Nsight Compute 中使用的术语

Question

两个问题：

根据 Nsight Compute 的说法，我的内核受计算限制。相对于峰值性能的 SM 利用率为 74%，内存利用率为 47%。但是，当我查看每个管道的利用率时，LSU 利用率远高于其他管道（75% 对 10-15%）。这是否表明我的内核受内存限制？如果计算和内存资源的利用率与管道利用率不对应，我不知道如何解释这些术语。
调度程序每 4 个周期才发出一次，这是否意味着我的内核受延迟限制？人们通常根据计算和内存资源的利用率来定义它。两者有什么关系？

Answer 1

在 CC7.5 GPU 上的 Nsight Compute 中

SM% 由 sm__throughput 定义，并且内存百分比由 gpu__compute_memory_throughtput

定义

sm_throughput 是以下指标的最大值：

sm__instruction_throughput
- sm__inst_executed
- sm__issue_active
- sm__mio_inst_issued
- sm__pipe_alu_cycles_active
- sm__inst_executed_pipe_cbu_pred_on_any
- sm__pipe_fp64_cycles_active
- sm__pipe_tensor_cycles_active
- sm__inst_executed_pipe_xu
- sm__pipe_fma_cycles_active
- sm__inst_executed_pipe_fp16
- sm__pipe_shared_cycles_active
- sm__inst_executed_pipe_uniform
- sm__instruction_throughput_internal_activity
sm__memory_throughput
- idc__request_cycles_active
- sm__inst_executed_pipe_adu
- sm__inst_executed_pipe_ipa
- sm__inst_executed_pipe_lsu
- sm__inst_executed_pipe_tex
- sm__mio_pq_read_cycles_active
- sm__mio_pq_write_cycles_active
- sm__mio2rf_writeback_active
- sm__memory_throughput_internal_activity

gpu__compute_memory_throughput 是以下指标的最大值：

gpu__compute_memory_access_throughput
- l1tex__data_bank_reads
- l1tex__data_bank_writes
- l1tex__data_pipe_lsu_wavefronts
- l1tex__data_pipe_tex_wavefronts
- l1tex__f_wavefronts
- lts__d_atomic_input_cycles_active
- lts__d_sectors
- lts__t_sectors
- lts__t_tag_requests
- gpu__compute_memory_access_throughput_internal_activity
gpu__compute_memory_access_throughput
l1tex__lsuin_requests
- l1tex__texin_sm2tex_req_cycles_active
- l1tex__lsu_writeback_active
- l1tex__tex_writeback_active
- l1tex__m_l1tex2xbar_req_cycles_active
- l1tex__m_xbar2l1tex_read_sectors
- lts__lts2xbar_cycles_active
- lts__xbar2lts_cycles_active
- lts__d_sectors_fill_device
- lts__d_sectors_fill_sysmem
- gpu__dram_throughput
- gpu__compute_memory_request_throughput_internal_activity

在您的情况下，限制器是 sm__inst_executed_pipe_lsu，它是指令吞吐量。如果您查看 sections/SpeedOfLight.py 延迟限制被定义为 sm__throughput 和 gpu__compute_memory_throuhgput < 60%.

一些指令流水线的吞吐量较低，例如 fp64、xu 和 lsu（因芯片而异）。管道利用率是 sm__throughput 的一部分。为了提高性能，选项是：

减少超额订阅管道的指令，或者
发出不同类型的指令以使用空发出周期。

产生故障

从 Nsight Compute 2020.1 开始，没有一个简单的命令行可以在没有运行分析会话的情况下生成列表。现在，您可以使用 breakdown:<throughput metric>avg.pct_of_peak_sustained.elapsed 收集一个吞吐量指标并解析输出以获取 sub-metric 名称。

例如：

ncu.exe --csv --metrics breakdown:sm__throughput.avg.pct_of_peak_sustained_elapsed --details-all -c 1 cuda_application.exe

生成：

"ID","Process ID","Process Name","Host Name","Kernel Name","Kernel Time","Context","Stream","Section Name","Metric Name","Metric Unit","Metric Value"
"0","33396","cuda_application.exe","127.0.0.1","kernel()","2020-Aug-20 13:26:26","1","7","Command line profiler metrics","gpu__dram_throughput.avg.pct_of_peak_sustained_elapsed","%","0.38"
"0","33396","cuda_application.exe","127.0.0.1","kernel()","2020-Aug-20 13:26:26","1","7","Command line profiler metrics","l1tex__data_bank_reads.avg.pct_of_peak_sustained_elapsed","%","0.05"
"0","33396","cuda_application.exe","127.0.0.1","kernel()","2020-Aug-20 13:26:26","1","7","Command line profiler metrics","l1tex__data_bank_writes.avg.pct_of_peak_sustained_elapsed","%","0.05"
...

关键字breakdown可以在Nsight Compute部分文件中使用来扩展吞吐量指标。这用于 SpeedOfLight.section.

Nsight Compute 中使用的术语

Terminology used in Nsight Compute

optimization

cuda

nsight-compute