## num_cores=1：数值检查通过；kernel 段 37 个 bundle；vld 8、vmul 8、vst 8；vsyncadd.remote 1
bundle    0: prelude-prefix-start
bundle    9: routine:wait-for-active-dma
bundle   16: routine:program-exit
bundle   19: routine:allocation-table-find
bundle   28: routine:allocation-table-get-or-add
bundle   54: routine:allocation-table-cut-and-add
bundle   71: routine:branch-impl
bundle  122: routine:branch
bundle  128: trampoline:program-start
bundle  503: function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [503, 507)
bundle  504: entry bundle: %scale.1 = custom-call(%x_hbm.1)
bundle  507: end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
bundle  508: function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [508, 525)
bundle  525: end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
bundle  525: exit bundle: %scale.1 = custom-call(%x_hbm.1)
bundle  526: function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [526, 542)
bundle  542: end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
bundle  660: 清单结束（共 660 个 bundle）
# function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [503, 507)
{ misc: vtrace 0x80000000 }
# entry bundle: %scale.1 = custom-call(%x_hbm.1)
{ s1: sld s6, [smem:0x1] }
{ s0: sne.s32 p0, s6, 0 }
{ s0: @p0 sbr.rel L_020d }
# end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
# ... 省略 runtime 代码 1 个 bundle
# function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [508, 525)
{ s0: simm.s32 s7, 0 ;  # materialized constant
  .encoding { s0.y = lo(imm0) } }
{ s0: dma.simple [vmem:s7], [hbm:s0], length=64, dst_flag=[sflag:52] }
{ misc: vwait.ge [sflag:52], 64 ;  # chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:66 [dma_wait; LLO 65]
  .encoding { misc.value = imm2 } }
{ vld: vld.8x128 v0, [vmem:0x0] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 27]
  misc: vsyncadd.s32 [sflag:52], -64 ;  # chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:66 [dma_wait; LLO 66]
  .encoding { misc.value = imm2 } }
{ va0: vmul.8x128.f32 v1, 2.0, v0 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 35]
  vld: vld.8x128 v2, [vmem:0x8] }  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 28]
{ va0: vmul.8x128.f32 v3, 2.0, v2 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 36]
  vld: vld.8x128 v4, [vmem:0x10] }  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 29]
{ va0: vmul.8x128.f32 v5, 2.0, v4 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 37]
  vst: vst.8x128 [vmem:0x0], v1 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 43]
  vld: vld.8x128 v6, [vmem:0x18] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 30]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v7, 2.0, v6 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 38]
  vst: vst.8x128 [vmem:0x8], v3 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 44]
  vld: vld.8x128 v8, [vmem:0x20] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 31]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v9, 2.0, v8 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 39]
  vst: vst.8x128 [vmem:0x10], v5 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 45]
  vld: vld.8x128 v10, [vmem:0x28] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 32]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v11, 2.0, v10 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 40]
  vst: vst.8x128 [vmem:0x18], v7 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 46]
  vld: vld.8x128 v12, [vmem:0x30] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 33]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v13, 2.0, v12 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 41]
  vst: vst.8x128 [vmem:0x20], v9 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 47]
  vld: vld.8x128 v14, [vmem:0x38] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 34]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v15, 2.0, v14 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 42]
  vst: vst.8x128 [vmem:0x28], v11 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 48]
{ vst: vst.8x128 [vmem:0x30], v13 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 49]
{ vst: vst.8x128 [vmem:0x38], v15 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 50]
{ s0: dma.simple [hbm:s1], [vmem:s7], length=64, dst_flag=[sflag:52] }
{ misc: vwait.ge [sflag:52], 64 ;  # chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:66 [dma_wait; LLO 67]
  .encoding { misc.value = imm2 } }
