POCV技术优势：减少芯片参数变异分析间的悲观估计

需积分: 20 39 浏览量更新于2024-07-16 收藏 1.22MB PDF 举报

本资源是Synopsys公司于2019年5月发布的《POCV（参数化芯片变异）应用指南》第三版（POCV AppNote Version 3）。POCV是一种与传统AOCV（架构级芯片变异）相比具有优势的芯片变异管理方法。它的核心在于通过参数化的手段来处理芯片设计中的变异问题，这有助于减少基于图分析和路径分析之间的过度悲观估计，从而提高设计的稳健性和性能预测的准确性。在POCV的应用中，关键知识点包括： 1. **概念理解**：POCV是针对芯片设计中的参数化变异模型，它允许设计师在设计初期就考虑变异因素，如工艺、温度、电压变化等，将这些影响纳入设计过程。这样做的目的是为了创建一个更为全面的设计，使得芯片在实际生产中的性能波动能够被更好地理解和控制。 2. **优势对比**：相较于AOCV，POCV的优势在于它能提供更精确的性能预测。AOCV可能过于悲观地估计了设计对于变异的敏感性，而POCV通过参数化的方法可以提供更加现实的性能范围，从而帮助工程师做出更合理的系统设计决策。 3. **版权和许可**：该文档受Synopsys公司所有，仅限于根据与Synopsys签署的书面许可协议使用。未经授权的复制、修改或分发都是严格禁止的。用户需注意遵守美国的出口管制法律，并确保不违反其他国家的相关法规。 4. **目的地控制声明**：所有技术数据受美国法律管辖，跨国家的信息披露必须符合美国法律，任何违反此规定的行动都是不被允许的。 5. **免责声明**：Synopsys公司及其许可方对本材料不做任何形式的保证，包括但不限于适销性和特定用途的适用性。用户在使用时应自行评估其适用性，并承担相应的责任。 6. **商标**：文档中提及的“Synopsys”公司及其部分产品名称均为Synopsys的商标。 POCV AppNote Version 3提供了对参数化芯片变异管理方法的深入探讨，为设计者提供了一种更为精确和实用的变异管理策略，以应对现代集成电路设计中的复杂性和不确定性。这对于提高设计效率，降低风险，以及优化芯片性能具有重要的实践指导意义。

table. For a net, the precedence from lowest to highest is the design POCV table and

hierarchical cell POCV table.

For POCV distance-based derating, the algorithm is exactly the same as distance-based

derating in AOCV analysis. For more details how bounding box is calculated in graph-

based analysis (GBA) and path-based analysis (PBA) in order to extract distance-based

derating in POCV analysis, please refer to AOCV Application Notes on SolvNet.

Advanced On-Chip Variation (AOCV) in PrimeTime

Guard-banding in POCV

Similar to AOCV flow, guard-banding allows you to model non-process related effects in

POCV flow. To perform guard-banding in POCV, specify the set_timing_derate

command with the -pocvm_guardband option.

For example, the following command applies a 5% guard-band POCV derating on both

early and late mode:

pt_shell> set_timing_derate –cell_delay \

-pocvm_guardband -early 0.95

pt_shell> set_timing_derate –cell_delay \

-pocvm_guardband -late 1.05

The POCV guard-band derating factor is applied on both the nominal cell delay (mean)

and cell delay variation (sigma).

To report only the guard-band derating factors, you must specify the

report_timing_derate command with the -pocvm_guardband option. To reset guard-

band derating factors, use the reset_timing_derate command with

the -pocvm_guardband option.

Scaling POCV Coefficient

In addition to guard-banding, the POCV flow also offers a way to scale only the variation

of cell delay without changing the coefficient in POCV tables. To scale the POCV

coefficient, specify the set_timing_derate command with

the -pocvm_coefficient_scale_factor option.

For example, the following command scales the POCV coefficient by 3% for both early

and late derating:

pt_shell> set_timing_derate –cell_delay \

-pocvm_coefficient_scale_factor -early 1.03

pt_shell> set_timing_derate –cell_delay \

-pocvm_coefficient -late 1.03

The POCV coefficient scaling factor is applied only on the variation of cell delay (sigma).

