llvm/test/CodeGen/LoongArch/psabi-restricted-scheduling.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc --mtriple=loongarch64 --code-model=medium --post-RA-scheduler=0 < %s \
; RUN:     | FileCheck %s --check-prefix=MEDIUM_NO_SCH
; RUN: llc --mtriple=loongarch64 --code-model=medium --post-RA-scheduler=1 < %s \
; RUN:     | FileCheck %s --check-prefix=MEDIUM_SCH
; RUN: llc --mtriple=loongarch64 --code-model=large --post-RA-scheduler=0 < %s \
; RUN:     | FileCheck %s --check-prefix=LARGE_NO_SCH
; RUN: llc --mtriple=loongarch64 --code-model=large --post-RA-scheduler=1 < %s \
; RUN:     | FileCheck %s --check-prefix=LARGE_SCH

;; FIXME: According to the description of the psABI v2.30, the code sequences
;; of `PseudoLA*_LARGE` instruction and Medium code model's function call must
;; be adjacent.

@g = dso_local global i64 zeroinitializer, align 4
@G = global i64 zeroinitializer, align 4
@gd = external thread_local global i64
@ld = external thread_local(localdynamic) global i64
@ie = external thread_local(initialexec) global i64

declare ptr @bar(i64)

define void @foo() nounwind {
; MEDIUM_NO_SCH-LABEL: foo:
; MEDIUM_NO_SCH:       # %bb.0:
; MEDIUM_NO_SCH-NEXT:    addi.d $sp, $sp, -16
; MEDIUM_NO_SCH-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; MEDIUM_NO_SCH-NEXT:    pcalau12i $a0, %got_pc_hi20(G)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, %got_pc_lo12(G)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, 0
; MEDIUM_NO_SCH-NEXT:    pcalau12i $a0, %pc_hi20(g)
; MEDIUM_NO_SCH-NEXT:    addi.d $a0, $a0, %pc_lo12(g)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, 0
; MEDIUM_NO_SCH-NEXT:    ori $a0, $zero, 1
; MEDIUM_NO_SCH-NEXT:    pcaddu18i $ra, %call36(bar)
; MEDIUM_NO_SCH-NEXT:    jirl $ra, $ra, 0
; MEDIUM_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(gd)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(gd)
; MEDIUM_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ld)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(ld)
; MEDIUM_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ie)
; MEDIUM_NO_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(ie)
; MEDIUM_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_NO_SCH-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; MEDIUM_NO_SCH-NEXT:    addi.d $sp, $sp, 16
; MEDIUM_NO_SCH-NEXT:    ret
;
; MEDIUM_SCH-LABEL: foo:
; MEDIUM_SCH:       # %bb.0:
; MEDIUM_SCH-NEXT:    addi.d $sp, $sp, -16
; MEDIUM_SCH-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; MEDIUM_SCH-NEXT:    pcalau12i $a0, %got_pc_hi20(G)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, %got_pc_lo12(G)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, 0
; MEDIUM_SCH-NEXT:    pcalau12i $a0, %pc_hi20(g)
; MEDIUM_SCH-NEXT:    addi.d $a0, $a0, %pc_lo12(g)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, 0
; MEDIUM_SCH-NEXT:    ori $a0, $zero, 1
; MEDIUM_SCH-NEXT:    pcaddu18i $ra, %call36(bar)
; MEDIUM_SCH-NEXT:    jirl $ra, $ra, 0
; MEDIUM_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(gd)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(gd)
; MEDIUM_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ld)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(ld)
; MEDIUM_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ie)
; MEDIUM_SCH-NEXT:    ld.d $a0, $a0, %ie_pc_lo12(ie)
; MEDIUM_SCH-NEXT:    ldx.d $a0, $a0, $tp
; MEDIUM_SCH-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; MEDIUM_SCH-NEXT:    addi.d $sp, $sp, 16
; MEDIUM_SCH-NEXT:    ret
;
; LARGE_NO_SCH-LABEL: foo:
; LARGE_NO_SCH:       # %bb.0:
; LARGE_NO_SCH-NEXT:    addi.d $sp, $sp, -16
; LARGE_NO_SCH-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LARGE_NO_SCH-NEXT:    pcalau12i $a0, %got_pc_hi20(G)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %got_pc_lo12(G)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %got64_pc_lo20(G)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %got64_pc_hi12(G)
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_NO_SCH-NEXT:    ld.d $a0, $a0, 0
