[RISCV] Explicitly select second operand of branch condition to X0.

[craig.topper]

Summary

At least based on the lit tests, the coalescer sometimes fail to
propagate the copy from X0 into the branch instruction. This patch
does it manually during isel. The majority of the changes are from
the select patterns.

Some of the changes are just register allocation changes. Only
the Select change affects the whether a b*z instruction is generated
in the tests. I changed the branch pattern for consistency.

[asb]

I'm obviously a big fan of the codegen improvements resulting from this patch...but it always feels unpleasant to add a target-specific workaround for what seems to be a bug/limitation elsewhere. I think it does make sense to land this workaround as it's not clear how feasible improving the coalescer would be, but perhaps you could file an issue to track this problem?

[reames]

LGTM as well. I think this is worth taking, but as noted, I think we may have more oddities to dig into here.

llvm/test/CodeGen/RISCV/fpclamptosat.ll

3493

I think there's still a problem with this code. This initialization of a1 appears to never be clobbered, and yet we have multiple register copies below instead of zero initialization. It's less immediate problematic than the branch interaction, but something odd is still going on here.

[asb]

I've filed https://github.com/llvm/llvm-project/issues/56872 to at least track this, as it would be great if target-independent improvements meant we could remove these patterns again.

llvm/lib/Target/RISCV/RISCVInstrInfo.td

 
 def riscv_selectcc_frag : PatFrag<(ops node:$lhs, node:$rhs, node:$cc,
                                        node:$truev, node:$falsev),
                                   (riscv_selectcc node:$lhs, node:$rhs,
                                                   node:$cc, node:$truev,
                                                   node:$falsev), [{}],
                                   IntCCtoRISCVCC>;
 
+multiclass SelectCC_GPR_rrirr<RegisterClass valty> {
   let usesCustomInserter = 1 in
-class SelectCC_rrirr<RegisterClass valty, RegisterClass cmpty>
-    : Pseudo<(outs valty:$dst),
-             (ins cmpty:$lhs, cmpty:$rhs, ixlenimm:$imm,
+  def _Using_CC_GPR : Pseudo<(outs valty:$dst),
+                             (ins GPR:$lhs, GPR:$rhs, ixlenimm:$cc,
                               valty:$truev, valty:$falsev),
                              [(set valty:$dst,
-               (riscv_selectcc_frag:$imm cmpty:$lhs, cmpty:$rhs, cond,
+                               (riscv_selectcc_frag:$cc GPR:$lhs, GPR:$rhs, cond,
                                                         valty:$truev, valty:$falsev))]>;
+  // Explicitly select 0 in the condition to X0. The register coalescer doesn't
+  // always do it.
+  def : Pat<(riscv_selectcc_frag:$cc GPR:$lhs, 0, cond, valty:$truev,
+                                     valty:$falsev),
+            (!cast<Instruction>(NAME#"_Using_CC_GPR") GPR:$lhs, X0,
+             (IntCCtoRISCVCC $cc), valty:$truev, valty:$falsev)>;
+}
 
-def Select_GPR_Using_CC_GPR : SelectCC_rrirr<GPR, GPR>;
+defm Select_GPR : SelectCC_GPR_rrirr<GPR>;
 
 /// Branches and jumps
 
 // Match `riscv_brcc` and lower to the appropriate RISC-V branch instruction.
-class BccPat<CondCode Cond, RVInstB Inst>
-    : Pat<(riscv_brcc GPR:$rs1, GPR:$rs2, Cond, bb:$imm12),
+multiclass BccPat<CondCode Cond, RVInstB Inst> {
+  def : Pat<(riscv_brcc GPR:$rs1, GPR:$rs2, Cond, bb:$imm12),
             (Inst GPR:$rs1, GPR:$rs2, simm13_lsb0:$imm12)>;
+  // Explicitly select 0 to X0. The register coalescer doesn't always do it.
+  def : Pat<(riscv_brcc GPR:$rs1, 0, Cond, bb:$imm12),
+            (Inst GPR:$rs1, X0, simm13_lsb0:$imm12)>;
+}
 
-def : BccPat<SETEQ, BEQ>;
-def : BccPat<SETNE, BNE>;
-def : BccPat<SETLT, BLT>;
-def : BccPat<SETGE, BGE>;
-def : BccPat<SETULT, BLTU>;
-def : BccPat<SETUGE, BGEU>;
+defm : BccPat<SETEQ, BEQ>;
+defm : BccPat<SETNE, BNE>;
+defm : BccPat<SETLT, BLT>;
+defm : BccPat<SETGE, BGE>;
+defm : BccPat<SETULT, BLTU>;
+defm : BccPat<SETUGE, BGEU>;
 
 let isBarrier = 1, isBranch = 1, isTerminator = 1 in
 def PseudoBR : Pseudo<(outs), (ins simm21_lsb0_jal:$imm20), [(br bb:$imm20)]>,
                PseudoInstExpansion<(JAL X0, simm21_lsb0_jal:$imm20)>;
 
 let isBarrier = 1, isBranch = 1, isIndirectBranch = 1, isTerminator = 1 in
 def PseudoBRIND : Pseudo<(outs), (ins GPRJALR:$rs1, simm12:$imm12), []>,
                   PseudoInstExpansion<(JALR X0, GPR:$rs1, simm12:$imm12)>;

llvm/lib/Target/RISCV/RISCVInstrInfoD.td

 def : Pat<(strict_fsetccs FPR64:$rs1, FPR64:$rs1, SETOEQ),
           (FLE_D $rs1, $rs1)>;
 
 def : PatSetCC<FPR64, any_fsetccs, SETLT, FLT_D>;
 def : PatSetCC<FPR64, any_fsetccs, SETOLT, FLT_D>;
 def : PatSetCC<FPR64, any_fsetccs, SETLE, FLE_D>;
 def : PatSetCC<FPR64, any_fsetccs, SETOLE, FLE_D>;
 
-def Select_FPR64_Using_CC_GPR : SelectCC_rrirr<FPR64, GPR>;
+defm Select_FPR64 : SelectCC_GPR_rrirr<FPR64>;
 
 /// Loads
 
 defm : LdPat<load, FLD, f64>;
 
 /// Stores
 
 defm : StPat<store, FSD, FPR64, f64>;

llvm/lib/Target/RISCV/RISCVInstrInfoF.td

 def : Pat<(strict_fsetccs FPR32:$rs1, FPR32:$rs1, SETOEQ),
           (FLE_S $rs1, $rs1)>;
 
 def : PatSetCC<FPR32, any_fsetccs, SETLT, FLT_S>;
 def : PatSetCC<FPR32, any_fsetccs, SETOLT, FLT_S>;
 def : PatSetCC<FPR32, any_fsetccs, SETLE, FLE_S>;
 def : PatSetCC<FPR32, any_fsetccs, SETOLE, FLE_S>;
 
-def Select_FPR32_Using_CC_GPR : SelectCC_rrirr<FPR32, GPR>;
+defm Select_FPR32 : SelectCC_GPR_rrirr<FPR32>;
 
 /// Loads
 
 defm : LdPat<load, FLW, f32>;
 
 /// Stores
 
 defm : StPat<store, FSW, FPR32, f32>;

llvm/lib/Target/RISCV/RISCVInstrInfoZfh.td

 def : Pat<(strict_fsetccs FPR16:$rs1, FPR16:$rs1, SETOEQ),
           (FLE_H $rs1, $rs1)>;
 
 def : PatSetCC<FPR16, any_fsetccs, SETLT, FLT_H>;
 def : PatSetCC<FPR16, any_fsetccs, SETOLT, FLT_H>;
 def : PatSetCC<FPR16, any_fsetccs, SETLE, FLE_H>;
 def : PatSetCC<FPR16, any_fsetccs, SETOLE, FLE_H>;
 
-def Select_FPR16_Using_CC_GPR : SelectCC_rrirr<FPR16, GPR>;
+defm Select_FPR16 : SelectCC_GPR_rrirr<FPR16>;
 } // Predicates = [HasStdExtZfh]
 
 let Predicates = [HasStdExtZfhOrZfhmin] in {
 /// Loads
 
 defm : LdPat<load, FLH, f16>;
 
 /// Stores

llvm/test/CodeGen/RISCV/double-convert.ll

 ; RV32I-NEXT:    addi sp, sp, -32
 ; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s5, 4(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    mv s1, a1
-; RV32I-NEXT:    mv s2, a0
+; RV32I-NEXT:    mv s0, a1
+; RV32I-NEXT:    mv s1, a0
+; RV32I-NEXT:    lui a0, 269824
+; RV32I-NEXT:    addi a3, a0, -1
+; RV32I-NEXT:    lui a2, 1047552
+; RV32I-NEXT:    mv a0, s1
+; RV32I-NEXT:    call __gtdf2@plt
+; RV32I-NEXT:    mv s3, a0
 ; RV32I-NEXT:    lui a3, 794112
-; RV32I-NEXT:    li s0, 0
+; RV32I-NEXT:    li s2, 0
+; RV32I-NEXT:    mv a0, s1
+; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    li a2, 0
 ; RV32I-NEXT:    call __gedf2@plt
-; RV32I-NEXT:    mv s3, a0
-; RV32I-NEXT:    mv a0, s2
-; RV32I-NEXT:    mv a1, s1
+; RV32I-NEXT:    mv s4, a0
+; RV32I-NEXT:    mv a0, s1
+; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __fixdfsi@plt
+; RV32I-NEXT:    lui a1, 524288
 ; RV32I-NEXT:    lui s5, 524288
-; RV32I-NEXT:    lui s4, 524288
-; RV32I-NEXT:    blt s3, s0, .LBB3_2
+; RV32I-NEXT:    bltz s4, .LBB3_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s4, a0
+; RV32I-NEXT:    mv s5, a0
 ; RV32I-NEXT:  .LBB3_2: # %start
-; RV32I-NEXT:    lui a0, 269824
-; RV32I-NEXT:    addi a3, a0, -1
-; RV32I-NEXT:    lui a2, 1047552
-; RV32I-NEXT:    mv a0, s2
-; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    call __gtdf2@plt
-; RV32I-NEXT:    bge s0, a0, .LBB3_4
+; RV32I-NEXT:    bge s2, s3, .LBB3_4
 ; RV32I-NEXT:  # %bb.3:
-; RV32I-NEXT:    addi s4, s5, -1
+; RV32I-NEXT:    addi s5, a1, -1
 ; RV32I-NEXT:  .LBB3_4: # %start
-; RV32I-NEXT:    mv a0, s2
-; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    mv a2, s2
-; RV32I-NEXT:    mv a3, s1
+; RV32I-NEXT:    mv a0, s1
+; RV32I-NEXT:    mv a1, s0
+; RV32I-NEXT:    mv a2, s1
+; RV32I-NEXT:    mv a3, s0
 ; RV32I-NEXT:    call __unorddf2@plt
-; RV32I-NEXT:    bne a0, s0, .LBB3_6
+; RV32I-NEXT:    bnez a0, .LBB3_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s0, s4
+; RV32I-NEXT:    mv s2, s5
 ; RV32I-NEXT:  .LBB3_6: # %start
-; RV32I-NEXT:    mv a0, s0
+; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s5, 4(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_d_sat:
 ; RV64I:       # %bb.0: # %start
 ; RV64I-NEXT:    addi sp, sp, -48
 ; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s4, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    li a0, -497
 ; RV64I-NEXT:    slli a1, a0, 53
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gedf2@plt
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixdfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    lui s4, 524288
 ; RV64I-NEXT:    lui s3, 524288
-; RV64I-NEXT:    bltz s2, .LBB3_2
+; RV64I-NEXT:    lui s2, 524288
+; RV64I-NEXT:    bltz s1, .LBB3_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB3_2: # %start
 ; RV64I-NEXT:    li a0, 527
 ; RV64I-NEXT:    slli a0, a0, 31
 ; RV64I-NEXT:    addi a0, a0, -1
 ; RV64I-NEXT:    slli a1, a0, 22
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtdf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB3_4
+; RV64I-NEXT:    blez a0, .LBB3_4
 ; RV64I-NEXT:  # %bb.3:
-; RV64I-NEXT:    addiw s3, s4, -1
+; RV64I-NEXT:    addiw s2, s3, -1
 ; RV64I-NEXT:  .LBB3_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unorddf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB3_6
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB3_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv a0, s2
 ; RV64I-NEXT:  .LBB3_6: # %start
-; RV64I-NEXT:    mv a0, s1
 ; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s4, 0(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 48
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i32 @llvm.fptosi.sat.i32.f64(double %a)
   ret i32 %0
 }
 declare i32 @llvm.fptosi.sat.i32.f64(double)
 

 ; RV64IFD-NEXT:    beqz a0, .LBB12_2
 ; RV64IFD-NEXT:  # %bb.1:
 ; RV64IFD-NEXT:    fcvt.l.d a0, fa0, rtz
 ; RV64IFD-NEXT:  .LBB12_2: # %start
 ; RV64IFD-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_l_d_sat:
 ; RV32I:       # %bb.0: # %start
-; RV32I-NEXT:    addi sp, sp, -48
-; RV32I-NEXT:    sw ra, 44(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s0, 40(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s1, 36(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s2, 32(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s3, 28(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s4, 24(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s5, 20(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s6, 16(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s7, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    addi sp, sp, -32
+; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s5, 4(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s6, 0(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s0, a1
 ; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    lui a0, 278016
 ; RV32I-NEXT:    addi s3, a0, -1
 ; RV32I-NEXT:    li a2, -1
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a3, s3
 ; RV32I-NEXT:    call __gtdf2@plt
 ; RV32I-NEXT:    mv s4, a0
 ; RV32I-NEXT:    lui a3, 802304
 ; RV32I-NEXT:    li s2, 0
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    li a2, 0
 ; RV32I-NEXT:    call __gedf2@plt
 ; RV32I-NEXT:    mv s6, a0
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __fixdfdi@plt
 ; RV32I-NEXT:    mv s5, a1
 ; RV32I-NEXT:    mv a1, s2
-; RV32I-NEXT:    blt s6, s2, .LBB12_2
+; RV32I-NEXT:    bltz s6, .LBB12_2
 ; RV32I-NEXT:  # %bb.1: # %start
 ; RV32I-NEXT:    mv a1, a0
 ; RV32I-NEXT:  .LBB12_2: # %start
 ; RV32I-NEXT:    li s6, -1
 ; RV32I-NEXT:    blt s2, s4, .LBB12_4
 ; RV32I-NEXT:  # %bb.3: # %start
 ; RV32I-NEXT:    mv s6, a1
 ; RV32I-NEXT:  .LBB12_4: # %start
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    mv a2, s1
 ; RV32I-NEXT:    mv a3, s0
 ; RV32I-NEXT:    call __unorddf2@plt
 ; RV32I-NEXT:    mv s4, s2
-; RV32I-NEXT:    bne a0, s2, .LBB12_6
+; RV32I-NEXT:    bnez a0, .LBB12_6
 ; RV32I-NEXT:  # %bb.5: # %start
 ; RV32I-NEXT:    mv s4, s6
 ; RV32I-NEXT:  .LBB12_6: # %start
+; RV32I-NEXT:    li a2, -1
+; RV32I-NEXT:    mv a0, s1
+; RV32I-NEXT:    mv a1, s0
+; RV32I-NEXT:    mv a3, s3
+; RV32I-NEXT:    call __gtdf2@plt
+; RV32I-NEXT:    mv s3, a0
 ; RV32I-NEXT:    lui a3, 802304
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    mv a2, s2
 ; RV32I-NEXT:    call __gedf2@plt
-; RV32I-NEXT:    lui s7, 524288
+; RV32I-NEXT:    lui a1, 524288
 ; RV32I-NEXT:    lui s6, 524288
-; RV32I-NEXT:    blt a0, s2, .LBB12_8
+; RV32I-NEXT:    bltz a0, .LBB12_8
 ; RV32I-NEXT:  # %bb.7: # %start
 ; RV32I-NEXT:    mv s6, s5
 ; RV32I-NEXT:  .LBB12_8: # %start
-; RV32I-NEXT:    li a2, -1
-; RV32I-NEXT:    mv a0, s1
-; RV32I-NEXT:    mv a1, s0
-; RV32I-NEXT:    mv a3, s3
-; RV32I-NEXT:    call __gtdf2@plt
-; RV32I-NEXT:    bge s2, a0, .LBB12_10
+; RV32I-NEXT:    bge s2, s3, .LBB12_10
 ; RV32I-NEXT:  # %bb.9:
-; RV32I-NEXT:    addi s6, s7, -1
+; RV32I-NEXT:    addi s6, a1, -1
 ; RV32I-NEXT:  .LBB12_10: # %start
 ; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    mv a2, s1
 ; RV32I-NEXT:    mv a3, s0
 ; RV32I-NEXT:    call __unorddf2@plt
-; RV32I-NEXT:    bne a0, s2, .LBB12_12
+; RV32I-NEXT:    bnez a0, .LBB12_12
 ; RV32I-NEXT:  # %bb.11: # %start
 ; RV32I-NEXT:    mv s2, s6
 ; RV32I-NEXT:  .LBB12_12: # %start
 ; RV32I-NEXT:    mv a0, s4
 ; RV32I-NEXT:    mv a1, s2
-; RV32I-NEXT:    lw ra, 44(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s0, 40(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s1, 36(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s2, 32(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s3, 28(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s4, 24(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s5, 20(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s6, 16(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s7, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    addi sp, sp, 48
+; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s5, 4(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s6, 0(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_l_d_sat:
 ; RV64I:       # %bb.0: # %start
 ; RV64I-NEXT:    addi sp, sp, -48
 ; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s4, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    li a0, -481
 ; RV64I-NEXT:    slli a1, a0, 53
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gedf2@plt
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixdfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    li s4, -1
-; RV64I-NEXT:    bltz s3, .LBB12_2
+; RV64I-NEXT:    li s3, -1
+; RV64I-NEXT:    bltz s2, .LBB12_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    j .LBB12_3
 ; RV64I-NEXT:  .LBB12_2:
-; RV64I-NEXT:    slli s2, s4, 63
+; RV64I-NEXT:    slli s1, s3, 63
 ; RV64I-NEXT:  .LBB12_3: # %start
 ; RV64I-NEXT:    li a0, 543
 ; RV64I-NEXT:    slli a0, a0, 53
 ; RV64I-NEXT:    addi a1, a0, -1
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtdf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB12_5
+; RV64I-NEXT:    blez a0, .LBB12_5
 ; RV64I-NEXT:  # %bb.4:
-; RV64I-NEXT:    srli s2, s4, 1
+; RV64I-NEXT:    srli s1, s3, 1
 ; RV64I-NEXT:  .LBB12_5: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unorddf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB12_7
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB12_7
 ; RV64I-NEXT:  # %bb.6: # %start
-; RV64I-NEXT:    mv s1, s2
-; RV64I-NEXT:  .LBB12_7: # %start
 ; RV64I-NEXT:    mv a0, s1
+; RV64I-NEXT:  .LBB12_7: # %start
 ; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s4, 0(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 48
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i64 @llvm.fptosi.sat.i64.f64(double %a)
   ret i64 %0
 }
 declare i64 @llvm.fptosi.sat.i64.f64(double)
 

