[RISCV] Enable TLSDESC by default on RISC-V

[paulkirth]

Summary

This patch enables RISC-V TLSDESC by default and updates llvm tests.

Depends on D151462

[paulkirth]

Rebase

[paulkirth]

Update test to match psABI draft.

[paulkirth]

Rebase

llvm/lib/Target/RISCV/RISCVTargetMachine.cpp

 
 static cl::opt<bool> EnableRISCVCopyPropagation(
     "riscv-enable-copy-propagation",
     cl::desc("Enable the copy propagation with RISC-V copy instr"),
     cl::init(true), cl::Hidden);
 
 cl::opt<bool> EnableRISCVTLSDESC("riscv-enable-tlsdesc",
                                  cl::desc("Enable the tlsdesc for RISC-V"),
-                                 cl::Hidden);
+                                 cl::init(true), cl::Hidden);
 
 extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeRISCVTarget() {
   RegisterTargetMachine<RISCVTargetMachine> X(getTheRISCV32Target());
   RegisterTargetMachine<RISCVTargetMachine> Y(getTheRISCV64Target());
   auto *PR = PassRegistry::getPassRegistry();
   initializeGlobalISel(*PR);
   initializeRISCVMakeCompressibleOptPass(*PR);
   initializeRISCVGatherScatterLoweringPass(*PR);

llvm/test/CodeGen/RISCV/O3-pipeline.ll

 ; CHECK-NEXT:       Function Alias Analysis Results
 ; CHECK-NEXT:       Natural Loop Information
 ; CHECK-NEXT:       Post-Dominator Tree Construction
 ; CHECK-NEXT:       Branch Probability Analysis
 ; CHECK-NEXT:       Assignment Tracking Analysis
 ; CHECK-NEXT:       Lazy Branch Probability Analysis
 ; CHECK-NEXT:       Lazy Block Frequency Analysis
 ; CHECK-NEXT:       RISC-V DAG->DAG Pattern Instruction Selection
+; CHECK-NEXT:       MachineDominator Tree Construction
+; CHECK-NEXT:       RISCV Local Dynamic TLS Access Clean-up 
 ; CHECK-NEXT:       Finalize ISel and expand pseudo-instructions
 ; CHECK-NEXT:       Lazy Machine Block Frequency Analysis
 ; CHECK-NEXT:       Early Tail Duplication
 ; CHECK-NEXT:       Optimize machine instruction PHIs
 ; CHECK-NEXT:       Slot index numbering
 ; CHECK-NEXT:       Merge disjoint stack slots
 ; CHECK-NEXT:       Local Stack Slot Allocation
 ; CHECK-NEXT:       Remove dead machine instructions

