4 ms·
It'd be neat to instead so this with the latest 8th gen i7 mobile chips that have 6 cores or the Ryzen 2700. Those 6 cores will make a huge impact for most soft
by Improvotter 8y ago
It'd be neat to instead so this with the latest 8th gen i7 mobile chips that have 6 cores or the Ryzen 2700. Those 6 cores will make a huge impact for most software development people. Maybe 32GB is possible if you can change it yourself unless it's unsupported.
- loup-vaillant 8y ago> Those 6 cores will make a huge impact for most software development people. I'm a software developer, and I fail to foresee such an impact. For me personally, and for the vast majority of the people I worked with directly in the past. Could you explain how 6 cores (and possibly 32GB of memory) it would make a dev's life easier, as opposed to 2-4 cores and 8-16GB?
- blux 8y agoIndeed, an over generalization, it of course does not apply to every software developer. But as a C++ developer I would appreciate extra cores to cut down compilation times. And even this won’t be true for every C++ developer in case their code base is relatively small, very well modularized, or you can use something like distcc.
- loup-vaillant 8y agoAh, unacceptably slow compilation times… I agree C++ is really bad at this (at least by default). This is especially frustrating when you know of Jonathan Blow's language, Go, and Wirth's compilers.
- pjmlp 8y agoC++ is compilation is relatively fast with incremental building and linking on Visual C++. Maybe UNIX compilers need to do some improvements here. And when using the experimental modules support it is even faster.
- loup-vaillant 8y agoWe shouldn't need incremental builds to be fast. Incremental builds have their own problems, which I'd rather avoid if at all possible. (They're more complex and leave more room for error than rebuilding everything every time.) I want to compile 100K lines of code or more in less than one second. Compared to that, C++ is slow as molasses.
- briandear 8y agoCompiling and building. A really big C++ project can take hours to build. My 18 core/128GB iMac Pro builds in 2.25 hours what it takes a 64 GB, 12 core Mac Pro over 5 hours to build. I’m probably on the far extreme of normal, but more cores and more RAM make a huge difference in tasks like rendering, compiling, building projects with huge dependency chains. For writing some JavaScript web application, the serious horsepower makes a only marginal difference in actual productivity. But on “big” stuff, it can save hours per day, which means savings of real money.
- STRML 8y agoIt still can on JS toolchains. There is multiprocess webpack (happypack) now, Flow threads really well, and the extra cores can even help when restarting a bunch of processes at once and live reloading. Shaving off even a second helps productivity when you do it dozens/hundreds of times a day.
- loup-vaillant 8y ago> A really big C++ project can take hours to build. My 18 core/128GB iMac Pro builds in 2.25 hours what it takes a 64 GB, 12 core Mac Pro over 5 hours to build. While I understand such monsters do exist, I cannot help but balk in horror at the though of millions of lines of code written in such an unsafe language. Things must have gone wrong on several levels to get to this point.