 ; RV32I:       # %bb.0: # %start
 ; RV32I-NEXT:    addi sp, sp, -32
 ; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s5, 4(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s1, a1
 ; RV32I-NEXT:    mv s2, a0
-; RV32I-NEXT:    lui a3, 790016
+; RV32I-NEXT:    lui a0, 265728
+; RV32I-NEXT:    addi a3, a0, -64
 ; RV32I-NEXT:    li s0, 0
+; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    li a2, 0
-; RV32I-NEXT:    call __gedf2@plt
+; RV32I-NEXT:    call __gtdf2@plt
 ; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    lui a3, 790016
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    call __fixdfsi@plt
-; RV32I-NEXT:    lui s4, 1048568
-; RV32I-NEXT:    blt s3, s0, .LBB26_2
-; RV32I-NEXT:  # %bb.1: # %start
+; RV32I-NEXT:    mv a2, s0
+; RV32I-NEXT:    call __gedf2@plt
 ; RV32I-NEXT:    mv s4, a0
-; RV32I-NEXT:  .LBB26_2: # %start
-; RV32I-NEXT:    lui a0, 265728
-; RV32I-NEXT:    addi a3, a0, -64
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    mv a2, s0
-; RV32I-NEXT:    call __gtdf2@plt
-; RV32I-NEXT:    bge s0, a0, .LBB26_4
+; RV32I-NEXT:    call __fixdfsi@plt
+; RV32I-NEXT:    lui s5, 1048568
+; RV32I-NEXT:    bltz s4, .LBB26_2
+; RV32I-NEXT:  # %bb.1: # %start
+; RV32I-NEXT:    mv s5, a0
+; RV32I-NEXT:  .LBB26_2: # %start
+; RV32I-NEXT:    bge s0, s3, .LBB26_4
 ; RV32I-NEXT:  # %bb.3:
 ; RV32I-NEXT:    lui a0, 8
-; RV32I-NEXT:    addi s4, a0, -1
+; RV32I-NEXT:    addi s5, a0, -1
 ; RV32I-NEXT:  .LBB26_4: # %start
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
 ; RV32I-NEXT:    mv a2, s2
 ; RV32I-NEXT:    mv a3, s1
 ; RV32I-NEXT:    call __unorddf2@plt
-; RV32I-NEXT:    bne a0, s0, .LBB26_6
+; RV32I-NEXT:    bnez a0, .LBB26_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s0, s4
+; RV32I-NEXT:    mv s0, s5
 ; RV32I-NEXT:  .LBB26_6: # %start
 ; RV32I-NEXT:    slli a0, s0, 16
 ; RV32I-NEXT:    srai a0, a0, 16
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s5, 4(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_s_sat_i16:
 ; RV64I:       # %bb.0: # %start
-; RV64I-NEXT:    addi sp, sp, -48
-; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    addi sp, sp, -32
+; RV64I-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s2, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    li a0, -505
 ; RV64I-NEXT:    slli a1, a0, 53
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gedf2@plt
 ; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixdfdi@plt
-; RV64I-NEXT:    li s2, 0
-; RV64I-NEXT:    lui s3, 1048568
+; RV64I-NEXT:    lui s2, 1048568
 ; RV64I-NEXT:    bltz s1, .LBB26_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB26_2: # %start
 ; RV64I-NEXT:    lui a0, 4152
 ; RV64I-NEXT:    addiw a0, a0, -1
 ; RV64I-NEXT:    slli a1, a0, 38
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtdf2@plt
-; RV64I-NEXT:    bge s2, a0, .LBB26_4
+; RV64I-NEXT:    blez a0, .LBB26_4
 ; RV64I-NEXT:  # %bb.3:
 ; RV64I-NEXT:    lui a0, 8
-; RV64I-NEXT:    addiw s3, a0, -1
+; RV64I-NEXT:    addiw s2, a0, -1
 ; RV64I-NEXT:  .LBB26_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unorddf2@plt
-; RV64I-NEXT:    bne a0, s2, .LBB26_6
+; RV64I-NEXT:    li a1, 0
+; RV64I-NEXT:    bnez a0, .LBB26_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s2, s3
+; RV64I-NEXT:    mv a1, s2
 ; RV64I-NEXT:  .LBB26_6: # %start
-; RV64I-NEXT:    slli a0, s2, 48
+; RV64I-NEXT:    slli a0, a1, 48
 ; RV64I-NEXT:    srai a0, a0, 48
-; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    addi sp, sp, 48
+; RV64I-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 32
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i16 @llvm.fptosi.sat.i16.f64(double %a)
   ret i16 %0
 }
 declare i16 @llvm.fptosi.sat.i16.f64(double)
 
 define zeroext i16 @fcvt_wu_s_i16(double %a) nounwind {

 ; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s1, a1
 ; RV32I-NEXT:    mv s2, a0
-; RV32I-NEXT:    lui a3, 787968
+; RV32I-NEXT:    lui a3, 263676
 ; RV32I-NEXT:    li s0, 0
 ; RV32I-NEXT:    li a2, 0
-; RV32I-NEXT:    call __gedf2@plt
+; RV32I-NEXT:    call __gtdf2@plt
 ; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    lui a3, 787968
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    call __fixdfsi@plt
-; RV32I-NEXT:    li s4, -128
-; RV32I-NEXT:    blt s3, s0, .LBB30_2
-; RV32I-NEXT:  # %bb.1: # %start
+; RV32I-NEXT:    mv a2, s0
+; RV32I-NEXT:    call __gedf2@plt
 ; RV32I-NEXT:    mv s4, a0
-; RV32I-NEXT:  .LBB30_2: # %start
-; RV32I-NEXT:    lui a3, 263676
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
-; RV32I-NEXT:    mv a2, s0
-; RV32I-NEXT:    call __gtdf2@plt
-; RV32I-NEXT:    li s3, 127
-; RV32I-NEXT:    blt s0, a0, .LBB30_4
+; RV32I-NEXT:    call __fixdfsi@plt
+; RV32I-NEXT:    li a1, -128
+; RV32I-NEXT:    bltz s4, .LBB30_2
+; RV32I-NEXT:  # %bb.1: # %start
+; RV32I-NEXT:    mv a1, a0
+; RV32I-NEXT:  .LBB30_2: # %start
+; RV32I-NEXT:    li s4, 127
+; RV32I-NEXT:    blt s0, s3, .LBB30_4
 ; RV32I-NEXT:  # %bb.3: # %start
-; RV32I-NEXT:    mv s3, s4
+; RV32I-NEXT:    mv s4, a1
 ; RV32I-NEXT:  .LBB30_4: # %start
 ; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:    mv a1, s1
 ; RV32I-NEXT:    mv a2, s2
 ; RV32I-NEXT:    mv a3, s1
 ; RV32I-NEXT:    call __unorddf2@plt
-; RV32I-NEXT:    bne a0, s0, .LBB30_6
+; RV32I-NEXT:    bnez a0, .LBB30_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s0, s3
+; RV32I-NEXT:    mv s0, s4
 ; RV32I-NEXT:  .LBB30_6: # %start
 ; RV32I-NEXT:    slli a0, s0, 24
 ; RV32I-NEXT:    srai a0, a0, 24
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_s_sat_i8:
 ; RV64I:       # %bb.0: # %start
-; RV64I-NEXT:    addi sp, sp, -48
-; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    addi sp, sp, -32
+; RV64I-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s2, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    li a0, -509
 ; RV64I-NEXT:    slli a1, a0, 53
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gedf2@plt
 ; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixdfdi@plt
-; RV64I-NEXT:    li s2, 0
-; RV64I-NEXT:    li s3, -128
+; RV64I-NEXT:    li s2, -128
 ; RV64I-NEXT:    bltz s1, .LBB30_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB30_2: # %start
 ; RV64I-NEXT:    lui a0, 65919
 ; RV64I-NEXT:    slli a1, a0, 34
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtdf2@plt
 ; RV64I-NEXT:    li s1, 127
-; RV64I-NEXT:    blt s2, a0, .LBB30_4
+; RV64I-NEXT:    bgtz a0, .LBB30_4
 ; RV64I-NEXT:  # %bb.3: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv s1, s2
 ; RV64I-NEXT:  .LBB30_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unorddf2@plt
-; RV64I-NEXT:    bne a0, s2, .LBB30_6
+; RV64I-NEXT:    li a1, 0
+; RV64I-NEXT:    bnez a0, .LBB30_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s2, s1
+; RV64I-NEXT:    mv a1, s1
 ; RV64I-NEXT:  .LBB30_6: # %start
-; RV64I-NEXT:    slli a0, s2, 56
+; RV64I-NEXT:    slli a0, a1, 56
 ; RV64I-NEXT:    srai a0, a0, 56
-; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    addi sp, sp, 48
+; RV64I-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 32
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i8 @llvm.fptosi.sat.i8.f64(double %a)
   ret i8 %0
 }
 declare i8 @llvm.fptosi.sat.i8.f64(double)
 
 define zeroext i8 @fcvt_wu_s_i8(double %a) nounwind {

llvm/test/CodeGen/RISCV/float-convert.ll

 ; RV32I-LABEL: fcvt_w_s_sat:
 ; RV32I:       # %bb.0: # %start
 ; RV32I-NEXT:    addi sp, sp, -32
 ; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s0, a0
 ; RV32I-NEXT:    lui a1, 847872
 ; RV32I-NEXT:    call __gesf2@plt
-; RV32I-NEXT:    mv s2, a0
+; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __fixsfsi@plt
-; RV32I-NEXT:    li s1, 0
-; RV32I-NEXT:    lui s4, 524288
 ; RV32I-NEXT:    lui s3, 524288
-; RV32I-NEXT:    bltz s2, .LBB1_2
+; RV32I-NEXT:    lui s2, 524288
+; RV32I-NEXT:    bltz s1, .LBB1_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    mv s2, a0
 ; RV32I-NEXT:  .LBB1_2: # %start
 ; RV32I-NEXT:    lui a0, 323584
 ; RV32I-NEXT:    addi a1, a0, -1
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gtsf2@plt
-; RV32I-NEXT:    bge s1, a0, .LBB1_4
+; RV32I-NEXT:    blez a0, .LBB1_4
 ; RV32I-NEXT:  # %bb.3:
-; RV32I-NEXT:    addi s3, s4, -1
+; RV32I-NEXT:    addi s2, s3, -1
 ; RV32I-NEXT:  .LBB1_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s1, .LBB1_6
+; RV32I-NEXT:    mv a1, a0
+; RV32I-NEXT:    li a0, 0
+; RV32I-NEXT:    bnez a1, .LBB1_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s1, s3
+; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:  .LBB1_6: # %start
-; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_s_sat:
 ; RV64I:       # %bb.0: # %start
 ; RV64I-NEXT:    addi sp, sp, -48
 ; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s4, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 847872
 ; RV64I-NEXT:    call __gesf2@plt
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    lui s4, 524288
 ; RV64I-NEXT:    lui s3, 524288
-; RV64I-NEXT:    bltz s2, .LBB1_2
+; RV64I-NEXT:    lui s2, 524288
+; RV64I-NEXT:    bltz s1, .LBB1_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB1_2: # %start
 ; RV64I-NEXT:    lui a0, 323584
 ; RV64I-NEXT:    addiw a1, a0, -1
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB1_4
+; RV64I-NEXT:    blez a0, .LBB1_4
 ; RV64I-NEXT:  # %bb.3:
-; RV64I-NEXT:    addiw s3, s4, -1
+; RV64I-NEXT:    addiw s2, s3, -1
 ; RV64I-NEXT:  .LBB1_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB1_6
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB1_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv a0, s2
 ; RV64I-NEXT:  .LBB1_6: # %start
-; RV64I-NEXT:    mv a0, s1
 ; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s4, 0(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 48
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i32 @llvm.fptosi.sat.i32.f32(float %a)
   ret i32 %0
 }
 declare i32 @llvm.fptosi.sat.i32.f32(float)
 

 ; RV32I-NEXT:    blt s1, a0, .LBB12_4
 ; RV32I-NEXT:  # %bb.3: # %start
 ; RV32I-NEXT:    mv s6, s5
 ; RV32I-NEXT:  .LBB12_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
 ; RV32I-NEXT:    mv s3, s1
-; RV32I-NEXT:    bne a0, s1, .LBB12_6
+; RV32I-NEXT:    bnez a0, .LBB12_6
 ; RV32I-NEXT:  # %bb.5: # %start
 ; RV32I-NEXT:    mv s3, s6
 ; RV32I-NEXT:  .LBB12_6: # %start
 ; RV32I-NEXT:    lui a1, 913408
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gesf2@plt
 ; RV32I-NEXT:    lui s6, 524288
 ; RV32I-NEXT:    lui s5, 524288
-; RV32I-NEXT:    blt a0, s1, .LBB12_8
+; RV32I-NEXT:    bltz a0, .LBB12_8
 ; RV32I-NEXT:  # %bb.7: # %start
 ; RV32I-NEXT:    mv s5, s2
 ; RV32I-NEXT:  .LBB12_8: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s4
 ; RV32I-NEXT:    call __gtsf2@plt
 ; RV32I-NEXT:    bge s1, a0, .LBB12_10
 ; RV32I-NEXT:  # %bb.9:
 ; RV32I-NEXT:    addi s5, s6, -1
 ; RV32I-NEXT:  .LBB12_10: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s1, .LBB12_12
+; RV32I-NEXT:    bnez a0, .LBB12_12
 ; RV32I-NEXT:  # %bb.11: # %start
 ; RV32I-NEXT:    mv s1, s5
 ; RV32I-NEXT:  .LBB12_12: # %start
 ; RV32I-NEXT:    mv a0, s3
 ; RV32I-NEXT:    mv a1, s1
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s5, 4(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s6, 0(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_l_s_sat:
 ; RV64I:       # %bb.0: # %start
 ; RV64I-NEXT:    addi sp, sp, -48
 ; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s4, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 913408
 ; RV64I-NEXT:    call __gesf2@plt
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    li s4, -1
-; RV64I-NEXT:    bltz s3, .LBB12_2
+; RV64I-NEXT:    li s3, -1
+; RV64I-NEXT:    bltz s2, .LBB12_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    j .LBB12_3
 ; RV64I-NEXT:  .LBB12_2:
-; RV64I-NEXT:    slli s2, s4, 63
+; RV64I-NEXT:    slli s1, s3, 63
 ; RV64I-NEXT:  .LBB12_3: # %start
 ; RV64I-NEXT:    lui a0, 389120
 ; RV64I-NEXT:    addiw a1, a0, -1
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB12_5
+; RV64I-NEXT:    blez a0, .LBB12_5
 ; RV64I-NEXT:  # %bb.4:
-; RV64I-NEXT:    srli s2, s4, 1
+; RV64I-NEXT:    srli s1, s3, 1
 ; RV64I-NEXT:  .LBB12_5: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB12_7
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB12_7
 ; RV64I-NEXT:  # %bb.6: # %start
-; RV64I-NEXT:    mv s1, s2
-; RV64I-NEXT:  .LBB12_7: # %start
 ; RV64I-NEXT:    mv a0, s1
+; RV64I-NEXT:  .LBB12_7: # %start
 ; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s4, 0(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 48
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i64 @llvm.fptosi.sat.i64.f32(float %a)
   ret i64 %0
 }
 declare i64 @llvm.fptosi.sat.i64.f32(float)
 

 ; RV64IF-NEXT:    fmax.s ft0, fa0, ft0
 ; RV64IF-NEXT:    fmin.s ft0, ft0, ft1
 ; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
 ; RV64IF-NEXT:  .LBB24_2: # %start
 ; RV64IF-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i16:
 ; RV32I:       # %bb.0: # %start
-; RV32I-NEXT:    addi sp, sp, -32
-; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    addi sp, sp, -16
+; RV32I-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s0, a0
 ; RV32I-NEXT:    lui a1, 815104
 ; RV32I-NEXT:    call __gesf2@plt
 ; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __fixsfsi@plt
-; RV32I-NEXT:    li s2, 0
-; RV32I-NEXT:    lui s3, 1048568
+; RV32I-NEXT:    lui s2, 1048568
 ; RV32I-NEXT:    bltz s1, .LBB24_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    mv s2, a0
 ; RV32I-NEXT:  .LBB24_2: # %start
 ; RV32I-NEXT:    lui a0, 290816
 ; RV32I-NEXT:    addi a1, a0, -512
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gtsf2@plt
-; RV32I-NEXT:    bge s2, a0, .LBB24_4
+; RV32I-NEXT:    blez a0, .LBB24_4
 ; RV32I-NEXT:  # %bb.3:
 ; RV32I-NEXT:    lui a0, 8
-; RV32I-NEXT:    addi s3, a0, -1
+; RV32I-NEXT:    addi s2, a0, -1
 ; RV32I-NEXT:  .LBB24_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s2, .LBB24_6
+; RV32I-NEXT:    li a1, 0
+; RV32I-NEXT:    bnez a0, .LBB24_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s2, s3
+; RV32I-NEXT:    mv a1, s2
 ; RV32I-NEXT:  .LBB24_6: # %start
-; RV32I-NEXT:    slli a0, s2, 16
+; RV32I-NEXT:    slli a0, a1, 16
 ; RV32I-NEXT:    srai a0, a0, 16
-; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    addi sp, sp, 32
+; RV32I-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_s_sat_i16:
 ; RV64I:       # %bb.0: # %start
-; RV64I-NEXT:    addi sp, sp, -48
-; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    addi sp, sp, -32
+; RV64I-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s2, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 815104
 ; RV64I-NEXT:    call __gesf2@plt
 ; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s2, 0
-; RV64I-NEXT:    lui s3, 1048568
+; RV64I-NEXT:    lui s2, 1048568
 ; RV64I-NEXT:    bltz s1, .LBB24_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB24_2: # %start
 ; RV64I-NEXT:    lui a0, 290816
 ; RV64I-NEXT:    addiw a1, a0, -512
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
-; RV64I-NEXT:    bge s2, a0, .LBB24_4
+; RV64I-NEXT:    blez a0, .LBB24_4
 ; RV64I-NEXT:  # %bb.3:
 ; RV64I-NEXT:    lui a0, 8
-; RV64I-NEXT:    addiw s3, a0, -1
+; RV64I-NEXT:    addiw s2, a0, -1
 ; RV64I-NEXT:  .LBB24_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s2, .LBB24_6
+; RV64I-NEXT:    li a1, 0
+; RV64I-NEXT:    bnez a0, .LBB24_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s2, s3
+; RV64I-NEXT:    mv a1, s2
 ; RV64I-NEXT:  .LBB24_6: # %start
-; RV64I-NEXT:    slli a0, s2, 48
+; RV64I-NEXT:    slli a0, a1, 48
 ; RV64I-NEXT:    srai a0, a0, 48
-; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    addi sp, sp, 48
+; RV64I-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 32
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i16 @llvm.fptosi.sat.i16.f32(float %a)
   ret i16 %0
 }
 declare i16 @llvm.fptosi.sat.i16.f32(float)
 
