Kuan-Wei Chiu comments

Results 15 comments of


                                            Kuan-Wei Chiu

Fix the microsecond-to-nanosecond conversion in rv_clock_gettime()

See: https://sourceware.org/git/?p=newlib-cygwin.git;a=commit;h=5f15d7c5817b07a6b18cbab17342c95cb7b42be4

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

I attempted to implement an LIRS cache and conducted performance testing. Unfortunately, I observed regressions in both the Coremark and Dhrystone benchmarks. Here are the data: - i5-10500 | Metric...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

> > I attempted to implement an LIRS cache and conducted performance testing. Unfortunately, I observed regressions in both the Coremark and Dhrystone benchmarks. > > Lately, we have encountered...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

> > Unfortunately, I observed regressions in both the Coremark and Dhrystone benchmarks. > > In addition to the typical benchmark suite, which includes CoreMark and Dhrystone, we also make...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

I've noticed that when using the code from the 'wip/jit' branch with MIR as the JIT backend and running the scimark2 benchmark with my implemented LIRS cache, I've encountered a...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

After conducting further testing, I have observed that regardless of the cache replacement algorithm used, the cache hit ratio consistently remains above 99.99%. Therefore, I believe that the primary factor...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

I acknowledge the possibility that CoreMark or Dhrystone may exhibit significant differences from real-world workloads. Perhaps, it would be prudent for us to explore alternative benchmarks that more accurately represent...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

> If you are exploring [semu](https://github.com/sysprog21/semu), please note that there is a fork called [semu-c64](https://github.com/onnokort/semu-c64) which includes additional features like `insn_count` for enhanced analysis and checkpoint functionality. Thanks for the...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

I noticed that the JIT compiler's code cache in the JVM doesn't utilize any cache replacement algorithm. Instead, when the code cache is full, it either disables the JIT compiler...

Evaluate block replacement algorithms which receive attention on improvement of hit ratio

My speculation is that in real workloads, the number of cache_get operations is typically much larger than cache_put operations. However, it's only during cache_put that the need to select a...