Largest number in ten bytes of code [closed]

Your goal is to print (to the standard output) the largest number possible, using just ten characters of code.

• You may use any features of your language, except built-in exponentiation functions.
• Similarly, you may not use scientific notation to enter a number. (Thus, no 9e+99.)
• The program must print the number without any input from the user. Similarly, no reading from other files, or from the Web, and so on.
• Your program must calculate a single number and print it. You can not print a string, nor can you print the same digit thousands of times.
• You may exclude from the 10-character limit any code necessary to print anything. For example, in Python 2 which uses the print x syntax, you can use up to 16 characters for your program.
• The program must actually succeed in the output. If it takes longer than an hour to run on the fastest computer in the world, it's invalid.
• The output may be in any format (so you can print 999, 5e+100, etc.)
• Infinity is an abstract concept, not a number. So it's not a valid output.
• Did you look at these two? codegolf.stackexchange.com/questions/185/… codegolf.stackexchange.com/questions/18028/… Jun 13 '14 at 13:35
• What do you exactly mean by "calculate". Also, If it takes longer than an hour to run on any computer in the world, it's invalid. is not objective. I could (theoretically) manufacture a computer that takes an hour to change one T-state Jun 13 '14 at 13:37
• Does bit-shift count as an exponentiation operator since it is equivalent to * 2^x ? Jun 13 '14 at 16:08
• The fact that the time limit depends on the performance of the fastest computer in the world makes it impossible for us to determine the set of valid answers... I don't really like that Jun 13 '14 at 19:45
• Exponentiation functions are not allowed, but ackermann function is allowed? That seems really arbitrary. It remains unclear why a symbolic manipulation is considered "calculating" the number. Jul 7 '15 at 13:12

Brainf**k 256 - 2147483647

>+[<+>+]<.


If you ignore the fact that most compilers & interpreters output data as it's ascii equivalent (be leanient, it is what it is ;) ), this will return the maximum value of the interpreter/compiler's datatype.

On some systems this is just 256, although on some (mine for example), this is the max value of a 32 bit integer, ie 2 147 483 647.

Edit:

-.


Will print the same thing in many fewer characters

• That second answer will print -1 on interpreters that use signed values for the tape Apr 13 '19 at 23:46

Braingolf, I don't even know [non-competing]

#~U&^^^^^^


Doesn't work on TIO as I need to get Dennis to pull, but as you can see here, U in Braingolf calculates a range from 1 to the last item on the stack, in this case ~ or 126

Then reduces the entire stack with exponentiation, meaning the stack now contains 1 value which is equal to 126^125^124...^2^1

We'll call this value n from now on.

Then I had some bytes left, so I filled them with some monadic ^ operators.

This means the final output is (((((n^n)^n)^n)^n)^n)

For reference, 126^125 is 3.518180682714907e+262 and (126^125)^124 is ~35000 digits long

• For all intended purposes, this is approximately a_127 in the exponential factorial. Nov 12 '17 at 21:01

Perl, non competing

I'm using this to highlight a little know corner of perl.

Perl can't really compete on this one because it doesn't have builtin bignums (of course you could load a bignum library).

But what everybody knows isn't completely true. One core function actually can handle big numbers.

The pack format w can actually convert any size natural number between base 10 and base 128. The base 128 integer is however represented as string bytes. The bitstring xxxxxxxyyyyyyyzzzzzzz become the bytes: 1xxxxxxx 1yyyyyyy 0zzzzzzz (every byte starts with 1 except the last one). And you can convert such a string to base 10 with unpack. So you can write code like:

unpack w,~A x 4**4 .A


which gives:

17440148077784539048602210552864286760481312243331966651657423831944908597692986131110771184688683631223604950868378426010091037391551287028966465246275171764867964902846884403624214574779667949236313638077978794791039372380746518407204456880869394123452212674801443116750853569815557532270825838757922217314748231826241930826238846175896997055564919425918463307658663171965135057749089077388054942032051553760309927468850847772989423963904144861205988704398838295854027686335454023567793114837657233481456867922127891951274737700618284015425


You can replace the 4**4 by larger values until you feel it takes too long or uses too much memory.

Unfortunately this is way too long for the limit of this challenge, and you can argue that the base 10 result is converted to a string before it becomes the result so the expression doesn't really produce a number. But internally perl really does the needed arithmetic to convert the input to base 10 which I always considered rather neat.

