## ncu_5120x17408.ncu-rep
served_M4    void exl3_gemm_kernel<4, 0, 2, 16, 32, 256, 4, 3> dur=  52.86us dram= 86.1% (  846GB/s rd=  44.7MB) sm= 39.4% issue= 39.4% tensor= 16.2% alu= 35.9% fma= 16.1% occ= 33.0/  33% l2hit=  8.8% regs=128 smem= 91.0K grid=128x512 bankcf=43699 stalls=barrier:3.0,math_pipe_throttle:1.6,wait:1.1,not_selected:1.1
mma_M4       void ag_mma_kernel<4, 2, 1, 1, 1>(const void *, const unsign dur=  55.01us dram= 82.7% (  813GB/s rd=  44.7MB) sm= 42.9% issue= 42.9% tensor=  7.8% alu= 40.9% fma= 16.5% occ= 59.0/  67% l2hit= 22.0% regs=64 smem=  9.0K grid=1360x256 bankcf=739840 stalls=long_scoreboard:9.0,wait:1.6,not_selected:1.1,selected:1.0
up_gemv_M4   void exl3_gemv_kernel<4, 0, 2, 1, 1, 0> dur=  55.07us dram= 82.6% (  811GB/s rd=  44.7MB) sm= 29.3% issue= 29.3% tensor= 15.4% alu= 23.1% fma= 15.4% occ= 35.3/  50% l2hit= 13.1% regs=80 smem= 17.0K grid=272x256 bankcf=104448 stalls=long_scoreboard:9.5,barrier:1.0,selected:1.0,wait:0.8
simt_M4      void ag_gemv_kernel<Exl3GemvLoader<4, 2, 8>, 4, 0, 4, 4>(con dur=  70.91us dram= 64.2% (  631GB/s rd=  44.7MB) sm= 61.0% issue= 52.7% tensor=  0.0% alu= 26.6% fma= 37.7% occ= 70.7/  83% l2hit= 16.0% regs=48 smem= 17.0K grid=680x256 bankcf=6001 stalls=long_scoreboard:6.3,not_selected:2.0,selected:1.0,math_pipe_throttle:1.0
i8_M4        void arbi_serve_exl3_a_prep::a_prep_kernel<1, 4, 1, 1, arbi_ dur=   2.69us dram=  2.7% (   26GB/s rd=   0.1MB) sm=  5.5% issue=  5.5% tensor=  0.0% alu=  8.0% fma=  3.2% occ= 18.6/ 100% l2hit= 88.7% regs=40 smem=  1.0K grid=320x128 bankcf=0 stalls=imc_miss:9.0,short_scoreboard:2.5,wait:1.6,selected:1.0
i8_M4        void exl3_i8_gemm_kernel<4, 2, 32, 128, 128, 1, 4, 2, 8, 0,  dur=  50.18us dram= 90.9% (  893GB/s rd=  44.8MB) sm= 23.1% issue= 23.1% tensor= 17.0% alu= 18.8% fma= 11.2% occ=  8.8/  25% l2hit= 34.3% regs=127 smem= 25.2K grid=136x128 bankcf=0 stalls=long_scoreboard:2.2,selected:1.0,wait:0.4,math_pipe_throttle:0.3
served_M8    void exl3_gemm_kernel<4, 0, 2, 16, 32, 256, 4, 3> dur=  53.57us dram= 85.0% (  835GB/s rd=  44.7MB) sm= 39.1% issue= 39.1% tensor= 16.2% alu= 36.0% fma= 16.2% occ= 33.1/  33% l2hit= 15.3% regs=128 smem= 91.0K grid=128x512 bankcf=87278 stalls=barrier:3.0,math_pipe_throttle:1.6,wait:1.1,not_selected:1.1
mma_M8       void ag_mma_kernel<4, 2, 1, 1, 1>(const void *, const unsign dur=  58.50us dram= 77.9% (  766GB/s rd=  44.8MB) sm= 44.2% issue= 44.2% tensor=  7.4% alu= 41.6% fma= 17.3% occ= 59.9/  67% l2hit= 33.3% regs=64 smem=  9.0K grid=1360x256 bankcf=739840 stalls=long_scoreboard:8.6,wait:1.6,not_selected:1.2,selected:1.0