 define zeroext i16 @fcvt_wu_s_i16(float %a) nounwind {

 ; RV64IF-NEXT:    fmax.s ft0, fa0, ft0
 ; RV64IF-NEXT:    fmin.s ft0, ft0, ft1
 ; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
 ; RV64IF-NEXT:  .LBB28_2: # %start
 ; RV64IF-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i8:
 ; RV32I:       # %bb.0: # %start
-; RV32I-NEXT:    addi sp, sp, -32
-; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    addi sp, sp, -16
+; RV32I-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    mv s0, a0
 ; RV32I-NEXT:    lui a1, 798720
 ; RV32I-NEXT:    call __gesf2@plt
 ; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __fixsfsi@plt
-; RV32I-NEXT:    li s2, 0
-; RV32I-NEXT:    li s3, -128
+; RV32I-NEXT:    li s2, -128
 ; RV32I-NEXT:    bltz s1, .LBB28_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    mv s2, a0
 ; RV32I-NEXT:  .LBB28_2: # %start
 ; RV32I-NEXT:    lui a1, 274400
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gtsf2@plt
 ; RV32I-NEXT:    li s1, 127
-; RV32I-NEXT:    blt s2, a0, .LBB28_4
+; RV32I-NEXT:    bgtz a0, .LBB28_4
 ; RV32I-NEXT:  # %bb.3: # %start
-; RV32I-NEXT:    mv s1, s3
+; RV32I-NEXT:    mv s1, s2
 ; RV32I-NEXT:  .LBB28_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s2, .LBB28_6
+; RV32I-NEXT:    li a1, 0
+; RV32I-NEXT:    bnez a0, .LBB28_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s2, s1
+; RV32I-NEXT:    mv a1, s1
 ; RV32I-NEXT:  .LBB28_6: # %start
-; RV32I-NEXT:    slli a0, s2, 24
+; RV32I-NEXT:    slli a0, a1, 24
 ; RV32I-NEXT:    srai a0, a0, 24
-; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    addi sp, sp, 32
+; RV32I-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_s_sat_i8:
 ; RV64I:       # %bb.0: # %start
-; RV64I-NEXT:    addi sp, sp, -48
-; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    addi sp, sp, -32
+; RV64I-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s2, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 798720
 ; RV64I-NEXT:    call __gesf2@plt
 ; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s2, 0
-; RV64I-NEXT:    li s3, -128
+; RV64I-NEXT:    li s2, -128
 ; RV64I-NEXT:    bltz s1, .LBB28_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB28_2: # %start
 ; RV64I-NEXT:    lui a1, 274400
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
 ; RV64I-NEXT:    li s1, 127
-; RV64I-NEXT:    blt s2, a0, .LBB28_4
+; RV64I-NEXT:    bgtz a0, .LBB28_4
 ; RV64I-NEXT:  # %bb.3: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv s1, s2
 ; RV64I-NEXT:  .LBB28_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s2, .LBB28_6
+; RV64I-NEXT:    li a1, 0
+; RV64I-NEXT:    bnez a0, .LBB28_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s2, s1
+; RV64I-NEXT:    mv a1, s1
 ; RV64I-NEXT:  .LBB28_6: # %start
-; RV64I-NEXT:    slli a0, s2, 56
+; RV64I-NEXT:    slli a0, a1, 56
 ; RV64I-NEXT:    srai a0, a0, 56
-; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    addi sp, sp, 48
+; RV64I-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 32
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i8 @llvm.fptosi.sat.i8.f32(float %a)
   ret i8 %0
 }
 declare i8 @llvm.fptosi.sat.i8.f32(float)
 
 define zeroext i8 @fcvt_wu_s_i8(float %a) nounwind {

llvm/test/CodeGen/RISCV/fpclamptosat.ll

 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    addi sp, sp, -16
 ; RV32-NEXT:    .cfi_def_cfa_offset 16
 ; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
 ; RV32-NEXT:    .cfi_offset ra, -4
 ; RV32-NEXT:    fmv.x.w a0, fa0
 ; RV32-NEXT:    call __extendhfsf2@plt
 ; RV32-NEXT:    call __fixsfdi@plt
-; RV32-NEXT:    mv a2, a0
-; RV32-NEXT:    bgez a1, .LBB35_7
+; RV32-NEXT:    mv a2, a1
+; RV32-NEXT:    mv a3, a0
+; RV32-NEXT:    bgez a1, .LBB35_6
 ; RV32-NEXT:  # %bb.1: # %entry
-; RV32-NEXT:    bnez a1, .LBB35_8
+; RV32-NEXT:    bnez a1, .LBB35_7
 ; RV32-NEXT:  .LBB35_2: # %entry
-; RV32-NEXT:    bgez a1, .LBB35_9
+; RV32-NEXT:    mv a1, a0
+; RV32-NEXT:    blez a2, .LBB35_8
 ; RV32-NEXT:  .LBB35_3: # %entry
-; RV32-NEXT:    mv a2, a0
-; RV32-NEXT:    blez a1, .LBB35_10
+; RV32-NEXT:    beqz a2, .LBB35_5
 ; RV32-NEXT:  .LBB35_4: # %entry
-; RV32-NEXT:    beqz a1, .LBB35_6
+; RV32-NEXT:    mv a0, a1
 ; RV32-NEXT:  .LBB35_5: # %entry
-; RV32-NEXT:    mv a0, a2
-; RV32-NEXT:  .LBB35_6: # %entry
 ; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 16
 ; RV32-NEXT:    ret
-; RV32-NEXT:  .LBB35_7: # %entry
-; RV32-NEXT:    li a2, -1
+; RV32-NEXT:  .LBB35_6: # %entry
+; RV32-NEXT:    li a2, 0
+; RV32-NEXT:    li a3, -1
 ; RV32-NEXT:    beqz a1, .LBB35_2
+; RV32-NEXT:  .LBB35_7: # %entry
+; RV32-NEXT:    mv a0, a3
+; RV32-NEXT:    mv a1, a0
+; RV32-NEXT:    bgtz a2, .LBB35_3
 ; RV32-NEXT:  .LBB35_8: # %entry
-; RV32-NEXT:    mv a0, a2
-; RV32-NEXT:    bltz a1, .LBB35_3
-; RV32-NEXT:  .LBB35_9: # %entry
 ; RV32-NEXT:    li a1, 0
-; RV32-NEXT:    mv a2, a0
-; RV32-NEXT:    bgtz a1, .LBB35_4
-; RV32-NEXT:  .LBB35_10: # %entry
-; RV32-NEXT:    li a2, 0
-; RV32-NEXT:    bnez a1, .LBB35_5
-; RV32-NEXT:    j .LBB35_6
+; RV32-NEXT:    bnez a2, .LBB35_4
+; RV32-NEXT:    j .LBB35_5
 ;
 ; RV64-LABEL: ustest_f16i32_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16
 ; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    .cfi_offset ra, -8
 ; RV64-NEXT:    fmv.x.w a0, fa0

 ; RV32IF-NEXT:    addi sp, sp, -32
 ; RV32IF-NEXT:    .cfi_def_cfa_offset 32
 ; RV32IF-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32IF-NEXT:    .cfi_offset ra, -4
 ; RV32IF-NEXT:    mv a2, a1
 ; RV32IF-NEXT:    mv a1, a0
 ; RV32IF-NEXT:    addi a0, sp, 8
 ; RV32IF-NEXT:    call __fixdfti@plt
-; RV32IF-NEXT:    lw a5, 8(sp)
-; RV32IF-NEXT:    lw a3, 20(sp)
-; RV32IF-NEXT:    lw a1, 12(sp)
+; RV32IF-NEXT:    lw a7, 8(sp)
+; RV32IF-NEXT:    lw a5, 12(sp)
+; RV32IF-NEXT:    lui a4, 524288
+; RV32IF-NEXT:    addi a0, a4, -1
 ; RV32IF-NEXT:    li a2, -1
-; RV32IF-NEXT:    mv a7, a5
-; RV32IF-NEXT:    bltz a3, .LBB45_2
+; RV32IF-NEXT:    mv a1, a7
+; RV32IF-NEXT:    bltu a5, a0, .LBB45_2
 ; RV32IF-NEXT:  # %bb.1: # %entry
-; RV32IF-NEXT:    li a7, -1
+; RV32IF-NEXT:    li a1, -1
 ; RV32IF-NEXT:  .LBB45_2: # %entry
-; RV32IF-NEXT:    lui a4, 524288
-; RV32IF-NEXT:    addi a6, a4, -1
-; RV32IF-NEXT:    mv t0, a5
-; RV32IF-NEXT:    bgeu a1, a6, .LBB45_19
+; RV32IF-NEXT:    lw a6, 20(sp)
+; RV32IF-NEXT:    mv a3, a7
+; RV32IF-NEXT:    bne a5, a0, .LBB45_19
 ; RV32IF-NEXT:  # %bb.3: # %entry
-; RV32IF-NEXT:    lw a0, 16(sp)
-; RV32IF-NEXT:    bne a1, a6, .LBB45_20
+; RV32IF-NEXT:    lw a1, 16(sp)
+; RV32IF-NEXT:    bgez a6, .LBB45_20
 ; RV32IF-NEXT:  .LBB45_4: # %entry
-; RV32IF-NEXT:    or t0, a0, a3
+; RV32IF-NEXT:    or t0, a1, a6
 ; RV32IF-NEXT:    bnez t0, .LBB45_21
 ; RV32IF-NEXT:  .LBB45_5: # %entry
-; RV32IF-NEXT:    mv a7, a1
-; RV32IF-NEXT:    bgez a3, .LBB45_22
+; RV32IF-NEXT:    mv a7, a5
+; RV32IF-NEXT:    bgez a6, .LBB45_22
 ; RV32IF-NEXT:  .LBB45_6: # %entry
-; RV32IF-NEXT:    bgeu a1, a6, .LBB45_23
+; RV32IF-NEXT:    bgeu a5, a0, .LBB45_23
 ; RV32IF-NEXT:  .LBB45_7: # %entry
 ; RV32IF-NEXT:    bnez t0, .LBB45_24
 ; RV32IF-NEXT:  .LBB45_8: # %entry
-; RV32IF-NEXT:    li a6, 0
-; RV32IF-NEXT:    bnez a3, .LBB45_25
+; RV32IF-NEXT:    li a0, 0
+; RV32IF-NEXT:    bnez a6, .LBB45_25
 ; RV32IF-NEXT:  .LBB45_9: # %entry
-; RV32IF-NEXT:    bgez a3, .LBB45_26
+; RV32IF-NEXT:    bgez a6, .LBB45_26
 ; RV32IF-NEXT:  .LBB45_10: # %entry
-; RV32IF-NEXT:    mv a7, a5
-; RV32IF-NEXT:    bgeu a4, a1, .LBB45_27
+; RV32IF-NEXT:    mv t0, a5
+; RV32IF-NEXT:    bltz a6, .LBB45_27
 ; RV32IF-NEXT:  .LBB45_11: # %entry
-; RV32IF-NEXT:    mv a0, a5
-; RV32IF-NEXT:    bne a1, a4, .LBB45_28
+; RV32IF-NEXT:    mv a1, a5
+; RV32IF-NEXT:    bgeu a4, a5, .LBB45_28
 ; RV32IF-NEXT:  .LBB45_12: # %entry
-; RV32IF-NEXT:    bltz a3, .LBB45_29
+; RV32IF-NEXT:    and a7, a0, a6
+; RV32IF-NEXT:    bne a7, a2, .LBB45_29
 ; RV32IF-NEXT:  .LBB45_13: # %entry
-; RV32IF-NEXT:    and a6, a6, a3
-; RV32IF-NEXT:    bne a6, a2, .LBB45_30
+; RV32IF-NEXT:    mv t0, a3
+; RV32IF-NEXT:    bgeu a4, a5, .LBB45_30
 ; RV32IF-NEXT:  .LBB45_14: # %entry
-; RV32IF-NEXT:    mv a5, a1
-; RV32IF-NEXT:    bltz a3, .LBB45_31
+; RV32IF-NEXT:    mv a0, a3
+; RV32IF-NEXT:    bne a5, a4, .LBB45_31
 ; RV32IF-NEXT:  .LBB45_15: # %entry
-; RV32IF-NEXT:    bgeu a4, a1, .LBB45_32
+; RV32IF-NEXT:    bltz a6, .LBB45_32
 ; RV32IF-NEXT:  .LBB45_16: # %entry
-; RV32IF-NEXT:    beq a6, a2, .LBB45_18
+; RV32IF-NEXT:    beq a7, a2, .LBB45_18
 ; RV32IF-NEXT:  .LBB45_17: # %entry
-; RV32IF-NEXT:    mv a1, a5
+; RV32IF-NEXT:    mv a0, a3
 ; RV32IF-NEXT:  .LBB45_18: # %entry
 ; RV32IF-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IF-NEXT:    addi sp, sp, 32
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB45_19: # %entry
-; RV32IF-NEXT:    li t0, -1
-; RV32IF-NEXT:    lw a0, 16(sp)
-; RV32IF-NEXT:    beq a1, a6, .LBB45_4
+; RV32IF-NEXT:    mv a3, a1
+; RV32IF-NEXT:    lw a1, 16(sp)
+; RV32IF-NEXT:    bltz a6, .LBB45_4
 ; RV32IF-NEXT:  .LBB45_20: # %entry
-; RV32IF-NEXT:    mv a5, t0
-; RV32IF-NEXT:    or t0, a0, a3
+; RV32IF-NEXT:    li a7, -1
+; RV32IF-NEXT:    or t0, a1, a6
 ; RV32IF-NEXT:    beqz t0, .LBB45_5
 ; RV32IF-NEXT:  .LBB45_21: # %entry
-; RV32IF-NEXT:    mv a5, a7
-; RV32IF-NEXT:    mv a7, a1
-; RV32IF-NEXT:    bltz a3, .LBB45_6
+; RV32IF-NEXT:    mv a3, a7
+; RV32IF-NEXT:    mv a7, a5
+; RV32IF-NEXT:    bltz a6, .LBB45_6
 ; RV32IF-NEXT:  .LBB45_22: # %entry
-; RV32IF-NEXT:    mv a7, a6
-; RV32IF-NEXT:    bltu a1, a6, .LBB45_7
+; RV32IF-NEXT:    mv a7, a0
+; RV32IF-NEXT:    bltu a5, a0, .LBB45_7
 ; RV32IF-NEXT:  .LBB45_23: # %entry
-; RV32IF-NEXT:    mv a1, a6
+; RV32IF-NEXT:    mv a5, a0
 ; RV32IF-NEXT:    beqz t0, .LBB45_8
 ; RV32IF-NEXT:  .LBB45_24: # %entry
-; RV32IF-NEXT:    mv a1, a7
-; RV32IF-NEXT:    li a6, 0
-; RV32IF-NEXT:    beqz a3, .LBB45_9
+; RV32IF-NEXT:    mv a5, a7
+; RV32IF-NEXT:    li a0, 0
+; RV32IF-NEXT:    beqz a6, .LBB45_9
 ; RV32IF-NEXT:  .LBB45_25: # %entry
-; RV32IF-NEXT:    srai a6, a3, 31
-; RV32IF-NEXT:    and a6, a6, a0
-; RV32IF-NEXT:    bltz a3, .LBB45_10
+; RV32IF-NEXT:    srai a0, a6, 31
+; RV32IF-NEXT:    and a0, a0, a1
+; RV32IF-NEXT:    bltz a6, .LBB45_10
 ; RV32IF-NEXT:  .LBB45_26: # %entry
-; RV32IF-NEXT:    li a3, 0
-; RV32IF-NEXT:    mv a7, a5
-; RV32IF-NEXT:    bltu a4, a1, .LBB45_11
+; RV32IF-NEXT:    li a6, 0
+; RV32IF-NEXT:    mv t0, a5
+; RV32IF-NEXT:    bgez a6, .LBB45_11
 ; RV32IF-NEXT:  .LBB45_27: # %entry
-; RV32IF-NEXT:    li a7, 0
-; RV32IF-NEXT:    mv a0, a5
-; RV32IF-NEXT:    beq a1, a4, .LBB45_12
+; RV32IF-NEXT:    lui t0, 524288
+; RV32IF-NEXT:    mv a1, a5
+; RV32IF-NEXT:    bltu a4, a5, .LBB45_12
 ; RV32IF-NEXT:  .LBB45_28: # %entry
-; RV32IF-NEXT:    mv a0, a7
-; RV32IF-NEXT:    bgez a3, .LBB45_13
+; RV32IF-NEXT:    lui a1, 524288
+; RV32IF-NEXT:    and a7, a0, a6
+; RV32IF-NEXT:    beq a7, a2, .LBB45_13
 ; RV32IF-NEXT:  .LBB45_29: # %entry
-; RV32IF-NEXT:    li a5, 0
-; RV32IF-NEXT:    and a6, a6, a3
-; RV32IF-NEXT:    beq a6, a2, .LBB45_14
+; RV32IF-NEXT:    mv a1, t0
+; RV32IF-NEXT:    mv t0, a3
+; RV32IF-NEXT:    bltu a4, a5, .LBB45_14
 ; RV32IF-NEXT:  .LBB45_30: # %entry
-; RV32IF-NEXT:    mv a0, a5
-; RV32IF-NEXT:    mv a5, a1
-; RV32IF-NEXT:    bgez a3, .LBB45_15
+; RV32IF-NEXT:    li t0, 0
+; RV32IF-NEXT:    mv a0, a3
+; RV32IF-NEXT:    beq a5, a4, .LBB45_15
 ; RV32IF-NEXT:  .LBB45_31: # %entry
-; RV32IF-NEXT:    lui a5, 524288
-; RV32IF-NEXT:    bltu a4, a1, .LBB45_16
+; RV32IF-NEXT:    mv a0, t0
+; RV32IF-NEXT:    bgez a6, .LBB45_16
 ; RV32IF-NEXT:  .LBB45_32: # %entry
-; RV32IF-NEXT:    lui a1, 524288
-; RV32IF-NEXT:    bne a6, a2, .LBB45_17
+; RV32IF-NEXT:    li a3, 0
+; RV32IF-NEXT:    bne a7, a2, .LBB45_17
 ; RV32IF-NEXT:    j .LBB45_18
 ;
 ; RV64IF-LABEL: stest_f64i64_mm:
 ; RV64IF:       # %bb.0: # %entry
 ; RV64IF-NEXT:    addi sp, sp, -16
 ; RV64IF-NEXT:    .cfi_def_cfa_offset 16
 ; RV64IF-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
 ; RV64IF-NEXT:    .cfi_offset ra, -8