llvm/test/CodeGen/RISCV/machinelicm-address-pseudos.ll

ret:
   ret void
 }
 
 @gd = external thread_local global i32
 
 define void @test_la_tls_gd(i32 signext %n) nounwind {
 ; RV32I-LABEL: test_la_tls_gd:
 ; RV32I:       # %bb.0: # %entry
-; 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:    li s2, 0
-; RV32I-NEXT:  .Lpcrel_hi3:
-; RV32I-NEXT:    auipc a0, %tls_gd_pcrel_hi(gd)
-; RV32I-NEXT:    addi s1, a0, %pcrel_lo(.Lpcrel_hi3)
+; RV32I-NEXT:    mv a1, a0
+; RV32I-NEXT:    li a2, 0
 ; RV32I-NEXT:  .LBB3_1: # %loop
 ; RV32I-NEXT:    # =>This Inner Loop Header: Depth=1
-; RV32I-NEXT:    mv a0, s1
-; RV32I-NEXT:    call __tls_get_addr@plt
+; RV32I-NEXT:  .Ltlsdesc_hi0:
+; RV32I-NEXT:    auipc a0, %tlsdesc_hi(gd)
+; RV32I-NEXT:    lw a3, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
+; RV32I-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
+; RV32I-NEXT:    jalr t0, 0(a3), %tlsdesc_call(.Ltlsdesc_hi0)
+; RV32I-NEXT:    add a0, a0, tp
 ; RV32I-NEXT:    lw a0, 0(a0)
-; RV32I-NEXT:    addi s2, s2, 1
-; RV32I-NEXT:    blt s2, s0, .LBB3_1
+; RV32I-NEXT:    addi a2, a2, 1
+; RV32I-NEXT:    blt a2, a1, .LBB3_1
 ; RV32I-NEXT:  # %bb.2: # %ret
-; 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: test_la_tls_gd:
 ; RV64I:       # %bb.0: # %entry
-; 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 s2, 0
-; RV64I-NEXT:  .Lpcrel_hi3:
-; RV64I-NEXT:    auipc a0, %tls_gd_pcrel_hi(gd)
-; RV64I-NEXT:    addi s1, a0, %pcrel_lo(.Lpcrel_hi3)
+; RV64I-NEXT:    mv a1, a0
+; RV64I-NEXT:    li a2, 0
 ; RV64I-NEXT:  .LBB3_1: # %loop
 ; RV64I-NEXT:    # =>This Inner Loop Header: Depth=1
-; RV64I-NEXT:    mv a0, s1
-; RV64I-NEXT:    call __tls_get_addr@plt
+; RV64I-NEXT:  .Ltlsdesc_hi0:
+; RV64I-NEXT:    auipc a0, %tlsdesc_hi(gd)
+; RV64I-NEXT:    ld a3, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
+; RV64I-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
+; RV64I-NEXT:    jalr t0, 0(a3), %tlsdesc_call(.Ltlsdesc_hi0)
+; RV64I-NEXT:    add a0, a0, tp
 ; RV64I-NEXT:    lw a0, 0(a0)
-; RV64I-NEXT:    addiw s2, s2, 1
-; RV64I-NEXT:    blt s2, s0, .LBB3_1
+; RV64I-NEXT:    addiw a2, a2, 1
+; RV64I-NEXT:    blt a2, a1, .LBB3_1
 ; RV64I-NEXT:  # %bb.2: # %ret
-; 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
 entry:
   br label %loop
 
 loop:
   %i = phi i32 [ %inc, %loop ], [ 0, %entry ]
   %0 = load volatile i32, ptr @gd, align 4
   %inc = add nuw nsw i32 %i, 1
   %cmp = icmp slt i32 %inc, %n
   br i1 %cmp, label %loop, label %ret
 
 ret:
   ret void
 }

llvm/test/CodeGen/RISCV/mir-target-flags.ll

 ; RV32-SMALL-NEXT: target-flags(riscv-tprel-lo) @t_le
 ; RV32-SMALL:      target-flags(riscv-plt) @callee
 ;
 ; RV32-MED-LABEL: name: caller
 ; RV32-MED:      target-flags(riscv-got-hi) @g_e
 ; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi0>
 ; RV32-MED:      target-flags(riscv-pcrel-hi) @g_i
 ; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi1>
-; RV32-MED:      target-flags(riscv-tls-gd-hi) @t_un
-; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi2>
-; RV32-MED:      target-flags(riscv-plt) &__tls_get_addr
-; RV32-MED:      target-flags(riscv-tls-gd-hi) @t_ld
-; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi3>
-; RV32-MED:      target-flags(riscv-plt) &__tls_get_addr
+; RV32-MED:      target-flags(riscv-tlsdesc-hi) @t_un
+; RV32-MED-NEXT: target-flags(riscv-tlsdesc-load-lo) <mcsymbol .Ltlsdesc_hi0>
+; RV32-MED-NEXT: target-flags(riscv-tlsdesc-add-lo) <mcsymbol .Ltlsdesc_hi0>
+; RV32-MED:      target-flags(riscv-tlsdesc-call) <mcsymbol .Ltlsdesc_hi0>
+; RV32-MED:      target-flags(riscv-tlsdesc-hi) &_TLS_MODULE_BASE_
+; RV32-MED-NEXT: target-flags(riscv-tlsdesc-load-lo) <mcsymbol .Ltlsdesc_hi1>
+; RV32-MED:      target-flags(riscv-tlsdesc-add-lo) <mcsymbol .Ltlsdesc_hi1>
+; RV32-MED:      target-flags(riscv-tlsdesc-call) <mcsymbol .Ltlsdesc_hi1>
 ; RV32-MED:      target-flags(riscv-tls-got-hi) @t_ie
-; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi4>
+; RV32-MED-NEXT: target-flags(riscv-pcrel-lo) <mcsymbol .Lpcrel_hi2>
 ; RV32-MED:      target-flags(riscv-tprel-hi) @t_le
 ; RV32-MED-NEXT: target-flags(riscv-tprel-add) @t_le
 ; RV32-MED-NEXT: target-flags(riscv-tprel-lo) @t_le
 ; RV32-MED:      target-flags(riscv-plt) @callee
 ;
   %b = load i32, ptr @g_e
   %c = load i32, ptr @g_i
   %d = load i32, ptr @t_un