{ misc: vsyncadd.s32 [sflag:52], -64 ;  # chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:66 [dma_wait; LLO 68]
  .encoding { misc.value = imm2 } }
# end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
# ... 省略 runtime 代码 1 个 bundle
# function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [526, 542)
{ 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(scale)/scale/mpmd_map [symbol 4, scale.1]

## num_cores=2：数值检查通过；kernel 段 43 个 bundle；vld 4、vmul 4、vst 4；vsyncadd.remote 2
bundle    0: prelude-prefix-start
bundle    9: routine:wait-for-active-dma
bundle   16: routine:program-exit
bundle   19: routine:allocation-table-find
bundle   28: routine:allocation-table-get-or-add
bundle   54: routine:allocation-table-cut-and-add
bundle   71: routine:branch-impl
bundle  122: routine:branch
bundle  128: trampoline:program-start
bundle  503: function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [503, 546)
bundle  504: entry bundle: %scale.1 = custom-call(%x_hbm.1)
bundle  529: exit bundle: %scale.1 = custom-call(%x_hbm.1)
bundle  546: end function jit(scale)/scale/mpmd_map [symbol 4, scale.1]
bundle  660: 清单结束（共 660 个 bundle）
# function jit(scale)/scale/mpmd_map [symbol 4, scale.1]; image bundles [503, 546)
{ misc: vtrace 0x80000000 }
# entry bundle: %scale.1 = custom-call(%x_hbm.1)
{ s1: sld s6, [smem:0x1] }
{ s1: sld s7, [smem:0x0] }
{ s0: sadd.s32 s8, 1, s6 }
{ s0: sand.u32 s9, 0x1, s8 ;  # smod.u32 w/div 2
  s1: sand.u32 s10, 0xfff, s7 }
{ s0: sand.u32 s11, 0x3, s9 ;
  s1: sshll.u32 s13, s10, 0x12 }
{ s0: sshll.u32 s12, s11, 0x10 }
{ s0: sadd.s32 s14, 131072, s12 }
{ s0: sshrl.u32 s15, s14, 0x2 }
{ s0: sor.u32 s16, 0x2d, s15 }
{ s0: sor.u32 s17, s16, s13 }
{ misc: vsyncadd.remote.s32 [sflag:s17], 1 }
{ misc: vwait.ge [sflag:45], 1 ;
  .encoding { misc.value = imm2 } }
{ s0: sshll.u32 s18, s6, 0x5 ;  # chapter02/01_megacore/01_pallas_two_cores.py:41:27-41:38 [mul; LLO 62]
  s1: simm.s32 s22, 0 ;  # materialized constant
  misc: vsyncadd.s32 [sflag:45], -1 ;
  .encoding { misc.value = imm2 ; s1.y = lo(imm0) } }
{ s0: sadd.s32 s21, s18, s0 }  # chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:59 [dma_start; LLO 21] | chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:66 [dma_wait; LLO 21]
{ s0: dma.simple [vmem:s22], [hbm:s21], length=32, dst_flag=[sflag:52] }
{ misc: vwait.ge [sflag:52], 32 ;  # chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:66 [dma_wait; LLO 64]
  .encoding { misc.value = imm2 } }
{ s0: sadd.s32 s25, s18, s1 ;  # chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:59 [dma_start; LLO 46] | chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:66 [dma_wait; LLO 46]
  vld: vld.8x128 v0, [vmem:0x0] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 34]
  misc: vsyncadd.s32 [sflag:52], -32 ;  # chapter02/01_megacore/01_pallas_two_cores.py:42:12-42:66 [dma_wait; LLO 65]
  .encoding { misc.value = imm2 } }
{ va0: vmul.8x128.f32 v1, 2.0, v0 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 38]
  vld: vld.8x128 v2, [vmem:0x8] }  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 35]
{ va0: vmul.8x128.f32 v3, 2.0, v2 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 39]
  vld: vld.8x128 v4, [vmem:0x10] }  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 36]
{ va0: vmul.8x128.f32 v5, 2.0, v4 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 40]
  vst: vst.8x128 [vmem:0x0], v1 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 42]
  vld: vld.8x128 v6, [vmem:0x18] ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:37 [get; LLO 37]
  .encoding { vst.offset = imm2 } }
{ va0: vmul.8x128.f32 v7, 2.0, v6 ;  # chapter02/01_megacore/01_pallas_two_cores.py:43:26-43:43 [mul; LLO 41]
  vst: vst.8x128 [vmem:0x8], v3 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 43]
{ vst: vst.8x128 [vmem:0x10], v5 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 44]
{ vst: vst.8x128 [vmem:0x18], v7 }  # chapter02/01_megacore/01_pallas_two_cores.py:43:12-43:23 [swap; LLO 45]
{ s0: dma.simple [hbm:s25], [vmem:s22], length=32, dst_flag=[sflag:52] }
{ misc: vwait.ge [sflag:52], 32 ;  # chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:66 [dma_wait; LLO 66]
  .encoding { misc.value = imm2 } }
# exit bundle: %scale.1 = custom-call(%x_hbm.1)
{ misc: vsyncadd.s32 [sflag:52], -32 ;  # chapter02/01_megacore/01_pallas_two_cores.py:44:12-44:66 [dma_wait; LLO 67]
  .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(scale)/scale/mpmd_map [symbol 4, scale.1]