000_0000, which is correct if subsequently used or stored as a 32 bit value (though non-canonical), but it needs explicit sign extending if it will be used as a 64 bit value to be the same as the result of DIVW.

[efriedma]

lib/Target/RISCV/RISCVInstrInfoM.td

57 ↗

(On Diff #180031)

The overall point is that I think the pattern miscompiles the following: int64_t f(int32_t x, int32_t y) { return (int64_t)x/y; }.

[asb]

lib/Target/RISCV/RISCVInstrInfoM.td

57 ↗

(On Diff #180031)

Thanks for the catch, I agree this pattern is incorrect. I had reasoned that i32 INT_MIN/-1 is undefined behaviour. As you rightly point out, there's no guarantee that the sdiv started life as an i32 sdiv (as opposed to an i64 sdiv that happens to have sexti32 operands) - so this is basically the same problem we've discussed before.

[asb]

Many thanks to @efriedma for catching the issue with one of the sdiv patterns. I've removed that pattern, and added similar logic to that used for udiv/urem so that anyext is converted to sext if this is likely to result in one of the *w instructions being selected (and thus avoiding unnecessary sext/zext of the input operands).

I've added a comment indicating that srem is believed to be safe (I'd appreciate more eyes, but I believe it's not possible for sexti32 operands to produce a result where res[63:32]=0 and res[31]=1.

[efriedma]

Please make sure to include my example as a regression test, so someone doesn't accidentally add the wrong pattern in the future.

Otherwise LGTM.

[asb]

Thanks for the review @efriedma. Updated to add the suggested regression test.

If you're able to give D56264 a final LGTM I can get both merged. I could in theory unpick this patch from its predecessor, but it looks like it's basically ready to land anyway.

llvm/trunk/lib/Target/RISCV/RISCVISelLowering.cpp

default:
     return false;
   case ISD::SHL:
   case ISD::SRA:
   case ISD::SRL:
     return Val.getOperand(1).getOpcode() != ISD::Constant;
   }
 }
 
+// Returns true if the given node is an sdiv, udiv, or urem with non-constant
+// operands.
+static bool isVariableSDivUDivURem(SDValue Val) {
+  switch (Val.getOpcode()) {
+  default:
+    return false;
+  case ISD::SDIV:
+  case ISD::UDIV:
+  case ISD::UREM:
+    return Val.getOperand(0).getOpcode() != ISD::Constant &&
+           Val.getOperand(1).getOpcode() != ISD::Constant;
+  }
+}
+
 SDValue RISCVTargetLowering::PerformDAGCombine(SDNode *N,
                                                DAGCombinerInfo &DCI) const {
   SelectionDAG &DAG = DCI.DAG;
 
   switch (N->getOpcode()) {
   default:
     break;
   case ISD::SHL:
case ISD::SRA: {
     SDLoc DL(N);
     SDValue NewRHS =
         DAG.getNode(ISD::AssertZext, DL, RHS.getValueType(), RHS,
                     DAG.getValueType(EVT::getIntegerVT(*DAG.getContext(), 5)));
     return DCI.CombineTo(
         N, DAG.getNode(N->getOpcode(), DL, LHS.getValueType(), LHS, NewRHS));
   }
   case ISD::ANY_EXTEND: {
-    // If any-extending an i32 variable-length shift to i64, then instead
-    // sign-extend in order to increase the chance of being able to select the
-    // sllw/srlw/sraw instruction.
+    // If any-extending an i32 variable-length shift or sdiv/udiv/urem to i64,
+    // then instead sign-extend in order to increase the chance of being able
+    // to select the sllw/srlw/sraw/divw/divuw/remuw instructions.
     SDValue Src = N->getOperand(0);
-    if (N->getValueType(0) != MVT::i64 || Src.getValueType() != MVT::i32 ||
-        !isVariableShift(Src))
+    if (N->getValueType(0) != MVT::i64 || Src.getValueType() != MVT::i32)
+      break;
+    if (!isVariableShift(Src) &&
+        !(Subtarget.hasStdExtM() && isVariableSDivUDivURem(Src)))
       break;
     SDLoc DL(N);
     return DCI.CombineTo(N, DAG.getNode(ISD::SIGN_EXTEND, DL, MVT::i64, Src));
   }
   case RISCVISD::SplitF64: {
     // If the input to SplitF64 is just BuildPairF64 then the operation is
     // redundant. Instead, use BuildPairF64's operands directly.
     SDValue Op0 = N->getOperand(0);

