intel vtune9 是否没有动态分析功能
我在学习代码优化
上面介绍vtune有动态分析功能 ,就是展开Hotspot看函数的汇编代码 ,然后可以设定需要分析从哪里开始的几行汇编代码 ,
运行动态分析功能 ,vtune就能够出一个报告,内容有: L1cache miss次数 L2cache miss次数, dataalignment 没对齐的访问次数等
可是我在vtune9下面一直没有找到
amd的codeanalize 有这个功能 ,难道现在只能使用amd cpu来做这部分的代码优化了么
vtune 动态分析的报告样例
Dynamic Analysis Results for the Chosen Hotspot
Decoder Minimum Clocks = 0, ; // Minimum decoding time: 0 clocks
Decoder Average Clocks = 0.7 ; // Average decoding time: 0.7 clock
Decoder Maximum Clocks = 14 ; // Maximum decoding time: 14 clocks
Retirement Minimum Clocks = 0, ; // Minimum retirement time: 0 clocks
Retirement Average Clocks = 6.9 ; // Average retirement time: 6.9 clocks
Retirement Maximum Clocks = 104 ; // Maximum retirement time: 104 clocks
Total Cycles = 20117 (35.88%) ; // Total execution time: 20,117 clocks
Micro-Ops for this instruction = 1 ; // Decoding occurs in one micro-operation
The instruction had to wait (0,0.1,2) cycles for its sources to be ready.
Warnings: 3*decode_slow:0 ; // No decoder conflicts
Dynamic Penalty: DC_rd_miss
The operand of this load instruction was not in the data cache. The
instruction stalls while the processor loads the specified address location
from L2 cache or the main memory.
Occurrences = 1 ; // This happened once.
Dynamic Penalty: DC_misalign
The instruction stalls because it accessed data that was split across two
data-cache lines.
Occurrences = 2000 ; // This happened 2,000 times.
Dynamic Penalty: L2data_rd_miss
The operand of this load instruction was not in L2 cache. The instruction
stalls while the processor loads the specified address location from the
main memory.
Occurrences = 1 ; // This happened once.
Dynamic Penalty: No_BTB_info
The BTB (Branch Target Buffer) does not contain information about this branch.
The branch was predicted using the static branch prediction algorithm.
Occurrences = 1 ; // This happened once.