[LoongArch] Expand atomicrmw fadd/fsub/fmin/fmax with CmpXChg

[SixWeining]

LGTM and wait others.

llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp

case Intrinsic::loongarch_masked_atomicrmw_nand_i32:
     return true;
     // TODO: Add more Intrinsics later.
   }
 }
 
 TargetLowering::AtomicExpansionKind
 LoongArchTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
   // TODO: Add more AtomicRMWInst that needs to be extended.
+
+  // Since floating-point operation requires a non-trivial set of data
+  // operations, use CmpXChg to expand.
+  if (AI->isFloatingPointOperation())
+    return AtomicExpansionKind::CmpXChg;
+
   unsigned Size = AI->getType()->getPrimitiveSizeInBits();
   if (Size == 8 || Size == 16)
     return AtomicExpansionKind::MaskedIntrinsic;
   return AtomicExpansionKind::None;
 }
 
 static Intrinsic::ID
 getIntrinsicForMaskedAtomicRMWBinOp(unsigned GRLen,

llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-fp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch64 --mattr=+f,-d < %s | FileCheck %s --check-prefix=LA64F
+; RUN: llc --mtriple=loongarch64 --mattr=+d < %s | FileCheck %s --check-prefix=LA64D
+
+;; Add more test cases after supporting different AtomicOrdering.
+
+define float @float_fadd_acquire(ptr %p) nounwind {
+; LA64F-LABEL: float_fadd_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    fld.s $fa0, $a0, 0
+; LA64F-NEXT:    addi.w $a1, $zero, 1
+; LA64F-NEXT:    movgr2fr.w $fa1, $a1
+; LA64F-NEXT:  .LBB0_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Loop Header: Depth=1
+; LA64F-NEXT:    # Child Loop BB0_3 Depth 2
+; LA64F-NEXT:    ffint.s.w $fa2, $fa1
+; LA64F-NEXT:    fadd.s $fa2, $fa0, $fa2
+; LA64F-NEXT:    movfr2gr.s $a1, $fa2
+; LA64F-NEXT:    movfr2gr.s $a2, $fa0
+; LA64F-NEXT:  .LBB0_3: # %atomicrmw.start
+; LA64F-NEXT:    # Parent Loop BB0_1 Depth=1
+; LA64F-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64F-NEXT:    ll.w $a3, $a0, 0
+; LA64F-NEXT:    bne $a3, $a2, .LBB0_5
+; LA64F-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB0_3 Depth=2
+; LA64F-NEXT:    dbar 0
+; LA64F-NEXT:    move $a4, $a1
+; LA64F-NEXT:    sc.w $a4, $a0, 0
+; LA64F-NEXT:    beqz $a4, .LBB0_3
+; LA64F-NEXT:    b .LBB0_6
+; LA64F-NEXT:  .LBB0_5: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LA64F-NEXT:    dbar 1792
+; LA64F-NEXT:  .LBB0_6: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LA64F-NEXT:    movgr2fr.w $fa0, $a3
+; LA64F-NEXT:    addi.w $a1, $a2, 0
+; LA64F-NEXT:    bne $a3, $a1, .LBB0_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: float_fadd_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fld.s $fa0, $a0, 0
+; LA64D-NEXT:    addi.w $a1, $zero, 1
+; LA64D-NEXT:    movgr2fr.w $fa1, $a1
+; LA64D-NEXT:  .LBB0_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Loop Header: Depth=1
+; LA64D-NEXT:    # Child Loop BB0_3 Depth 2
+; LA64D-NEXT:    ffint.s.w $fa2, $fa1
+; LA64D-NEXT:    fadd.s $fa2, $fa0, $fa2
+; LA64D-NEXT:    movfr2gr.s $a1, $fa2
+; LA64D-NEXT:    movfr2gr.s $a2, $fa0
+; LA64D-NEXT:  .LBB0_3: # %atomicrmw.start
+; LA64D-NEXT:    # Parent Loop BB0_1 Depth=1
+; LA64D-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64D-NEXT:    ll.w $a3, $a0, 0
+; LA64D-NEXT:    bne $a3, $a2, .LBB0_5
+; LA64D-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB0_3 Depth=2
+; LA64D-NEXT:    dbar 0
+; LA64D-NEXT:    move $a4, $a1
+; LA64D-NEXT:    sc.w $a4, $a0, 0
+; LA64D-NEXT:    beqz $a4, .LBB0_3
+; LA64D-NEXT:    b .LBB0_6
+; LA64D-NEXT:  .LBB0_5: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LA64D-NEXT:    dbar 1792
+; LA64D-NEXT:  .LBB0_6: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB0_1 Depth=1
+; LA64D-NEXT:    movgr2fr.w $fa0, $a3
+; LA64D-NEXT:    addi.w $a1, $a2, 0
+; LA64D-NEXT:    bne $a3, $a1, .LBB0_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    ret
+  %v = atomicrmw fadd ptr %p, float 1.0 acquire, align 4
+  ret float %v
+}
+
+define float @float_fsub_acquire(ptr %p) nounwind {
+; LA64F-LABEL: float_fsub_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    fld.s $fa0, $a0, 0
+; LA64F-NEXT:    pcalau12i $a1, %pc_hi20(.LCPI1_0)
+; LA64F-NEXT:    addi.d $a1, $a1, %pc_lo12(.LCPI1_0)
+; LA64F-NEXT:    fld.s $fa1, $a1, 0
+; LA64F-NEXT:  .LBB1_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Loop Header: Depth=1
+; LA64F-NEXT:    # Child Loop BB1_3 Depth 2
