Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    We did a lot of testings of BLIS/OpenBLAS, the issue is mainly stability (both in terms of numerical and functional) across a large hardware and OS variations we have to support. That said, we do not exclude the possibility of allowing users to supply and use their own BLAS libraries; of course, in the a future Stata.

    Comment


    • #17
      Thanks for the response, Hua Peng (StataCorp). That would be a fantastic feature. Either way, I hope that you all continue to test the stability of BLIS/OpenBLAS.

      AMD over the last few years has been beating up Intel in the CPU space and a lot of folks, myself included, have switched over to AMD CPUs for our computing needs.

      Comment


      • #18
        At least for my current project, Stata 17 is much faster than Stata 16, which I presume is due to the Intel MKL. I am simulating a small VAR model and obtaining bootstrap forecast intervals within each round of simulation, and all of the code is written in Mata.

        Stata 17 is using one core, and its output log reads

        Code:
        Sim     5   Time 12.97   Cumulative    64.37   ETA 2021-04-28 00:29:15 AM
        Sim    10   Time 13.35   Cumulative   129.70   ETA 2021-04-28 01:29:40 AM
        Sim    15   Time 13.05   Cumulative   196.05   ETA 2021-04-28 02:32:10 AM
        I set up Stata 16 to run the exact same do-file, Stata is using one (different) core, and its output log reads

        Code:
        Sim     5   Time 20.82   Cumulative   103.00   ETA 2021-05-01 08:58:19 AM
        Sim    10   Time 20.68   Cumulative   208.22   ETA 2021-05-01 11:16:51 AM
        Sim    15   Time 20.80   Cumulative   312.18   ETA 2021-05-01 11:10:57 AM
        Since many estimators are written in Mata, I suspect they will run faster as well. Note that I am not working with a particularly large dataset or matrices -- just average sized matrices typical of time-series analysis.

        Good job StataCorp!

        Comment


        • #19
          Originally posted by Joseph Coveney View Post
          During previous installations, I remember being prompted to uninstall the prior release's installation before completing installation of the new one. That didn't happen this time and I didn't notice any mention of it in the online guidance. I take it that it's become moot.

          I plan to uninstall the previous release, anyway, but now that I've got the new release installed, I trust that uninstalling the prior one won't mess up the new one's installation, right?
          There is no harm in uninstalling an older version of Stata before installing a newer version of Stata. We recommend it but it's not required. The Stata for Windows installer has never prompted the user to uninstall an older version of Stata when installing a newer version of Stata.

          When installing Stata for Mac, we recommend uninstalling an older version of Stata before installing a newer version of Stata. This is because on older versions of macOS, double-clicking a dataset or do-file could potentially launch the wrong version of Stata if multiple versions of Stata are installed so the installer would prompt the user to delete an existing Stata folder if one was detected. In more recent releases, the Stata for Mac installer will simply rename an existing Stata folder if one is detected rather than delete it. Modern versions of macOS will launch the most recently installed version of Stata if multiple copies are installed so it's less of an issue.
          Last edited by Chinh Nguyen (StataCorp); 22 Apr 2021, 14:54.
          -Chinh Nguyen

          Comment


          • #20
            Congratulations!
            When upgrading STATA, is there a way to copy the commands installed by ssc install from the previous version?

            Comment


            • #21
              No need to copy. Unless you specifically change the adopath, they're still there in the same place and Stata 17 knows to look for them there.

              Comment


              • #22
                Richard, Chinh,Thank you for replying to my question. And apologies for misremembering things.

                Comment


                • #23
                  On the trivial side,Stata/IC (Intercooled) has been replaced by Stata/BE (Basic Edition). I always thought Intercooled was a really neat name but I had no idea what it meant, and I suspect most beginning users had no idea either. The new name isn't as neat but it probably has a clearer meaning.
                  -------------------------------------------
                  Richard Williams, Notre Dame Dept of Sociology
                  Stata Version: 17.0 MP (2 processor)

                  EMAIL: [email protected]
                  WWW: https://www3.nd.edu/~rwilliam

                  Comment


                  • #24
                    Just made a HUGE jump from Stata 13 to 17.

                    I use the table command a lot and was used to be able to use string variables as rownames when, for example, I was getting some summary statistics of a given variable which could be categorised (by my string var).
                    It seems this is not possible anymore, although because of my jump maybe it hasn't been for quite a while already.

                    Has anyone else had the same issue?

                    I can obviously get around it by encoding my categorical string variable, but it will take some time to do it having a big number of old datasets around.

                    Comment


                    • #25
                      The old -table- should still be available under version control. Prepend your command with -version 13: table ....whatever .... -

                      Originally posted by Nicolas Alejandro Rey gonzalez View Post
                      Just made a HUGE jump from Stata 13 to 17.

                      I use the table command a lot and was used to be able to use string variables as rownames when, for example, I was getting some summary statistics of a given variable which could be categorised (by my string var).
                      It seems this is not possible anymore, although because of my jump maybe it hasn't been for quite a while already.

                      Has anyone else had the same issue?

                      I can obviously get around it by encoding my categorical string variable, but it will take some time to do it having a big number of old datasets around.

                      Comment


                      • #26
                        Originally posted by Leonardo Guizzetti View Post
                        The old -table- should still be available under version control. Prepend your command with -version 13: table ....whatever .... -


                        Hello,
                        Thank you very much.
                        Works like charm.
                        I guess I never had to downgrade versions while using Stata 13 for a long time.

                        Regards

                        Comment


                        • #27
                          The other thing you can do is -encode- that string variable and use the encoded version in the current version of table instead.

                          Comment


                          • #28
                            Originally posted by Clyde Schechter View Post
                            The other thing you can do is -encode- that string variable and use the encoded version in the current version of table instead.
                            Hello,
                            Yes of course. And its what I'll be doing from now on (encode categorical vars in the data pre-processing).
                            Nevertheless it is useful to be able to call on version control when I need to rerun do files from older projects.

                            Comment


                            • #29
                              Is it possible to upgrade from Stata 16 to Stata 17? If that's the case, how much does it costs and how? Thanks!

                              Comment


                              • #30
                                Go to www.stata.com and click on Purchase. Follow the instructions to branch through the various options that apply to you. There are different prices depending on your specific situation.

                                Comment

                                Working...
                                X