Perl 6, 456,574 digits

[*] 1..ↈ


No TIO because it takes 2 minutes to run.

Aceto

kτ\p******u


Prints out

1 * 10846 (2017)

1.1 * 10846 (2018)

1.3 * 10846 (2019)

6.89 * 1013537 (2019)

Explanation

k - makes the stack 'sticky' so that values when popped are only copied,
not removed
τ - pushes the date on the stack (the year is on the top, 2017 right now...)
\ - escapes the next character (print)
p - (ignored on first go-around)
*
*
*
*        Multiplies top element by second, but stack has one element so it does it by itself
*
*
*
*
u - reverses instruction pointer direction
*
*
*
*
*
*
*
*
p - prints
\ - escapes the date
k - makes stack sticky


Try it online!

• actually you can add another * since p is excluded from bytecount Apr 4 '19 at 0:47

Keg, 260144641↑260144641

*:*(:|:*


Surprised I was able to get it this high, probably can be improved

Edit: Improved by A__

How it works

   Pushes 127 to the stack twice, I used 127 since it is the highest number that can be pushed in 1 byte (Note there is two characters there, they're just unprintables)
*    Multiplies them
:*   Squares that
(    Begin a for loop
:|   That will run 260144641 times (since that is the top element of the stack)
:*   That squares the top element of the stack
At the end it will automatically get printed

• No, z isn't the largest number possible with 1 byte; 0x7F is. I usually use unprintable characters when I golf in Keg, to represent specific constants.
– user85052
Aug 10 '19 at 12:57
• Oh wow, I was struggling to paste that into TIO, thanks for the better solution :) Aug 10 '19 at 13:03
• Now that I think about it, I'm pretty sure this is way higher, but I have no clue how to calculate what that would be Aug 10 '19 at 13:18
• Wait nvm, I forgot the | on the first bracket Aug 10 '19 at 13:19
• If you were using that, you would need to close out the brackets at the end in order to output it, so I don't think it could be higher Aug 10 '19 at 13:25

><>, ~$$\4.24 \times 10^{1177545600} \$$

'****l(?;n


Try it online!

Outputs the ordinal value of * (42) around 4088700 times. Every loop it only terminates if the length of the stack is higher than multiplying the ordinal values of l(?;n = 108*40*63*59*110 = 1766318400, but since it only pushes 3 elements to the stack each time, this takes a long while.

Perl, 1.84467422290351e+26 (On a 64-bit machine)

perl -le "print ~0*9999999"

– Zaid
Jun 15 '14 at 14:25

$txr -p '(mask 999)' 535754303593133660474212524530000905280702405852766803721875194185175525562468061246 599189407847929063797336458776573412593572642846157810728751889829846429852761096554 990320661140395677219337642394922319490470301292036210344653556258987007434741839952 7286296858625998634149561158533358569939198279680  • What is this doing? Jun 16 '14 at 0:15 • @g.rocket It makes a bit mask in which bit 999 is set. It takes multiple arguments, e.g (mask 1 2) yields 6. – Kaz Jun 16 '14 at 2:12 C++ prints 18446744073709551615 # ULLONG_MAX is 10 characters. #include <iostream> #include <limits.h> using namespace std; int main() { cout<<ULLONG_MAX<<endl; return 0; }  • And what value does that print? Jun 14 '14 at 2:15 • @KyleKanos Fixed it, thanks Jun 14 '14 at 5:07 Bash / shell You may exclude from the 10-character limit any code necessary to print anything. ... you can use up to 16 characters for your program. (6 chars for code + 7 chars for printing = 26 digits) = 6888888 digits $ seq -s9 999999


This number is bigger, but people may argue that it's in the wrong format.

The output may be in any format (so you can print 999, 5e+100, etc.).