+; LA64F-NEXT:    fadd.s $fa2, $fa0, $fa1
+; LA64F-NEXT:    movfr2gr.s $a1, $fa2
+; LA64F-NEXT:    movfr2gr.s $a2, $fa0
+; LA64F-NEXT:  .LBB1_3: # %atomicrmw.start
+; LA64F-NEXT:    # Parent Loop BB1_1 Depth=1
+; LA64F-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64F-NEXT:    ll.w $a3, $a0, 0
+; LA64F-NEXT:    bne $a3, $a2, .LBB1_5
+; LA64F-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB1_3 Depth=2
+; LA64F-NEXT:    dbar 0
+; LA64F-NEXT:    move $a4, $a1
+; LA64F-NEXT:    sc.w $a4, $a0, 0
+; LA64F-NEXT:    beqz $a4, .LBB1_3
+; LA64F-NEXT:    b .LBB1_6
+; LA64F-NEXT:  .LBB1_5: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LA64F-NEXT:    dbar 1792
+; LA64F-NEXT:  .LBB1_6: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LA64F-NEXT:    movgr2fr.w $fa0, $a3
+; LA64F-NEXT:    addi.w $a1, $a2, 0
+; LA64F-NEXT:    bne $a3, $a1, .LBB1_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: float_fsub_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fld.s $fa0, $a0, 0
+; LA64D-NEXT:    pcalau12i $a1, %pc_hi20(.LCPI1_0)
+; LA64D-NEXT:    addi.d $a1, $a1, %pc_lo12(.LCPI1_0)
+; LA64D-NEXT:    fld.s $fa1, $a1, 0
+; LA64D-NEXT:  .LBB1_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Loop Header: Depth=1
+; LA64D-NEXT:    # Child Loop BB1_3 Depth 2
+; LA64D-NEXT:    fadd.s $fa2, $fa0, $fa1
+; LA64D-NEXT:    movfr2gr.s $a1, $fa2
+; LA64D-NEXT:    movfr2gr.s $a2, $fa0
+; LA64D-NEXT:  .LBB1_3: # %atomicrmw.start
+; LA64D-NEXT:    # Parent Loop BB1_1 Depth=1
+; LA64D-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64D-NEXT:    ll.w $a3, $a0, 0
+; LA64D-NEXT:    bne $a3, $a2, .LBB1_5
+; LA64D-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB1_3 Depth=2
+; LA64D-NEXT:    dbar 0
+; LA64D-NEXT:    move $a4, $a1
+; LA64D-NEXT:    sc.w $a4, $a0, 0
+; LA64D-NEXT:    beqz $a4, .LBB1_3
+; LA64D-NEXT:    b .LBB1_6
+; LA64D-NEXT:  .LBB1_5: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LA64D-NEXT:    dbar 1792
+; LA64D-NEXT:  .LBB1_6: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB1_1 Depth=1
+; LA64D-NEXT:    movgr2fr.w $fa0, $a3
+; LA64D-NEXT:    addi.w $a1, $a2, 0
+; LA64D-NEXT:    bne $a3, $a1, .LBB1_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    ret
+  %v = atomicrmw fsub ptr %p, float 1.0 acquire, align 4
+  ret float %v
+}
+
+define float @float_fmin_acquire(ptr %p) nounwind {
+; LA64F-LABEL: float_fmin_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    fld.s $fa0, $a0, 0
+; LA64F-NEXT:    addi.w $a1, $zero, 1
+; LA64F-NEXT:    movgr2fr.w $fa1, $a1
+; LA64F-NEXT:  .LBB2_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Loop Header: Depth=1
+; LA64F-NEXT:    # Child Loop BB2_3 Depth 2
+; LA64F-NEXT:    ffint.s.w $fa2, $fa1
+; LA64F-NEXT:    fmax.s $fa3, $fa0, $fa0
+; LA64F-NEXT:    fmin.s $fa2, $fa3, $fa2
+; LA64F-NEXT:    movfr2gr.s $a1, $fa2
+; LA64F-NEXT:    movfr2gr.s $a2, $fa0
+; LA64F-NEXT:  .LBB2_3: # %atomicrmw.start
+; LA64F-NEXT:    # Parent Loop BB2_1 Depth=1
+; LA64F-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64F-NEXT:    ll.w $a3, $a0, 0
+; LA64F-NEXT:    bne $a3, $a2, .LBB2_5
+; LA64F-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB2_3 Depth=2
+; LA64F-NEXT:    dbar 0
+; LA64F-NEXT:    move $a4, $a1
+; LA64F-NEXT:    sc.w $a4, $a0, 0
+; LA64F-NEXT:    beqz $a4, .LBB2_3
+; LA64F-NEXT:    b .LBB2_6
+; LA64F-NEXT:  .LBB2_5: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LA64F-NEXT:    dbar 1792
+; LA64F-NEXT:  .LBB2_6: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LA64F-NEXT:    movgr2fr.w $fa0, $a3
+; LA64F-NEXT:    addi.w $a1, $a2, 0
+; LA64F-NEXT:    bne $a3, $a1, .LBB2_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: float_fmin_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fld.s $fa0, $a0, 0
+; LA64D-NEXT:    addi.w $a1, $zero, 1
+; LA64D-NEXT:    movgr2fr.w $fa1, $a1
+; LA64D-NEXT:  .LBB2_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Loop Header: Depth=1
+; LA64D-NEXT:    # Child Loop BB2_3 Depth 2
+; LA64D-NEXT:    ffint.s.w $fa2, $fa1
+; LA64D-NEXT:    fmax.s $fa3, $fa0, $fa0
+; LA64D-NEXT:    fmin.s $fa2, $fa3, $fa2
+; LA64D-NEXT:    movfr2gr.s $a1, $fa2
+; LA64D-NEXT:    movfr2gr.s $a2, $fa0
+; LA64D-NEXT:  .LBB2_3: # %atomicrmw.start
+; LA64D-NEXT:    # Parent Loop BB2_1 Depth=1
+; LA64D-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64D-NEXT:    ll.w $a3, $a0, 0
