Half precision arithmetic (float16) is something which must be “emulated” by numpy I guess, as there are no corresponding types in the underlying C language (and in the appropriate processor instructions) for it. On the other hand, single precision (float32) and double precision (float64) operations can be done very efficiently using native data types. As … Read more
It’s called setting CPU affinity, and it’s an OS setting for processes, not specific to Java. On Linux: http://www.cyberciti.biz/tips/setting-processor-affinity-certain-task-or-process.html On Windows: http://www.addictivetips.com/windows-tips/how-to-set-processor-affinity-to-an-application-in-windows/ On Mac it doesn’t look like you can set it: https://superuser.com/questions/149312/how-to-set-processor-affinity-on-os-x
The short answer is that only new data gets sent down the wire. Here’s how it works. There are three important parts of the Meteor server that manage subscriptions: the publish function, which defines the logic for what data the subscription provides; the Mongo driver, which watches the database for changes; and the merge box, … Read more
It has to do with the fact that SQLite has a page cache. According to the Documentation, the default page cache is 2000 1K pages or about 2Mb. Since this is about 75% to 90% of your data, it isn’t surprising that the two number are very similar. My guess is that in addition to … Read more
The following code uses a windows function that is more accurate than Excel. It is taken from http://msdn.microsoft.com/en-us/library/aa730921.aspx#Office2007excelPerf_MakingWorkbooksCalculateFaster. The same page also contains some great tips on improving performance in Excel 2007. Private Declare Function getFrequency Lib “kernel32” _ Alias “QueryPerformanceFrequency” (cyFrequency As Currency) As Long Private Declare Function getTickCount Lib “kernel32” _ Alias “QueryPerformanceCounter” … Read more
There is no strict reason why a bytecode based language like C# or Java that has a JIT cannot be as fast as C++ code. However C++ code used to be significantly faster for a long time, and also today still is in many cases. This is mainly due to the more advanced JIT optimizations … Read more
The difference is due to memory pages of array not being resident in process address space (global scope array is stored in .bss section of the executable that hasn’t been paged in, it is zero-initialized). Whereas vector has just been allocated and zero-filled, so its memory pages are already present. If you add std::fill_n(v.data(), n, … Read more
Micro-benchmarking aside, the performance of the GCM implementation in JDK 8 (at least up to 1.8.0_25) is crippled. I can consistently reproduce the 3MB/s (on a Haswell i7 laptop) with a more mature micro-benchmark. From a code dive, this appears to be due to a naive multiplier implementation and no hardware acceleration for the GCM … Read more
Running any code for the first time may have some (significant) overhead, e.g. related code may be loaded, many things may be deferred until they are needed (e.g. internal buffers). Running the same thing again may take significantly less time, the difference may even be several orders of magnitude. Whenever you want to measure execution … Read more
The message “intermediate result is being cached” is just a blind guess in the canned message reported by %timeit. It may or may not be true, and you should not assume it is correct. In particular, one of the most common reasons for the first run being slowest is that the array is in the … Read more