 ; RV32IFD-LABEL: stest_f64i64_mm:
 ; RV32IFD:       # %bb.0: # %entry
 ; RV32IFD-NEXT:    addi sp, sp, -32
 ; RV32IFD-NEXT:    .cfi_def_cfa_offset 32
 ; RV32IFD-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32IFD-NEXT:    .cfi_offset ra, -4
 ; RV32IFD-NEXT:    addi a0, sp, 8
 ; RV32IFD-NEXT:    call __fixdfti@plt
-; RV32IFD-NEXT:    lw a5, 8(sp)
-; RV32IFD-NEXT:    lw a3, 20(sp)
-; RV32IFD-NEXT:    lw a1, 12(sp)
+; RV32IFD-NEXT:    lw a7, 8(sp)
+; RV32IFD-NEXT:    lw a5, 12(sp)
+; RV32IFD-NEXT:    lui a4, 524288
+; RV32IFD-NEXT:    addi a0, a4, -1
 ; RV32IFD-NEXT:    li a2, -1
-; RV32IFD-NEXT:    mv a7, a5
-; RV32IFD-NEXT:    bltz a3, .LBB45_2
+; RV32IFD-NEXT:    mv a1, a7
+; RV32IFD-NEXT:    bltu a5, a0, .LBB45_2
 ; RV32IFD-NEXT:  # %bb.1: # %entry
-; RV32IFD-NEXT:    li a7, -1
+; RV32IFD-NEXT:    li a1, -1
 ; RV32IFD-NEXT:  .LBB45_2: # %entry
-; RV32IFD-NEXT:    lui a4, 524288
-; RV32IFD-NEXT:    addi a6, a4, -1
-; RV32IFD-NEXT:    mv t0, a5
-; RV32IFD-NEXT:    bgeu a1, a6, .LBB45_19
+; RV32IFD-NEXT:    lw a6, 20(sp)
+; RV32IFD-NEXT:    mv a3, a7
+; RV32IFD-NEXT:    bne a5, a0, .LBB45_19
 ; RV32IFD-NEXT:  # %bb.3: # %entry
-; RV32IFD-NEXT:    lw a0, 16(sp)
-; RV32IFD-NEXT:    bne a1, a6, .LBB45_20
+; RV32IFD-NEXT:    lw a1, 16(sp)
+; RV32IFD-NEXT:    bgez a6, .LBB45_20
 ; RV32IFD-NEXT:  .LBB45_4: # %entry
-; RV32IFD-NEXT:    or t0, a0, a3
+; RV32IFD-NEXT:    or t0, a1, a6
 ; RV32IFD-NEXT:    bnez t0, .LBB45_21
 ; RV32IFD-NEXT:  .LBB45_5: # %entry
-; RV32IFD-NEXT:    mv a7, a1
-; RV32IFD-NEXT:    bgez a3, .LBB45_22
+; RV32IFD-NEXT:    mv a7, a5
+; RV32IFD-NEXT:    bgez a6, .LBB45_22
 ; RV32IFD-NEXT:  .LBB45_6: # %entry
-; RV32IFD-NEXT:    bgeu a1, a6, .LBB45_23
+; RV32IFD-NEXT:    bgeu a5, a0, .LBB45_23
 ; RV32IFD-NEXT:  .LBB45_7: # %entry
 ; RV32IFD-NEXT:    bnez t0, .LBB45_24
 ; RV32IFD-NEXT:  .LBB45_8: # %entry
-; RV32IFD-NEXT:    li a6, 0
-; RV32IFD-NEXT:    bnez a3, .LBB45_25
+; RV32IFD-NEXT:    li a0, 0
+; RV32IFD-NEXT:    bnez a6, .LBB45_25
 ; RV32IFD-NEXT:  .LBB45_9: # %entry
-; RV32IFD-NEXT:    bgez a3, .LBB45_26
+; RV32IFD-NEXT:    bgez a6, .LBB45_26
 ; RV32IFD-NEXT:  .LBB45_10: # %entry
-; RV32IFD-NEXT:    mv a7, a5
-; RV32IFD-NEXT:    bgeu a4, a1, .LBB45_27
+; RV32IFD-NEXT:    mv t0, a5
+; RV32IFD-NEXT:    bltz a6, .LBB45_27
 ; RV32IFD-NEXT:  .LBB45_11: # %entry
-; RV32IFD-NEXT:    mv a0, a5
-; RV32IFD-NEXT:    bne a1, a4, .LBB45_28
+; RV32IFD-NEXT:    mv a1, a5
+; RV32IFD-NEXT:    bgeu a4, a5, .LBB45_28
 ; RV32IFD-NEXT:  .LBB45_12: # %entry
-; RV32IFD-NEXT:    bltz a3, .LBB45_29
+; RV32IFD-NEXT:    and a7, a0, a6
+; RV32IFD-NEXT:    bne a7, a2, .LBB45_29
 ; RV32IFD-NEXT:  .LBB45_13: # %entry
-; RV32IFD-NEXT:    and a6, a6, a3
-; RV32IFD-NEXT:    bne a6, a2, .LBB45_30
+; RV32IFD-NEXT:    mv t0, a3
+; RV32IFD-NEXT:    bgeu a4, a5, .LBB45_30
 ; RV32IFD-NEXT:  .LBB45_14: # %entry
-; RV32IFD-NEXT:    mv a5, a1
-; RV32IFD-NEXT:    bltz a3, .LBB45_31
+; RV32IFD-NEXT:    mv a0, a3
+; RV32IFD-NEXT:    bne a5, a4, .LBB45_31
 ; RV32IFD-NEXT:  .LBB45_15: # %entry
-; RV32IFD-NEXT:    bgeu a4, a1, .LBB45_32
+; RV32IFD-NEXT:    bltz a6, .LBB45_32
 ; RV32IFD-NEXT:  .LBB45_16: # %entry
-; RV32IFD-NEXT:    beq a6, a2, .LBB45_18
+; RV32IFD-NEXT:    beq a7, a2, .LBB45_18
 ; RV32IFD-NEXT:  .LBB45_17: # %entry
-; RV32IFD-NEXT:    mv a1, a5
+; RV32IFD-NEXT:    mv a0, a3
 ; RV32IFD-NEXT:  .LBB45_18: # %entry
 ; RV32IFD-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IFD-NEXT:    addi sp, sp, 32
 ; RV32IFD-NEXT:    ret
 ; RV32IFD-NEXT:  .LBB45_19: # %entry
-; RV32IFD-NEXT:    li t0, -1
-; RV32IFD-NEXT:    lw a0, 16(sp)
-; RV32IFD-NEXT:    beq a1, a6, .LBB45_4
+; RV32IFD-NEXT:    mv a3, a1
+; RV32IFD-NEXT:    lw a1, 16(sp)
+; RV32IFD-NEXT:    bltz a6, .LBB45_4
 ; RV32IFD-NEXT:  .LBB45_20: # %entry
-; RV32IFD-NEXT:    mv a5, t0
-; RV32IFD-NEXT:    or t0, a0, a3
+; RV32IFD-NEXT:    li a7, -1
+; RV32IFD-NEXT:    or t0, a1, a6
 ; RV32IFD-NEXT:    beqz t0, .LBB45_5
 ; RV32IFD-NEXT:  .LBB45_21: # %entry
-; RV32IFD-NEXT:    mv a5, a7
-; RV32IFD-NEXT:    mv a7, a1
-; RV32IFD-NEXT:    bltz a3, .LBB45_6
+; RV32IFD-NEXT:    mv a3, a7
+; RV32IFD-NEXT:    mv a7, a5
+; RV32IFD-NEXT:    bltz a6, .LBB45_6
 ; RV32IFD-NEXT:  .LBB45_22: # %entry
-; RV32IFD-NEXT:    mv a7, a6
-; RV32IFD-NEXT:    bltu a1, a6, .LBB45_7
+; RV32IFD-NEXT:    mv a7, a0
+; RV32IFD-NEXT:    bltu a5, a0, .LBB45_7
 ; RV32IFD-NEXT:  .LBB45_23: # %entry
-; RV32IFD-NEXT:    mv a1, a6
+; RV32IFD-NEXT:    mv a5, a0
 ; RV32IFD-NEXT:    beqz t0, .LBB45_8
 ; RV32IFD-NEXT:  .LBB45_24: # %entry
-; RV32IFD-NEXT:    mv a1, a7
-; RV32IFD-NEXT:    li a6, 0
-; RV32IFD-NEXT:    beqz a3, .LBB45_9
+; RV32IFD-NEXT:    mv a5, a7
+; RV32IFD-NEXT:    li a0, 0
+; RV32IFD-NEXT:    beqz a6, .LBB45_9
 ; RV32IFD-NEXT:  .LBB45_25: # %entry
-; RV32IFD-NEXT:    srai a6, a3, 31
-; RV32IFD-NEXT:    and a6, a6, a0
-; RV32IFD-NEXT:    bltz a3, .LBB45_10
+; RV32IFD-NEXT:    srai a0, a6, 31
+; RV32IFD-NEXT:    and a0, a0, a1
+; RV32IFD-NEXT:    bltz a6, .LBB45_10
 ; RV32IFD-NEXT:  .LBB45_26: # %entry
-; RV32IFD-NEXT:    li a3, 0
-; RV32IFD-NEXT:    mv a7, a5
-; RV32IFD-NEXT:    bltu a4, a1, .LBB45_11
+; RV32IFD-NEXT:    li a6, 0
+; RV32IFD-NEXT:    mv t0, a5
+; RV32IFD-NEXT:    bgez a6, .LBB45_11
 ; RV32IFD-NEXT:  .LBB45_27: # %entry
-; RV32IFD-NEXT:    li a7, 0
-; RV32IFD-NEXT:    mv a0, a5
-; RV32IFD-NEXT:    beq a1, a4, .LBB45_12
+; RV32IFD-NEXT:    lui t0, 524288
+; RV32IFD-NEXT:    mv a1, a5
+; RV32IFD-NEXT:    bltu a4, a5, .LBB45_12
 ; RV32IFD-NEXT:  .LBB45_28: # %entry
-; RV32IFD-NEXT:    mv a0, a7
-; RV32IFD-NEXT:    bgez a3, .LBB45_13
+; RV32IFD-NEXT:    lui a1, 524288
+; RV32IFD-NEXT:    and a7, a0, a6
+; RV32IFD-NEXT:    beq a7, a2, .LBB45_13
 ; RV32IFD-NEXT:  .LBB45_29: # %entry
-; RV32IFD-NEXT:    li a5, 0
-; RV32IFD-NEXT:    and a6, a6, a3
-; RV32IFD-NEXT:    beq a6, a2, .LBB45_14
+; RV32IFD-NEXT:    mv a1, t0
+; RV32IFD-NEXT:    mv t0, a3
+; RV32IFD-NEXT:    bltu a4, a5, .LBB45_14
 ; RV32IFD-NEXT:  .LBB45_30: # %entry
-; RV32IFD-NEXT:    mv a0, a5
-; RV32IFD-NEXT:    mv a5, a1
-; RV32IFD-NEXT:    bgez a3, .LBB45_15
+; RV32IFD-NEXT:    li t0, 0
+; RV32IFD-NEXT:    mv a0, a3
+; RV32IFD-NEXT:    beq a5, a4, .LBB45_15
 ; RV32IFD-NEXT:  .LBB45_31: # %entry
-; RV32IFD-NEXT:    lui a5, 524288
-; RV32IFD-NEXT:    bltu a4, a1, .LBB45_16
+; RV32IFD-NEXT:    mv a0, t0
+; RV32IFD-NEXT:    bgez a6, .LBB45_16
 ; RV32IFD-NEXT:  .LBB45_32: # %entry
-; RV32IFD-NEXT:    lui a1, 524288
-; RV32IFD-NEXT:    bne a6, a2, .LBB45_17
+; RV32IFD-NEXT:    li a3, 0
+; RV32IFD-NEXT:    bne a7, a2, .LBB45_17
 ; RV32IFD-NEXT:    j .LBB45_18
 ;
 ; RV64IFD-LABEL: stest_f64i64_mm:
 ; RV64IFD:       # %bb.0: # %entry
 ; RV64IFD-NEXT:    feq.d a0, fa0, fa0
 ; RV64IFD-NEXT:    beqz a0, .LBB45_2
 ; RV64IFD-NEXT:  # %bb.1:
 ; RV64IFD-NEXT:    fcvt.l.d a0, fa0, rtz

 ; RV32IF-NEXT:    addi a0, sp, 8
 ; RV32IF-NEXT:    call __fixunsdfti@plt
 ; RV32IF-NEXT:    lw a0, 20(sp)
 ; RV32IF-NEXT:    lw a3, 16(sp)
 ; RV32IF-NEXT:    li a1, 0
 ; RV32IF-NEXT:    beqz a0, .LBB46_3
 ; RV32IF-NEXT:  # %bb.1: # %entry
 ; RV32IF-NEXT:    mv a2, a1
-; RV32IF-NEXT:    beq a2, a1, .LBB46_4
+; RV32IF-NEXT:    beqz a2, .LBB46_4
 ; RV32IF-NEXT:  .LBB46_2:
 ; RV32IF-NEXT:    lw a4, 8(sp)
 ; RV32IF-NEXT:    j .LBB46_5
 ; RV32IF-NEXT:  .LBB46_3:
 ; RV32IF-NEXT:    seqz a2, a3
-; RV32IF-NEXT:    bne a2, a1, .LBB46_2
+; RV32IF-NEXT:    bnez a2, .LBB46_2
 ; RV32IF-NEXT:  .LBB46_4: # %entry
 ; RV32IF-NEXT:    mv a4, a1
 ; RV32IF-NEXT:  .LBB46_5: # %entry
 ; RV32IF-NEXT:    xori a3, a3, 1
 ; RV32IF-NEXT:    or a3, a3, a0
 ; RV32IF-NEXT:    mv a0, a1
-; RV32IF-NEXT:    beq a3, a1, .LBB46_7
+; RV32IF-NEXT:    beqz a3, .LBB46_7
 ; RV32IF-NEXT:  # %bb.6: # %entry
 ; RV32IF-NEXT:    mv a0, a4
 ; RV32IF-NEXT:  .LBB46_7: # %entry
-; RV32IF-NEXT:    bne a2, a1, .LBB46_9
+; RV32IF-NEXT:    bnez a2, .LBB46_9
 ; RV32IF-NEXT:  # %bb.8: # %entry
 ; RV32IF-NEXT:    mv a2, a1
-; RV32IF-NEXT:    bne a3, a1, .LBB46_10
+; RV32IF-NEXT:    bnez a3, .LBB46_10
 ; RV32IF-NEXT:    j .LBB46_11
 ; RV32IF-NEXT:  .LBB46_9:
 ; RV32IF-NEXT:    lw a2, 12(sp)
-; RV32IF-NEXT:    beq a3, a1, .LBB46_11
+; RV32IF-NEXT:    beqz a3, .LBB46_11
 ; RV32IF-NEXT:  .LBB46_10: # %entry
 ; RV32IF-NEXT:    mv a1, a2
 ; RV32IF-NEXT:  .LBB46_11: # %entry
 ; RV32IF-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IF-NEXT:    addi sp, sp, 32
 ; RV32IF-NEXT:    ret
 ;
 ; RV64-LABEL: utest_f64i64_mm:

 ; RV32IFD-NEXT:    addi sp, sp, -32
 ; RV32IFD-NEXT:    .cfi_def_cfa_offset 32
 ; RV32IFD-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32IFD-NEXT:    .cfi_offset ra, -4
 ; RV32IFD-NEXT:    addi a0, sp, 8
 ; RV32IFD-NEXT:    call __fixunsdfti@plt
 ; RV32IFD-NEXT:    lw a0, 20(sp)
 ; RV32IFD-NEXT:    lw a3, 16(sp)
 ; RV32IFD-NEXT:    li a1, 0
 ; RV32IFD-NEXT:    beqz a0, .LBB46_3
 ; RV32IFD-NEXT:  # %bb.1: # %entry
 ; RV32IFD-NEXT:    mv a2, a1
-; RV32IFD-NEXT:    beq a2, a1, .LBB46_4
+; RV32IFD-NEXT:    beqz a2, .LBB46_4
 ; RV32IFD-NEXT:  .LBB46_2:
 ; RV32IFD-NEXT:    lw a4, 8(sp)
 ; RV32IFD-NEXT:    j .LBB46_5
 ; RV32IFD-NEXT:  .LBB46_3:
 ; RV32IFD-NEXT:    seqz a2, a3
-; RV32IFD-NEXT:    bne a2, a1, .LBB46_2
+; RV32IFD-NEXT:    bnez a2, .LBB46_2
 ; RV32IFD-NEXT:  .LBB46_4: # %entry
 ; RV32IFD-NEXT:    mv a4, a1
 ; RV32IFD-NEXT:  .LBB46_5: # %entry
 ; RV32IFD-NEXT:    xori a3, a3, 1
 ; RV32IFD-NEXT:    or a3, a3, a0
 ; RV32IFD-NEXT:    mv a0, a1
-; RV32IFD-NEXT:    beq a3, a1, .LBB46_7
+; RV32IFD-NEXT:    beqz a3, .LBB46_7
 ; RV32IFD-NEXT:  # %bb.6: # %entry
 ; RV32IFD-NEXT:    mv a0, a4
 ; RV32IFD-NEXT:  .LBB46_7: # %entry
-; RV32IFD-NEXT:    bne a2, a1, .LBB46_9
+; RV32IFD-NEXT:    bnez a2, .LBB46_9
 ; RV32IFD-NEXT:  # %bb.8: # %entry
 ; RV32IFD-NEXT:    mv a2, a1
-; RV32IFD-NEXT:    bne a3, a1, .LBB46_10
+; RV32IFD-NEXT:    bnez a3, .LBB46_10
 ; RV32IFD-NEXT:    j .LBB46_11
 ; RV32IFD-NEXT:  .LBB46_9:
 ; RV32IFD-NEXT:    lw a2, 12(sp)
-; RV32IFD-NEXT:    beq a3, a1, .LBB46_11
+; RV32IFD-NEXT:    beqz a3, .LBB46_11
 ; RV32IFD-NEXT:  .LBB46_10: # %entry
 ; RV32IFD-NEXT:    mv a1, a2
 ; RV32IFD-NEXT:  .LBB46_11: # %entry
 ; RV32IFD-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IFD-NEXT:    addi sp, sp, 32
 ; RV32IFD-NEXT:    ret
 entry:
   %conv = fptoui double %x to i128
   %spec.store.select = call i128 @llvm.umin.i128(i128 %conv, i128 18446744073709551616)
   %conv6 = trunc i128 %spec.store.select to i64
   ret i64 %conv6
 }
 
