matrixone
matrixone copied to clipboard
matrixorigin
[Bug]: perf of sysbench random_ranges 1000 threads on 10w data decrease than before
Is there an existing issue for the same bug?
- [X] I have checked the existing issues.
Branch Name
main
Commit ID
8415938eea7fec24ae4446591207e3cbd6428c8b
Other Environment Information
- Hardware parameters: Standlone
- OS type:
- Others:
Actual Behavior
commit : 8415938eea7fec24ae4446591207e3cbd6428c8b
cpu profile: cpu_lower.gar.gz
commit: c23a51fdb54087fc945ce7d0cc82f4ecec0a045c
cpu profile: cpu_higher.gar.gz
after binsearch, the issue is caused by pr：
commit: ac76417fdad34c78f71ea72817b38a52177ebad9
Expected Behavior
No response
Steps to Reproduce
run sysbench random range test：
1、git clone mo-sysbench
2、sysbench --mysql-host=127.0.0.1 --mysql-port=6001 --mysql-user=dump --mysql-password=111 oltp_common.lua --mysql-db=sysbench_db --tables=10 --table_size=100000 --threads=1000 --time=30 --report-interval=10 cleanup
sysbench --mysql-host=127.0.0.1 --mysql-port=6001 --mysql-user=dump --mysql-password=111 select_random_ranges.lua --mysql-db=sysbench_db --tables=10 --table_size=100000 --threads=1000 --time=30 --report-interval=10 prepare;
sysbench --mysql-host=127.0.0.1 --mysql-port=6001 --mysql-user=dump --mysql-password=111 select_random_ranges.lua --mysql-db=sysbench_db --tables=10 --table_size=100000 --time=30 --report-interval=10 --range_selects=off --point_selects=1 --threads=1000 run
Additional information
No response
我今天在看1000仓1000并发 random ranges的问题， 从profile里看，93%的耗时在处理cast谓词上。但是据 @m-schen 所说，cast谓词执行一次以后应该会自动折叠掉。 麻烦 @ouyuanning 确认一下是不是折叠的逻辑改坏了？