Figure 5 Example the output of report_ocvm –type pocvm command with POCV side file

pt_shell> report_ocvm -type pocvm [get_cell I]

******************************************************

POCV Table Set : *Default*

******************************************************

POCV coefficient: 0.0500

Name Object Process Voltage Sense Path Delay Inherited

-------------------------------------------------------------------------------

I cell early * rise clock cell lib1/INV1

I cell early * fall clock cell lib1/INV1

I cell early * rise data cell lib1/INV1

I cell early * fall data cell lib1/INV1

******************************************************

POCV Table Set : *Default*

******************************************************

POCV coefficient: 0.0500

Name Object Process Voltage Sense Path Delay Inherited

-------------------------------------------------------------------------------

I cell late * rise clock cell lib1/INV1

I cell late * fall clock cell lib1/INV1

I cell late * rise data cell lib1/INV1

I cell late * fall data cell lib1/INV1

To report all cells missing POCV coefficient and distance-based derating data, you can

do the following:

pt_shell> report_ocvm –type pocvm –cell_delay –list_not_annotated

To report all cells only missing POCV coefficient data only (when you did not apply

distance-based derating table in POCV analysis), you can do the following:

pt_shell> report_ocvm –type pocvm –cell_delay -list_not_annotated \

–coefficient

You can use the report_ocvm command before update_timing to check if there are

any cells missing POCV coefficient or distance-based derating table.

Reporting POCV Analysis Results

POCV analysis is supported by all reporting commands in PrimeTime. By default,

PrimeTime reports the timing at the 3-sigma value.

To display the variation in the timing report for POCV analysis, you can do

report_timing –derate command, which displays the “Mean” column and the

“Sensit” column. The “Mean” column is the nominal cell delay, i.e. the delay without

POCV or without variation. The “Sensit” column is the sensitivity or the variation of cell

delay due to the POCV coefficient.

pt_shell> report_timing -derate

By default, report_timing –derate command does not show “Mean” and “Sensit”

column. To see these columns, set the variable

variation_report_timing_increment_format to delay_variation.

pt_shell> set_app_var variation_report_timing_increment_format \

delay_variation

Figure 6 shows an example of the output of the report_timing –derate command for

PrimeTime version prior to M-2016.12.

Figure 6 The output of the report_timing –derate command

pt_shell> report_timing -derate -sign 5 -nosplit

Startpoint: i (input port clocked by CLK)

Endpoint: o (output port clocked by CLK)

Path Group: CLK

Path Type: max

Point Derate Mean Sensit Incr Path

--------------------------------------------------------------------------

clock CLK (rise edge) 0.00000 0.00000

clock network delay (ideal) 0.00000 0.00000

input external delay 0.00000 0.00000 r

i (in) 0.00000 0.00000 0.00000 0.00000 r

I/I (INV1) 1.00000 0.00000 0.00000 0.00000 0.00000 r

I/ZN (INV1) 1.02000 0.00975 0.00049 0.01121 & 0.01121 f

o (out) 1.00000 0.00007 0.00000 0.00007 & 0.01128 f

data arrival time 0.01128

clock CLK (rise edge) 1.00000 1.00000

clock network delay (ideal) 0.00000 1.00000

output external delay 0.00000 1.00000

data required time 1.00000 0.00000 1.00000

--------------------------------------------------------------------------

data required time 1.00000 0.00000 1.00000

data arrival time -0.00982 0.00049 -0.01128

--------------------------------------------------------------------------

slack (MET) 0.99018 0.00049 0.98872

The “Incr” column is calculated from the “Mean” and “Sensit” columns using the

equation below.

     

󰇛



󰇜







where K is the value of timing_pocvm_report_sigma. The default is 3. If slack sensit

is zero, the Incr value equals the Mean value.

剩余77页未读，继续阅读

xiaocanmeng

粉丝: 0
资源: 1

POCV技术优势：减少芯片参数变异分析间的悲观估计

POCV.zip_Omega

dpm相关资源文档

siliconsmart

解释这段代码def getPoly2Params(a,b): # x=np.array(df.loc[a:b,0].astype('float64')) # y=np.array(df.loc[a:b,1].astype('float64')) x = np.array(range(0, b-a+1)) y = np.array(df.loc[a:b,1]) # trend=np.polyfit(x,y,4) trend,pocv=curve_fit(Poly2func,x,y) trend=list(trend) return trend

pocv与aocv的区别

socv pocv aocv的区别

【人脸识别】基于matlab GUI PCA人脸识别（识别率）【含Matlab源码 802期】.md

2023-04-06-项目笔记 - 第二百八十七阶段 - 4.4.2.285全局变量的作用域-285 -2025.10.15

SSM+JSP手办周边商城答辩PPT.ppt

Ruby语言教程知识点详解 ruby重点知识点总结掌握.docx

最新资源