 define i64 @ustest_f64i64_mm(double %x) {
 ; RV32IF-LABEL: ustest_f64i64_mm:
 ; RV32IF:       # %bb.0: # %entry
 ; RV32IF-NEXT:    addi sp, sp, -32
 ; RV32IF-NEXT:    .cfi_def_cfa_offset 32
 ; RV32IF-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32IF-NEXT:    .cfi_offset ra, -4
 ; RV32IF-NEXT:    mv a2, a1
 ; RV32IF-NEXT:    mv a1, a0
 ; RV32IF-NEXT:    addi a0, sp, 8
 ; RV32IF-NEXT:    call __fixdfti@plt
-; RV32IF-NEXT:    lw a0, 16(sp)
 ; RV32IF-NEXT:    lw a2, 20(sp)
-; RV32IF-NEXT:    li a1, 1
-; RV32IF-NEXT:    mv a4, a0
-; RV32IF-NEXT:    bgez a2, .LBB47_5
+; RV32IF-NEXT:    lw a3, 16(sp)
+; RV32IF-NEXT:    beqz a2, .LBB47_3
 ; RV32IF-NEXT:  # %bb.1: # %entry
-; RV32IF-NEXT:    mv a3, a0
-; RV32IF-NEXT:    bgeu a0, a1, .LBB47_6
-; RV32IF-NEXT:  .LBB47_2: # %entry
-; RV32IF-NEXT:    beqz a2, .LBB47_7
-; RV32IF-NEXT:  .LBB47_3: # %entry
-; RV32IF-NEXT:    slti a1, a2, 0
-; RV32IF-NEXT:    mv a3, a4
-; RV32IF-NEXT:    beqz a1, .LBB47_8
-; RV32IF-NEXT:  .LBB47_4:
-; RV32IF-NEXT:    lw a5, 8(sp)
-; RV32IF-NEXT:    j .LBB47_9
+; RV32IF-NEXT:    slti a0, a2, 0
+; RV32IF-NEXT:    beqz a0, .LBB47_4
+; RV32IF-NEXT:  .LBB47_2:
+; RV32IF-NEXT:    lw a5, 12(sp)
+; RV32IF-NEXT:    j .LBB47_5
+; RV32IF-NEXT:  .LBB47_3:
+; RV32IF-NEXT:    seqz a0, a3
+; RV32IF-NEXT:    bnez a0, .LBB47_2
+; RV32IF-NEXT:  .LBB47_4: # %entry
+; RV32IF-NEXT:    li a5, 0
 ; RV32IF-NEXT:  .LBB47_5: # %entry
-; RV32IF-NEXT:    li a4, 1
-; RV32IF-NEXT:    mv a3, a0
-; RV32IF-NEXT:    bltu a0, a1, .LBB47_2
-; RV32IF-NEXT:  .LBB47_6: # %entry
-; RV32IF-NEXT:    li a3, 1
-; RV32IF-NEXT:    bnez a2, .LBB47_3
-; RV32IF-NEXT:  .LBB47_7:
-; RV32IF-NEXT:    seqz a1, a0
-; RV32IF-NEXT:    bnez a1, .LBB47_4
-; RV32IF-NEXT:  .LBB47_8: # %entry
+; RV32IF-NEXT:    xori a1, a3, 1
+; RV32IF-NEXT:    or a4, a1, a2
+; RV32IF-NEXT:    li a1, 0
+; RV32IF-NEXT:    beqz a4, .LBB47_7
+; RV32IF-NEXT:  # %bb.6: # %entry
+; RV32IF-NEXT:    mv a1, a5
+; RV32IF-NEXT:  .LBB47_7: # %entry
+; RV32IF-NEXT:    bnez a0, .LBB47_9
+; RV32IF-NEXT:  # %bb.8: # %entry
 ; RV32IF-NEXT:    li a5, 0
-; RV32IF-NEXT:  .LBB47_9: # %entry
-; RV32IF-NEXT:    xori a0, a0, 1
-; RV32IF-NEXT:    or a0, a0, a2
-; RV32IF-NEXT:    li a4, 0
-; RV32IF-NEXT:    beqz a0, .LBB47_11
-; RV32IF-NEXT:  # %bb.10: # %entry
-; RV32IF-NEXT:    mv a4, a5
+; RV32IF-NEXT:    li a0, 0
+; RV32IF-NEXT:    bnez a4, .LBB47_10
+; RV32IF-NEXT:    j .LBB47_11
+; RV32IF-NEXT:  .LBB47_9:
+; RV32IF-NEXT:    lw a5, 8(sp)
+; RV32IF-NEXT:    li a0, 0
+; RV32IF-NEXT:    beqz a4, .LBB47_11
+; RV32IF-NEXT:  .LBB47_10: # %entry
+; RV32IF-NEXT:    mv a0, a5
 ; RV32IF-NEXT:  .LBB47_11: # %entry
-; RV32IF-NEXT:    bnez a1, .LBB47_13
+; RV32IF-NEXT:    li a5, 1
+; RV32IF-NEXT:    mv a4, a3
+; RV32IF-NEXT:    bgez a2, .LBB47_17
 ; RV32IF-NEXT:  # %bb.12: # %entry
-; RV32IF-NEXT:    li a5, 0
-; RV32IF-NEXT:    li a1, 0
-; RV32IF-NEXT:    bnez a0, .LBB47_14
-; RV32IF-NEXT:    j .LBB47_15
-; RV32IF-NEXT:  .LBB47_13:
-; RV32IF-NEXT:    lw a5, 12(sp)
-; RV32IF-NEXT:    li a1, 0
-; RV32IF-NEXT:    beqz a0, .LBB47_15
+; RV32IF-NEXT:    bgeu a3, a5, .LBB47_18
+; RV32IF-NEXT:  .LBB47_13: # %entry
+; RV32IF-NEXT:    bnez a2, .LBB47_19
 ; RV32IF-NEXT:  .LBB47_14: # %entry
-; RV32IF-NEXT:    mv a1, a5
-; RV32IF-NEXT:  .LBB47_15: # %entry
 ; RV32IF-NEXT:    bgez a2, .LBB47_20
-; RV32IF-NEXT:  # %bb.16: # %entry
-; RV32IF-NEXT:    mv a5, a4
-; RV32IF-NEXT:    beqz a1, .LBB47_21
+; RV32IF-NEXT:  .LBB47_15: # %entry
+; RV32IF-NEXT:    beqz a2, .LBB47_21
+; RV32IF-NEXT:  .LBB47_16: # %entry
+; RV32IF-NEXT:    sgtz a4, a2
+; RV32IF-NEXT:    mv a5, a0
+; RV32IF-NEXT:    beqz a4, .LBB47_22
+; RV32IF-NEXT:    j .LBB47_23
 ; RV32IF-NEXT:  .LBB47_17: # %entry
-; RV32IF-NEXT:    mv a0, a4
-; RV32IF-NEXT:    bnez a1, .LBB47_22
+; RV32IF-NEXT:    li a4, 1
+; RV32IF-NEXT:    bltu a3, a5, .LBB47_13
 ; RV32IF-NEXT:  .LBB47_18: # %entry
-; RV32IF-NEXT:    beqz a2, .LBB47_23
+; RV32IF-NEXT:    li a3, 1
+; RV32IF-NEXT:    beqz a2, .LBB47_14
 ; RV32IF-NEXT:  .LBB47_19: # %entry
-; RV32IF-NEXT:    sgtz a5, a2
-; RV32IF-NEXT:    beqz a5, .LBB47_24
-; RV32IF-NEXT:    j .LBB47_25
+; RV32IF-NEXT:    mv a3, a4
+; RV32IF-NEXT:    bltz a2, .LBB47_15
 ; RV32IF-NEXT:  .LBB47_20: # %entry
 ; RV32IF-NEXT:    li a2, 0
-; RV32IF-NEXT:    mv a5, a4
-; RV32IF-NEXT:    bnez a1, .LBB47_17
-; RV32IF-NEXT:  .LBB47_21: # %entry
-; RV32IF-NEXT:    li a5, 0
-; RV32IF-NEXT:    mv a0, a4
-; RV32IF-NEXT:    beqz a1, .LBB47_18
+; RV32IF-NEXT:    bnez a2, .LBB47_16
+; RV32IF-NEXT:  .LBB47_21:
+; RV32IF-NEXT:    snez a4, a3
+; RV32IF-NEXT:    mv a5, a0
+; RV32IF-NEXT:    bnez a4, .LBB47_23
 ; RV32IF-NEXT:  .LBB47_22: # %entry
-; RV32IF-NEXT:    mv a0, a5
-; RV32IF-NEXT:    bnez a2, .LBB47_19
-; RV32IF-NEXT:  .LBB47_23:
-; RV32IF-NEXT:    snez a5, a3
-; RV32IF-NEXT:    bnez a5, .LBB47_25
-; RV32IF-NEXT:  .LBB47_24: # %entry
-; RV32IF-NEXT:    li a4, 0
+; RV32IF-NEXT:    li a5, 0
+; RV32IF-NEXT:  .LBB47_23: # %entry
+; RV32IF-NEXT:    mv a6, a0
+; RV32IF-NEXT:    beqz a1, .LBB47_30
+; RV32IF-NEXT:  # %bb.24: # %entry
+; RV32IF-NEXT:    bnez a1, .LBB47_31
 ; RV32IF-NEXT:  .LBB47_25: # %entry
 ; RV32IF-NEXT:    or a2, a3, a2
-; RV32IF-NEXT:    bnez a2, .LBB47_30
-; RV32IF-NEXT:  # %bb.26: # %entry
+; RV32IF-NEXT:    bnez a2, .LBB47_32
+; RV32IF-NEXT:  .LBB47_26: # %entry
 ; RV32IF-NEXT:    mv a3, a1
-; RV32IF-NEXT:    beqz a5, .LBB47_31
+; RV32IF-NEXT:    beqz a4, .LBB47_33
 ; RV32IF-NEXT:  .LBB47_27: # %entry
 ; RV32IF-NEXT:    beqz a2, .LBB47_29
 ; RV32IF-NEXT:  .LBB47_28: # %entry
 ; RV32IF-NEXT:    mv a1, a3
 ; RV32IF-NEXT:  .LBB47_29: # %entry
 ; RV32IF-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IF-NEXT:    addi sp, sp, 32
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB47_30: # %entry
-; RV32IF-NEXT:    mv a0, a4
-; RV32IF-NEXT:    mv a3, a1
-; RV32IF-NEXT:    bnez a5, .LBB47_27
+; RV32IF-NEXT:    li a6, 0
+; RV32IF-NEXT:    beqz a1, .LBB47_25
 ; RV32IF-NEXT:  .LBB47_31: # %entry
+; RV32IF-NEXT:    mv a0, a6
+; RV32IF-NEXT:    or a2, a3, a2
+; RV32IF-NEXT:    beqz a2, .LBB47_26
+; RV32IF-NEXT:  .LBB47_32: # %entry
+; RV32IF-NEXT:    mv a0, a5
+; RV32IF-NEXT:    mv a3, a1
+; RV32IF-NEXT:    bnez a4, .LBB47_27
+; RV32IF-NEXT:  .LBB47_33: # %entry
 ; RV32IF-NEXT:    li a3, 0
 ; RV32IF-NEXT:    bnez a2, .LBB47_28
 ; RV32IF-NEXT:    j .LBB47_29
 ;
 ; RV64-LABEL: ustest_f64i64_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16

 ; RV32IFD-LABEL: ustest_f64i64_mm:
 ; RV32IFD:       # %bb.0: # %entry
 ; RV32IFD-NEXT:    addi sp, sp, -32
 ; RV32IFD-NEXT:    .cfi_def_cfa_offset 32
 ; RV32IFD-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32IFD-NEXT:    .cfi_offset ra, -4
 ; RV32IFD-NEXT:    addi a0, sp, 8
 ; RV32IFD-NEXT:    call __fixdfti@plt
-; RV32IFD-NEXT:    lw a0, 16(sp)
 ; RV32IFD-NEXT:    lw a2, 20(sp)
-; RV32IFD-NEXT:    li a1, 1
-; RV32IFD-NEXT:    mv a4, a0
-; RV32IFD-NEXT:    bgez a2, .LBB47_5
+; RV32IFD-NEXT:    lw a3, 16(sp)
+; RV32IFD-NEXT:    beqz a2, .LBB47_3
 ; RV32IFD-NEXT:  # %bb.1: # %entry
-; RV32IFD-NEXT:    mv a3, a0
-; RV32IFD-NEXT:    bgeu a0, a1, .LBB47_6
-; RV32IFD-NEXT:  .LBB47_2: # %entry
-; RV32IFD-NEXT:    beqz a2, .LBB47_7
-; RV32IFD-NEXT:  .LBB47_3: # %entry
-; RV32IFD-NEXT:    slti a1, a2, 0
-; RV32IFD-NEXT:    mv a3, a4
-; RV32IFD-NEXT:    beqz a1, .LBB47_8
-; RV32IFD-NEXT:  .LBB47_4:
-; RV32IFD-NEXT:    lw a5, 8(sp)
-; RV32IFD-NEXT:    j .LBB47_9
+; RV32IFD-NEXT:    slti a0, a2, 0
+; RV32IFD-NEXT:    beqz a0, .LBB47_4
+; RV32IFD-NEXT:  .LBB47_2:
+; RV32IFD-NEXT:    lw a5, 12(sp)
+; RV32IFD-NEXT:    j .LBB47_5
+; RV32IFD-NEXT:  .LBB47_3:
+; RV32IFD-NEXT:    seqz a0, a3
+; RV32IFD-NEXT:    bnez a0, .LBB47_2
+; RV32IFD-NEXT:  .LBB47_4: # %entry
+; RV32IFD-NEXT:    li a5, 0
 ; RV32IFD-NEXT:  .LBB47_5: # %entry
-; RV32IFD-NEXT:    li a4, 1
-; RV32IFD-NEXT:    mv a3, a0
-; RV32IFD-NEXT:    bltu a0, a1, .LBB47_2
-; RV32IFD-NEXT:  .LBB47_6: # %entry
-; RV32IFD-NEXT:    li a3, 1
-; RV32IFD-NEXT:    bnez a2, .LBB47_3
-; RV32IFD-NEXT:  .LBB47_7:
-; RV32IFD-NEXT:    seqz a1, a0
-; RV32IFD-NEXT:    bnez a1, .LBB47_4
-; RV32IFD-NEXT:  .LBB47_8: # %entry
+; RV32IFD-NEXT:    xori a1, a3, 1
+; RV32IFD-NEXT:    or a4, a1, a2
+; RV32IFD-NEXT:    li a1, 0
+; RV32IFD-NEXT:    beqz a4, .LBB47_7
+; RV32IFD-NEXT:  # %bb.6: # %entry
+; RV32IFD-NEXT:    mv a1, a5
+; RV32IFD-NEXT:  .LBB47_7: # %entry
+; RV32IFD-NEXT:    bnez a0, .LBB47_9
+; RV32IFD-NEXT:  # %bb.8: # %entry
 ; RV32IFD-NEXT:    li a5, 0
-; RV32IFD-NEXT:  .LBB47_9: # %entry
-; RV32IFD-NEXT:    xori a0, a0, 1
-; RV32IFD-NEXT:    or a0, a0, a2
-; RV32IFD-NEXT:    li a4, 0
-; RV32IFD-NEXT:    beqz a0, .LBB47_11
-; RV32IFD-NEXT:  # %bb.10: # %entry
-; RV32IFD-NEXT:    mv a4, a5
+; RV32IFD-NEXT:    li a0, 0
+; RV32IFD-NEXT:    bnez a4, .LBB47_10
+; RV32IFD-NEXT:    j .LBB47_11
+; RV32IFD-NEXT:  .LBB47_9:
+; RV32IFD-NEXT:    lw a5, 8(sp)
+; RV32IFD-NEXT:    li a0, 0
+; RV32IFD-NEXT:    beqz a4, .LBB47_11
+; RV32IFD-NEXT:  .LBB47_10: # %entry
+; RV32IFD-NEXT:    mv a0, a5
 ; RV32IFD-NEXT:  .LBB47_11: # %entry
-; RV32IFD-NEXT:    bnez a1, .LBB47_13
+; RV32IFD-NEXT:    li a5, 1
+; RV32IFD-NEXT:    mv a4, a3
+; RV32IFD-NEXT:    bgez a2, .LBB47_17
 ; RV32IFD-NEXT:  # %bb.12: # %entry
-; RV32IFD-NEXT:    li a5, 0
-; RV32IFD-NEXT:    li a1, 0
-; RV32IFD-NEXT:    bnez a0, .LBB47_14
-; RV32IFD-NEXT:    j .LBB47_15
-; RV32IFD-NEXT:  .LBB47_13:
-; RV32IFD-NEXT:    lw a5, 12(sp)
-; RV32IFD-NEXT:    li a1, 0
-; RV32IFD-NEXT:    beqz a0, .LBB47_15
+; RV32IFD-NEXT:    bgeu a3, a5, .LBB47_18
+; RV32IFD-NEXT:  .LBB47_13: # %entry
+; RV32IFD-NEXT:    bnez a2, .LBB47_19
 ; RV32IFD-NEXT:  .LBB47_14: # %entry
-; RV32IFD-NEXT:    mv a1, a5
-; RV32IFD-NEXT:  .LBB47_15: # %entry
 ; RV32IFD-NEXT:    bgez a2, .LBB47_20
-; RV32IFD-NEXT:  # %bb.16: # %entry
-; RV32IFD-NEXT:    mv a5, a4
-; RV32IFD-NEXT:    beqz a1, .LBB47_21
+; RV32IFD-NEXT:  .LBB47_15: # %entry
+; RV32IFD-NEXT:    beqz a2, .LBB47_21
+; RV32IFD-NEXT:  .LBB47_16: # %entry
+; RV32IFD-NEXT:    sgtz a4, a2
+; RV32IFD-NEXT:    mv a5, a0
+; RV32IFD-NEXT:    beqz a4, .LBB47_22
+; RV32IFD-NEXT:    j .LBB47_23
 ; RV32IFD-NEXT:  .LBB47_17: # %entry
-; RV32IFD-NEXT:    mv a0, a4
-; RV32IFD-NEXT:    bnez a1, .LBB47_22
+; RV32IFD-NEXT:    li a4, 1
+; RV32IFD-NEXT:    bltu a3, a5, .LBB47_13
 ; RV32IFD-NEXT:  .LBB47_18: # %entry
-; RV32IFD-NEXT:    beqz a2, .LBB47_23
+; RV32IFD-NEXT:    li a3, 1
+; RV32IFD-NEXT:    beqz a2, .LBB47_14
 ; RV32IFD-NEXT:  .LBB47_19: # %entry
-; RV32IFD-NEXT:    sgtz a5, a2
-; RV32IFD-NEXT:    beqz a5, .LBB47_24
-; RV32IFD-NEXT:    j .LBB47_25
+; RV32IFD-NEXT:    mv a3, a4
+; RV32IFD-NEXT:    bltz a2, .LBB47_15
 ; RV32IFD-NEXT:  .LBB47_20: # %entry
 ; RV32IFD-NEXT:    li a2, 0
-; RV32IFD-NEXT:    mv a5, a4
-; RV32IFD-NEXT:    bnez a1, .LBB47_17
-; RV32IFD-NEXT:  .LBB47_21: # %entry
-; RV32IFD-NEXT:    li a5, 0
-; RV32IFD-NEXT:    mv a0, a4
-; RV32IFD-NEXT:    beqz a1, .LBB47_18
+; RV32IFD-NEXT:    bnez a2, .LBB47_16
+; RV32IFD-NEXT:  .LBB47_21:
+; RV32IFD-NEXT:    snez a4, a3
+; RV32IFD-NEXT:    mv a5, a0
+; RV32IFD-NEXT:    bnez a4, .LBB47_23
 ; RV32IFD-NEXT:  .LBB47_22: # %entry
-; RV32IFD-NEXT:    mv a0, a5
-; RV32IFD-NEXT:    bnez a2, .LBB47_19
-; RV32IFD-NEXT:  .LBB47_23:
-; RV32IFD-NEXT:    snez a5, a3
-; RV32IFD-NEXT:    bnez a5, .LBB47_25
-; RV32IFD-NEXT:  .LBB47_24: # %entry
-; RV32IFD-NEXT:    li a4, 0
+; RV32IFD-NEXT:    li a5, 0
+; RV32IFD-NEXT:  .LBB47_23: # %entry
+; RV32IFD-NEXT:    mv a6, a0
+; RV32IFD-NEXT:    beqz a1, .LBB47_30
+; RV32IFD-NEXT:  # %bb.24: # %entry
+; RV32IFD-NEXT:    bnez a1, .LBB47_31
 ; RV32IFD-NEXT:  .LBB47_25: # %entry
 ; RV32IFD-NEXT:    or a2, a3, a2
-; RV32IFD-NEXT:    bnez a2, .LBB47_30
-; RV32IFD-NEXT:  # %bb.26: # %entry
+; RV32IFD-NEXT:    bnez a2, .LBB47_32
+; RV32IFD-NEXT:  .LBB47_26: # %entry
 ; RV32IFD-NEXT:    mv a3, a1
-; RV32IFD-NEXT:    beqz a5, .LBB47_31
+; RV32IFD-NEXT:    beqz a4, .LBB47_33
 ; RV32IFD-NEXT:  .LBB47_27: # %entry
 ; RV32IFD-NEXT:    beqz a2, .LBB47_29
 ; RV32IFD-NEXT:  .LBB47_28: # %entry
 ; RV32IFD-NEXT:    mv a1, a3
 ; RV32IFD-NEXT:  .LBB47_29: # %entry
 ; RV32IFD-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32IFD-NEXT:    addi sp, sp, 32
 ; RV32IFD-NEXT:    ret
 ; RV32IFD-NEXT:  .LBB47_30: # %entry
-; RV32IFD-NEXT:    mv a0, a4
-; RV32IFD-NEXT:    mv a3, a1
-; RV32IFD-NEXT:    bnez a5, .LBB47_27
+; RV32IFD-NEXT:    li a6, 0
+; RV32IFD-NEXT:    beqz a1, .LBB47_25
 ; RV32IFD-NEXT:  .LBB47_31: # %entry
+; RV32IFD-NEXT:    mv a0, a6
+; RV32IFD-NEXT:    or a2, a3, a2
+; RV32IFD-NEXT:    beqz a2, .LBB47_26
+; RV32IFD-NEXT:  .LBB47_32: # %entry
+; RV32IFD-NEXT:    mv a0, a5
+; RV32IFD-NEXT:    mv a3, a1
+; RV32IFD-NEXT:    bnez a4, .LBB47_27
+; RV32IFD-NEXT:  .LBB47_33: # %entry
 ; RV32IFD-NEXT:    li a3, 0
 ; RV32IFD-NEXT:    bnez a2, .LBB47_28
 ; RV32IFD-NEXT:    j .LBB47_29
 entry:
   %conv = fptosi double %x to i128
   %spec.store.select = call i128 @llvm.smin.i128(i128 %conv, i128 18446744073709551616)
   %spec.store.select7 = call i128 @llvm.smax.i128(i128 %spec.store.select, i128 0)
   %conv6 = trunc i128 %spec.store.select7 to i64
   ret i64 %conv6
 }
 
