The current translation is perfectly grammatical and does carry the full meaning of the original quote. However, it is too literal, akin more to a Google translation than a native one. You wouldn't come across such a sentence in a book or newspaper for instance. The one I'm proposing in the other hand is more like a real-world rendition of the quote. I hope that people with a good grasp on Arabic would also weigh in on this.

This PR adds support for M1 Macs. I have extracted the process of obtaining the CPU name into a separate function to make a little bit more clean as the M1 doesn't need the loop. I also omitted the inclusion of cpuid.h in case of M1 as it was throwing a compiler error "This is only for x86". I guess that this has only inline asm for intel cpus.

Edit: I didn't know if I should open this PR against master or V3 so I went with master. If you want me to change that, feel free to tell me :)

I just wanted to share my benchmark results:

# gnome-terminal
CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
VT support: no
ManyLine: 82.4858s (0.0121gb/s)
LongLine: 71.3162s (0.0140gb/s)
FGPerChar: 5.1682s (0.0562gb/s)
FGBGPerChar: 9.6057s (0.0587gb/s)
TermMarkV2 Normal: 168.5759s (0.0169gb/s)

# alacritty
CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
VT support: no
ManyLine: 24.2993s (0.0412gb/s)
LongLine: 19.4805s (0.0513gb/s)
FGPerChar: 3.7274s (0.0779gb/s)
FGBGPerChar: 8.0369s (0.0702gb/s)
TermMarkV2 Normal: 55.5441s (0.0514gb/s)

# foot (https://codeberg.org/dnkl/foot)
CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
VT support: no
ManyLine: 19.0191s (0.0526gb/s)
LongLine: 6.5224s (0.1533gb/s)
FGPerChar: 2.8055s (0.1035gb/s)
FGBGPerChar: 5.2957s (0.1065gb/s)
TermMarkV2 Normal: 33.6426s (0.0848gb/s)

# Hyper (https://hyper.is/) - it had some issues with the color output
CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
VT support: no
ManyLine: 15.6713s (0.0638gb/s)
LongLine: 7.8218s (0.1278gb/s)
FGPerChar: 2.9988s (0.0968gb/s)
FGBGPerChar: 6.2595s (0.0901gb/s)
TermMarkV2 Normal: 32.7513s (0.0872gb/s)

According to my registry I have version 6.3..9600 (yes double dot) of the sdk.

So the ENABLE_VIRTUAL_TERMINAL_PROCESSING is not defined, I slapped in this after include windows.h and it worked fine.

#ifndef ENABLE_VIRTUAL_TERMINAL_PROCESSING
#define ENABLE_VIRTUAL_TERMINAL_PROCESSING 0x0004
#endif

Still slow though...

Is this something you are interested in?

other problems i see is that one may not always be interested in having all tests ran but only a select set of tests.

You can wake up the one who is sleeping, but not one who is pretending to sleep.

Îl poți trezi pe omul care doarme, dar nu și pe cel care se preface că doarme.

This change prints the Apple product name in place of CPU on "Apple Silicon" machines, and "Unknown" when the CPU is completely unknown.

Having the benchmark run with an unidentified CPU is probably more useful than having it fail to build on a new platform, so #warning is chosen over #error.

I'm not sure how to pull detailed ARM specs of the CPU, so there's potentially more work here, but this is enough to get things running on M1 machines.

Tested on a MacBook Air:

$  ./termbench small

CPU: MacBookAir10,1
VT support: yes
Fast pipes: no
ManyLine: 0.001gb / 0.038s (0.026gb/s)
LongLine: 0.001gb / 0.029s (0.034gb/s)
FGPerChar: 0.001gb / 0.045s (0.022gb/s)
FGBGPerChar: 0.001gb / 0.036s (0.027gb/s)
Binary: 0.001gb / 0.052s (0.019gb/s)
TermMarkV3 small: 0.201s (0.024gb/s)

Add a bunch of VT and Unicode stuff at the end just so you can see along with the benchmark whether they handle RTL, Arabic, Hebrew, blink, inverse video, whatever.

- Casey

It does not add many diacritics. I could formulate it in a way that adds more but it be less natural. "onog koji spava" would then be "spavajućeg".

1. Avoid bit/byte confusion IEC 60027 specifies that the symbol for binary digit should be 'bit', and this should be used in all multiples, such as 'kbit', for kilobit. However, the lower-case letter 'b' is widely used as well and was recommended by the IEEE 1541 Standard (2002).

2. Also avoid "giga" confusion. Giga according to SI is abbreviated as uppercase G and means 10^6. Termbench uses binary units (1024^n) therefore this should be written as Gi (gibi)

Running build.bat there are 2 warning and 2 errors. Sample error:

termbench.cpp:286:43: error: constant expression evaluates to 17179869184 which cannot be narrowed to type 'size_t' (aka 'unsigned int') [-Wc++11-narrowing]
    {"ManyLine", ManyLine, {1*Meg, 1*Gig, 16*Gig}},
                                          ^~~~~~
termbench.cpp:286:43: note: insert an explicit cast to silence this issue
    {"ManyLine", ManyLine, {1*Meg, 1*Gig, 16*Gig}},
                                          ^~~~~~
                                          static_cast<size_t>( )

How do you build termbench on Win11 + VS2022? Thx.