llvm/trunk/lib/Target/RISCV/RISCVInstrInfoM.td

 def : PatGprGpr<mulhs, MULH>;
 def : PatGprGpr<mulhu, MULHU>;
 // No ISDOpcode for mulhsu
 def : PatGprGpr<sdiv, DIV>;
 def : PatGprGpr<udiv, DIVU>;
 def : PatGprGpr<srem, REM>;
 def : PatGprGpr<urem, REMU>;
 } // Predicates = [HasStdExtM]
+
+let Predicates = [HasStdExtM, IsRV64] in {
+def : Pat<(sext_inreg (mul GPR:$rs1, GPR:$rs2), i32),
+          (MULW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(sext_inreg (sdiv (sexti32 GPR:$rs1),
+                            (sexti32 GPR:$rs2)), i32),
+          (DIVW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(zexti32 (sdiv (sexti32 GPR:$rs1),
+                         (sexti32 GPR:$rs2))),
+          (SRLI (SLLI (DIVW GPR:$rs1, GPR:$rs2), 32), 32)>;
+def : Pat<(sext_inreg (udiv (zexti32 GPR:$rs1), (zexti32 GPR:$rs2)), i32),
+          (DIVUW GPR:$rs1, GPR:$rs2)>;
+// It's cheaper to perform a divuw and zero-extend the result than to
+// zero-extend both inputs to a udiv.
+def : Pat<(udiv (and GPR:$rs1, 0xffffffff), (and GPR:$rs2, 0xffffffff)),
+          (SRLI (SLLI (DIVUW GPR:$rs1, GPR:$rs2), 32), 32)>;
+// Although the sexti32 operands may not have originated from an i32 srem,
+// this pattern is safe as it is impossible for two sign extended inputs to
+// produce a result where res[63:32]=0 and res[31]=1.
+def : Pat<(srem (sexti32 GPR:$rs1), (sexti32 GPR:$rs2)),
+          (REMW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(sext_inreg (srem (sexti32 GPR:$rs1),
+                            (sexti32 GPR:$rs2)), i32),
+          (REMW GPR:$rs1, GPR:$rs2)>;
+def : Pat<(sext_inreg (urem (zexti32 GPR:$rs1), (zexti32 GPR:$rs2)), i32),
+          (REMUW GPR:$rs1, GPR:$rs2)>;
+// It's cheaper to perform a remuw and zero-extend the result than to
+// zero-extend both inputs to a urem.
+def : Pat<(urem (and GPR:$rs1, 0xffffffff), (and GPR:$rs2, 0xffffffff)),
+          (SRLI (SLLI (REMUW GPR:$rs1, GPR:$rs2), 32), 32)>;
+} // Predicates = [HasStdExtM, IsRV64]

llvm/trunk/test/CodeGen/RISCV/div.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32I %s
 ; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IM %s
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64I %s
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IM %s
 
 define i32 @udiv(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: udiv:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __udivsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: udiv:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    divu a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: udiv:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    slli a0, a0, 32
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    slli a1, a1, 32
+; RV64I-NEXT:    srli a1, a1, 32
+; RV64I-NEXT:    call __udivdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: udiv:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = udiv i32 %a, %b
   ret i32 %1
 }
 
