liyuming1978
liyuming1978
in win10, I find the cpu cost is high when do infer (or wait infer-request), the clWaitForEvents seems to a busy wait.
for the long time of clCreateBuffer , got it, too much clCreateBuffer (>30000) (the speed is the same as CL_MEM_USE_HOST_PTR , but too much calls) , I will try drop...
A conclusion: 1. longtime of clCreatebuffer and clReleaseBuffer for mobilenet: openvino will call clCreatebuffer for each group, if group is 512, it will create 512 *2 buffer (weights and bias)...
@MichalMrozek cl_cache works! that good to my patch. for DECLARE_DEBUG_VARIABLE... dose it need re-build opencl driver? all just set register key? I use quick and small model to get best...
@MichalMrozek how to enable cl_cache in ubuntu? I use nuc in windows, so , always AC.. my email is [email protected] :)
MPP_ENC_BASE_CFG_CHANGE_LOW_DELAY 这个怎么用呢, 或者这个有用吗
我增加了 mpp_enc_cfg_set_s32(cfg_, "base:low_delay", 1); mpp_enc_cfg_set_u32(cfg_, "split:mode", 1); //或者2 mpp_enc_cfg_set_u32(cfg_, "split:out", 1); rk_api_->control(codec_ctx_, MPP_ENC_SET_CFG, cfg_); 然后, -- 编码器就没输出了。
static void check_low_delay_part_mode(MppEncImpl *enc) { MppEncCfgSet *cfg = &enc->cfg; enc->low_delay_part_mode = 0; if (!(cfg->base.low_delay)) return; if (!cfg->split.split_mode) return; if (mpp_enc_hal_check_part_mode(enc->enc_hal)) return; enc->low_delay_part_mode = 1; } 有什么样例, 去配置编码低延时模式吗? @HermanChen , 这个函数感觉base.low_delay需要配成1
rga.cpp 里面 request->mmu_info.mmu_en = 1; request->mmu_info.mmu_flag = 1; request->mmu_info.mmu_flag = ((2 & 0x3) mmu_info.mmu_flag |= (1
我把request->mmu_info.mmu_flag |= (1