search

The-OpenROAD-Project / OpenSTA

OpenSTA engine
GNU General Public License v3.0
399 stars 172 forks source link

report_checks with -from -through -to options #16

Closed guannan-git closed 5 years ago

guannan-git commented 5 years ago

Dear developer,

I am using the report_checks command with -through and -to options to show top k critical paths involving through pins and destination pin. I can get k paths by using -endpoints k, but slacks of these paths are always INF if I choose to use the register input (commonly considered as a type of endpoint) as -to pin. I experimented this on a classic s27 design. The command I used is

report_checks -unconstrained -format end -endpoint_count 4 -through inst_6/A1 -to inst_16/D

I obtained the following result.

inst_16/D (DFFR_X2) INF 202.22 INF (MET) inst_16/D (DFFR_X2) INF 181.02 INF (MET) inst_16/D (DFFR_X2) INF 177.97 INF (MET) inst_16/D (DFFR_X2) INF 153.33 INF (MET)

If I choose to use primary output pin G17 as -to pin. The slack values are valid.

G17 (output) -0.20 177.68 -177.88 (VIOLATED) G17 (output) -0.20 174.15 -174.35 (VIOLATED) G17 (output) -0.20 153.04 -153.24 (VIOLATED) G17 (output) -0.20 138.33 -138.53 (VIOLATED)

I am not sure if I missed any steps or I used the wrong pin as -to pin. I have attached the entire design below. "s27.tcl" is the file I used as command file to OpenSTA and "opensta_prev.log" is the logfile I had when I ran the input command file.

s27.zip

Thanks

jjcherry56 commented 5 years ago

Dear User,

You have some problems with your libraries. s27_Early.lib is missing setup_rising/falling timing groups s27_Late.lib is missing hold_rising/falling timing groups

The structure of the libraries must match exactly.

On Monday, March 18, 2019, Guannan Guo notifications@github.com wrote:

Dear developer,

I am using the report_checks command with -though and -to options to show top k critical paths involving though pins and destination pin. I can get k paths by using -endpoints k, but slacks of these paths are always INF if I choose to use the register input (commonly considered as a type of endpoint) as -to pin. I experimented this on a classic s27 design. The command I used is

report_checks -unconstrained -format end -endpoint_count 4 -through inst_6/A1 -to inst_16/D

I obtained the following result.

inst_16/D (DFFR_X2) INF 202.22 INF (MET) inst_16/D (DFFR_X2) INF 181.02 INF (MET) inst_16/D (DFFR_X2) INF 177.97 INF (MET) inst_16/D (DFFR_X2) INF 153.33 INF (MET)

If I choose to use primary output pin G17 as -to pin. The slack values are valid.

G17 (output) -0.20 177.68 -177.88 (VIOLATED) G17 (output) -0.20 174.15 -174.35 (VIOLATED) G17 (output) -0.20 153.04 -153.24 (VIOLATED) G17 (output) -0.20 138.33 -138.53 (VIOLATED)

I am not sure if I missed any steps or I used the wrong pin as -to pin. I have attached the entire design below. "s27.tcl" is the file I used as command file to OpenSTA and "opensta_prev.log" is the logfile I had when I ran the input command file.

s27.zip https://github.com/abk-openroad/OpenSTA/files/2980157/s27.zip

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/abk-openroad/OpenSTA/issues/16, or mute the thread https://github.com/notifications/unsubscribe-auth/AhI8lcJOeHB5Ad5eSl_LwErrSqZQ1iBiks5vX_uJgaJpZM4b6vpf .

jjcherry56 commented 5 years ago

Looking closer the command script is pretty confused. The read_liberty commands redefine the same library cells without distinguishing the min and max libraries. Do you think it is going to use the file name to understand which is which? It should look like this:

read_liberty -min $early_lib read_liberty -max $late_lib

Again, the structure of the libraries file must be identical. The only difference should be the numbers. Perhaps an error or warning should be printed, but I've never seen this sort of confusion.

There is no point in calling "find_timing -full". That is implicit in all of the reporting commands.

It makes no sense to call "log_begin" twice.

On Monday, March 18, 2019, James Cherry cherry@parallaxsw.com wrote:

Dear User,

You have some problems with your libraries. s27_Early.lib is missing setup_rising/falling timing groups s27_Late.lib is missing hold_rising/falling timing groups

The structure of the libraries must match exactly.

On Monday, March 18, 2019, Guannan Guo notifications@github.com wrote:

Dear developer,

I am using the report_checks command with -though and -to options to show top k critical paths involving though pins and destination pin. I can get k paths by using -endpoints k, but slacks of these paths are always INF if I choose to use the register input (commonly considered as a type of endpoint) as -to pin. I experimented this on a classic s27 design. The command I used is

report_checks -unconstrained -format end -endpoint_count 4 -through inst_6/A1 -to inst_16/D

I obtained the following result.

inst_16/D (DFFR_X2) INF 202.22 INF (MET) inst_16/D (DFFR_X2) INF 181.02 INF (MET) inst_16/D (DFFR_X2) INF 177.97 INF (MET) inst_16/D (DFFR_X2) INF 153.33 INF (MET)

If I choose to use primary output pin G17 as -to pin. The slack values are valid.

G17 (output) -0.20 177.68 -177.88 (VIOLATED) G17 (output) -0.20 174.15 -174.35 (VIOLATED) G17 (output) -0.20 153.04 -153.24 (VIOLATED) G17 (output) -0.20 138.33 -138.53 (VIOLATED)

I am not sure if I missed any steps or I used the wrong pin as -to pin. I have attached the entire design below. "s27.tcl" is the file I used as command file to OpenSTA and "opensta_prev.log" is the logfile I had when I ran the input command file.

s27.zip https://github.com/abk-openroad/OpenSTA/files/2980157/s27.zip

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/abk-openroad/OpenSTA/issues/16, or mute the thread https://github.com/notifications/unsubscribe-auth/AhI8lcJOeHB5Ad5eSl_LwErrSqZQ1iBiks5vX_uJgaJpZM4b6vpf .

guannan-git commented 5 years ago

Thanks a lot for your reply! I am using read_liberty -min $early_lib read_liberty -max $late_lib to distinguish between min and max libraries. However, I still get INF on slacks. Should I merge both min and max libraries so that the merged file has both setup_rising/falling and hold_rising/falling groups?

jjcherry56 commented 5 years ago

I'll say it a third time. The libraries much be structurally the same. Every port. Every timing group. I am adding error messages to check them. Normally libraries are generated by scripts that know better.

On Monday, March 18, 2019, Guannan Guo notifications@github.com wrote:

Thanks a lot for your reply! I am using read_liberty -min $early_lib read_liberty -max $late_lib to distinguish between min and max libraries. However, I still get INF on slacks. Should I merge both min and max libraries so that the merged file has both setup_rising/falling and hold_rising/falling groups.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/abk-openroad/OpenSTA/issues/16#issuecomment-474204123, or mute the thread https://github.com/notifications/unsubscribe-auth/AhI8lbKdIUn2r_GblngLwTvm-S8pZk7bks5vYHaEgaJpZM4b6vpf .

guannan-git commented 5 years ago

I can get slack values now. Thanks!

jjcherry56 commented 5 years ago

I commited checks that report warnings for libraries without matching timing groups. Unfortunately, tau contest data is based on IBM's broken libraries that suffer from this problem. I also added some example command files and libraries that will be used in future tau contests that use the same cell names and are correctly formed.