llvm/test/CodeGen/RISCV/tls-models.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -relocation-model=pic < %s \
 ; RUN:     | FileCheck -check-prefix=RV32-PIC %s
 ; RUN: llc -mtriple=riscv64 -relocation-model=pic < %s \
 ; RUN:     | FileCheck -check-prefix=RV64-PIC %s
 ; RUN: llc -mtriple=riscv32 < %s | FileCheck -check-prefix=RV32-NOPIC %s
 ; RUN: llc -mtriple=riscv64 < %s | FileCheck -check-prefix=RV64-NOPIC %s
 
-; RUN: llc -mtriple=riscv32 -relocation-model=pic -riscv-enable-tlsdesc < %s \
-; RUN:     | FileCheck -check-prefix=RV32-PIC-TLSDESC %s
-; RUN: llc -mtriple=riscv64 -relocation-model=pic -riscv-enable-tlsdesc < %s \
-; RUN:     | FileCheck -check-prefix=RV64-PIC-TLSDESC %s
-; RUN: llc -mtriple=riscv32 < %s  -riscv-enable-tlsdesc | FileCheck -check-prefix=RV32-NOPIC-TLSDESC %s
-; RUN: llc -mtriple=riscv64 < %s  -riscv-enable-tlsdesc | FileCheck -check-prefix=RV64-NOPIC-TLSDESC %s
-
-
 ; Check that TLS symbols are lowered correctly based on the specified
 ; model. Make sure they're external to avoid them all being optimised to Local
 ; Exec for the executable.
 
 @unspecified = external thread_local global i32
 @ld = external thread_local(localdynamic) global i32
 @ld2 = external thread_local(localdynamic) global i32
 @ld3 = external thread_local(localdynamic) global i32
 @ie = external thread_local(initialexec) global i32
 @le = external thread_local(localexec) global i32
 
 
 ; No model specified
 
 define ptr @f1() nounwind {
 ; RV32-PIC-LABEL: f1:
 ; RV32-PIC:       # %bb.0: # %entry
-; RV32-PIC-NEXT:    addi sp, sp, -16
-; RV32-PIC-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32-PIC-NEXT:  .Lpcrel_hi0:
-; RV32-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(unspecified)
-; RV32-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi0)
-; RV32-PIC-NEXT:    call __tls_get_addr@plt
-; RV32-PIC-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
-; RV32-PIC-NEXT:    addi sp, sp, 16
+; RV32-PIC-NEXT:  .Ltlsdesc_hi0:
+; RV32-PIC-NEXT:    auipc a0, %tlsdesc_hi(unspecified)
+; RV32-PIC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
+; RV32-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
+; RV32-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi0)
+; RV32-PIC-NEXT:    add a0, a0, tp
 ; RV32-PIC-NEXT:    ret
 ;
 ; RV64-PIC-LABEL: f1:
 ; RV64-PIC:       # %bb.0: # %entry
-; RV64-PIC-NEXT:    addi sp, sp, -16
-; RV64-PIC-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
-; RV64-PIC-NEXT:  .Lpcrel_hi0:
-; RV64-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(unspecified)
-; RV64-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi0)
-; RV64-PIC-NEXT:    call __tls_get_addr@plt
-; RV64-PIC-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
-; RV64-PIC-NEXT:    addi sp, sp, 16
+; RV64-PIC-NEXT:  .Ltlsdesc_hi0:
+; RV64-PIC-NEXT:    auipc a0, %tlsdesc_hi(unspecified)
+; RV64-PIC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
+; RV64-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
+; RV64-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi0)
+; RV64-PIC-NEXT:    add a0, a0, tp
 ; RV64-PIC-NEXT:    ret
 ;
 ; RV32-NOPIC-LABEL: f1:
 ; RV32-NOPIC:       # %bb.0: # %entry
 ; RV32-NOPIC-NEXT:  .Lpcrel_hi0:
 ; RV32-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(unspecified)
 ; RV32-NOPIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV32-NOPIC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-NEXT:    ret
 ;
 ; RV64-NOPIC-LABEL: f1:
 ; RV64-NOPIC:       # %bb.0: # %entry
 ; RV64-NOPIC-NEXT:  .Lpcrel_hi0:
 ; RV64-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(unspecified)
 ; RV64-NOPIC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV64-NOPIC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-NEXT:    ret
-;
 ; RV32-PIC-TLSDESC-LABEL: f1:
 ; RV32-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi0:
 ; RV32-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(unspecified)
 ; RV32-PIC-TLSDESC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
 ; RV32-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
 ; RV32-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi0)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-PIC-TLSDESC-NEXT:    ret
-;
 ; RV64-PIC-TLSDESC-LABEL: f1:
 ; RV64-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi0:
 ; RV64-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(unspecified)
 ; RV64-PIC-TLSDESC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi0)(a0)
 ; RV64-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi0)
 ; RV64-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi0)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-PIC-TLSDESC-NEXT:    ret
-;
 ; RV32-NOPIC-TLSDESC-LABEL: f1:
 ; RV32-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi0:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(unspecified)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    ret
-;
 ; RV64-NOPIC-TLSDESC-LABEL: f1:
 ; RV64-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi0:
 ; RV64-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(unspecified)
 ; RV64-NOPIC-TLSDESC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV64-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-TLSDESC-NEXT:    ret
 entry:
   ret ptr @unspecified
 }
 
