Tracked topic
… Bytedance has been working on Reserved THP as a potential stepping stone or goal of unifying HugeTLB and Transparent Huge Page THP support within the Linux kernel for kernel huge pages. …
… Bytedance engineers thus took to reworking the kernel SMP code to allow preemption during IPI completion waiting in order to avoid scheduling latency spikes. Bytedance engineers found the existing Linux SMP behavior to be particularly bad for latency-sensitive workloads like DPDK. …
… That optimization comes from Bytedance engineer Fengnan Chang after discovering the wasted memory write bandwidth due to the caller just discarding the iterator when it's done. …
… Bytedance engineer Fengnan Chang noticed a "significant bottleneck" when carrying out 4K random reads on Gen5 NVMe SSDs. …