pt的timing 比icc的反而好

icc route ok后，然后抽spef到pt，这个时候看起来pt的timing比icc要差些的。然后我用pt fix leakage后写出eco file 到icc， 这个时候icc的timing 比起pt反而要差了。
举例一条路径，基本上从clock的source 端到 路径起始端的cp，两边差距很小，不到1p，
然后经过第一个UHVT cell， pt的timing 就表现的比icc好了。我用report_delay_calculation
做比较 ， icc的如下：
From pin:apbb_bj/apb1_master/U101/I
To pin:apbb_bj/apb1_master/U101/ZN
Main Library Units:1ns1pF1kOhm

From Pin Operating Voltage, Temperature, Process: 0.81:0.00:1.00
To Pin Operating Voltage, Temperature,Process: 0.81:0.00:1.00
----------------------------------------------------------------
ThresholdsRise_DelayRiseFall_DelayFallRail
i/po/pSlewi/po/pSlewVolt
----------------------------------------------------------------
FromPin50.0 50.030.0->70.050.0 50.070.0->30.0 0.81
ToPin50.0 50.030.0->70.050.0 50.070.0->30.0 0.81
----------------------------------------------------------------
Library: 'tcbn28hpmbwp35ss0p81v0c_ccs'
Library Units:1ns1pF1kOhm
Library Cell: 'CKND1BWP35HVT'
arc sense:negative_unate
arc type:cell
Information: Using CCS timing info. (TIM-025)
Arnoldi-based Delay Calculation:
RC network on pin 'apbb_bj/apb1_master/U101/ZN' :
------------------------------------------------------
Number of elements = 32 Capacitances + 31 Resistances
Total capacitance= 0.005433 pF
Total capacitance= 0.005433 (in main library unit)
Total resistance= 0.279561 Kohm
Advanced driver-modeling used for rise and fall.
RiseFall
------------------------------------------------------------------
Input transition time= 0.0576840.055911(in main library unit)
Effective capacitance= 0.0055220.005512(in pF)
Effective capacitance= 0.0055220.005512(in main library unit)
Output transition time = 0.0893660.093764(in main library unit)
Cell delay= 0.0813170.090136(in main library unit)

pt的如下：
From pin: apbb_bj/apb1_master/U101/I
To pin:apbb_bj/apb1_master/U101/ZN
Main Library Units:1ns1pF1kOhm
Library: 'tcbn28hpmbwp35hvtss0p81v0c_ccs'
Library Units:1ns1pF1kOhm
Library Cell: 'CKND1BWP35HVT'
arc sense:negative_unate
arc type:cell
CalculationRiseRiseFallFallSlewRail
Thresholds:DelaySlewDelaySlewDerateVoltageTemp.
-----------------------------------------------------------------------
from-pin5030->705070->300.5000.8100.0
to-pin5030->705070->300.5000.8100.0
RC network on pin 'apbb_bj/apb1_master/U101/ZN' :
------------------------------------------------------
Number of elements = 44 Capacitances + 49 Resistances
Total capacitance= 0.005018 pF
Total capacitance= 0.005018 (in library unit)
Total resistance= 0.282406 Kohm
Advanced driver-modeling used for rise and fall.
RiseFall
------------------------------------------------------------------
Input transition time= 0.0599730.058407(in library unit)
Effective capacitance= 0.0048830.004871(in pF)
Effective capacitance= 0.0048830.004871(in library unit)
Output transition time = 0.0825450.086591(in library unit)
Cell delay= 0.0778490.086217(in library unit)
Annotated max rise delta transition: 0 pin transition: 0.0825446
Annotated max fall delta transition: 0 pin transition: 0.0865911
1

pt的初始transition 比icc的要差，但是别的看起来都比icc快了一些。
请教下大牛们是什么原因?

这个累积的效果非常惊人，整条路径的差距能到好几百p。
是我的icc的某些option设置的太悲观了?使得rc的抽取出来的数值比pt要悲观一些

我问个 问题： ICCLibrary: 'tcbn28hpmbwp35ss0p81v0c_ccs'和 PTLibrary: 'tcbn28hpmbwp35hvtss0p81v0c_ccs'有啥区别 ?我估计算timing的时候icc这个库的更悲观 我觉得是这个的差距 或者是你约束的不同