up_gemv_M8   void exl3_gemv_kernel<4, 0, 2, 1, 1, 0> dur=  55.30us dram= 82.3% (  809GB/s rd=  44.7MB) sm= 29.4% issue= 29.4% tensor= 15.3% alu= 23.0% fma= 15.3% occ= 35.0/  50% l2hit= 22.1% regs=80 smem= 17.0K grid=272x256 bankcf=104448 stalls=long_scoreboard:9.5,barrier:1.1,selected:1.0,wait:0.8
simt_M8      void ag_gemv_kernel<Exl3GemvLoader<4, 2, 8>, 4, 0, 8, 4>(con dur=  96.86us dram= 47.0% (  462GB/s rd=  44.7MB) sm= 82.1% issue= 57.5% tensor=  0.0% alu= 21.2% fma= 38.8% occ= 41.9/  50% l2hit= 21.4% regs=64 smem= 33.0K grid=680x256 bankcf=7179 stalls=mio_throttle:2.3,not_selected:1.8,selected:1.0,long_scoreboard:0.8
simt_M8      void ag_gemv_reduce_kernel<Exl3GemvLoader<4, 2, 8>, 8> dur=   5.89us dram= 50.2% (  490GB/s rd=   2.9MB) sm= 11.0% issue= 11.0% tensor=  0.0% alu= 11.0% fma=  4.6% occ= 54.4/  83% l2hit= 30.3% regs=44 smem=  5.0K grid=1088x256 bankcf=0 stalls=long_scoreboard:14.7,barrier:14.1,imc_miss:8.0,wait:2.2
i8_M8        void arbi_serve_exl3_a_prep::a_prep_kernel<1, 4, 1, 1, arbi_ dur=   2.66us dram=  4.3% (   41GB/s rd=   0.1MB) sm=  6.3% issue=  6.3% tensor=  0.0% alu=  8.7% fma=  3.6% occ= 19.0/ 100% l2hit= 84.7% regs=40 smem=  1.0K grid=320x128 bankcf=0 stalls=imc_miss:9.0,short_scoreboard:2.2,long_scoreboard:1.6,wait:1.6
i8_M8        void exl3_i8_gemm_kernel<4, 2, 32, 128, 128, 1, 4, 2, 8, 0,  dur=  50.69us dram= 90.0% (  884GB/s rd=  44.8MB) sm= 23.0% issue= 23.0% tensor= 17.0% alu= 18.8% fma= 11.2% occ=  8.8/  25% l2hit= 34.4% regs=127 smem= 25.2K grid=136x128 bankcf=0 stalls=long_scoreboard:2.2,selected:1.0,wait:0.4,math_pipe_throttle:0.3
served_M16   void exl3_gemm_kernel<4, 0, 2, 16, 32, 256, 4, 3> dur=  54.72us dram= 83.3% (  819GB/s rd=  44.8MB) sm= 38.6% issue= 38.6% tensor= 16.0% alu= 35.6% fma= 16.1% occ= 33.1/  33% l2hit= 26.1% regs=128 smem= 91.0K grid=128x512 bankcf=174372 stalls=barrier:3.1,math_pipe_throttle:1.6,wait:1.1,not_selected:1.1
mma_M16      void ag_mma_prestage_kernel<2, 1>(const void *, const __half dur=   2.30us dram=  8.2% (   79GB/s rd=   0.2MB) sm=  4.8% issue=  4.8% tensor=  0.0% alu= 11.0% fma=  5.2% occ= 16.4/ 100% l2hit= 62.0% regs=24 smem=  1.0K grid=80x256 bankcf=0 stalls=long_scoreboard:5.2,imc_miss:3.5,wait:1.4,selected:1.0