 define i32 @udiv_constant(i32 %a) nounwind {
 ; RV32I-LABEL: udiv_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    call __udivsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: udiv_constant:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    lui a1, 838861
 ; RV32IM-NEXT:    addi a1, a1, -819
 ; RV32IM-NEXT:    mulhu a0, a0, a1
 ; RV32IM-NEXT:    srli a0, a0, 2
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: udiv_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    slli a0, a0, 32
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __udivdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: udiv_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    lui a1, 1035469
+; RV64IM-NEXT:    addiw a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    mulhu a0, a0, a1
+; RV64IM-NEXT:    srli a0, a0, 2
+; RV64IM-NEXT:    ret
   %1 = udiv i32 %a, 5
   ret i32 %1
 }
 
 define i32 @udiv_pow2(i32 %a) nounwind {
 ; RV32I-LABEL: udiv_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    srli a0, a0, 3
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: udiv_pow2:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    srli a0, a0, 3
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: udiv_pow2:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srliw a0, a0, 3
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: udiv_pow2:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    srliw a0, a0, 3
+; RV64IM-NEXT:    ret
   %1 = udiv i32 %a, 8
   ret i32 %1
 }
 
 define i64 @udiv64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: udiv64:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __udivdi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: udiv64:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    addi sp, sp, -16
 ; RV32IM-NEXT:    sw ra, 12(sp)
 ; RV32IM-NEXT:    call __udivdi3
 ; RV32IM-NEXT:    lw ra, 12(sp)
 ; RV32IM-NEXT:    addi sp, sp, 16
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: udiv64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    call __udivdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: udiv64:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = udiv i64 %a, %b
   ret i64 %1
 }
 
 define i64 @udiv64_constant(i64 %a) nounwind {
 ; RV32I-LABEL: udiv64_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16

 ; RV32IM-NEXT:    addi sp, sp, -16
 ; RV32IM-NEXT:    sw ra, 12(sp)
 ; RV32IM-NEXT:    addi a2, zero, 5
 ; RV32IM-NEXT:    mv a3, zero
 ; RV32IM-NEXT:    call __udivdi3
 ; RV32IM-NEXT:    lw ra, 12(sp)
 ; RV32IM-NEXT:    addi sp, sp, 16
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: udiv64_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __udivdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: udiv64_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    lui a1, 1035469
+; RV64IM-NEXT:    addiw a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, -819
+; RV64IM-NEXT:    mulhu a0, a0, a1
+; RV64IM-NEXT:    srli a0, a0, 2
+; RV64IM-NEXT:    ret
   %1 = udiv i64 %a, 5
   ret i64 %1
 }
 
 define i32 @sdiv(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: sdiv:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __divsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: sdiv:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    div a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    sext.w a1, a1
+; RV64I-NEXT:    call __divdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = sdiv i32 %a, %b
   ret i32 %1
 }
 
 define i32 @sdiv_constant(i32 %a) nounwind {
 ; RV32I-LABEL: sdiv_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    call __divsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: sdiv_constant:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    lui a1, 419430
 ; RV32IM-NEXT:    addi a1, a1, 1639
 ; RV32IM-NEXT:    mulh a0, a0, a1
 ; RV32IM-NEXT:    srli a1, a0, 31
 ; RV32IM-NEXT:    srai a0, a0, 1
 ; RV32IM-NEXT:    add a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __divdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    sext.w a0, a0
+; RV64IM-NEXT:    lui a1, 13107
+; RV64IM-NEXT:    addiw a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 13
+; RV64IM-NEXT:    addi a1, a1, 1639
+; RV64IM-NEXT:    mulh a0, a0, a1
+; RV64IM-NEXT:    srli a1, a0, 63
+; RV64IM-NEXT:    srai a0, a0, 1
+; RV64IM-NEXT:    add a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = sdiv i32 %a, 5
   ret i32 %1
 }
 
