Speed tests: add test for largest data

skynowa commented 9 years ago

    #define str \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg" \
        "asdcf'awsrk.fw3[045im2,[4oc'wxe;/fzd1[-,c4rxj.wk;szlcgjx34-rx134rxqwetaertg"
...
    } else if (which == "tinyformat") {
        for (long i = 0; i < loops; ++ i) {
            tfm::printf("%0.10f" str "%04d" str "%+g" str "%s" str "%p" str "%c" str "%%\n",
                1.234, 42, 3.13, "string", (void *)1000, (int)'X');
        }
    }
...

c42f commented 9 years ago

That's certainly an odd value for str, is there any reasoning behind it?

skynowa commented 9 years ago

Soory. I mean, that add test for longest string. Instead "string" use str.

#define str "longest_string...................................................abc"

tfm::printf("%0.10f:%04d:%+g:%s:%p:%c:%%\n",
    1.234, 42, 3.13, str, (void *)1000, (int)'X');

c42f commented 9 years ago

Sorry for the delay, I forgot about this.

Your second example is testing something quite different from the first, so I'm still not quite sure what the intention is here.

Your first suggestion tests the speed at which tinyformat processes a really long format string. This may be of some interest in particular use cases.
Your second suggestion tests the speed at which a string argument can be handled when printing with "%s". This is really testing the underlying implementation of operator<<(std::ostream&, const char*), but I guess it might be of interest to someone.

Neither of these seem like particularly typical use cases, so I probably wouldn't put them in the main speed test - it is meant to be a simple "smoke test" to check that nothing has gone horribly wrong. However, if you'd care about performance, I'd encourage you to submit a pull request expanding the speed test into a full test suite with various different cases of interest. In that case I think it would make sense to compare against both libc printf and iostreams (boost::format is so far behind that we may as well ignore it)

c42f / tinyformat

Speed tests: add test for largest data #22