 
 ; localdynamic specified
 
 define ptr @f2() nounwind {
 ; RV32-PIC-LABEL: f2:
 ; RV32-PIC:       # %bb.0: # %entry
-; RV32-PIC-NEXT:    addi sp, sp, -16
-; RV32-PIC-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32-PIC-NEXT:  .Lpcrel_hi1:
-; RV32-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld)
-; RV32-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi1)
-; RV32-PIC-NEXT:    call __tls_get_addr@plt
-; RV32-PIC-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
-; RV32-PIC-NEXT:    addi sp, sp, 16
+; RV32-PIC-NEXT:  .Ltlsdesc_hi1:
+; RV32-PIC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
+; RV32-PIC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi1)(a0)
+; RV32-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi1)
+; RV32-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi1)
+; RV32-PIC-NEXT:    add a0, a0, tp
+; RV32-PIC-NEXT:    lui a1, %tprel_hi(ld)
+; RV32-PIC-NEXT:    add a0, a1, a0
+; RV32-PIC-NEXT:    addi a0, a0, %tprel_lo(ld)
 ; RV32-PIC-NEXT:    ret
 ;
 ; RV64-PIC-LABEL: f2:
 ; RV64-PIC:       # %bb.0: # %entry
-; RV64-PIC-NEXT:    addi sp, sp, -16
-; RV64-PIC-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
-; RV64-PIC-NEXT:  .Lpcrel_hi1:
-; RV64-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld)
-; RV64-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi1)
-; RV64-PIC-NEXT:    call __tls_get_addr@plt
-; RV64-PIC-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
-; RV64-PIC-NEXT:    addi sp, sp, 16
+; RV64-PIC-NEXT:  .Ltlsdesc_hi1:
+; RV64-PIC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
+; RV64-PIC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi1)(a0)
+; RV64-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi1)
+; RV64-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi1)
+; RV64-PIC-NEXT:    add a0, a0, tp
+; RV64-PIC-NEXT:    lui a1, %tprel_hi(ld)
+; RV64-PIC-NEXT:    add a0, a1, a0
+; RV64-PIC-NEXT:    addi a0, a0, %tprel_lo(ld)
 ; RV64-PIC-NEXT:    ret
 ;
 ; RV32-NOPIC-LABEL: f2:
 ; RV32-NOPIC:       # %bb.0: # %entry
 ; RV32-NOPIC-NEXT:  .Lpcrel_hi1:
 ; RV32-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV32-NOPIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi1)(a0)
 ; RV32-NOPIC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-NEXT:    ret
 ;
 ; RV64-NOPIC-LABEL: f2:
 ; RV64-NOPIC:       # %bb.0: # %entry
 ; RV64-NOPIC-NEXT:  .Lpcrel_hi1:
 ; RV64-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV64-NOPIC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi1)(a0)
 ; RV64-NOPIC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-NEXT:    ret
-;
 ; RV32-PIC-TLSDESC-LABEL: f2:
 ; RV32-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi1:
 ; RV32-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
 ; RV32-PIC-TLSDESC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi1)(a0)
 ; RV32-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi1)
 ; RV32-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi1)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a1, a0
 ; RV32-PIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(ld)
 ; RV32-PIC-TLSDESC-NEXT:    ret
-;
 ; RV64-PIC-TLSDESC-LABEL: f2:
 ; RV64-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi1:
 ; RV64-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
 ; RV64-PIC-TLSDESC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi1)(a0)
 ; RV64-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi1)
 ; RV64-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi1)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a1, a0
 ; RV64-PIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(ld)
 ; RV64-PIC-TLSDESC-NEXT:    ret
-;
 ; RV32-NOPIC-TLSDESC-LABEL: f2:
 ; RV32-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi1:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi1)(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    ret
-;
 ; RV64-NOPIC-TLSDESC-LABEL: f2:
 ; RV64-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi1:
 ; RV64-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV64-NOPIC-TLSDESC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi1)(a0)
 ; RV64-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-TLSDESC-NEXT:    ret
 entry:
   ret ptr @ld
 }
 