 define i32 @sdiv_pow2(i32 %a) nounwind {
 ; RV32I-LABEL: sdiv_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    srai a1, a0, 31
 ; RV32I-NEXT:    srli a1, a1, 29
 ; RV32I-NEXT:    add a0, a0, a1
 ; RV32I-NEXT:    srai a0, a0, 3
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: sdiv_pow2:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    srai a1, a0, 31
 ; RV32IM-NEXT:    srli a1, a1, 29
 ; RV32IM-NEXT:    add a0, a0, a1
 ; RV32IM-NEXT:    srai a0, a0, 3
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv_pow2:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    sext.w a1, a0
+; RV64I-NEXT:    srli a1, a1, 60
+; RV64I-NEXT:    andi a1, a1, 7
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    sraiw a0, a0, 3
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv_pow2:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    sext.w a1, a0
+; RV64IM-NEXT:    srli a1, a1, 60
+; RV64IM-NEXT:    andi a1, a1, 7
+; RV64IM-NEXT:    add a0, a0, a1
+; RV64IM-NEXT:    sraiw a0, a0, 3
+; RV64IM-NEXT:    ret
   %1 = sdiv i32 %a, 8
   ret i32 %1
 }
 
 define i64 @sdiv64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: sdiv64:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __divdi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: sdiv64:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    addi sp, sp, -16
 ; RV32IM-NEXT:    sw ra, 12(sp)
 ; RV32IM-NEXT:    call __divdi3
 ; RV32IM-NEXT:    lw ra, 12(sp)
 ; RV32IM-NEXT:    addi sp, sp, 16
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    call __divdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv64:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    div a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = sdiv i64 %a, %b
   ret i64 %1
 }
 
 define i64 @sdiv64_constant(i64 %a) nounwind {
 ; RV32I-LABEL: sdiv64_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16

 ; RV32IM-NEXT:    addi sp, sp, -16
 ; RV32IM-NEXT:    sw ra, 12(sp)
 ; RV32IM-NEXT:    addi a2, zero, 5
 ; RV32IM-NEXT:    mv a3, zero
 ; RV32IM-NEXT:    call __divdi3
 ; RV32IM-NEXT:    lw ra, 12(sp)
 ; RV32IM-NEXT:    addi sp, sp, 16
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv64_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __divdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv64_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    lui a1, 13107
+; RV64IM-NEXT:    addiw a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 12
+; RV64IM-NEXT:    addi a1, a1, 819
+; RV64IM-NEXT:    slli a1, a1, 13
+; RV64IM-NEXT:    addi a1, a1, 1639
+; RV64IM-NEXT:    mulh a0, a0, a1
+; RV64IM-NEXT:    srli a1, a0, 63
+; RV64IM-NEXT:    srai a0, a0, 1
+; RV64IM-NEXT:    add a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = sdiv i64 %a, 5
   ret i64 %1
 }
+
+; Although this sdiv has two sexti32 operands, it shouldn't compile to divw on
+; RV64M as that wouldn't produce the correct result for e.g. INT_MIN/-1.
+
+define i64 @sdiv64_sext_operands(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: sdiv64_sext_operands:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    addi sp, sp, -16
+; RV32I-NEXT:    sw ra, 12(sp)
+; RV32I-NEXT:    mv a2, a1
+; RV32I-NEXT:    srai a1, a0, 31
+; RV32I-NEXT:    srai a3, a2, 31
+; RV32I-NEXT:    call __divdi3
+; RV32I-NEXT:    lw ra, 12(sp)
+; RV32I-NEXT:    addi sp, sp, 16
+; RV32I-NEXT:    ret
+;
+; RV32IM-LABEL: sdiv64_sext_operands:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    addi sp, sp, -16
+; RV32IM-NEXT:    sw ra, 12(sp)
+; RV32IM-NEXT:    mv a2, a1
+; RV32IM-NEXT:    srai a1, a0, 31
+; RV32IM-NEXT:    srai a3, a2, 31
+; RV32IM-NEXT:    call __divdi3
+; RV32IM-NEXT:    lw ra, 12(sp)
+; RV32IM-NEXT:    addi sp, sp, 16
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: sdiv64_sext_operands:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    sext.w a1, a1
+; RV64I-NEXT:    call __divdi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: sdiv64_sext_operands:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    sext.w a1, a1
+; RV64IM-NEXT:    sext.w a0, a0
+; RV64IM-NEXT:    div a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sext i32 %a to i64
+  %2 = sext i32 %b to i64
+  %3 = sdiv i64 %1, %2
+  ret i64 %3
+}

llvm/trunk/test/CodeGen/RISCV/mul.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32I %s
 ; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IM %s
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64I %s
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IM %s
 