+; LA64D-NEXT:    bne $a3, $a2, .LBB2_5
+; LA64D-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB2_3 Depth=2
+; LA64D-NEXT:    dbar 0
+; LA64D-NEXT:    move $a4, $a1
+; LA64D-NEXT:    sc.w $a4, $a0, 0
+; LA64D-NEXT:    beqz $a4, .LBB2_3
+; LA64D-NEXT:    b .LBB2_6
+; LA64D-NEXT:  .LBB2_5: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LA64D-NEXT:    dbar 1792
+; LA64D-NEXT:  .LBB2_6: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB2_1 Depth=1
+; LA64D-NEXT:    movgr2fr.w $fa0, $a3
+; LA64D-NEXT:    addi.w $a1, $a2, 0
+; LA64D-NEXT:    bne $a3, $a1, .LBB2_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    ret
+  %v = atomicrmw fmin ptr %p, float 1.0 acquire, align 4
+  ret float %v
+}
+
+define float @float_fmax_acquire(ptr %p) nounwind {
+; LA64F-LABEL: float_fmax_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    fld.s $fa0, $a0, 0
+; LA64F-NEXT:    addi.w $a1, $zero, 1
+; LA64F-NEXT:    movgr2fr.w $fa1, $a1
+; LA64F-NEXT:  .LBB3_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Loop Header: Depth=1
+; LA64F-NEXT:    # Child Loop BB3_3 Depth 2
+; LA64F-NEXT:    ffint.s.w $fa2, $fa1
+; LA64F-NEXT:    fmax.s $fa3, $fa0, $fa0
+; LA64F-NEXT:    fmax.s $fa2, $fa3, $fa2
+; LA64F-NEXT:    movfr2gr.s $a1, $fa2
+; LA64F-NEXT:    movfr2gr.s $a2, $fa0
+; LA64F-NEXT:  .LBB3_3: # %atomicrmw.start
+; LA64F-NEXT:    # Parent Loop BB3_1 Depth=1
+; LA64F-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64F-NEXT:    ll.w $a3, $a0, 0
+; LA64F-NEXT:    bne $a3, $a2, .LBB3_5
+; LA64F-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB3_3 Depth=2
+; LA64F-NEXT:    dbar 0
+; LA64F-NEXT:    move $a4, $a1
+; LA64F-NEXT:    sc.w $a4, $a0, 0
+; LA64F-NEXT:    beqz $a4, .LBB3_3
+; LA64F-NEXT:    b .LBB3_6
+; LA64F-NEXT:  .LBB3_5: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LA64F-NEXT:    dbar 1792
+; LA64F-NEXT:  .LBB3_6: # %atomicrmw.start
+; LA64F-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LA64F-NEXT:    movgr2fr.w $fa0, $a3
+; LA64F-NEXT:    addi.w $a1, $a2, 0
+; LA64F-NEXT:    bne $a3, $a1, .LBB3_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: float_fmax_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fld.s $fa0, $a0, 0
+; LA64D-NEXT:    addi.w $a1, $zero, 1
+; LA64D-NEXT:    movgr2fr.w $fa1, $a1
+; LA64D-NEXT:  .LBB3_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Loop Header: Depth=1
+; LA64D-NEXT:    # Child Loop BB3_3 Depth 2
+; LA64D-NEXT:    ffint.s.w $fa2, $fa1
+; LA64D-NEXT:    fmax.s $fa3, $fa0, $fa0
+; LA64D-NEXT:    fmax.s $fa2, $fa3, $fa2
+; LA64D-NEXT:    movfr2gr.s $a1, $fa2
+; LA64D-NEXT:    movfr2gr.s $a2, $fa0
+; LA64D-NEXT:  .LBB3_3: # %atomicrmw.start
+; LA64D-NEXT:    # Parent Loop BB3_1 Depth=1
+; LA64D-NEXT:    # => This Inner Loop Header: Depth=2
+; LA64D-NEXT:    ll.w $a3, $a0, 0
+; LA64D-NEXT:    bne $a3, $a2, .LBB3_5
+; LA64D-NEXT:  # %bb.4: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB3_3 Depth=2
+; LA64D-NEXT:    dbar 0
+; LA64D-NEXT:    move $a4, $a1
+; LA64D-NEXT:    sc.w $a4, $a0, 0
+; LA64D-NEXT:    beqz $a4, .LBB3_3
+; LA64D-NEXT:    b .LBB3_6
+; LA64D-NEXT:  .LBB3_5: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LA64D-NEXT:    dbar 1792
+; LA64D-NEXT:  .LBB3_6: # %atomicrmw.start
+; LA64D-NEXT:    # in Loop: Header=BB3_1 Depth=1
+; LA64D-NEXT:    movgr2fr.w $fa0, $a3
+; LA64D-NEXT:    addi.w $a1, $a2, 0
+; LA64D-NEXT:    bne $a3, $a1, .LBB3_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    ret
+  %v = atomicrmw fmax ptr %p, float 1.0 acquire, align 4
+  ret float %v
+}
+
+define double @double_fadd_acquire(ptr %p) nounwind {
+; LA64F-LABEL: double_fadd_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    addi.d $sp, $sp, -64
+; LA64F-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s0, $sp, 40 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s1, $sp, 32 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s2, $sp, 24 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s3, $sp, 16 # 8-byte Folded Spill
+; LA64F-NEXT:    move $fp, $a0
+; LA64F-NEXT:    ld.wu $a0, $a0, 0
+; LA64F-NEXT:    ld.wu $a1, $fp, 4
+; LA64F-NEXT:    slli.d $a1, $a1, 32
+; LA64F-NEXT:    or $a0, $a1, $a0
+; LA64F-NEXT:    ori $s0, $zero, 8
+; LA64F-NEXT:    addi.d $s1, $sp, 8
+; LA64F-NEXT:    addi.d $s2, $sp, 0
+; LA64F-NEXT:    ori $s3, $zero, 2