 
 ; initialexec specified
 
 define ptr @f3() nounwind {
 ; RV32-PIC-LABEL: f3:
 ; RV32-PIC:       # %bb.0: # %entry
-; RV32-PIC-NEXT:  .Lpcrel_hi2:
+; RV32-PIC-NEXT:  .Lpcrel_hi0:
 ; RV32-PIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
-; RV32-PIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi2)(a0)
+; RV32-PIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV32-PIC-NEXT:    add a0, a0, tp
 ; RV32-PIC-NEXT:    ret
 ;
 ; RV64-PIC-LABEL: f3:
 ; RV64-PIC:       # %bb.0: # %entry
-; RV64-PIC-NEXT:  .Lpcrel_hi2:
+; RV64-PIC-NEXT:  .Lpcrel_hi0:
 ; RV64-PIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
-; RV64-PIC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi2)(a0)
+; RV64-PIC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV64-PIC-NEXT:    add a0, a0, tp
 ; RV64-PIC-NEXT:    ret
 ;
 ; RV32-NOPIC-LABEL: f3:
 ; RV32-NOPIC:       # %bb.0: # %entry
 ; RV32-NOPIC-NEXT:  .Lpcrel_hi2:
 ; RV32-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV32-NOPIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi2)(a0)
 ; RV32-NOPIC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-NEXT:    ret
 ;
 ; RV64-NOPIC-LABEL: f3:
 ; RV64-NOPIC:       # %bb.0: # %entry
 ; RV64-NOPIC-NEXT:  .Lpcrel_hi2:
 ; RV64-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV64-NOPIC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi2)(a0)
 ; RV64-NOPIC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-NEXT:    ret
-;
 ; RV32-PIC-TLSDESC-LABEL: f3:
 ; RV32-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-PIC-TLSDESC-NEXT:  .Lpcrel_hi0:
 ; RV32-PIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV32-PIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-PIC-TLSDESC-NEXT:    ret
-;
 ; RV64-PIC-TLSDESC-LABEL: f3:
 ; RV64-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-PIC-TLSDESC-NEXT:  .Lpcrel_hi0:
 ; RV64-PIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV64-PIC-TLSDESC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi0)(a0)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-PIC-TLSDESC-NEXT:    ret
-;
 ; RV32-NOPIC-TLSDESC-LABEL: f3:
 ; RV32-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi2:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi2)(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    ret
-;
 ; RV64-NOPIC-TLSDESC-LABEL: f3:
 ; RV64-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi2:
 ; RV64-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ie)
 ; RV64-NOPIC-TLSDESC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi2)(a0)
 ; RV64-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-TLSDESC-NEXT:    ret
 entry:

