setOperationAction(ISD::SUBC, XLenVT, Expand);
setOperationAction(ISD::SUBE, XLenVT, Expand);
- setOperationAction(ISD::SREM, XLenVT, Expand);
+ if (!Subtarget.hasStdExtM()) {
+ setOperationAction(ISD::MUL, XLenVT, Expand);
+ setOperationAction(ISD::MULHS, XLenVT, Expand);
+ setOperationAction(ISD::MULHU, XLenVT, Expand);
+ setOperationAction(ISD::SDIV, XLenVT, Expand);
+ setOperationAction(ISD::UDIV, XLenVT, Expand);
+ setOperationAction(ISD::SREM, XLenVT, Expand);
+ setOperationAction(ISD::UREM, XLenVT, Expand);
+ }
+
setOperationAction(ISD::SDIVREM, XLenVT, Expand);
- setOperationAction(ISD::SDIV, XLenVT, Expand);
- setOperationAction(ISD::UREM, XLenVT, Expand);
setOperationAction(ISD::UDIVREM, XLenVT, Expand);
- setOperationAction(ISD::UDIV, XLenVT, Expand);
-
- setOperationAction(ISD::MUL, XLenVT, Expand);
setOperationAction(ISD::SMUL_LOHI, XLenVT, Expand);
setOperationAction(ISD::UMUL_LOHI, XLenVT, Expand);
- setOperationAction(ISD::MULHS, XLenVT, Expand);
- setOperationAction(ISD::MULHU, XLenVT, Expand);
setOperationAction(ISD::SHL_PARTS, XLenVT, Expand);
setOperationAction(ISD::SRL_PARTS, XLenVT, Expand);
def REMW : ALUW_rr<0b0000001, 0b110, "remw">;
def REMUW : ALUW_rr<0b0000001, 0b111, "remuw">;
} // Predicates = [HasStdExtM, IsRV64]
+
+//===----------------------------------------------------------------------===//
+// Pseudo-instructions and codegen patterns
+//===----------------------------------------------------------------------===//
+
+let Predicates = [HasStdExtM] in {
+def : PatGprGpr<mul, MUL>;
+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]
; 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: | FileCheck -check-prefix=RV32I %s
+; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
+; RUN: | FileCheck -check-prefix=RV32IM %s
define i32 @udiv(i32 %a, i32 %b) nounwind {
; RV32I-LABEL: udiv:
; 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
%1 = udiv i32 %a, %b
ret i32 %1
}
; 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
%1 = udiv i32 %a, 5
ret i32 %1
}
; 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
%1 = udiv i32 %a, 8
ret i32 %1
}
; 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: lui a4, %hi(__udivdi3)
+; RV32IM-NEXT: addi a4, a4, %lo(__udivdi3)
+; RV32IM-NEXT: jalr a4
+; RV32IM-NEXT: lw ra, 12(sp)
+; RV32IM-NEXT: addi sp, sp, 16
+; RV32IM-NEXT: ret
%1 = udiv i64 %a, %b
ret i64 %1
}
; RV32I-NEXT: lw ra, 12(sp)
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
+;
+; RV32IM-LABEL: udiv64_constant:
+; RV32IM: # %bb.0:
+; RV32IM-NEXT: addi sp, sp, -16
+; RV32IM-NEXT: sw ra, 12(sp)
+; RV32IM-NEXT: lui a2, %hi(__udivdi3)
+; RV32IM-NEXT: addi a4, a2, %lo(__udivdi3)
+; RV32IM-NEXT: addi a2, zero, 5
+; RV32IM-NEXT: mv a3, zero
+; RV32IM-NEXT: jalr a4
+; RV32IM-NEXT: lw ra, 12(sp)
+; RV32IM-NEXT: addi sp, sp, 16
+; RV32IM-NEXT: ret
%1 = udiv i64 %a, 5
ret i64 %1
}
; 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
%1 = sdiv i32 %a, %b
ret i32 %1
}
; 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
%1 = sdiv i32 %a, 5
ret i32 %1
}
; 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
%1 = sdiv i32 %a, 8
ret i32 %1
}
; 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: lui a4, %hi(__divdi3)
+; RV32IM-NEXT: addi a4, a4, %lo(__divdi3)
+; RV32IM-NEXT: jalr a4
+; RV32IM-NEXT: lw ra, 12(sp)
+; RV32IM-NEXT: addi sp, sp, 16
+; RV32IM-NEXT: ret
%1 = sdiv i64 %a, %b
ret i64 %1
}
; RV32I-NEXT: lw ra, 12(sp)
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
+;
+; RV32IM-LABEL: sdiv64_constant:
+; RV32IM: # %bb.0:
+; RV32IM-NEXT: addi sp, sp, -16
+; RV32IM-NEXT: sw ra, 12(sp)
+; RV32IM-NEXT: lui a2, %hi(__divdi3)
+; RV32IM-NEXT: addi a4, a2, %lo(__divdi3)
+; RV32IM-NEXT: addi a2, zero, 5
+; RV32IM-NEXT: mv a3, zero
+; RV32IM-NEXT: jalr a4
+; RV32IM-NEXT: lw ra, 12(sp)
+; RV32IM-NEXT: addi sp, sp, 16
+; RV32IM-NEXT: ret
%1 = sdiv i64 %a, 5
ret i64 %1
}
; 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: | FileCheck -check-prefix=RV32I %s
+; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
+; RUN: | FileCheck -check-prefix=RV32IM %s
define i32 @square(i32 %a) nounwind {
; RV32I-LABEL: square:
; 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
%1 = mul i32 %a, %a
ret i32 %1
}
; 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
%1 = mul i32 %a, %b
ret i32 %1
}
; 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
%1 = mul i32 %a, 5
ret i32 %1
}
; 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
%1 = mul i32 %a, 8
ret i32 %1
}
; 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
%1 = mul i64 %a, %b
ret i64 %1
}
; 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
%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: lui a1, %hi(__muldi3)
+; RV32I-NEXT: addi a4, a1, %lo(__muldi3)
+; RV32I-NEXT: srai a1, a0, 31
+; RV32I-NEXT: srai a3, a2, 31
+; RV32I-NEXT: jalr a4
+; 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
+ %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 {
+; 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: lui a1, %hi(__muldi3)
+; RV32I-NEXT: addi a4, a1, %lo(__muldi3)
+; RV32I-NEXT: mv a1, zero
+; RV32I-NEXT: mv a3, zero
+; RV32I-NEXT: jalr a4
+; 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
+ %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
+}
; 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: | FileCheck -check-prefix=RV32I %s
+; RUN: llc -mtriple=riscv32 -mattr=+m -verify-machineinstrs < %s \
+; RUN: | FileCheck -check-prefix=RV32IM %s
define i32 @urem(i32 %a, i32 %b) nounwind {
; RV32I-LABEL: urem:
; 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
%1 = urem i32 %a, %b
ret i32 %1
}
; 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
%1 = srem i32 %a, %b
ret i32 %1
}