(8 chars for code + 8 chars for printing = 26 digits) = more than 1183888008 digits

$seq -s9 99999999  It usually takes around a minute to print the number and it contains the scientific representation of it. Some funny example: (10 chars = 26 digits) echo $$2760127601276012760127601  • Your program must calculate a single number and print it. You can not print a string, nor can you print the same digit thousands of times. Jun 14 '14 at 14:55 Python 2.7 - 8794643931199480236 15616093818140822 print hash('zz')  • Um, that's 12 chars. Jun 13 '14 at 19:48 • Ehh valid point @user80551. Guess I was slightly distracted by the footy match on TV :-) Jun 14 '14 at 6:17 Perl: 6.27710173538668e+57 on cygwin with a Perl 32 bit perl -e 'print ~0*~0*~0, "\n"'  Ozone, 9e+90000 (1)n1n1s2  Though it depends on dialect. JavaScript, more than 13 003 624 633 896 (but it reads a clock) 9*new Date  Reading from a clock is effectively taking an input. Feels like cheating. However, the technique seems to be well-received in the answers here. If clocks are not allowed, then: JavaScript, 186 025 771 008 99*(7<<28)  Trigonometry takes too many bytes (Math.sin, Math.PI). Bit-shift goes near the integer maximum, then multiplication yields a floating point. • Bit shift operation are limited to 32 bit integers. 60 == 28 used in a shift operation (try in console: [7<<28, 7<<60]) Jun 13 '14 at 23:07 Brain-Flak, 100 (non-competing) (()@lt99)  Note that this code must be run with the -d flag which adds 3 bytes. Try it online! This uses the @lt debug flag in the ruby interpreter. This flag simply adds the following number to the value which is then pushed by the enclosing parentheses. The enclosing parentheses do not modify the number and are required for it to be printed so they are not counted towards then 10 bytes. The () adds one to the value and causes the interpreter to recognize the enclosing parentheses as pushing a value (parentheses will only push a value if they enclose non-debug flag Brain-Flak code. Tcl, number in the Demo Output because it is very extense (has 30104 digits) namespace path tcl::mathop puts [<< 9 99999]  The relevant part is << 9 99999  , the rest is boilerplate. Try it online! Acc!!, 999999 19*3 Write _ Write _ Write _ Write _ Write _ Write _  As any code necessary to output anything is excluded, all 'Write ' statements are excluded. Try it online! • > Your program must calculate a single number and print it. You can not print a string, nor can you print the same digit thousands of times. Aug 10 '19 at 13:50 PHP, 9 bytes, MAX INT (9223372036854775807) echo ~0^1<<63;  or echo -1^1<<63;  Set an int of all bits on (-1) then XOR it with an int of only the first bit on to flip the sign bit. Pyt, (only 6 bytes) 9!ḞɳƖ*  Try it online! Link to number I don't really see any point in getting any bigger than this. Anything else exceeds the runtime limit on TIO 9 pushes 9 ! factorial of 9 (362880) Ḟ push xth Fibonacci Number (really big) ɳ push 0123456789 ₫ reverse ("987654321") Ɩ cast to int ("987654321") * multiplies  Runic This answer is only works, if the following things are true: 1. I implement the Ackermann function using a non stack overflowing implementation. • Which I did briefly last night, but removed as there is no useful reason to keep it • It did successfully calculate Ack(4,3) over the course of 20-30 seconds, a number with 19729 digits. 2. If there are no arbitrary execution limits (e.g. TIO's 60 second timeout) 3. The executing computer has an arbitrary amount of memory to store exceedingly large stacks of BigIntegers But if given those, then this program... 'ÿY:A:A:A@  ...prints a number so large simply calculating how many digits it has is impossible, due to the explosive nature of the Ackermann sequence. Heck the inputs to the first call are both 255000! The output (and thus input to the next call) is (in Knuth up-arrow notation): 2 ↑254998255003 - 3 (that's 254 BILLION arrows!) Explanation > Implicit entry 'ÿ Pushes the character ÿ onto the stack (which will be implicitly converted to an int whenever a math operator is called on it). Y Multiply by 1000 (the 'biggest' thing we can apply to a single stack value in 1 byte) : Duplicate top item on the stack A Pop x and y, call Ack(x,y) : Duplicate top item on the stack A Pop x and y, call Ack(x,y) : Duplicate top item on the stack A Pop x and y, call Ack(x,y) @ Print the entire stack, top to bottom (there's only 1 value) and terminate  Larger still? AAAAAAAA@  Push the character sequence A,A,A,A,A,A,A,A,@ onto the stack (9 items) as the instruction pointer begins a character sequence on  and loops around until it sees  again. Each A then invokes Ack(x,y) (where x is the top of the stack) until the stack contains 1 entry, printing it. Not sure if this is actually larger or not. However... The closest we can get by abiding by the limitations is this: ':pba,*@  (Note the non-printing character 0x8F in position 2) Try it online! Which prints 1.79657789596691E+308 (143143 * 1.1). The multiplier of 1.1 is the largest that can be applied without rolling over to infinity due to the limits on a double: even multiplying by a9, (1.11) is sufficient to trigger the infinity bit. Perl 6, 730 quadrillion digits 9+<now*now  This builds the Perl 5 solution (time*time) and the Javascript (bit shifting). Time is currently around 1.56 billion, so we square it (~2.44 quintillion), and then bitshift 9 (because > 1 ha) a fairly insane number of digits to the right. Perl 6 uses arbitrary precision integers, and that applies to bitshifting as well. Bit shifting in and of itself doesn't take hardly any time at all, it's just a question of memory restraint and the largest super computers ought to have sufficient, but I don't on my puny laptop. Given 210x ≈ 103x, we can surmise that there should be around some 730 quadrillion digits if I did my math right. Times out on TIO for 9+<$x where \$x > 1000000, but handles 100000 in sub 1 secs ao no TIO link. Answer of course will continue to grow exponentially