; LARGE_NO_SCH-NEXT:    pcalau12i $a0, %pc_hi20(g)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %pc_lo12(g)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %pc64_lo20(g)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %pc64_hi12(g)
; LARGE_NO_SCH-NEXT:    add.d $a0, $t8, $a0
; LARGE_NO_SCH-NEXT:    ld.d $a0, $a0, 0
; LARGE_NO_SCH-NEXT:    ori $a0, $zero, 1
; LARGE_NO_SCH-NEXT:    pcalau12i $ra, %got_pc_hi20(bar)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %got_pc_lo12(bar)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %got64_pc_lo20(bar)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %got64_pc_hi12(bar)
; LARGE_NO_SCH-NEXT:    ldx.d $ra, $t8, $ra
; LARGE_NO_SCH-NEXT:    jirl $ra, $ra, 0
; LARGE_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(gd)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(gd)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(gd)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(gd)
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ld)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(ld)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(ld)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(ld)
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_NO_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ie)
; LARGE_NO_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(ie)
; LARGE_NO_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(ie)
; LARGE_NO_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(ie)
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_NO_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_NO_SCH-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LARGE_NO_SCH-NEXT:    addi.d $sp, $sp, 16
; LARGE_NO_SCH-NEXT:    ret
;
; LARGE_SCH-LABEL: foo:
; LARGE_SCH:       # %bb.0:
; LARGE_SCH-NEXT:    addi.d $sp, $sp, -16
; LARGE_SCH-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LARGE_SCH-NEXT:    pcalau12i $a0, %got_pc_hi20(G)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %got_pc_lo12(G)
; LARGE_SCH-NEXT:    lu32i.d $t8, %got64_pc_lo20(G)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %got64_pc_hi12(G)
; LARGE_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_SCH-NEXT:    ld.d $a0, $a0, 0
; LARGE_SCH-NEXT:    pcalau12i $a0, %pc_hi20(g)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %pc_lo12(g)
; LARGE_SCH-NEXT:    lu32i.d $t8, %pc64_lo20(g)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %pc64_hi12(g)
; LARGE_SCH-NEXT:    add.d $a0, $t8, $a0
; LARGE_SCH-NEXT:    ld.d $a0, $a0, 0
; LARGE_SCH-NEXT:    ori $a0, $zero, 1
; LARGE_SCH-NEXT:    pcalau12i $ra, %got_pc_hi20(bar)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %got_pc_lo12(bar)
; LARGE_SCH-NEXT:    lu32i.d $t8, %got64_pc_lo20(bar)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %got64_pc_hi12(bar)
; LARGE_SCH-NEXT:    ldx.d $ra, $t8, $ra
; LARGE_SCH-NEXT:    jirl $ra, $ra, 0
; LARGE_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(gd)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(gd)
; LARGE_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(gd)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(gd)
; LARGE_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ld)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(ld)
; LARGE_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(ld)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(ld)
; LARGE_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_SCH-NEXT:    pcalau12i $a0, %ie_pc_hi20(ie)
; LARGE_SCH-NEXT:    addi.d $t8, $zero, %ie_pc_lo12(ie)
; LARGE_SCH-NEXT:    lu32i.d $t8, %ie64_pc_lo20(ie)
; LARGE_SCH-NEXT:    lu52i.d $t8, $t8, %ie64_pc_hi12(ie)
; LARGE_SCH-NEXT:    ldx.d $a0, $t8, $a0
; LARGE_SCH-NEXT:    ldx.d $a0, $a0, $tp
; LARGE_SCH-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LARGE_SCH-NEXT:    addi.d $sp, $sp, 16
; LARGE_SCH-NEXT:    ret
  %V = load volatile i64, ptr @G
  %v = load volatile i64, ptr @g
  call void @bar(i64 1)
  %v_gd = load volatile i64, ptr @gd
  %v_ld = load volatile i64, ptr @ld
  %v_ie = load volatile i64, ptr @ie
  ret void
}