+; LA64F-NEXT:  .LBB4_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64F-NEXT:    st.d $a0, $sp, 8
+; LA64F-NEXT:    lu52i.d $a1, $zero, 1023
+; LA64F-NEXT:    bl %plt(__adddf3)
+; LA64F-NEXT:    st.d $a0, $sp, 0
+; LA64F-NEXT:    move $a0, $s0
+; LA64F-NEXT:    move $a1, $fp
+; LA64F-NEXT:    move $a2, $s1
+; LA64F-NEXT:    move $a3, $s2
+; LA64F-NEXT:    move $a4, $s3
+; LA64F-NEXT:    move $a5, $s3
+; LA64F-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64F-NEXT:    move $a1, $a0
+; LA64F-NEXT:    ld.d $a0, $sp, 8
+; LA64F-NEXT:    beqz $a1, .LBB4_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ld.d $s3, $sp, 16 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s2, $sp, 24 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s1, $sp, 32 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s0, $sp, 40 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64F-NEXT:    addi.d $sp, $sp, 64
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: double_fadd_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    addi.d $sp, $sp, -80
+; LA64D-NEXT:    st.d $ra, $sp, 72 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $fp, $sp, 64 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s0, $sp, 56 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s1, $sp, 48 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s2, $sp, 40 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s3, $sp, 32 # 8-byte Folded Spill
+; LA64D-NEXT:    fst.d $fs0, $sp, 24 # 8-byte Folded Spill
+; LA64D-NEXT:    move $fp, $a0
+; LA64D-NEXT:    ld.wu $a0, $a0, 0
+; LA64D-NEXT:    ld.wu $a1, $fp, 4
+; LA64D-NEXT:    slli.d $a1, $a1, 32
+; LA64D-NEXT:    or $a0, $a1, $a0
+; LA64D-NEXT:    movgr2fr.d $fa0, $a0
+; LA64D-NEXT:    addi.d $a0, $zero, 1
+; LA64D-NEXT:    movgr2fr.d $fs0, $a0
+; LA64D-NEXT:    ori $s0, $zero, 8
+; LA64D-NEXT:    addi.d $s1, $sp, 16
+; LA64D-NEXT:    addi.d $s2, $sp, 8
+; LA64D-NEXT:    ori $s3, $zero, 2
+; LA64D-NEXT:  .LBB4_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64D-NEXT:    fst.d $fa0, $sp, 16
+; LA64D-NEXT:    ffint.d.l $fa1, $fs0
+; LA64D-NEXT:    fadd.d $fa0, $fa0, $fa1
+; LA64D-NEXT:    fst.d $fa0, $sp, 8
+; LA64D-NEXT:    move $a0, $s0
+; LA64D-NEXT:    move $a1, $fp
+; LA64D-NEXT:    move $a2, $s1
+; LA64D-NEXT:    move $a3, $s2
+; LA64D-NEXT:    move $a4, $s3
+; LA64D-NEXT:    move $a5, $s3
+; LA64D-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64D-NEXT:    fld.d $fa0, $sp, 16
+; LA64D-NEXT:    beqz $a0, .LBB4_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    fld.d $fs0, $sp, 24 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s3, $sp, 32 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s2, $sp, 40 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s1, $sp, 48 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s0, $sp, 56 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $fp, $sp, 64 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $ra, $sp, 72 # 8-byte Folded Reload
+; LA64D-NEXT:    addi.d $sp, $sp, 80
+; LA64D-NEXT:    ret
+  %v = atomicrmw fadd ptr %p, double 1.0 acquire, align 4
+  ret double %v
+}
+
+define double @double_fsub_acquire(ptr %p) nounwind {
+; LA64F-LABEL: double_fsub_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    addi.d $sp, $sp, -64
+; LA64F-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s0, $sp, 40 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s1, $sp, 32 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s2, $sp, 24 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s3, $sp, 16 # 8-byte Folded Spill
+; LA64F-NEXT:    move $fp, $a0
+; LA64F-NEXT:    ld.wu $a0, $a0, 0
+; LA64F-NEXT:    ld.wu $a1, $fp, 4
+; LA64F-NEXT:    slli.d $a1, $a1, 32
+; LA64F-NEXT:    or $a0, $a1, $a0
+; LA64F-NEXT:    ori $s0, $zero, 8
+; LA64F-NEXT:    addi.d $s1, $sp, 8
+; LA64F-NEXT:    addi.d $s2, $sp, 0
+; LA64F-NEXT:    ori $s3, $zero, 2
+; LA64F-NEXT:  .LBB5_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64F-NEXT:    st.d $a0, $sp, 8
+; LA64F-NEXT:    lu52i.d $a1, $zero, -1025
+; LA64F-NEXT:    bl %plt(__adddf3)
+; LA64F-NEXT:    st.d $a0, $sp, 0
+; LA64F-NEXT:    move $a0, $s0
+; LA64F-NEXT:    move $a1, $fp
+; LA64F-NEXT:    move $a2, $s1
+; LA64F-NEXT:    move $a3, $s2
+; LA64F-NEXT:    move $a4, $s3
+; LA64F-NEXT:    move $a5, $s3