(if that's considered out of bounds, though, can just make it 9+<now which should still fetch a fairly high number, or any other technique to generate the largest in-bounds number at in 7 digits)

• Wouldn't a bitshift of 1 quadrillion (base 2) be 300 trillion digits (base 10)? Jul 28 '19 at 3:56
• @Draco18s erm yes. Should be 730 quadrillion off of 2.4 quintillion digits (a billion squared nets a quintillion not a quadrillion) Jul 28 '19 at 6:49
• I knew one of the two was wrong. I just didn't validate the billion-squared. ;) Jul 28 '19 at 16:36
• @Draco18s I blame my lack of coffee at the time lol. In any case, result = "really big", certainly bigger than anything not using a built-in ack() function Jul 28 '19 at 18:42

Keg, 1.85302E+15

It is definitely the worst answer here...

9:*:*:*:*.

• Got it up to 221533456↑221533456 by using a for loop with the fact that alpha-numeric characters get auto-pushed to the stack Aug 10 '19 at 12:53

BBC BASIC 2

7 bytes

P.2^127


Result

1.70141181E38


As commented the previous answer may violate the exponation rule, so here's something using the EXP (exponent) function instead.

P.EXP(88)


which returns

1.65163622E38


• Is this using scientific notation? Aug 16 '19 at 14:52
• It is 2 raised to the power of 127, so if that is consider "exponentiation" I'll retract. The example above was 9e+99. Aug 16 '19 at 15:33
• Okay my misunderstanding then. Aug 20 '19 at 9:20

{{{iss}}}o

Crashes TIO, but in theory works.

This means:

Repeat 1000 times: add one, than square twice.

First iteration: 0 => 1 => 1 => 1 Second iteration: 1 => 2 => 4 => 16 Third iteration: 16 => 17 => 289 => 83521 And after this it grows without bound.

CSASM v2.3, 9222471316929301708

func main:
push 1.9
bits
asl
print
ret
end


Explanation

The minimum code needed to print something in CSASM is as follows:

func main:
push <something>
print
ret
end


Thus, I took the liberty of having the 10-byte limit be on the <something> rather than the entire program, as is mentioned in the challenge:

You may exclude from the 10-character limit any code necessary to print anything.

Otherwise, this language wouldn't be able to even have a submission for this challenge.
(The boilerplate for an empty main function is 16 bytes.)

So, that leaves me with figuring out how to best use those 10 bytes. Hence, I abused one of the instructions in the language: bits

According to the syntax.txt file:

Pops A (an <f32> or <f64>), then pushes an <i32> or <i64> set to the bit representation of A

... where A stands for "the top value on the stack".

In short, a single- or double-precision floating-point value is popped from the stack. The value's bit representation (sign, exponent, mantissa) is then stored into a 32- or 64-bit integer value and that integer value is pushed to the stack.

Why use 1.9?

The bit representation of 1.9 is: $$00111111 11111110 01100110 01100110 01100110 01100110 01100110 01100110$$ This value interpreted as an i64 (or long) would be 4611235658464650854. However, this can be better. Using the last 3 available bytes, we can perform an Arithmetic Shift Left (asl) instruction, to get the following bit representation instead: $$01111111 11111100 11001100 11001100 11001100 11001100 11001100 11001100$$ ... which ends up being 9222471316929301708 instead.