-define i32 @square(i32 %a) nounwind {
+define signext i32 @square(i32 %a) nounwind {
 ; RV32I-LABEL: square:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    mv a1, a0
 ; RV32I-NEXT:    call __mulsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: square:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    mul a0, a0, a0
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: square:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: square:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a0
+; RV64IM-NEXT:    ret
   %1 = mul i32 %a, %a
   ret i32 %1
 }
 
-define i32 @mul(i32 %a, i32 %b) nounwind {
+define signext i32 @mul(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: mul:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __mulsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mul:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    mul a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mul:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mul:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = mul i32 %a, %b
   ret i32 %1
 }
 
-define i32 @mul_constant(i32 %a) nounwind {
+define signext i32 @mul_constant(i32 %a) nounwind {
 ; RV32I-LABEL: mul_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    call __mulsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mul_constant:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    addi a1, zero, 5
 ; RV32IM-NEXT:    mul a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mul_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mul_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    addi a1, zero, 5
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = mul i32 %a, 5
   ret i32 %1
 }
 
 define i32 @mul_pow2(i32 %a) nounwind {
 ; RV32I-LABEL: mul_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    slli a0, a0, 3
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mul_pow2:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    slli a0, a0, 3
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mul_pow2:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a0, a0, 3
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mul_pow2:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 3
+; RV64IM-NEXT:    ret
   %1 = mul i32 %a, 8
   ret i32 %1
 }
 
 define i64 @mul64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: mul64:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __muldi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mul64:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    mul a3, a0, a3
 ; RV32IM-NEXT:    mulhu a4, a0, a2
 ; RV32IM-NEXT:    add a3, a4, a3
 ; RV32IM-NEXT:    mul a1, a1, a2
 ; RV32IM-NEXT:    add a1, a3, a1
 ; RV32IM-NEXT:    mul a0, a0, a2
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mul64:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mul64:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = mul i64 %a, %b
   ret i64 %1
 }
 
 define i64 @mul64_constant(i64 %a) nounwind {
 ; RV32I-LABEL: mul64_constant:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a2, zero, 5
 ; RV32I-NEXT:    mv a3, zero
 ; RV32I-NEXT:    call __muldi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mul64_constant:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    addi a2, zero, 5
 ; RV32IM-NEXT:    mul a1, a1, a2
 ; RV32IM-NEXT:    mulhu a3, a0, a2
 ; RV32IM-NEXT:    add a1, a3, a1
 ; RV32IM-NEXT:    mul a0, a0, a2
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mul64_constant:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    addi a1, zero, 5
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mul64_constant:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    addi a1, zero, 5
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = mul i64 %a, 5
   ret i64 %1
 }
 
 define i32 @mulhs(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: mulhs:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    mv a2, a1
 ; RV32I-NEXT:    srai a1, a0, 31
 ; RV32I-NEXT:    srai a3, a2, 31
 ; RV32I-NEXT:    call __muldi3
 ; RV32I-NEXT:    mv a0, a1
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mulhs:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    mulh a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mulhs:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    sext.w a1, a1
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mulhs:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    sext.w a1, a1
+; RV64IM-NEXT:    sext.w a0, a0
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
   %1 = sext i32 %a to i64
   %2 = sext i32 %b to i64
   %3 = mul i64 %1, %2
   %4 = lshr i64 %3, 32
   %5 = trunc i64 %4 to i32
   ret i32 %5
 }
 