+; LA64F-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64F-NEXT:    move $a1, $a0
+; LA64F-NEXT:    ld.d $a0, $sp, 8
+; LA64F-NEXT:    beqz $a1, .LBB5_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ld.d $s3, $sp, 16 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s2, $sp, 24 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s1, $sp, 32 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s0, $sp, 40 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64F-NEXT:    addi.d $sp, $sp, 64
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: double_fsub_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    addi.d $sp, $sp, -80
+; LA64D-NEXT:    st.d $ra, $sp, 72 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $fp, $sp, 64 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s0, $sp, 56 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s1, $sp, 48 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s2, $sp, 40 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s3, $sp, 32 # 8-byte Folded Spill
+; LA64D-NEXT:    fst.d $fs0, $sp, 24 # 8-byte Folded Spill
+; LA64D-NEXT:    move $fp, $a0
+; LA64D-NEXT:    ld.wu $a0, $a0, 0
+; LA64D-NEXT:    ld.wu $a1, $fp, 4
+; LA64D-NEXT:    slli.d $a1, $a1, 32
+; LA64D-NEXT:    or $a0, $a1, $a0
+; LA64D-NEXT:    movgr2fr.d $fa0, $a0
+; LA64D-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI5_0)
+; LA64D-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI5_0)
+; LA64D-NEXT:    fld.d $fs0, $a0, 0
+; LA64D-NEXT:    ori $s0, $zero, 8
+; LA64D-NEXT:    addi.d $s1, $sp, 16
+; LA64D-NEXT:    addi.d $s2, $sp, 8
+; LA64D-NEXT:    ori $s3, $zero, 2
+; LA64D-NEXT:  .LBB5_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64D-NEXT:    fst.d $fa0, $sp, 16
+; LA64D-NEXT:    fadd.d $fa0, $fa0, $fs0
+; LA64D-NEXT:    fst.d $fa0, $sp, 8
+; LA64D-NEXT:    move $a0, $s0
+; LA64D-NEXT:    move $a1, $fp
+; LA64D-NEXT:    move $a2, $s1
+; LA64D-NEXT:    move $a3, $s2
+; LA64D-NEXT:    move $a4, $s3
+; LA64D-NEXT:    move $a5, $s3
+; LA64D-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64D-NEXT:    fld.d $fa0, $sp, 16
+; LA64D-NEXT:    beqz $a0, .LBB5_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    fld.d $fs0, $sp, 24 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s3, $sp, 32 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s2, $sp, 40 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s1, $sp, 48 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s0, $sp, 56 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $fp, $sp, 64 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $ra, $sp, 72 # 8-byte Folded Reload
+; LA64D-NEXT:    addi.d $sp, $sp, 80
+; LA64D-NEXT:    ret
+  %v = atomicrmw fsub ptr %p, double 1.0 acquire, align 4
+  ret double %v
+}
+
+define double @double_fmin_acquire(ptr %p) nounwind {
+; LA64F-LABEL: double_fmin_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    addi.d $sp, $sp, -64
+; LA64F-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s0, $sp, 40 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s1, $sp, 32 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s2, $sp, 24 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s3, $sp, 16 # 8-byte Folded Spill
+; LA64F-NEXT:    move $fp, $a0
+; LA64F-NEXT:    ld.wu $a0, $a0, 0
+; LA64F-NEXT:    ld.wu $a1, $fp, 4
+; LA64F-NEXT:    slli.d $a1, $a1, 32
+; LA64F-NEXT:    or $a0, $a1, $a0
+; LA64F-NEXT:    ori $s0, $zero, 8
+; LA64F-NEXT:    addi.d $s1, $sp, 8
+; LA64F-NEXT:    addi.d $s2, $sp, 0
+; LA64F-NEXT:    ori $s3, $zero, 2
+; LA64F-NEXT:  .LBB6_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64F-NEXT:    st.d $a0, $sp, 8
+; LA64F-NEXT:    lu52i.d $a1, $zero, 1023
+; LA64F-NEXT:    bl %plt(fmin)
+; LA64F-NEXT:    st.d $a0, $sp, 0
+; LA64F-NEXT:    move $a0, $s0
+; LA64F-NEXT:    move $a1, $fp
+; LA64F-NEXT:    move $a2, $s1
+; LA64F-NEXT:    move $a3, $s2
+; LA64F-NEXT:    move $a4, $s3
+; LA64F-NEXT:    move $a5, $s3
+; LA64F-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64F-NEXT:    move $a1, $a0
+; LA64F-NEXT:    ld.d $a0, $sp, 8
+; LA64F-NEXT:    beqz $a1, .LBB6_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ld.d $s3, $sp, 16 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s2, $sp, 24 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s1, $sp, 32 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s0, $sp, 40 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64F-NEXT:    addi.d $sp, $sp, 64