 define i64 @stest_f32i64_mm(float %x) {
 ; RV32-LABEL: stest_f32i64_mm:
 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    addi sp, sp, -32
 ; RV32-NEXT:    .cfi_def_cfa_offset 32
 ; RV32-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32-NEXT:    .cfi_offset ra, -4
 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixsfti@plt
-; RV32-NEXT:    lw a5, 8(sp)
-; RV32-NEXT:    lw a3, 20(sp)
-; RV32-NEXT:    lw a1, 12(sp)
+; RV32-NEXT:    lw a7, 8(sp)
+; RV32-NEXT:    lw a5, 12(sp)
+; RV32-NEXT:    lui a4, 524288
+; RV32-NEXT:    addi a0, a4, -1
 ; RV32-NEXT:    li a2, -1
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bltz a3, .LBB48_2
+; RV32-NEXT:    mv a1, a7
+; RV32-NEXT:    bltu a5, a0, .LBB48_2
 ; RV32-NEXT:  # %bb.1: # %entry
-; RV32-NEXT:    li a7, -1
+; RV32-NEXT:    li a1, -1
 ; RV32-NEXT:  .LBB48_2: # %entry
-; RV32-NEXT:    lui a4, 524288
-; RV32-NEXT:    addi a6, a4, -1
-; RV32-NEXT:    mv t0, a5
-; RV32-NEXT:    bgeu a1, a6, .LBB48_19
+; RV32-NEXT:    lw a6, 20(sp)
+; RV32-NEXT:    mv a3, a7
+; RV32-NEXT:    bne a5, a0, .LBB48_19
 ; RV32-NEXT:  # %bb.3: # %entry
-; RV32-NEXT:    lw a0, 16(sp)
-; RV32-NEXT:    bne a1, a6, .LBB48_20
+; RV32-NEXT:    lw a1, 16(sp)
+; RV32-NEXT:    bgez a6, .LBB48_20
 ; RV32-NEXT:  .LBB48_4: # %entry
-; RV32-NEXT:    or t0, a0, a3
+; RV32-NEXT:    or t0, a1, a6
 ; RV32-NEXT:    bnez t0, .LBB48_21
 ; RV32-NEXT:  .LBB48_5: # %entry
-; RV32-NEXT:    mv a7, a1
-; RV32-NEXT:    bgez a3, .LBB48_22
+; RV32-NEXT:    mv a7, a5
+; RV32-NEXT:    bgez a6, .LBB48_22
 ; RV32-NEXT:  .LBB48_6: # %entry
-; RV32-NEXT:    bgeu a1, a6, .LBB48_23
+; RV32-NEXT:    bgeu a5, a0, .LBB48_23
 ; RV32-NEXT:  .LBB48_7: # %entry
 ; RV32-NEXT:    bnez t0, .LBB48_24
 ; RV32-NEXT:  .LBB48_8: # %entry
-; RV32-NEXT:    li a6, 0
-; RV32-NEXT:    bnez a3, .LBB48_25
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    bnez a6, .LBB48_25
 ; RV32-NEXT:  .LBB48_9: # %entry
-; RV32-NEXT:    bgez a3, .LBB48_26
+; RV32-NEXT:    bgez a6, .LBB48_26
 ; RV32-NEXT:  .LBB48_10: # %entry
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bgeu a4, a1, .LBB48_27
+; RV32-NEXT:    mv t0, a5
+; RV32-NEXT:    bltz a6, .LBB48_27
 ; RV32-NEXT:  .LBB48_11: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    bne a1, a4, .LBB48_28
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    bgeu a4, a5, .LBB48_28
 ; RV32-NEXT:  .LBB48_12: # %entry
-; RV32-NEXT:    bltz a3, .LBB48_29
+; RV32-NEXT:    and a7, a0, a6
+; RV32-NEXT:    bne a7, a2, .LBB48_29
 ; RV32-NEXT:  .LBB48_13: # %entry
-; RV32-NEXT:    and a6, a6, a3
-; RV32-NEXT:    bne a6, a2, .LBB48_30
+; RV32-NEXT:    mv t0, a3
+; RV32-NEXT:    bgeu a4, a5, .LBB48_30
 ; RV32-NEXT:  .LBB48_14: # %entry
-; RV32-NEXT:    mv a5, a1
-; RV32-NEXT:    bltz a3, .LBB48_31
+; RV32-NEXT:    mv a0, a3
+; RV32-NEXT:    bne a5, a4, .LBB48_31
 ; RV32-NEXT:  .LBB48_15: # %entry
-; RV32-NEXT:    bgeu a4, a1, .LBB48_32
+; RV32-NEXT:    bltz a6, .LBB48_32
 ; RV32-NEXT:  .LBB48_16: # %entry
-; RV32-NEXT:    beq a6, a2, .LBB48_18
+; RV32-NEXT:    beq a7, a2, .LBB48_18
 ; RV32-NEXT:  .LBB48_17: # %entry
-; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    mv a0, a3
 ; RV32-NEXT:  .LBB48_18: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ; RV32-NEXT:  .LBB48_19: # %entry
-; RV32-NEXT:    li t0, -1
-; RV32-NEXT:    lw a0, 16(sp)
-; RV32-NEXT:    beq a1, a6, .LBB48_4
+; RV32-NEXT:    mv a3, a1
+; RV32-NEXT:    lw a1, 16(sp)
+; RV32-NEXT:    bltz a6, .LBB48_4
 ; RV32-NEXT:  .LBB48_20: # %entry
-; RV32-NEXT:    mv a5, t0
-; RV32-NEXT:    or t0, a0, a3
+; RV32-NEXT:    li a7, -1
+; RV32-NEXT:    or t0, a1, a6
 ; RV32-NEXT:    beqz t0, .LBB48_5
 ; RV32-NEXT:  .LBB48_21: # %entry
-; RV32-NEXT:    mv a5, a7
-; RV32-NEXT:    mv a7, a1
-; RV32-NEXT:    bltz a3, .LBB48_6
+; RV32-NEXT:    mv a3, a7
+; RV32-NEXT:    mv a7, a5
+; RV32-NEXT:    bltz a6, .LBB48_6
 ; RV32-NEXT:  .LBB48_22: # %entry
-; RV32-NEXT:    mv a7, a6
-; RV32-NEXT:    bltu a1, a6, .LBB48_7
+; RV32-NEXT:    mv a7, a0
+; RV32-NEXT:    bltu a5, a0, .LBB48_7
 ; RV32-NEXT:  .LBB48_23: # %entry
-; RV32-NEXT:    mv a1, a6
+; RV32-NEXT:    mv a5, a0
 ; RV32-NEXT:    beqz t0, .LBB48_8
 ; RV32-NEXT:  .LBB48_24: # %entry
-; RV32-NEXT:    mv a1, a7
-; RV32-NEXT:    li a6, 0
-; RV32-NEXT:    beqz a3, .LBB48_9
+; RV32-NEXT:    mv a5, a7
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    beqz a6, .LBB48_9
 ; RV32-NEXT:  .LBB48_25: # %entry
-; RV32-NEXT:    srai a6, a3, 31
-; RV32-NEXT:    and a6, a6, a0
-; RV32-NEXT:    bltz a3, .LBB48_10
+; RV32-NEXT:    srai a0, a6, 31
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    bltz a6, .LBB48_10
 ; RV32-NEXT:  .LBB48_26: # %entry
-; RV32-NEXT:    li a3, 0
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bltu a4, a1, .LBB48_11
+; RV32-NEXT:    li a6, 0
+; RV32-NEXT:    mv t0, a5
+; RV32-NEXT:    bgez a6, .LBB48_11
 ; RV32-NEXT:  .LBB48_27: # %entry
-; RV32-NEXT:    li a7, 0
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    beq a1, a4, .LBB48_12
+; RV32-NEXT:    lui t0, 524288
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    bltu a4, a5, .LBB48_12
 ; RV32-NEXT:  .LBB48_28: # %entry
-; RV32-NEXT:    mv a0, a7
-; RV32-NEXT:    bgez a3, .LBB48_13
+; RV32-NEXT:    lui a1, 524288
+; RV32-NEXT:    and a7, a0, a6
+; RV32-NEXT:    beq a7, a2, .LBB48_13
 ; RV32-NEXT:  .LBB48_29: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    and a6, a6, a3
-; RV32-NEXT:    beq a6, a2, .LBB48_14
+; RV32-NEXT:    mv a1, t0
+; RV32-NEXT:    mv t0, a3
+; RV32-NEXT:    bltu a4, a5, .LBB48_14
 ; RV32-NEXT:  .LBB48_30: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    mv a5, a1
-; RV32-NEXT:    bgez a3, .LBB48_15
+; RV32-NEXT:    li t0, 0
+; RV32-NEXT:    mv a0, a3
+; RV32-NEXT:    beq a5, a4, .LBB48_15
 ; RV32-NEXT:  .LBB48_31: # %entry
-; RV32-NEXT:    lui a5, 524288
-; RV32-NEXT:    bltu a4, a1, .LBB48_16
+; RV32-NEXT:    mv a0, t0
+; RV32-NEXT:    bgez a6, .LBB48_16
 ; RV32-NEXT:  .LBB48_32: # %entry
-; RV32-NEXT:    lui a1, 524288
-; RV32-NEXT:    bne a6, a2, .LBB48_17
+; RV32-NEXT:    li a3, 0
+; RV32-NEXT:    bne a7, a2, .LBB48_17
 ; RV32-NEXT:    j .LBB48_18
 ;
 ; RV64-LABEL: stest_f32i64_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    feq.s a0, fa0, fa0
 ; RV64-NEXT:    beqz a0, .LBB48_2
 ; RV64-NEXT:  # %bb.1:
 ; RV64-NEXT:    fcvt.l.s a0, fa0, rtz

 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixunssfti@plt
 ; RV32-NEXT:    lw a0, 20(sp)
 ; RV32-NEXT:    lw a3, 16(sp)
 ; RV32-NEXT:    li a1, 0
 ; RV32-NEXT:    beqz a0, .LBB49_3
 ; RV32-NEXT:  # %bb.1: # %entry
 ; RV32-NEXT:    mv a2, a1
-; RV32-NEXT:    beq a2, a1, .LBB49_4
+; RV32-NEXT:    beqz a2, .LBB49_4
 ; RV32-NEXT:  .LBB49_2:
 ; RV32-NEXT:    lw a4, 8(sp)
 ; RV32-NEXT:    j .LBB49_5
 ; RV32-NEXT:  .LBB49_3:
 ; RV32-NEXT:    seqz a2, a3
-; RV32-NEXT:    bne a2, a1, .LBB49_2
+; RV32-NEXT:    bnez a2, .LBB49_2
 ; RV32-NEXT:  .LBB49_4: # %entry
 ; RV32-NEXT:    mv a4, a1
 ; RV32-NEXT:  .LBB49_5: # %entry
 ; RV32-NEXT:    xori a3, a3, 1
 ; RV32-NEXT:    or a3, a3, a0
 ; RV32-NEXT:    mv a0, a1
-; RV32-NEXT:    beq a3, a1, .LBB49_7
+; RV32-NEXT:    beqz a3, .LBB49_7
 ; RV32-NEXT:  # %bb.6: # %entry
 ; RV32-NEXT:    mv a0, a4
 ; RV32-NEXT:  .LBB49_7: # %entry
-; RV32-NEXT:    bne a2, a1, .LBB49_9
+; RV32-NEXT:    bnez a2, .LBB49_9
 ; RV32-NEXT:  # %bb.8: # %entry
 ; RV32-NEXT:    mv a2, a1
-; RV32-NEXT:    bne a3, a1, .LBB49_10
+; RV32-NEXT:    bnez a3, .LBB49_10
 ; RV32-NEXT:    j .LBB49_11
 ; RV32-NEXT:  .LBB49_9:
 ; RV32-NEXT:    lw a2, 12(sp)
-; RV32-NEXT:    beq a3, a1, .LBB49_11
+; RV32-NEXT:    beqz a3, .LBB49_11
 ; RV32-NEXT:  .LBB49_10: # %entry
 ; RV32-NEXT:    mv a1, a2
 ; RV32-NEXT:  .LBB49_11: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: utest_f32i64_mm:

 ; RV32-LABEL: ustest_f32i64_mm:
 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    addi sp, sp, -32
 ; RV32-NEXT:    .cfi_def_cfa_offset 32
 ; RV32-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32-NEXT:    .cfi_offset ra, -4
 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixsfti@plt
-; RV32-NEXT:    lw a0, 16(sp)
 ; RV32-NEXT:    lw a2, 20(sp)
-; RV32-NEXT:    li a1, 1
-; RV32-NEXT:    mv a4, a0
-; RV32-NEXT:    bgez a2, .LBB50_5
+; RV32-NEXT:    lw a3, 16(sp)
+; RV32-NEXT:    beqz a2, .LBB50_3
 ; RV32-NEXT:  # %bb.1: # %entry
-; RV32-NEXT:    mv a3, a0
-; RV32-NEXT:    bgeu a0, a1, .LBB50_6
-; RV32-NEXT:  .LBB50_2: # %entry
-; RV32-NEXT:    beqz a2, .LBB50_7
-; RV32-NEXT:  .LBB50_3: # %entry
-; RV32-NEXT:    slti a1, a2, 0
-; RV32-NEXT:    mv a3, a4
-; RV32-NEXT:    beqz a1, .LBB50_8
-; RV32-NEXT:  .LBB50_4:
-; RV32-NEXT:    lw a5, 8(sp)
-; RV32-NEXT:    j .LBB50_9
+; RV32-NEXT:    slti a0, a2, 0
+; RV32-NEXT:    beqz a0, .LBB50_4
+; RV32-NEXT:  .LBB50_2:
+; RV32-NEXT:    lw a5, 12(sp)
+; RV32-NEXT:    j .LBB50_5
+; RV32-NEXT:  .LBB50_3:
+; RV32-NEXT:    seqz a0, a3
+; RV32-NEXT:    bnez a0, .LBB50_2
+; RV32-NEXT:  .LBB50_4: # %entry
+; RV32-NEXT:    li a5, 0
 ; RV32-NEXT:  .LBB50_5: # %entry
-; RV32-NEXT:    li a4, 1
-; RV32-NEXT:    mv a3, a0
-; RV32-NEXT:    bltu a0, a1, .LBB50_2
-; RV32-NEXT:  .LBB50_6: # %entry
-; RV32-NEXT:    li a3, 1
-; RV32-NEXT:    bnez a2, .LBB50_3
-; RV32-NEXT:  .LBB50_7:
-; RV32-NEXT:    seqz a1, a0
-; RV32-NEXT:    bnez a1, .LBB50_4
-; RV32-NEXT:  .LBB50_8: # %entry
+; RV32-NEXT:    xori a1, a3, 1
+; RV32-NEXT:    or a4, a1, a2
+; RV32-NEXT:    li a1, 0
+; RV32-NEXT:    beqz a4, .LBB50_7
+; RV32-NEXT:  # %bb.6: # %entry
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:  .LBB50_7: # %entry
+; RV32-NEXT:    bnez a0, .LBB50_9
+; RV32-NEXT:  # %bb.8: # %entry
 ; RV32-NEXT:    li a5, 0
-; RV32-NEXT:  .LBB50_9: # %entry
-; RV32-NEXT:    xori a0, a0, 1
-; RV32-NEXT:    or a0, a0, a2
-; RV32-NEXT:    li a4, 0
-; RV32-NEXT:    beqz a0, .LBB50_11
-; RV32-NEXT:  # %bb.10: # %entry
-; RV32-NEXT:    mv a4, a5
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    bnez a4, .LBB50_10
+; RV32-NEXT:    j .LBB50_11
+; RV32-NEXT:  .LBB50_9:
+; RV32-NEXT:    lw a5, 8(sp)
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    beqz a4, .LBB50_11
+; RV32-NEXT:  .LBB50_10: # %entry
+; RV32-NEXT:    mv a0, a5
 ; RV32-NEXT:  .LBB50_11: # %entry
-; RV32-NEXT:    bnez a1, .LBB50_13
+; RV32-NEXT:    li a5, 1
+; RV32-NEXT:    mv a4, a3
+; RV32-NEXT:    bgez a2, .LBB50_17
 ; RV32-NEXT:  # %bb.12: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    li a1, 0
-; RV32-NEXT:    bnez a0, .LBB50_14
-; RV32-NEXT:    j .LBB50_15
-; RV32-NEXT:  .LBB50_13:
-; RV32-NEXT:    lw a5, 12(sp)
-; RV32-NEXT:    li a1, 0
-; RV32-NEXT:    beqz a0, .LBB50_15
+; RV32-NEXT:    bgeu a3, a5, .LBB50_18
+; RV32-NEXT:  .LBB50_13: # %entry
+; RV32-NEXT:    bnez a2, .LBB50_19
 ; RV32-NEXT:  .LBB50_14: # %entry
-; RV32-NEXT:    mv a1, a5
-; RV32-NEXT:  .LBB50_15: # %entry
 ; RV32-NEXT:    bgez a2, .LBB50_20
-; RV32-NEXT:  # %bb.16: # %entry
-; RV32-NEXT:    mv a5, a4
-; RV32-NEXT:    beqz a1, .LBB50_21
+; RV32-NEXT:  .LBB50_15: # %entry
+; RV32-NEXT:    beqz a2, .LBB50_21
+; RV32-NEXT:  .LBB50_16: # %entry
+; RV32-NEXT:    sgtz a4, a2
+; RV32-NEXT:    mv a5, a0
+; RV32-NEXT:    beqz a4, .LBB50_22
+; RV32-NEXT:    j .LBB50_23
 ; RV32-NEXT:  .LBB50_17: # %entry
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    bnez a1, .LBB50_22
+; RV32-NEXT:    li a4, 1
+; RV32-NEXT:    bltu a3, a5, .LBB50_13
 ; RV32-NEXT:  .LBB50_18: # %entry
-; RV32-NEXT:    beqz a2, .LBB50_23
+; RV32-NEXT:    li a3, 1
+; RV32-NEXT:    beqz a2, .LBB50_14
 ; RV32-NEXT:  .LBB50_19: # %entry
-; RV32-NEXT:    sgtz a5, a2
-; RV32-NEXT:    beqz a5, .LBB50_24
-; RV32-NEXT:    j .LBB50_25
+; RV32-NEXT:    mv a3, a4
+; RV32-NEXT:    bltz a2, .LBB50_15
 ; RV32-NEXT:  .LBB50_20: # %entry
 ; RV32-NEXT:    li a2, 0
-; RV32-NEXT:    mv a5, a4
-; RV32-NEXT:    bnez a1, .LBB50_17
-; RV32-NEXT:  .LBB50_21: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    beqz a1, .LBB50_18
+; RV32-NEXT:    bnez a2, .LBB50_16
+; RV32-NEXT:  .LBB50_21:
+; RV32-NEXT:    snez a4, a3
+; RV32-NEXT:    mv a5, a0
+; RV32-NEXT:    bnez a4, .LBB50_23
 ; RV32-NEXT:  .LBB50_22: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    bnez a2, .LBB50_19
-; RV32-NEXT:  .LBB50_23:
-; RV32-NEXT:    snez a5, a3
-; RV32-NEXT:    bnez a5, .LBB50_25
-; RV32-NEXT:  .LBB50_24: # %entry
-; RV32-NEXT:    li a4, 0
+; RV32-NEXT:    li a5, 0
+; RV32-NEXT:  .LBB50_23: # %entry
+; RV32-NEXT:    mv a6, a0
+; RV32-NEXT:    beqz a1, .LBB50_30
+; RV32-NEXT:  # %bb.24: # %entry
+; RV32-NEXT:    bnez a1, .LBB50_31
 ; RV32-NEXT:  .LBB50_25: # %entry
 ; RV32-NEXT:    or a2, a3, a2
-; RV32-NEXT:    bnez a2, .LBB50_30
-; RV32-NEXT:  # %bb.26: # %entry
+; RV32-NEXT:    bnez a2, .LBB50_32
+; RV32-NEXT:  .LBB50_26: # %entry
 ; RV32-NEXT:    mv a3, a1
-; RV32-NEXT:    beqz a5, .LBB50_31
+; RV32-NEXT:    beqz a4, .LBB50_33
 ; RV32-NEXT:  .LBB50_27: # %entry
 ; RV32-NEXT:    beqz a2, .LBB50_29
 ; RV32-NEXT:  .LBB50_28: # %entry
 ; RV32-NEXT:    mv a1, a3
 ; RV32-NEXT:  .LBB50_29: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ; RV32-NEXT:  .LBB50_30: # %entry
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    mv a3, a1
-; RV32-NEXT:    bnez a5, .LBB50_27
+; RV32-NEXT:    li a6, 0
+; RV32-NEXT:    beqz a1, .LBB50_25
 ; RV32-NEXT:  .LBB50_31: # %entry
+; RV32-NEXT:    mv a0, a6
+; RV32-NEXT:    or a2, a3, a2
+; RV32-NEXT:    beqz a2, .LBB50_26
+; RV32-NEXT:  .LBB50_32: # %entry
+; RV32-NEXT:    mv a0, a5
+; RV32-NEXT:    mv a3, a1
+; RV32-NEXT:    bnez a4, .LBB50_27
+; RV32-NEXT:  .LBB50_33: # %entry
 ; RV32-NEXT:    li a3, 0
 ; RV32-NEXT:    bnez a2, .LBB50_28
 ; RV32-NEXT:    j .LBB50_29
 ;
 ; RV64-LABEL: ustest_f32i64_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16