mma_M16      void ag_mma_kernel<4, 2, 2, 1, 1>(const void *, const unsign dur=  56.86us dram= 80.3% (  796GB/s rd=  44.8MB) sm= 35.6% issue= 35.6% tensor= 16.0% alu= 34.7% fma= 14.1% occ= 59.5/  67% l2hit= 51.1% regs=64 smem= 17.0K grid=1360x256 bankcf=296749 stalls=long_scoreboard:10.4,wait:1.6,math_pipe_throttle:1.4,barrier:1.3
i8_M16       void arbi_serve_exl3_a_prep::a_prep_kernel<1, 4, 1, 1, arbi_ dur=   2.72us dram=  7.4% (   70GB/s rd=   0.2MB) sm=  7.9% issue=  7.9% tensor=  0.0% alu=  9.4% fma=  4.3% occ= 18.1/ 100% l2hit= 80.0% regs=40 smem=  1.0K grid=320x128 bankcf=0 stalls=imc_miss:5.8,long_scoreboard:2.4,short_scoreboard:1.7,wait:1.5
i8_M16       void exl3_i8_gemm_kernel<4, 2, 32, 128, 128, 1, 4, 2, 8, 0,  dur=  51.01us dram= 89.4% (  878GB/s rd=  44.8MB) sm= 23.0% issue= 23.0% tensor= 16.9% alu= 18.8% fma= 11.3% occ=  8.9/  25% l2hit= 34.7% regs=127 smem= 25.2K grid=136x128 bankcf=19 stalls=long_scoreboard:2.2,selected:1.0,wait:0.4,math_pipe_throttle:0.3
served_M32   void exl3_gemm_kernel<4, 0, 2, 16, 32, 256, 4, 3> dur=  92.38us dram= 49.5% (  487GB/s rd=  45.0MB) sm= 45.4% issue= 45.4% tensor= 18.1% alu= 40.3% fma= 18.2% occ= 33.4/  33% l2hit= 62.7% regs=128 smem= 91.0K grid=128x512 bankcf=348264 stalls=barrier:2.4,math_pipe_throttle:1.6,wait:1.2,not_selected:1.1
mma_M32      void ag_mma_prestage_kernel<4, 1>(const void *, const __half dur=   2.53us dram= 14.3% (  137GB/s rd=   0.3MB) sm=  8.8% issue=  8.8% tensor=  0.0% alu= 13.0% fma=  6.0% occ= 20.1/ 100% l2hit= 59.0% regs=24 smem=  1.0K grid=160x256 bankcf=0 stalls=long_scoreboard:6.3,imc_miss:3.2,wait:1.4,selected:1.0
mma_M32      void ag_mma_kernel<4, 2, 4, 1, 1>(const void *, const unsign dur=  63.46us dram= 72.2% (  739GB/s rd=  45.0MB) sm= 35.9% issue= 35.9% tensor= 28.6% alu= 33.0% fma= 13.3% occ= 43.9/  50% l2hit= 67.2% regs=64 smem= 33.0K grid=1360x256 bankcf=634015 stalls=long_scoreboard:4.6,math_pipe_throttle:3.1,wait:1.7,barrier:1.0
i8_M32       void arbi_serve_exl3_a_prep::a_prep_kernel<1, 4, 1, 1, arbi_ dur=   2.85us dram= 12.9% (  125GB/s rd=   0.4MB) sm= 10.7% issue= 10.7% tensor=  0.0% alu= 11.9% fma=  5.8% occ= 18.3/ 100% l2hit= 70.9% regs=40 smem=  1.0K grid=320x128 bankcf=0 stalls=long_scoreboard:4.2,imc_miss:2.9,wait:1.4,short_scoreboard:1.2
i8_M32       void exl3_i8_gemm_kernel<4, 2, 32, 128, 128, 1, 4, 2, 8, 0,  dur=  51.65us dram= 88.3% (  867GB/s rd=  44.8MB) sm= 23.1% issue= 23.1% tensor= 16.8% alu= 18.8% fma= 11.3% occ=  8.8/  25% l2hit= 35.2% regs=127 smem= 25.2K grid=136x128 bankcf=0 stalls=long_scoreboard:2.1,selected:1.0,wait:0.4,math_pipe_throttle:0.3