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: double_fmin_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    addi.d $sp, $sp, -80
+; LA64D-NEXT:    st.d $ra, $sp, 72 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $fp, $sp, 64 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s0, $sp, 56 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s1, $sp, 48 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s2, $sp, 40 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s3, $sp, 32 # 8-byte Folded Spill
+; LA64D-NEXT:    fst.d $fs0, $sp, 24 # 8-byte Folded Spill
+; LA64D-NEXT:    move $fp, $a0
+; LA64D-NEXT:    ld.wu $a0, $a0, 0
+; LA64D-NEXT:    ld.wu $a1, $fp, 4
+; LA64D-NEXT:    slli.d $a1, $a1, 32
+; LA64D-NEXT:    or $a0, $a1, $a0
+; LA64D-NEXT:    movgr2fr.d $fa0, $a0
+; LA64D-NEXT:    addi.d $a0, $zero, 1
+; LA64D-NEXT:    movgr2fr.d $fs0, $a0
+; LA64D-NEXT:    ori $s0, $zero, 8
+; LA64D-NEXT:    addi.d $s1, $sp, 16
+; LA64D-NEXT:    addi.d $s2, $sp, 8
+; LA64D-NEXT:    ori $s3, $zero, 2
+; LA64D-NEXT:  .LBB6_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64D-NEXT:    fst.d $fa0, $sp, 16
+; LA64D-NEXT:    fmax.d $fa0, $fa0, $fa0
+; LA64D-NEXT:    ffint.d.l $fa1, $fs0
+; LA64D-NEXT:    fmin.d $fa0, $fa0, $fa1
+; LA64D-NEXT:    fst.d $fa0, $sp, 8
+; LA64D-NEXT:    move $a0, $s0
+; LA64D-NEXT:    move $a1, $fp
+; LA64D-NEXT:    move $a2, $s1
+; LA64D-NEXT:    move $a3, $s2
+; LA64D-NEXT:    move $a4, $s3
+; LA64D-NEXT:    move $a5, $s3
+; LA64D-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64D-NEXT:    fld.d $fa0, $sp, 16
+; LA64D-NEXT:    beqz $a0, .LBB6_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    fld.d $fs0, $sp, 24 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s3, $sp, 32 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s2, $sp, 40 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s1, $sp, 48 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s0, $sp, 56 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $fp, $sp, 64 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $ra, $sp, 72 # 8-byte Folded Reload
+; LA64D-NEXT:    addi.d $sp, $sp, 80
+; LA64D-NEXT:    ret
+  %v = atomicrmw fmin ptr %p, double 1.0 acquire, align 4
+  ret double %v
+}
+
+define double @double_fmax_acquire(ptr %p) nounwind {
+; LA64F-LABEL: double_fmax_acquire:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    addi.d $sp, $sp, -64
+; LA64F-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s0, $sp, 40 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s1, $sp, 32 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s2, $sp, 24 # 8-byte Folded Spill
+; LA64F-NEXT:    st.d $s3, $sp, 16 # 8-byte Folded Spill
+; LA64F-NEXT:    move $fp, $a0
+; LA64F-NEXT:    ld.wu $a0, $a0, 0
+; LA64F-NEXT:    ld.wu $a1, $fp, 4
+; LA64F-NEXT:    slli.d $a1, $a1, 32
+; LA64F-NEXT:    or $a0, $a1, $a0
+; LA64F-NEXT:    ori $s0, $zero, 8
+; LA64F-NEXT:    addi.d $s1, $sp, 8
+; LA64F-NEXT:    addi.d $s2, $sp, 0
+; LA64F-NEXT:    ori $s3, $zero, 2
+; LA64F-NEXT:  .LBB7_1: # %atomicrmw.start
+; LA64F-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64F-NEXT:    st.d $a0, $sp, 8
+; LA64F-NEXT:    lu52i.d $a1, $zero, 1023
+; LA64F-NEXT:    bl %plt(fmax)
+; LA64F-NEXT:    st.d $a0, $sp, 0
+; LA64F-NEXT:    move $a0, $s0
+; LA64F-NEXT:    move $a1, $fp
+; LA64F-NEXT:    move $a2, $s1
+; LA64F-NEXT:    move $a3, $s2
+; LA64F-NEXT:    move $a4, $s3
+; LA64F-NEXT:    move $a5, $s3
+; LA64F-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64F-NEXT:    move $a1, $a0
+; LA64F-NEXT:    ld.d $a0, $sp, 8
+; LA64F-NEXT:    beqz $a1, .LBB7_1
+; LA64F-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64F-NEXT:    ld.d $s3, $sp, 16 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s2, $sp, 24 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s1, $sp, 32 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $s0, $sp, 40 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64F-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64F-NEXT:    addi.d $sp, $sp, 64
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: double_fmax_acquire:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    addi.d $sp, $sp, -80