 ; RV32-NEXT:    addi sp, sp, -32
 ; RV32-NEXT:    .cfi_def_cfa_offset 32
 ; RV32-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32-NEXT:    .cfi_offset ra, -4
 ; RV32-NEXT:    fmv.x.w a0, fa0
 ; RV32-NEXT:    call __extendhfsf2@plt
 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixsfti@plt
-; RV32-NEXT:    lw a5, 8(sp)
-; RV32-NEXT:    lw a3, 20(sp)
-; RV32-NEXT:    lw a1, 12(sp)
+; RV32-NEXT:    lw a7, 8(sp)
+; RV32-NEXT:    lw a5, 12(sp)
+; RV32-NEXT:    lui a4, 524288
+; RV32-NEXT:    addi a0, a4, -1
 ; RV32-NEXT:    li a2, -1
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bltz a3, .LBB51_2
+; RV32-NEXT:    mv a1, a7
+; RV32-NEXT:    bltu a5, a0, .LBB51_2
 ; RV32-NEXT:  # %bb.1: # %entry
-; RV32-NEXT:    li a7, -1
+; RV32-NEXT:    li a1, -1
 ; RV32-NEXT:  .LBB51_2: # %entry
-; RV32-NEXT:    lui a4, 524288
-; RV32-NEXT:    addi a6, a4, -1
-; RV32-NEXT:    mv t0, a5
-; RV32-NEXT:    bgeu a1, a6, .LBB51_19
+; RV32-NEXT:    lw a6, 20(sp)
+; RV32-NEXT:    mv a3, a7
+; RV32-NEXT:    bne a5, a0, .LBB51_19
 ; RV32-NEXT:  # %bb.3: # %entry
-; RV32-NEXT:    lw a0, 16(sp)
-; RV32-NEXT:    bne a1, a6, .LBB51_20
+; RV32-NEXT:    lw a1, 16(sp)
+; RV32-NEXT:    bgez a6, .LBB51_20
 ; RV32-NEXT:  .LBB51_4: # %entry
-; RV32-NEXT:    or t0, a0, a3
+; RV32-NEXT:    or t0, a1, a6
 ; RV32-NEXT:    bnez t0, .LBB51_21
 ; RV32-NEXT:  .LBB51_5: # %entry
-; RV32-NEXT:    mv a7, a1
-; RV32-NEXT:    bgez a3, .LBB51_22
+; RV32-NEXT:    mv a7, a5
+; RV32-NEXT:    bgez a6, .LBB51_22
 ; RV32-NEXT:  .LBB51_6: # %entry
-; RV32-NEXT:    bgeu a1, a6, .LBB51_23
+; RV32-NEXT:    bgeu a5, a0, .LBB51_23
 ; RV32-NEXT:  .LBB51_7: # %entry
 ; RV32-NEXT:    bnez t0, .LBB51_24
 ; RV32-NEXT:  .LBB51_8: # %entry
-; RV32-NEXT:    li a6, 0
-; RV32-NEXT:    bnez a3, .LBB51_25
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    bnez a6, .LBB51_25
 ; RV32-NEXT:  .LBB51_9: # %entry
-; RV32-NEXT:    bgez a3, .LBB51_26
+; RV32-NEXT:    bgez a6, .LBB51_26
 ; RV32-NEXT:  .LBB51_10: # %entry
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bgeu a4, a1, .LBB51_27
+; RV32-NEXT:    mv t0, a5
+; RV32-NEXT:    bltz a6, .LBB51_27
 ; RV32-NEXT:  .LBB51_11: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    bne a1, a4, .LBB51_28
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    bgeu a4, a5, .LBB51_28
 ; RV32-NEXT:  .LBB51_12: # %entry
-; RV32-NEXT:    bltz a3, .LBB51_29
+; RV32-NEXT:    and a7, a0, a6
+; RV32-NEXT:    bne a7, a2, .LBB51_29
 ; RV32-NEXT:  .LBB51_13: # %entry
-; RV32-NEXT:    and a6, a6, a3
-; RV32-NEXT:    bne a6, a2, .LBB51_30
+; RV32-NEXT:    mv t0, a3
+; RV32-NEXT:    bgeu a4, a5, .LBB51_30
 ; RV32-NEXT:  .LBB51_14: # %entry
-; RV32-NEXT:    mv a5, a1
-; RV32-NEXT:    bltz a3, .LBB51_31
+; RV32-NEXT:    mv a0, a3
+; RV32-NEXT:    bne a5, a4, .LBB51_31
 ; RV32-NEXT:  .LBB51_15: # %entry
-; RV32-NEXT:    bgeu a4, a1, .LBB51_32
+; RV32-NEXT:    bltz a6, .LBB51_32
 ; RV32-NEXT:  .LBB51_16: # %entry
-; RV32-NEXT:    beq a6, a2, .LBB51_18
+; RV32-NEXT:    beq a7, a2, .LBB51_18
 ; RV32-NEXT:  .LBB51_17: # %entry
-; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    mv a0, a3
 ; RV32-NEXT:  .LBB51_18: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ; RV32-NEXT:  .LBB51_19: # %entry
-; RV32-NEXT:    li t0, -1
-; RV32-NEXT:    lw a0, 16(sp)
-; RV32-NEXT:    beq a1, a6, .LBB51_4
+; RV32-NEXT:    mv a3, a1
+; RV32-NEXT:    lw a1, 16(sp)
+; RV32-NEXT:    bltz a6, .LBB51_4
 ; RV32-NEXT:  .LBB51_20: # %entry
-; RV32-NEXT:    mv a5, t0
-; RV32-NEXT:    or t0, a0, a3
+; RV32-NEXT:    li a7, -1
+; RV32-NEXT:    or t0, a1, a6
 ; RV32-NEXT:    beqz t0, .LBB51_5
 ; RV32-NEXT:  .LBB51_21: # %entry
-; RV32-NEXT:    mv a5, a7
-; RV32-NEXT:    mv a7, a1
-; RV32-NEXT:    bltz a3, .LBB51_6
+; RV32-NEXT:    mv a3, a7
+; RV32-NEXT:    mv a7, a5
+; RV32-NEXT:    bltz a6, .LBB51_6
 ; RV32-NEXT:  .LBB51_22: # %entry
-; RV32-NEXT:    mv a7, a6
-; RV32-NEXT:    bltu a1, a6, .LBB51_7
+; RV32-NEXT:    mv a7, a0
+; RV32-NEXT:    bltu a5, a0, .LBB51_7
 ; RV32-NEXT:  .LBB51_23: # %entry
-; RV32-NEXT:    mv a1, a6
+; RV32-NEXT:    mv a5, a0
 ; RV32-NEXT:    beqz t0, .LBB51_8
 ; RV32-NEXT:  .LBB51_24: # %entry
-; RV32-NEXT:    mv a1, a7
-; RV32-NEXT:    li a6, 0
-; RV32-NEXT:    beqz a3, .LBB51_9
+; RV32-NEXT:    mv a5, a7
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    beqz a6, .LBB51_9
 ; RV32-NEXT:  .LBB51_25: # %entry
-; RV32-NEXT:    srai a6, a3, 31
-; RV32-NEXT:    and a6, a6, a0
-; RV32-NEXT:    bltz a3, .LBB51_10
+; RV32-NEXT:    srai a0, a6, 31
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    bltz a6, .LBB51_10
 ; RV32-NEXT:  .LBB51_26: # %entry
-; RV32-NEXT:    li a3, 0
-; RV32-NEXT:    mv a7, a5
-; RV32-NEXT:    bltu a4, a1, .LBB51_11
+; RV32-NEXT:    li a6, 0
+; RV32-NEXT:    mv t0, a5
+; RV32-NEXT:    bgez a6, .LBB51_11
 ; RV32-NEXT:  .LBB51_27: # %entry
-; RV32-NEXT:    li a7, 0
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    beq a1, a4, .LBB51_12
+; RV32-NEXT:    lui t0, 524288
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:    bltu a4, a5, .LBB51_12
 ; RV32-NEXT:  .LBB51_28: # %entry
-; RV32-NEXT:    mv a0, a7
-; RV32-NEXT:    bgez a3, .LBB51_13
+; RV32-NEXT:    lui a1, 524288
+; RV32-NEXT:    and a7, a0, a6
+; RV32-NEXT:    beq a7, a2, .LBB51_13
 ; RV32-NEXT:  .LBB51_29: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    and a6, a6, a3
-; RV32-NEXT:    beq a6, a2, .LBB51_14
+; RV32-NEXT:    mv a1, t0
+; RV32-NEXT:    mv t0, a3
+; RV32-NEXT:    bltu a4, a5, .LBB51_14
 ; RV32-NEXT:  .LBB51_30: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    mv a5, a1
-; RV32-NEXT:    bgez a3, .LBB51_15
+; RV32-NEXT:    li t0, 0
+; RV32-NEXT:    mv a0, a3
+; RV32-NEXT:    beq a5, a4, .LBB51_15
 ; RV32-NEXT:  .LBB51_31: # %entry
-; RV32-NEXT:    lui a5, 524288
-; RV32-NEXT:    bltu a4, a1, .LBB51_16
+; RV32-NEXT:    mv a0, t0
+; RV32-NEXT:    bgez a6, .LBB51_16
 ; RV32-NEXT:  .LBB51_32: # %entry
-; RV32-NEXT:    lui a1, 524288
-; RV32-NEXT:    bne a6, a2, .LBB51_17
+; RV32-NEXT:    li a3, 0
+; RV32-NEXT:    bne a7, a2, .LBB51_17
 ; RV32-NEXT:    j .LBB51_18
 ;
 ; RV64-LABEL: stest_f16i64_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16
 ; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    .cfi_offset ra, -8

 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixunssfti@plt
 ; RV32-NEXT:    lw a0, 20(sp)
 ; RV32-NEXT:    lw a3, 16(sp)
 ; RV32-NEXT:    li a1, 0
 ; RV32-NEXT:    beqz a0, .LBB52_3
 ; RV32-NEXT:  # %bb.1: # %entry
 ; RV32-NEXT:    mv a2, a1
-; RV32-NEXT:    beq a2, a1, .LBB52_4
+; RV32-NEXT:    beqz a2, .LBB52_4
 ; RV32-NEXT:  .LBB52_2:
 ; RV32-NEXT:    lw a4, 8(sp)
 ; RV32-NEXT:    j .LBB52_5
 ; RV32-NEXT:  .LBB52_3:
 ; RV32-NEXT:    seqz a2, a3
-; RV32-NEXT:    bne a2, a1, .LBB52_2
+; RV32-NEXT:    bnez a2, .LBB52_2
 ; RV32-NEXT:  .LBB52_4: # %entry
 ; RV32-NEXT:    mv a4, a1
 ; RV32-NEXT:  .LBB52_5: # %entry
 ; RV32-NEXT:    xori a3, a3, 1
 ; RV32-NEXT:    or a3, a3, a0
 ; RV32-NEXT:    mv a0, a1
-; RV32-NEXT:    beq a3, a1, .LBB52_7
+; RV32-NEXT:    beqz a3, .LBB52_7
 ; RV32-NEXT:  # %bb.6: # %entry
 ; RV32-NEXT:    mv a0, a4
 ; RV32-NEXT:  .LBB52_7: # %entry
-; RV32-NEXT:    bne a2, a1, .LBB52_9
+; RV32-NEXT:    bnez a2, .LBB52_9
 ; RV32-NEXT:  # %bb.8: # %entry
 ; RV32-NEXT:    mv a2, a1
-; RV32-NEXT:    bne a3, a1, .LBB52_10
+; RV32-NEXT:    bnez a3, .LBB52_10
 ; RV32-NEXT:    j .LBB52_11
 ; RV32-NEXT:  .LBB52_9:
 ; RV32-NEXT:    lw a2, 12(sp)
-; RV32-NEXT:    beq a3, a1, .LBB52_11
+; RV32-NEXT:    beqz a3, .LBB52_11
 ; RV32-NEXT:  .LBB52_10: # %entry
 ; RV32-NEXT:    mv a1, a2
 ; RV32-NEXT:  .LBB52_11: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: utesth_f16i64_mm:

 ; RV32-NEXT:    addi sp, sp, -32
 ; RV32-NEXT:    .cfi_def_cfa_offset 32
 ; RV32-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32-NEXT:    .cfi_offset ra, -4
 ; RV32-NEXT:    fmv.x.w a0, fa0
 ; RV32-NEXT:    call __extendhfsf2@plt
 ; RV32-NEXT:    addi a0, sp, 8
 ; RV32-NEXT:    call __fixsfti@plt
-; RV32-NEXT:    lw a0, 16(sp)
 ; RV32-NEXT:    lw a2, 20(sp)
-; RV32-NEXT:    li a1, 1
-; RV32-NEXT:    mv a4, a0
-; RV32-NEXT:    bgez a2, .LBB53_5
+; RV32-NEXT:    lw a3, 16(sp)
+; RV32-NEXT:    beqz a2, .LBB53_3
 ; RV32-NEXT:  # %bb.1: # %entry
-; RV32-NEXT:    mv a3, a0
-; RV32-NEXT:    bgeu a0, a1, .LBB53_6
-; RV32-NEXT:  .LBB53_2: # %entry
-; RV32-NEXT:    beqz a2, .LBB53_7
-; RV32-NEXT:  .LBB53_3: # %entry
-; RV32-NEXT:    slti a1, a2, 0
-; RV32-NEXT:    mv a3, a4
-; RV32-NEXT:    beqz a1, .LBB53_8
-; RV32-NEXT:  .LBB53_4:
-; RV32-NEXT:    lw a5, 8(sp)
-; RV32-NEXT:    j .LBB53_9
+; RV32-NEXT:    slti a0, a2, 0
+; RV32-NEXT:    beqz a0, .LBB53_4
+; RV32-NEXT:  .LBB53_2:
+; RV32-NEXT:    lw a5, 12(sp)
+; RV32-NEXT:    j .LBB53_5
+; RV32-NEXT:  .LBB53_3:
+; RV32-NEXT:    seqz a0, a3
+; RV32-NEXT:    bnez a0, .LBB53_2
+; RV32-NEXT:  .LBB53_4: # %entry
+; RV32-NEXT:    li a5, 0
 ; RV32-NEXT:  .LBB53_5: # %entry
-; RV32-NEXT:    li a4, 1
-; RV32-NEXT:    mv a3, a0
-; RV32-NEXT:    bltu a0, a1, .LBB53_2
-; RV32-NEXT:  .LBB53_6: # %entry
-; RV32-NEXT:    li a3, 1
-; RV32-NEXT:    bnez a2, .LBB53_3
-; RV32-NEXT:  .LBB53_7:
-; RV32-NEXT:    seqz a1, a0
-; RV32-NEXT:    bnez a1, .LBB53_4
-; RV32-NEXT:  .LBB53_8: # %entry
+; RV32-NEXT:    xori a1, a3, 1
+; RV32-NEXT:    or a4, a1, a2
+; RV32-NEXT:    li a1, 0
+; RV32-NEXT:    beqz a4, .LBB53_7
+; RV32-NEXT:  # %bb.6: # %entry
+; RV32-NEXT:    mv a1, a5
+; RV32-NEXT:  .LBB53_7: # %entry
+; RV32-NEXT:    bnez a0, .LBB53_9
+; RV32-NEXT:  # %bb.8: # %entry
 ; RV32-NEXT:    li a5, 0
-; RV32-NEXT:  .LBB53_9: # %entry
-; RV32-NEXT:    xori a0, a0, 1
-; RV32-NEXT:    or a0, a0, a2
-; RV32-NEXT:    li a4, 0
-; RV32-NEXT:    beqz a0, .LBB53_11
-; RV32-NEXT:  # %bb.10: # %entry
-; RV32-NEXT:    mv a4, a5
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    bnez a4, .LBB53_10
+; RV32-NEXT:    j .LBB53_11
+; RV32-NEXT:  .LBB53_9:
+; RV32-NEXT:    lw a5, 8(sp)
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    beqz a4, .LBB53_11
+; RV32-NEXT:  .LBB53_10: # %entry
+; RV32-NEXT:    mv a0, a5
 ; RV32-NEXT:  .LBB53_11: # %entry
-; RV32-NEXT:    bnez a1, .LBB53_13
+; RV32-NEXT:    li a5, 1
+; RV32-NEXT:    mv a4, a3
+; RV32-NEXT:    bgez a2, .LBB53_17
 ; RV32-NEXT:  # %bb.12: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    li a1, 0
-; RV32-NEXT:    bnez a0, .LBB53_14
-; RV32-NEXT:    j .LBB53_15
-; RV32-NEXT:  .LBB53_13:
-; RV32-NEXT:    lw a5, 12(sp)
-; RV32-NEXT:    li a1, 0
-; RV32-NEXT:    beqz a0, .LBB53_15
+; RV32-NEXT:    bgeu a3, a5, .LBB53_18
+; RV32-NEXT:  .LBB53_13: # %entry
+; RV32-NEXT:    bnez a2, .LBB53_19
 ; RV32-NEXT:  .LBB53_14: # %entry
-; RV32-NEXT:    mv a1, a5
-; RV32-NEXT:  .LBB53_15: # %entry
 ; RV32-NEXT:    bgez a2, .LBB53_20
-; RV32-NEXT:  # %bb.16: # %entry
-; RV32-NEXT:    mv a5, a4
-; RV32-NEXT:    beqz a1, .LBB53_21
+; RV32-NEXT:  .LBB53_15: # %entry
+; RV32-NEXT:    beqz a2, .LBB53_21
+; RV32-NEXT:  .LBB53_16: # %entry
+; RV32-NEXT:    sgtz a4, a2
+; RV32-NEXT:    mv a5, a0
+; RV32-NEXT:    beqz a4, .LBB53_22
+; RV32-NEXT:    j .LBB53_23
 ; RV32-NEXT:  .LBB53_17: # %entry
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    bnez a1, .LBB53_22
+; RV32-NEXT:    li a4, 1
+; RV32-NEXT:    bltu a3, a5, .LBB53_13
 ; RV32-NEXT:  .LBB53_18: # %entry
-; RV32-NEXT:    beqz a2, .LBB53_23
+; RV32-NEXT:    li a3, 1
+; RV32-NEXT:    beqz a2, .LBB53_14
 ; RV32-NEXT:  .LBB53_19: # %entry
-; RV32-NEXT:    sgtz a5, a2
-; RV32-NEXT:    beqz a5, .LBB53_24
-; RV32-NEXT:    j .LBB53_25
+; RV32-NEXT:    mv a3, a4
+; RV32-NEXT:    bltz a2, .LBB53_15
 ; RV32-NEXT:  .LBB53_20: # %entry
 ; RV32-NEXT:    li a2, 0
-; RV32-NEXT:    mv a5, a4
-; RV32-NEXT:    bnez a1, .LBB53_17
-; RV32-NEXT:  .LBB53_21: # %entry
-; RV32-NEXT:    li a5, 0
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    beqz a1, .LBB53_18
+; RV32-NEXT:    bnez a2, .LBB53_16
+; RV32-NEXT:  .LBB53_21:
+; RV32-NEXT:    snez a4, a3
+; RV32-NEXT:    mv a5, a0
+; RV32-NEXT:    bnez a4, .LBB53_23
 ; RV32-NEXT:  .LBB53_22: # %entry
-; RV32-NEXT:    mv a0, a5
-; RV32-NEXT:    bnez a2, .LBB53_19
-; RV32-NEXT:  .LBB53_23:
-; RV32-NEXT:    snez a5, a3
-; RV32-NEXT:    bnez a5, .LBB53_25
-; RV32-NEXT:  .LBB53_24: # %entry
-; RV32-NEXT:    li a4, 0
+; RV32-NEXT:    li a5, 0
+; RV32-NEXT:  .LBB53_23: # %entry
+; RV32-NEXT:    mv a6, a0
+; RV32-NEXT:    beqz a1, .LBB53_30
+; RV32-NEXT:  # %bb.24: # %entry
+; RV32-NEXT:    bnez a1, .LBB53_31
 ; RV32-NEXT:  .LBB53_25: # %entry
 ; RV32-NEXT:    or a2, a3, a2
-; RV32-NEXT:    bnez a2, .LBB53_30
-; RV32-NEXT:  # %bb.26: # %entry
+; RV32-NEXT:    bnez a2, .LBB53_32
+; RV32-NEXT:  .LBB53_26: # %entry
 ; RV32-NEXT:    mv a3, a1
-; RV32-NEXT:    beqz a5, .LBB53_31
+; RV32-NEXT:    beqz a4, .LBB53_33
 ; RV32-NEXT:  .LBB53_27: # %entry
 ; RV32-NEXT:    beqz a2, .LBB53_29
 ; RV32-NEXT:  .LBB53_28: # %entry
 ; RV32-NEXT:    mv a1, a3
 ; RV32-NEXT:  .LBB53_29: # %entry
 ; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 32
 ; RV32-NEXT:    ret
 ; RV32-NEXT:  .LBB53_30: # %entry
-; RV32-NEXT:    mv a0, a4
-; RV32-NEXT:    mv a3, a1
-; RV32-NEXT:    bnez a5, .LBB53_27
+; RV32-NEXT:    li a6, 0
+; RV32-NEXT:    beqz a1, .LBB53_25
 ; RV32-NEXT:  .LBB53_31: # %entry
+; RV32-NEXT:    mv a0, a6
+; RV32-NEXT:    or a2, a3, a2
+; RV32-NEXT:    beqz a2, .LBB53_26
+; RV32-NEXT:  .LBB53_32: # %entry
+; RV32-NEXT:    mv a0, a5
+; RV32-NEXT:    mv a3, a1
+; RV32-NEXT:    bnez a4, .LBB53_27
+; RV32-NEXT:  .LBB53_33: # %entry
 ; RV32-NEXT:    li a3, 0
 ; RV32-NEXT:    bnez a2, .LBB53_28
 ; RV32-NEXT:    j .LBB53_29
 ;
 ; RV64-LABEL: ustest_f16i64_mm:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16

llvm/test/CodeGen/RISCV/fpclamptosat_vec.ll

 ; CHECK-NEXT:    beqz a3, .LBB46_2
 ; CHECK-NEXT:  # %bb.1: # %entry
 ; CHECK-NEXT:    mv a2, a1
 ; CHECK-NEXT:  .LBB46_2: # %entry
 ; CHECK-NEXT:    li a4, 1
 ; CHECK-NEXT:    mv a0, a1
 ; CHECK-NEXT:    bne a3, a4, .LBB46_7
 ; CHECK-NEXT:  # %bb.3: # %entry
-; CHECK-NEXT:    bne s1, a1, .LBB46_8
+; CHECK-NEXT:    bnez s1, .LBB46_8
 ; CHECK-NEXT:  .LBB46_4: # %entry
 ; CHECK-NEXT:    beq s1, a4, .LBB46_6
 ; CHECK-NEXT:  .LBB46_5: # %entry
 ; CHECK-NEXT:    mv a1, s0
 ; CHECK-NEXT:  .LBB46_6: # %entry
 ; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    fld fs0, 0(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    addi sp, sp, 32
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB46_7: # %entry
 ; CHECK-NEXT:    mv a0, a2
-; CHECK-NEXT:    beq s1, a1, .LBB46_4
+; CHECK-NEXT:    beqz s1, .LBB46_4
 ; CHECK-NEXT:  .LBB46_8: # %entry
 ; CHECK-NEXT:    mv s0, a1
 ; CHECK-NEXT:    bne s1, a4, .LBB46_5
 ; CHECK-NEXT:    j .LBB46_6
 entry:
   %conv = fptoui <2 x double> %x to <2 x i128>
   %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
   %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>

 ; CHECK-NEXT:    beqz a3, .LBB49_2
 ; CHECK-NEXT:  # %bb.1: # %entry
 ; CHECK-NEXT:    mv a2, a1
 ; CHECK-NEXT:  .LBB49_2: # %entry
 ; CHECK-NEXT:    li a4, 1
 ; CHECK-NEXT:    mv a0, a1
 ; CHECK-NEXT:    bne a3, a4, .LBB49_7
 ; CHECK-NEXT:  # %bb.3: # %entry
-; CHECK-NEXT:    bne s1, a1, .LBB49_8
+; CHECK-NEXT:    bnez s1, .LBB49_8
 ; CHECK-NEXT:  .LBB49_4: # %entry
 ; CHECK-NEXT:    beq s1, a4, .LBB49_6
 ; CHECK-NEXT:  .LBB49_5: # %entry
 ; CHECK-NEXT:    mv a1, s0
 ; CHECK-NEXT:  .LBB49_6: # %entry
 ; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    fld fs0, 0(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    addi sp, sp, 32
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB49_7: # %entry
 ; CHECK-NEXT:    mv a0, a2
-; CHECK-NEXT:    beq s1, a1, .LBB49_4
+; CHECK-NEXT:    beqz s1, .LBB49_4
 ; CHECK-NEXT:  .LBB49_8: # %entry
 ; CHECK-NEXT:    mv s0, a1
 ; CHECK-NEXT:    bne s1, a4, .LBB49_5
 ; CHECK-NEXT:    j .LBB49_6
 entry:
   %conv = fptoui <2 x float> %x to <2 x i128>
   %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
   %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>

 ; CHECK-NEXT:    beqz a3, .LBB52_2
 ; CHECK-NEXT:  # %bb.1: # %entry
 ; CHECK-NEXT:    mv a2, a1
 ; CHECK-NEXT:  .LBB52_2: # %entry
 ; CHECK-NEXT:    li a4, 1
 ; CHECK-NEXT:    mv a0, a1
 ; CHECK-NEXT:    bne a3, a4, .LBB52_7
 ; CHECK-NEXT:  # %bb.3: # %entry
-; CHECK-NEXT:    bne s1, a1, .LBB52_8
+; CHECK-NEXT:    bnez s1, .LBB52_8
 ; CHECK-NEXT:  .LBB52_4: # %entry
 ; CHECK-NEXT:    beq s1, a4, .LBB52_6
 ; CHECK-NEXT:  .LBB52_5: # %entry
 ; CHECK-NEXT:    mv a1, s0
 ; CHECK-NEXT:  .LBB52_6: # %entry
 ; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    addi sp, sp, 32
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB52_7: # %entry
 ; CHECK-NEXT:    mv a0, a2
-; CHECK-NEXT:    beq s1, a1, .LBB52_4
+; CHECK-NEXT:    beqz s1, .LBB52_4
 ; CHECK-NEXT:  .LBB52_8: # %entry
 ; CHECK-NEXT:    mv s0, a1
 ; CHECK-NEXT:    bne s1, a4, .LBB52_5
 ; CHECK-NEXT:    j .LBB52_6
 entry:
   %conv = fptoui <2 x half> %x to <2 x i128>
   %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
   %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>

llvm/test/CodeGen/RISCV/half-convert.ll

 ; RV64IDZFH-NEXT:    fmax.s ft0, ft0, ft1
 ; RV64IDZFH-NEXT:    fmin.s ft0, ft0, ft2
 ; RV64IDZFH-NEXT:    fcvt.l.s a0, ft0, rtz
 ; RV64IDZFH-NEXT:  .LBB1_2: # %start
 ; RV64IDZFH-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_si_h_sat:
 ; RV32I:       # %bb.0: # %start
-; RV32I-NEXT:    addi sp, sp, -32
-; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    addi sp, sp, -16
+; RV32I-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32I-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    slli a0, a0, 16
 ; RV32I-NEXT:    srli a0, a0, 16
 ; RV32I-NEXT:    call __extendhfsf2@plt
 ; RV32I-NEXT:    mv s0, a0
 ; RV32I-NEXT:    lui a1, 815104
 ; RV32I-NEXT:    call __gesf2@plt
-; RV32I-NEXT:    mv s2, a0
+; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __fixsfsi@plt
-; RV32I-NEXT:    li s1, 0
-; RV32I-NEXT:    lui s3, 1048568
-; RV32I-NEXT:    bltz s2, .LBB1_2
+; RV32I-NEXT:    lui s2, 1048568
+; RV32I-NEXT:    bltz s1, .LBB1_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    mv s2, a0
 ; RV32I-NEXT:  .LBB1_2: # %start
 ; RV32I-NEXT:    lui a0, 290816
 ; RV32I-NEXT:    addi a1, a0, -512
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gtsf2@plt
-; RV32I-NEXT:    bge s1, a0, .LBB1_4
+; RV32I-NEXT:    blez a0, .LBB1_4
 ; RV32I-NEXT:  # %bb.3:
 ; RV32I-NEXT:    lui a0, 8
-; RV32I-NEXT:    addi s3, a0, -1
+; RV32I-NEXT:    addi s2, a0, -1
 ; RV32I-NEXT:  .LBB1_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s1, .LBB1_6
+; RV32I-NEXT:    mv a1, a0
+; RV32I-NEXT:    li a0, 0
+; RV32I-NEXT:    bnez a1, .LBB1_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s1, s3
+; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:  .LBB1_6: # %start
-; RV32I-NEXT:    mv a0, s1
-; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    addi sp, sp, 32
+; RV32I-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
+; RV32I-NEXT:    addi sp, sp, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_si_h_sat:
 ; RV64I:       # %bb.0: # %start
-; RV64I-NEXT:    addi sp, sp, -48
-; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    addi sp, sp, -32
+; RV64I-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    sd s2, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    slli a0, a0, 48
 ; RV64I-NEXT:    srli a0, a0, 48
 ; RV64I-NEXT:    call __extendhfsf2@plt
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 815104
 ; RV64I-NEXT:    call __gesf2@plt
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    lui s3, 1048568
-; RV64I-NEXT:    bltz s2, .LBB1_2
+; RV64I-NEXT:    lui s2, 1048568
+; RV64I-NEXT:    bltz s1, .LBB1_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB1_2: # %start
 ; RV64I-NEXT:    lui a0, 290816
 ; RV64I-NEXT:    addiw a1, a0, -512
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB1_4
+; RV64I-NEXT:    blez a0, .LBB1_4
 ; RV64I-NEXT:  # %bb.3:
 ; RV64I-NEXT:    lui a0, 8
-; RV64I-NEXT:    addiw s3, a0, -1
+; RV64I-NEXT:    addiw s2, a0, -1
 ; RV64I-NEXT:  .LBB1_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB1_6
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB1_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv a0, s2
 ; RV64I-NEXT:  .LBB1_6: # %start
-; RV64I-NEXT:    mv a0, s1
-; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    addi sp, sp, 48
+; RV64I-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    ld s2, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 32
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i16 @llvm.fptosi.sat.i16.f16(half %a)
   ret i16 %0
 }
 declare i16 @llvm.fptosi.sat.i16.f16(half)
 
 define i16 @fcvt_ui_h(half %a) nounwind {

 ; RV32I-LABEL: fcvt_w_h_sat:
 ; RV32I:       # %bb.0: # %start
 ; RV32I-NEXT:    addi sp, sp, -32
 ; RV32I-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s0, 24(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s1, 20(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s2, 16(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    sw s3, 12(sp) # 4-byte Folded Spill
-; RV32I-NEXT:    sw s4, 8(sp) # 4-byte Folded Spill
 ; RV32I-NEXT:    slli a0, a0, 16
 ; RV32I-NEXT:    srli a0, a0, 16
 ; RV32I-NEXT:    call __extendhfsf2@plt
 ; RV32I-NEXT:    mv s0, a0
 ; RV32I-NEXT:    lui a1, 847872
 ; RV32I-NEXT:    call __gesf2@plt
-; RV32I-NEXT:    mv s2, a0
+; RV32I-NEXT:    mv s1, a0
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __fixsfsi@plt
-; RV32I-NEXT:    li s1, 0
-; RV32I-NEXT:    lui s4, 524288
 ; RV32I-NEXT:    lui s3, 524288
-; RV32I-NEXT:    bltz s2, .LBB5_2
+; RV32I-NEXT:    lui s2, 524288
+; RV32I-NEXT:    bltz s1, .LBB5_2
 ; RV32I-NEXT:  # %bb.1: # %start
-; RV32I-NEXT:    mv s3, a0
+; RV32I-NEXT:    mv s2, a0
 ; RV32I-NEXT:  .LBB5_2: # %start
 ; RV32I-NEXT:    lui a0, 323584
 ; RV32I-NEXT:    addi a1, a0, -1
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    call __gtsf2@plt
-; RV32I-NEXT:    bge s1, a0, .LBB5_4
+; RV32I-NEXT:    blez a0, .LBB5_4
 ; RV32I-NEXT:  # %bb.3:
-; RV32I-NEXT:    addi s3, s4, -1
+; RV32I-NEXT:    addi s2, s3, -1
 ; RV32I-NEXT:  .LBB5_4: # %start
 ; RV32I-NEXT:    mv a0, s0
 ; RV32I-NEXT:    mv a1, s0
 ; RV32I-NEXT:    call __unordsf2@plt
-; RV32I-NEXT:    bne a0, s1, .LBB5_6
+; RV32I-NEXT:    mv a1, a0
+; RV32I-NEXT:    li a0, 0
+; RV32I-NEXT:    bnez a1, .LBB5_6
 ; RV32I-NEXT:  # %bb.5: # %start
-; RV32I-NEXT:    mv s1, s3
+; RV32I-NEXT:    mv a0, s2
 ; RV32I-NEXT:  .LBB5_6: # %start
-; RV32I-NEXT:    mv a0, s1
 ; RV32I-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s0, 24(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s1, 20(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s2, 16(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    lw s3, 12(sp) # 4-byte Folded Reload
-; RV32I-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32I-NEXT:    addi sp, sp, 32
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: fcvt_w_h_sat:
 ; RV64I:       # %bb.0: # %start
 ; RV64I-NEXT:    addi sp, sp, -48
 ; RV64I-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
-; RV64I-NEXT:    sd s4, 0(sp) # 8-byte Folded Spill
 ; RV64I-NEXT:    slli a0, a0, 48
 ; RV64I-NEXT:    srli a0, a0, 48
 ; RV64I-NEXT:    call __extendhfsf2@plt
 ; RV64I-NEXT:    mv s0, a0
 ; RV64I-NEXT:    lui a1, 847872
 ; RV64I-NEXT:    call __gesf2@plt
-; RV64I-NEXT:    mv s2, a0
+; RV64I-NEXT:    mv s1, a0
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __fixsfdi@plt
-; RV64I-NEXT:    li s1, 0
-; RV64I-NEXT:    lui s4, 524288
 ; RV64I-NEXT:    lui s3, 524288
-; RV64I-NEXT:    bltz s2, .LBB5_2
+; RV64I-NEXT:    lui s2, 524288
+; RV64I-NEXT:    bltz s1, .LBB5_2
 ; RV64I-NEXT:  # %bb.1: # %start
-; RV64I-NEXT:    mv s3, a0
+; RV64I-NEXT:    mv s2, a0
 ; RV64I-NEXT:  .LBB5_2: # %start
 ; RV64I-NEXT:    lui a0, 323584
 ; RV64I-NEXT:    addiw a1, a0, -1
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    call __gtsf2@plt
-; RV64I-NEXT:    bge s1, a0, .LBB5_4
+; RV64I-NEXT:    blez a0, .LBB5_4
 ; RV64I-NEXT:  # %bb.3:
-; RV64I-NEXT:    addiw s3, s4, -1
+; RV64I-NEXT:    addiw s2, s3, -1
 ; RV64I-NEXT:  .LBB5_4: # %start
 ; RV64I-NEXT:    mv a0, s0
 ; RV64I-NEXT:    mv a1, s0
 ; RV64I-NEXT:    call __unordsf2@plt
-; RV64I-NEXT:    bne a0, s1, .LBB5_6
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a0, 0
+; RV64I-NEXT:    bnez a1, .LBB5_6
 ; RV64I-NEXT:  # %bb.5: # %start
-; RV64I-NEXT:    mv s1, s3
+; RV64I-NEXT:    mv a0, s2
 ; RV64I-NEXT:  .LBB5_6: # %start
-; RV64I-NEXT:    mv a0, s1
 ; RV64I-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s2, 16(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    ld s3, 8(sp) # 8-byte Folded Reload
-; RV64I-NEXT:    ld s4, 0(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 48
 ; RV64I-NEXT:    ret
 start:
   %0 = tail call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %0
 }
 declare i32 @llvm.fptosi.sat.i32.f16(half)
 

 ; RV32I-NEXT:    blt s1, a0, .LBB10_4
 ; RV32I-NEXT:  # %bb.3: # %start
 ; RV32I-NEXT:    mv s6, s5