## 经 TC VMEM：数值检查通过
cfence: 1
dma.simple: 2
sadd.s32: 2
sand.u32: 3
seq.s32: 1
simm.s32: 1
sld: 3
sor.u32: 2
sshll.u32: 2
sshrl.u32: 1
vsyncadd.remote.s32: 1
vsyncadd.s32: 3
vtrace: 2
vwait.ge: 3
# function jit(through_vmem)/through_vmem/mpmd_map [symbol 4, through_vmem.1]; image bundles [503, 529)
{ misc: vtrace 0x80000000 }
# entry bundle: %through_vmem.1 = custom-call(%x_hbm.1)
{ s1: sld s6, [smem:0x1] }
{ s0: seq.s32 p0, s6, 0 }
{ s0: simm.s32 s7, 0 ;  # materialized constant
  .encoding { s0.y = lo(imm0) } }
{ s0: @p0 dma.simple [vmem:s7], [hbm:s0], length=64, dst_flag=[sflag:52] }
{ misc: @p0 vwait.ge [sflag:52], 64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:38:12-38:55
  .encoding { misc.value = imm2 } }
{ misc: @p0 vsyncadd.s32 [sflag:52], -64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:38:12-38:55
  .encoding { misc.value = imm2 } }
{ s0: @p0 dma.simple [hbm:s1], [vmem:s7], length=64, dst_flag=[sflag:52] }
{ misc: @p0 vwait.ge [sflag:52], 64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:39:12-39:55
  .encoding { misc.value = imm2 } }
# exit bundle: %through_vmem.1 = custom-call(%x_hbm.1)
{ misc: @p0 vsyncadd.s32 [sflag:52], -64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:39:12-39:55
  .encoding { misc.value = imm2 } }
{ s1: sld s20, [smem:0x0] }
{ s1: sld s21, [smem:0x1] }
{ s0: sand.u32 s24, 0xfff, s20 }
{ s0: sadd.s32 s22, 1, s21 ;
  s1: sshll.u32 s27, s24, 0x12 }
{ s0: sand.u32 s23, 0x1, s22 }  # smod.u32 w/div 2
{ s0: sand.u32 s25, 0x3, s23 }
{ s0: sshll.u32 s26, s25, 0x10 }
{ s0: sadd.s32 s28, 131072, s26 }
{ s0: sshrl.u32 s29, s28, 0x2 }
{ s0: sor.u32 s30, 0x2d, s29 }
{ s0: sor.u32 s2, s30, s27 }
{ misc: vsyncadd.remote.s32 [sflag:s2], 1 }
{ misc: vwait.ge [sflag:45], 1 ;
  .encoding { misc.value = imm2 } }
{ misc: vsyncadd.s32 [sflag:45], -1 ;
  .encoding { misc.value = imm2 } }
{ misc: cfence }
{ misc: vtrace 0x90000000 }
# end function jit(through_vmem)/through_vmem/mpmd_map [symbol 4, through_vmem.1]

## HBM→HBM：数值检查通过
cfence: 1
dma.general: 1
sadd.s32: 2
sand.u32: 3
seq.s32: 1
simm.s32: 3
sld: 3
sor.u32: 2
sshll.u32: 2
sshrl.u32: 1
vsyncadd.remote.s32: 1
vsyncadd.s32: 2
vtrace: 2
vwait.ge: 2
# function jit(hbm_to_hbm)/hbm_to_hbm/mpmd_map [symbol 4, hbm_to_hbm.1]; image bundles [503, 527)
{ misc: vtrace 0x80000000 }
# entry bundle: %hbm_to_hbm.1 = custom-call(%x_hbm.1)
{ s1: sld s6, [smem:0x1] }
{ s0: seq.s32 p0, s6, 0 }
{ s0: simm.s32 s7, 52 ;  # materialized constant
  s1: simm.s32 s8, 0 }  # materialized constant
{ s0: simm.s32 s9, 0 }  # materialized constant
{ s0: @p0 dma.general [hbm:s1], [hbm:s0], length=64, stride_descriptor=[smem:0x0], stride_count=0, src_flag=[sflag:s8], dst_flag=[sflag:s7], ici_dest=s9 }  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:58:12-58:47 | hbm-to-hbm
{ misc: @p0 vwait.ge [sflag:52], 64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:58:12-58:54
  .encoding { misc.value = imm2 } }
# exit bundle: %hbm_to_hbm.1 = custom-call(%x_hbm.1)
{ misc: @p0 vsyncadd.s32 [sflag:52], -64 ;  # chapter01/04_local_dma/01_pallas_f32_copy_64x128.py:58:12-58:54
  .encoding { misc.value = imm2 } }
{ s1: sld s20, [smem:0x0] }
{ s1: sld s21, [smem:0x1] }
{ s0: sand.u32 s24, 0xfff, s20 }
{ s0: sadd.s32 s22, 1, s21 ;
  s1: sshll.u32 s27, s24, 0x12 }
{ s0: sand.u32 s23, 0x1, s22 }  # smod.u32 w/div 2
{ s0: sand.u32 s25, 0x3, s23 }
{ s0: sshll.u32 s26, s25, 0x10 }
{ s0: sadd.s32 s28, 131072, s26 }
{ s0: sshrl.u32 s29, s28, 0x2 }
{ s0: sor.u32 s30, 0x2d, s29 }
{ s0: sor.u32 s2, s30, s27 }
{ misc: vsyncadd.remote.s32 [sflag:s2], 1 }
{ misc: vwait.ge [sflag:45], 1 ;
  .encoding { misc.value = imm2 } }
{ misc: vsyncadd.s32 [sflag:45], -1 ;
  .encoding { misc.value = imm2 } }
{ misc: cfence }
{ misc: vtrace 0x90000000 }
# end function jit(hbm_to_hbm)/hbm_to_hbm/mpmd_map [symbol 4, hbm_to_hbm.1]