+; LA64D-NEXT:    st.d $ra, $sp, 72 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $fp, $sp, 64 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s0, $sp, 56 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s1, $sp, 48 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s2, $sp, 40 # 8-byte Folded Spill
+; LA64D-NEXT:    st.d $s3, $sp, 32 # 8-byte Folded Spill
+; LA64D-NEXT:    fst.d $fs0, $sp, 24 # 8-byte Folded Spill
+; LA64D-NEXT:    move $fp, $a0
+; LA64D-NEXT:    ld.wu $a0, $a0, 0
+; LA64D-NEXT:    ld.wu $a1, $fp, 4
+; LA64D-NEXT:    slli.d $a1, $a1, 32
+; LA64D-NEXT:    or $a0, $a1, $a0
+; LA64D-NEXT:    movgr2fr.d $fa0, $a0
+; LA64D-NEXT:    addi.d $a0, $zero, 1
+; LA64D-NEXT:    movgr2fr.d $fs0, $a0
+; LA64D-NEXT:    ori $s0, $zero, 8
+; LA64D-NEXT:    addi.d $s1, $sp, 16
+; LA64D-NEXT:    addi.d $s2, $sp, 8
+; LA64D-NEXT:    ori $s3, $zero, 2
+; LA64D-NEXT:  .LBB7_1: # %atomicrmw.start
+; LA64D-NEXT:    # =>This Inner Loop Header: Depth=1
+; LA64D-NEXT:    fst.d $fa0, $sp, 16
+; LA64D-NEXT:    fmax.d $fa0, $fa0, $fa0
+; LA64D-NEXT:    ffint.d.l $fa1, $fs0
+; LA64D-NEXT:    fmax.d $fa0, $fa0, $fa1
+; LA64D-NEXT:    fst.d $fa0, $sp, 8
+; LA64D-NEXT:    move $a0, $s0
+; LA64D-NEXT:    move $a1, $fp
+; LA64D-NEXT:    move $a2, $s1
+; LA64D-NEXT:    move $a3, $s2
+; LA64D-NEXT:    move $a4, $s3
+; LA64D-NEXT:    move $a5, $s3
+; LA64D-NEXT:    bl %plt(__atomic_compare_exchange)
+; LA64D-NEXT:    fld.d $fa0, $sp, 16
+; LA64D-NEXT:    beqz $a0, .LBB7_1
+; LA64D-NEXT:  # %bb.2: # %atomicrmw.end
+; LA64D-NEXT:    fld.d $fs0, $sp, 24 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s3, $sp, 32 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s2, $sp, 40 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s1, $sp, 48 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $s0, $sp, 56 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $fp, $sp, 64 # 8-byte Folded Reload
+; LA64D-NEXT:    ld.d $ra, $sp, 72 # 8-byte Folded Reload
+; LA64D-NEXT:    addi.d $sp, $sp, 80
+; LA64D-NEXT:    ret
+  %v = atomicrmw fmax ptr %p, double 1.0 acquire, align 4
+  ret double %v
+}

llvm/test/Transforms/AtomicExpand/LoongArch/atomicrmw-fp.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S --mtriple=loongarch64 --atomic-expand --mattr=+d %s | FileCheck %s
+
+define float @atomicrmw_fadd_float(ptr %ptr, float %value) {
+; CHECK-LABEL: @atomicrmw_fadd_float(
+; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[PTR:%.*]], align 4
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CHECK-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret float [[TMP5]]
+;
+  %res = atomicrmw fadd ptr %ptr, float %value seq_cst
+  ret float %res
+}
+
+define float @atomicrmw_fsub_float(ptr %ptr, float %value) {
+; CHECK-LABEL: @atomicrmw_fsub_float(
+; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[PTR:%.*]], align 4
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[NEW:%.*]] = fsub float [[LOADED]], [[VALUE:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CHECK-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret float [[TMP5]]
+;
+  %res = atomicrmw fsub ptr %ptr, float %value seq_cst
+  ret float %res
+}
+
+define float @atomicrmw_fmin_float(ptr %ptr, float %value) {
+; CHECK-LABEL: @atomicrmw_fmin_float(
+; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[PTR:%.*]], align 4
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP6:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = call float @llvm.minnum.f32(float [[LOADED]], float [[VALUE:%.*]])
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast float [[LOADED]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = cmpxchg ptr [[PTR]], i32 [[TMP4]], i32 [[TMP3]] seq_cst seq_cst, align 4
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
+; CHECK-NEXT:    [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret float [[TMP6]]
+;
+  %res = atomicrmw fmin ptr %ptr, float %value seq_cst
+  ret float %res
+}
+
+define float @atomicrmw_fmax_float(ptr %ptr, float %value) {
+; CHECK-LABEL: @atomicrmw_fmax_float(
+; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[PTR:%.*]], align 4
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP6:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = call float @llvm.maxnum.f32(float [[LOADED]], float [[VALUE:%.*]])