 ; RV32-NOPIC-NEXT:    ret
 ;
 ; RV64-NOPIC-LABEL: f4:
 ; RV64-NOPIC:       # %bb.0: # %entry
 ; RV64-NOPIC-NEXT:    lui a0, %tprel_hi(le)
 ; RV64-NOPIC-NEXT:    add a0, a0, tp, %tprel_add(le)
 ; RV64-NOPIC-NEXT:    addi a0, a0, %tprel_lo(le)
 ; RV64-NOPIC-NEXT:    ret
-;
 ; RV32-PIC-TLSDESC-LABEL: f4:
 ; RV32-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-PIC-TLSDESC-NEXT:    lui a0, %tprel_hi(le)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a0, tp, %tprel_add(le)
 ; RV32-PIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(le)
 ; RV32-PIC-TLSDESC-NEXT:    ret
-;
 ; RV64-PIC-TLSDESC-LABEL: f4:
 ; RV64-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-PIC-TLSDESC-NEXT:    lui a0, %tprel_hi(le)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a0, tp, %tprel_add(le)
 ; RV64-PIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(le)
 ; RV64-PIC-TLSDESC-NEXT:    ret
-;
 ; RV32-NOPIC-TLSDESC-LABEL: f4:
 ; RV32-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-NOPIC-TLSDESC-NEXT:    lui a0, %tprel_hi(le)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp, %tprel_add(le)
 ; RV32-NOPIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(le)
 ; RV32-NOPIC-TLSDESC-NEXT:    ret
-;
 ; RV64-NOPIC-TLSDESC-LABEL: f4:
 ; RV64-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-NOPIC-TLSDESC-NEXT:    lui a0, %tprel_hi(le)
 ; RV64-NOPIC-TLSDESC-NEXT:    add a0, a0, tp, %tprel_add(le)
 ; RV64-NOPIC-TLSDESC-NEXT:    addi a0, a0, %tprel_lo(le)
 ; RV64-NOPIC-TLSDESC-NEXT:    ret
 entry:
   ret ptr @le
 }
 
 ; localdynamic specified
 
 define i32 @dedup_localdynamic() nounwind {
 ; RV32-PIC-LABEL: dedup_localdynamic:
 ; RV32-PIC:       # %bb.0: # %entry
-; RV32-PIC-NEXT:    addi sp, sp, -16
-; RV32-PIC-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32-PIC-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32-PIC-NEXT:  .Lpcrel_hi3:
-; RV32-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld)
-; RV32-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi3)
-; RV32-PIC-NEXT:    call __tls_get_addr@plt
-; RV32-PIC-NEXT:    lw s0, 0(a0)
-; RV32-PIC-NEXT:    bnez s0, .LBB4_2
+; RV32-PIC-NEXT:  .Ltlsdesc_hi2:
+; RV32-PIC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
+; RV32-PIC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi2)(a0)
+; RV32-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi2)
+; RV32-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi2)
+; RV32-PIC-NEXT:    add a0, a0, tp
+; RV32-PIC-NEXT:    lui a1, %tprel_hi(ld)
+; RV32-PIC-NEXT:    add a1, a1, a0
+; RV32-PIC-NEXT:    lw a1, %tprel_lo(ld)(a1)
+; RV32-PIC-NEXT:    bnez a1, .LBB4_2
 ; RV32-PIC-NEXT:  # %bb.1:
-; RV32-PIC-NEXT:  .Lpcrel_hi4:
-; RV32-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld2)
-; RV32-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi4)
-; RV32-PIC-NEXT:    call __tls_get_addr@plt
-; RV32-PIC-NEXT:    lw s0, 0(a0)
+; RV32-PIC-NEXT:    lui a1, %tprel_hi(ld2)
+; RV32-PIC-NEXT:    add a1, a1, a0
+; RV32-PIC-NEXT:    lw a1, %tprel_lo(ld2)(a1)
