使用hipprof --pmc得到的数据只到L2缓存，无法看到核函数对全局内存的访问情况，这对于分析核函数是memory-bound还是compute-bound十分麻烦
在性能分析培训视频
https://developer.sourcefind.cn/wp-content/uploads/2023/03/%E5%BC%82%E6%9E%84%E8%AE%A1%E7%AE%97%E5%B9%B3%E5%8F%B0%E6%80%A7%E8%83%BD%E5%88%86%E6%9E%90%E6%8A%80%E6%9C%AF.mp4
中提到了TA、TCC参数，但是在hipprof --pmc的输出中并没有看到

我的hipprof --pmc输出为
root@worker-0:/work/home/lets_daze/conv3d# head -n 30 pmc_results_61.txt
kernel-name:"Conv3d"
kernel dispatch index 0
gpu id 0
dispatch queue id 0
dispatch queue index 6
thread id 63
grid size 55115776
work group size 256
shared memory size 2048
scratch size 0
vgpr count 16
sgpr count 48
fbarrier count 0
signal handle 0x0
kernel time 0.087216(s)
processed ALU instructions 96.852296(%)
performance 9843.297374(Gflops)
shared memory operation 651056388
shared memory bank conflict 0.000000(%)
L1 cache unit is active 24.133045(%)
L1 cache unit is stalled 0.605108(%)
L2 cache write unit is stalled 0.000000(%)
L2 cache hit rate 96.177484(%)
请问如何看到核函数访问全局内存的大小和带宽？