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast float [[LOADED]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = cmpxchg ptr [[PTR]], i32 [[TMP4]], i32 [[TMP3]] seq_cst seq_cst, align 4
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
+; CHECK-NEXT:    [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret float [[TMP6]]
+;
+  %res = atomicrmw fmax ptr %ptr, float %value seq_cst
+  ret float %res
+}
+
+define double @atomicrmw_fadd_double(ptr %ptr, double %value) {
+; CHECK-LABEL: @atomicrmw_fadd_double(
+; CHECK-NEXT:    [[TMP1:%.*]] = load double, ptr [[PTR:%.*]], align 8
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi double [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[NEW:%.*]] = fadd double [[LOADED]], [[VALUE:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast double [[NEW]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double [[LOADED]] to i64
+; CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[PTR]], i64 [[TMP3]], i64 [[TMP2]] seq_cst seq_cst, align 8
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP4]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP4]], 0
+; CHECK-NEXT:    [[TMP5]] = bitcast i64 [[NEWLOADED]] to double
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret double [[TMP5]]
+;
+  %res = atomicrmw fadd ptr %ptr, double %value seq_cst
+  ret double %res
+}
+
+define double @atomicrmw_fsub_double(ptr %ptr, double %value) {
+; CHECK-LABEL: @atomicrmw_fsub_double(
+; CHECK-NEXT:    [[TMP1:%.*]] = load double, ptr [[PTR:%.*]], align 8
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi double [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[NEW:%.*]] = fsub double [[LOADED]], [[VALUE:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast double [[NEW]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double [[LOADED]] to i64
+; CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[PTR]], i64 [[TMP3]], i64 [[TMP2]] seq_cst seq_cst, align 8
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP4]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP4]], 0
+; CHECK-NEXT:    [[TMP5]] = bitcast i64 [[NEWLOADED]] to double
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret double [[TMP5]]
+;
+  %res = atomicrmw fsub ptr %ptr, double %value seq_cst
+  ret double %res
+}
+
+define double @atomicrmw_fmin_double(ptr %ptr, double %value) {
+; CHECK-LABEL: @atomicrmw_fmin_double(
+; CHECK-NEXT:    [[TMP1:%.*]] = load double, ptr [[PTR:%.*]], align 8
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi double [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP6:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.minnum.f64(double [[LOADED]], double [[VALUE:%.*]])
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double [[TMP2]] to i64
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast double [[LOADED]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = cmpxchg ptr [[PTR]], i64 [[TMP4]], i64 [[TMP3]] seq_cst seq_cst, align 8
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
+; CHECK-NEXT:    [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret double [[TMP6]]
+;
+  %res = atomicrmw fmin ptr %ptr, double %value seq_cst
+  ret double %res
+}
+
+define double @atomicrmw_fmax_double(ptr %ptr, double %value) {
+; CHECK-LABEL: @atomicrmw_fmax_double(
+; CHECK-NEXT:    [[TMP1:%.*]] = load double, ptr [[PTR:%.*]], align 8
+; CHECK-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CHECK:       atomicrmw.start:
+; CHECK-NEXT:    [[LOADED:%.*]] = phi double [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP6:%.*]], [[ATOMICRMW_START]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.maxnum.f64(double [[LOADED]], double [[VALUE:%.*]])
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double [[TMP2]] to i64
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast double [[LOADED]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = cmpxchg ptr [[PTR]], i64 [[TMP4]], i64 [[TMP3]] seq_cst seq_cst, align 8
+; CHECK-NEXT:    [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
+; CHECK-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
+; CHECK-NEXT:    [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
+; CHECK-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK:       atomicrmw.end:
+; CHECK-NEXT:    ret double [[TMP6]]
+;
+  %res = atomicrmw fmax ptr %ptr, double %value seq_cst
+  ret double %res
+}