 ; RV32-PIC-NEXT:  .LBB4_2:
-; RV32-PIC-NEXT:  .Lpcrel_hi5:
-; RV32-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld3)
-; RV32-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi5)
-; RV32-PIC-NEXT:    call __tls_get_addr@plt
-; RV32-PIC-NEXT:    lw a0, 0(a0)
-; RV32-PIC-NEXT:    add a0, s0, a0
-; RV32-PIC-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
-; RV32-PIC-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
-; RV32-PIC-NEXT:    addi sp, sp, 16
+; RV32-PIC-NEXT:    lui a2, %tprel_hi(ld3)
+; RV32-PIC-NEXT:    add a0, a2, a0
+; RV32-PIC-NEXT:    lw a0, %tprel_lo(ld3)(a0)
+; RV32-PIC-NEXT:    add a0, a1, a0
 ; RV32-PIC-NEXT:    ret
 ;
 ; RV64-PIC-LABEL: dedup_localdynamic:
 ; RV64-PIC:       # %bb.0: # %entry
-; RV64-PIC-NEXT:    addi sp, sp, -16
-; RV64-PIC-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
-; RV64-PIC-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
-; RV64-PIC-NEXT:  .Lpcrel_hi3:
-; RV64-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld)
-; RV64-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi3)
-; RV64-PIC-NEXT:    call __tls_get_addr@plt
-; RV64-PIC-NEXT:    lw s0, 0(a0)
-; RV64-PIC-NEXT:    bnez s0, .LBB4_2
+; RV64-PIC-NEXT:  .Ltlsdesc_hi2:
+; RV64-PIC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
+; RV64-PIC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi2)(a0)
+; RV64-PIC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi2)
+; RV64-PIC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi2)
+; RV64-PIC-NEXT:    add a0, a0, tp
+; RV64-PIC-NEXT:    lui a1, %tprel_hi(ld)
+; RV64-PIC-NEXT:    add a1, a1, a0
+; RV64-PIC-NEXT:    lw a1, %tprel_lo(ld)(a1)
+; RV64-PIC-NEXT:    bnez a1, .LBB4_2
 ; RV64-PIC-NEXT:  # %bb.1:
-; RV64-PIC-NEXT:  .Lpcrel_hi4:
-; RV64-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld2)
-; RV64-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi4)
-; RV64-PIC-NEXT:    call __tls_get_addr@plt
-; RV64-PIC-NEXT:    lw s0, 0(a0)
+; RV64-PIC-NEXT:    lui a1, %tprel_hi(ld2)
+; RV64-PIC-NEXT:    add a1, a1, a0
+; RV64-PIC-NEXT:    lw a1, %tprel_lo(ld2)(a1)
 ; RV64-PIC-NEXT:  .LBB4_2:
-; RV64-PIC-NEXT:  .Lpcrel_hi5:
-; RV64-PIC-NEXT:    auipc a0, %tls_gd_pcrel_hi(ld3)
-; RV64-PIC-NEXT:    addi a0, a0, %pcrel_lo(.Lpcrel_hi5)
-; RV64-PIC-NEXT:    call __tls_get_addr@plt
-; RV64-PIC-NEXT:    lw a0, 0(a0)
-; RV64-PIC-NEXT:    addw a0, s0, a0
-; RV64-PIC-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
-; RV64-PIC-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
-; RV64-PIC-NEXT:    addi sp, sp, 16
+; RV64-PIC-NEXT:    lui a2, %tprel_hi(ld3)
+; RV64-PIC-NEXT:    add a0, a2, a0
+; RV64-PIC-NEXT:    lw a0, %tprel_lo(ld3)(a0)
+; RV64-PIC-NEXT:    addw a0, a1, a0
 ; RV64-PIC-NEXT:    ret
 ;
 ; RV32-NOPIC-LABEL: dedup_localdynamic:
 ; RV32-NOPIC:       # %bb.0: # %entry
 ; RV32-NOPIC-NEXT:  .Lpcrel_hi3:
 ; RV32-NOPIC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV32-NOPIC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi3)(a0)
 ; RV32-NOPIC-NEXT:    add a0, a0, tp