-define i32 @mulhu(i32 %a, i32 %b) nounwind {
+define zeroext i32 @mulhu(i32 zeroext %a, i32 zeroext %b) nounwind {
 ; RV32I-LABEL: mulhu:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    mv a2, a1
 ; RV32I-NEXT:    mv a1, zero
 ; RV32I-NEXT:    mv a3, zero
 ; RV32I-NEXT:    call __muldi3
 ; RV32I-NEXT:    mv a0, a1
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: mulhu:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    mulhu a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mulhu:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mulhu:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
   %1 = zext i32 %a to i64
   %2 = zext i32 %b to i64
   %3 = mul i64 %1, %2
   %4 = lshr i64 %3, 32
   %5 = trunc i64 %4 to i32
   ret i32 %5
 }

llvm/trunk/test/CodeGen/RISCV/rem.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32I %s
 ; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IM %s
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64I %s
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IM %s
 
 define i32 @urem(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: urem:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __umodsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: urem:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    remu a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: urem:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    slli a0, a0, 32
+; RV64I-NEXT:    srli a0, a0, 32
+; RV64I-NEXT:    slli a1, a1, 32
+; RV64I-NEXT:    srli a1, a1, 32
+; RV64I-NEXT:    call __umoddi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: urem:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = urem i32 %a, %b
   ret i32 %1
 }
 
 define i32 @srem(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: srem:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi sp, sp, -16
 ; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    call __modsi3
 ; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV32IM-LABEL: srem:
 ; RV32IM:       # %bb.0:
 ; RV32IM-NEXT:    rem a0, a0, a1
 ; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: srem:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp)
+; RV64I-NEXT:    sext.w a0, a0
+; RV64I-NEXT:    sext.w a1, a1
+; RV64I-NEXT:    call __moddi3
+; RV64I-NEXT:    ld ra, 8(sp)
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: srem:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
   %1 = srem i32 %a, %b
   ret i32 %1
 }

llvm/trunk/test/CodeGen/RISCV/rv64m-exhaustive-w-insts.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IM
+
+; The patterns for the 'W' suffixed RV64M instructions have the potential of
+; missing cases. This file checks all the variants of
+; sign-extended/zero-extended/any-extended inputs and outputs.
+
+define i32 @aext_mulw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_mulw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_mulw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_mulw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_mulw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_mulw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_mulw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_mulw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_mulw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_mulw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_mulw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_mulw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_mulw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_mulw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_mulw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_mulw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    mul a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a1, a1, 32
+; RV64IM-NEXT:    srli a1, a1, 32
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a1, a1, 32
+; RV64IM-NEXT:    srli a1, a1, 32
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = udiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_divw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_divw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_divw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_divw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_divw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_divw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_divw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_divw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_divw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_divw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_divw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_divw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    divw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = sdiv i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = srem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: aext_remuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define i32 @aext_remuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: aext_remuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: sext_remuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define signext i32 @sext_remuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: sext_remuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_aext_aext(i32 %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remuw_aext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_aext_sext(i32 %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_aext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_aext_zext(i32 %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_aext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    remu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_sext_aext(i32 signext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remuw_sext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_sext_sext(i32 signext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_sext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remuw a0, a0, a1
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_sext_zext(i32 signext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_sext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a0, a0, 32
+; RV64IM-NEXT:    srli a0, a0, 32
+; RV64IM-NEXT:    remu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_zext_aext(i32 zeroext %a, i32 %b) nounwind {
+; RV64IM-LABEL: zext_remuw_zext_aext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a1, a1, 32
+; RV64IM-NEXT:    srli a1, a1, 32
+; RV64IM-NEXT:    remu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_zext_sext(i32 zeroext %a, i32 signext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_zext_sext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slli a1, a1, 32
+; RV64IM-NEXT:    srli a1, a1, 32
+; RV64IM-NEXT:    remu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}
+
+define zeroext i32 @zext_remuw_zext_zext(i32 zeroext %a, i32 zeroext %b) nounwind {
+; RV64IM-LABEL: zext_remuw_zext_zext:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    remu a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = urem i32 %a, %b
+  ret i32 %1
+}