If I were to use any value larger than 1.9 within the limits of the 3 bytes for the number (say, 2.0), the exponent part of the bit representation would be 10000000000 or some larger number.

The MSB in the exponent would get shifted left into the sign bit, resulting in a negative end result. Not good.

Furthermore, if I were to not use the asl instruction and instead opted to use 1.9999, the value would still be smaller than using 1.9 and the asl instruction.

TI-Basic (TI-83), 9.999999314e99

68!4032.2


very close to the soft limit of the calculator (-1e100<x<1e100). It can be overcome and get closer to the hard limit (1e128) but not in less than 10 characters.

69!58.4376 is a good contender too but not as close

TI-Basic (TI-89), 9.99999681785e999

449!259.61


Again, quite close to the limit (1e1000 gets automatically replaced with ∞)

KonamiCode, 1,000,000

v(^>>>>>>)<<< (Note: This excludes <<< because it is used for output.)

05AB1E, $$\\Large>1.0111110111×10^{146,337}\$$

žmbbbbbbbb


Try it online! This is the only competing one, the alternative breaks rule 5.

žmbbbbbbbb  # full program
b  # convert...
žm          # 9876543210...
b         # in binary...
b        # in binary...
b       # in binary...
b      # in binary...
b     # in binary...
b    # in binary...
b   # in binary...
b  # to binary
# implicit output


Alternative

This program scores: $$\large(r((x+1)^7-1)=…(r2=(r1=x^x)^{r1})^{r2}…)^{r((x+1)^7-1)}\\\Huge x=9{,}876{,}543{,}210$$Note: here, $$\\large r2\$$ doesn't mean $$\\large2r\$$ as it normally does. It means the second result. So $$\\large((x+1)^7-1)\$$ means the $$\\large((x+1)^7-1)^\text{th}\$$ result.

žmEEEEEEEm


žmEEEEEEEm  # full program
E         # for N in [0, 1, ...,
žm          # ..., 9876543210...
E         # ]...
E        # for N in [0, 1, ...,
žm          # ..., 9876543210...
E        # ]...
E       # for N in [0, 1, ...,
žm          # ..., 9876543210...
E       # ]...
E      # for N in [0, 1, ...,
žm          # ..., 9876543210...
E      # ]...
E     # for N in [0, 1, ...,
žm          # ..., 9876543210...
E     # ]...
E    # for N in [0, 1, ...,
žm          # ..., 9876543210...
E    # ]...
E   # for N in [0, 1, ...,
žm          # ..., 9876543210...
E   # ]...
m  # push...
žm          # 9876543210...
# (implicit) or top of stack if not first iteration...
m  # to the power of...
žm          # 9876543210...
# (implicit) or top of stack if not first iteration

• I've removed most of the \large elements in the $\LaTeX$ as they're unnecessarily obtrusive Jul 12 at 22:59

CJam - 40.4 million digits

1X27m<m<
`

Assuming bit shift is ok, this calculates 1<<(1<<27). It takes about 15 minutes on my laptop, using java 8 to run the interpreter.

• Hey, aditsu, did you design CJam?
– user19785
Jun 14 '14 at 8:57
• @404NotFound yeah, but a lot of it is like GolfScript Jun 14 '14 at 14:01
• Cool, where did you learn parsing to write an interpreter? I'm really interested in parsing etc., but I'm only 12 (no CS courses ;() I'm currently reading "Parsing Techniques: a Practical Guide" and trying to write my toy compiler in C++
– user19785
Jun 14 '14 at 17:59
• @404NotFound I'm not sure I can answer exactly... my first experience with parsing was probably writing an arithmetic expression evaluator, and I learned that from a book or maybe somebody else's source code. In university I learned something about formal languages - finite automata, formal grammars (of different types) and parsing techniques. Then I read a few things on the net about some other types of parsers, and wrote some code to try things out, but I can't say I learned a lot, and I'm not really good at this. I designed CJam to be easy to parse, something like C++ is MUCH MUCH harder. Jun 15 '14 at 8:18
• Of course I'm not trying to parse C++, I mean I write simple parsers in C++... In retrospect, was the university's class really crucial? Or did it have a more mild effect on your learning?
– user19785
Jun 15 '14 at 9:26