 ; RV64-NOPIC-NEXT:  .LBB4_2:
 ; RV64-NOPIC-NEXT:  .Lpcrel_hi5:
 ; RV64-NOPIC-NEXT:    auipc a1, %tls_ie_pcrel_hi(ld3)
 ; RV64-NOPIC-NEXT:    ld a1, %pcrel_lo(.Lpcrel_hi5)(a1)
 ; RV64-NOPIC-NEXT:    add a1, a1, tp
 ; RV64-NOPIC-NEXT:    lw a1, 0(a1)
 ; RV64-NOPIC-NEXT:    addw a0, a0, a1
 ; RV64-NOPIC-NEXT:    ret
-;
 ; RV32-PIC-TLSDESC-LABEL: dedup_localdynamic:
 ; RV32-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi2:
 ; RV32-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
 ; RV32-PIC-TLSDESC-NEXT:    lw a1, %tlsdesc_load_lo(.Ltlsdesc_hi2)(a0)
 ; RV32-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi2)
 ; RV32-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi2)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld)
 ; RV32-PIC-TLSDESC-NEXT:    add a1, a1, a0
 ; RV32-PIC-TLSDESC-NEXT:    lw a1, %tprel_lo(ld)(a1)
 ; RV32-PIC-TLSDESC-NEXT:    bnez a1, .LBB4_2
 ; RV32-PIC-TLSDESC-NEXT:  # %bb.1:
 ; RV32-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld2)
 ; RV32-PIC-TLSDESC-NEXT:    add a1, a1, a0
 ; RV32-PIC-TLSDESC-NEXT:    lw a1, %tprel_lo(ld2)(a1)
 ; RV32-PIC-TLSDESC-NEXT:  .LBB4_2:
 ; RV32-PIC-TLSDESC-NEXT:    lui a2, %tprel_hi(ld3)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a2, a0
 ; RV32-PIC-TLSDESC-NEXT:    lw a0, %tprel_lo(ld3)(a0)
 ; RV32-PIC-TLSDESC-NEXT:    add a0, a1, a0
 ; RV32-PIC-TLSDESC-NEXT:    ret
-;
 ; RV64-PIC-TLSDESC-LABEL: dedup_localdynamic:
 ; RV64-PIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-PIC-TLSDESC-NEXT:  .Ltlsdesc_hi2:
 ; RV64-PIC-TLSDESC-NEXT:    auipc a0, %tlsdesc_hi(_TLS_MODULE_BASE_)
 ; RV64-PIC-TLSDESC-NEXT:    ld a1, %tlsdesc_load_lo(.Ltlsdesc_hi2)(a0)
 ; RV64-PIC-TLSDESC-NEXT:    addi a0, a0, %tlsdesc_add_lo(.Ltlsdesc_hi2)
 ; RV64-PIC-TLSDESC-NEXT:    jalr t0, 0(a1), %tlsdesc_call(.Ltlsdesc_hi2)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld)
 ; RV64-PIC-TLSDESC-NEXT:    add a1, a1, a0
 ; RV64-PIC-TLSDESC-NEXT:    lw a1, %tprel_lo(ld)(a1)
 ; RV64-PIC-TLSDESC-NEXT:    bnez a1, .LBB4_2
 ; RV64-PIC-TLSDESC-NEXT:  # %bb.1:
 ; RV64-PIC-TLSDESC-NEXT:    lui a1, %tprel_hi(ld2)
 ; RV64-PIC-TLSDESC-NEXT:    add a1, a1, a0
 ; RV64-PIC-TLSDESC-NEXT:    lw a1, %tprel_lo(ld2)(a1)
 ; RV64-PIC-TLSDESC-NEXT:  .LBB4_2:
 ; RV64-PIC-TLSDESC-NEXT:    lui a2, %tprel_hi(ld3)
 ; RV64-PIC-TLSDESC-NEXT:    add a0, a2, a0
 ; RV64-PIC-TLSDESC-NEXT:    lw a0, %tprel_lo(ld3)(a0)
 ; RV64-PIC-TLSDESC-NEXT:    addw a0, a1, a0
 ; RV64-PIC-TLSDESC-NEXT:    ret
-;
 ; RV32-NOPIC-TLSDESC-LABEL: dedup_localdynamic:
 ; RV32-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi3:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi3)(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, 0(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    bnez a0, .LBB4_2
 ; RV32-NOPIC-TLSDESC-NEXT:  # %bb.1:
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi4:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld2)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, %pcrel_lo(.Lpcrel_hi4)(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a0, 0(a0)
 ; RV32-NOPIC-TLSDESC-NEXT:  .LBB4_2:
 ; RV32-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi5:
 ; RV32-NOPIC-TLSDESC-NEXT:    auipc a1, %tls_ie_pcrel_hi(ld3)
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a1, %pcrel_lo(.Lpcrel_hi5)(a1)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a1, a1, tp
 ; RV32-NOPIC-TLSDESC-NEXT:    lw a1, 0(a1)
 ; RV32-NOPIC-TLSDESC-NEXT:    add a0, a0, a1
 ; RV32-NOPIC-TLSDESC-NEXT:    ret
-;
 ; RV64-NOPIC-TLSDESC-LABEL: dedup_localdynamic:
 ; RV64-NOPIC-TLSDESC:       # %bb.0: # %entry
 ; RV64-NOPIC-TLSDESC-NEXT:  .Lpcrel_hi3:
 ; RV64-NOPIC-TLSDESC-NEXT:    auipc a0, %tls_ie_pcrel_hi(ld)
 ; RV64-NOPIC-TLSDESC-NEXT:    ld a0, %pcrel_lo(.Lpcrel_hi3)(a0)
 ; RV64-NOPIC-TLSDESC-NEXT:    add a0, a0, tp
 ; RV64-NOPIC-TLSDESC-NEXT:    lw a0, 0(a0)
 ; RV64-NOPIC-TLSDESC-NEXT:    bnez a0, .LBB4_2