That's something we finally agree upon. And you'll understand what this means. It's very grim outlook for the industry. We're already seeing that a handful of people rule the markets. And this will only get worse. Only the people with the deepest pockets will build their own dedicated hardware and take all business. It's not a nice picture for the future, that's for sure ...Quote from amazingIndustry:
If you have all the resources, top ASM or C programmers, and TONS of development time then you should program everything in ASM and C. You may live in a perfect world, most hedge funds, investment banks, certainly I, do not!!!
You know, that's just theoretically not possible. C# uses .NET libraries. This means it is NOT code that is tailored to your specific needs. With C++ you can nail it down and only use code you really want.Quote from dewton:
So given the same algorithms used in C++ and C#, is it true that the .NET JIT would produce code that's faster than C++ code compiled with the Intel C++ compiler?
Quote from luckybastard:
That's something we finally agree upon. And you'll understand what this means. It's very grim outlook for the industry. We're already seeing that a handful of people rule the markets. And this will only get worse. Only the people with the deepest pockets will build their own dedicated hardware and take all business. It's not a nice picture for the future, that's for sure ...
Quote from luckybastard:
You know, that's just theoretically not possible. C# uses .NET libraries. This means it is NOT code that is tailored to your specific needs. With C++ you can nail it down and only use code you really want.
Quote from luckybastard:
But anyway, it IS kind of depressing. To be among the fastest is getting more and more expensive by the day. a 40gbe connection (which offers several microseconds improvement over 10 gbe) to arca and nasdaq alone, together with data and some FGPA's will set you back around $50,000 per month. And then, you'll have a commercial FPGA. The big houses will have their own hardware programmers and will have FPGA's that are faster than yours. This race is getting more and more expensive and only a very few firms will be able to keep going like this ...
Quote from luckybastard:
But anyway, it IS kind of depressing. To be among the fastest is getting more and more expensive by the day. a 40gbe connection (which offers several microseconds improvement over 10 gbe) to arca and nasdaq alone, together with data and some FGPA's will set you back around $50,000 per month. And then, you'll have a commercial FPGA. The big houses will have their own hardware programmers and will have FPGA's that are faster than yours. This race is getting more and more expensive and only a very few firms will be able to keep going like this ...
Maybe the key to our differences is that we're talking about different things here. If you need to be THE fastest, then you don't have another option than C++ (or ASM, but I still think C++ would be the right choice). However this also comes with the need for steep monthly costs (easily more than $50,000/month). A firm can decide 'ok, we're not going for the really 'free' money that only the fastest out there will be able to pick up, but we're going to differentiate ourselves in the kind of strategy we run'. In that case, not every millisecond will count and C# probably is the right choice for your needs.Quote from amazingIndustry:
There are trading houses that run hft algos on windows boxes, linux boxes, they run C# code, C++ code, Erlang, and many others. They all compete for microseconds and they would not run C# or in your opinion "worthless" languages if they lost money. Hence, your analysis just is not correct.
Thats all I have to say to this.