##### shape=40 linear=down_proj M=8192 set=src rot=8 -> /w/_logs/bank_k5_s40.ncu-rep
ops=1460.3 GOP  k=17408 n=5120 M=8192 arm=ours shape=40 (128, 128, 128, 1, 4, 2, 8)
excessive shared wavefronts (= bank conflicts): 89,128,960

by SASS form:
  form                          conflicts     wavefronts       executed  per inst
  LDS                          89,128,960    178,257,920     89,128,960   1.000000
  LDGSTS.E.BYPASS.128                   0     72,417,280     18,104,320   0.000000
  LDS.128                               0    178,257,920     44,564,480   0.000000
  @!PT (never taken)                    0              0      4,147,200   0.000000

top 15 instruction addresses:
  0x79b9ef29aaa0  LDS                           1,392,640
  0x79b9ef29aa60  LDS                           1,392,640
  0x79b9ef29a980  LDS                           1,392,640
  0x79b9ef29a900  LDS                           1,392,640
  0x79b9ef29a8d0  LDS                           1,392,640
  0x79b9ef29a880  LDS                           1,392,640
  0x79b9ef29a870  LDS                           1,392,640
  0x79b9ef29a810  LDS                           1,392,640
  0x79b9ef29a7e0  LDS                           1,392,640
  0x79b9ef29a7d0  LDS                           1,392,640
  0x79b9ef29a790  LDS                           1,392,640
  0x79b9ef29a780  LDS                           1,392,640
  0x79b9ef29a730  LDS                           1,392,640
  0x79b9ef29a720  LDS                           1,392,640
  0x79b9ef29a6b0  LDS                           1,392,640
