Mapping [-1,+1] floats to Q31 fixed-point - c

I need to convert float to Q31 fixed-point, Q31 meaning 1 sign bit, 0 bits for integer part, and 31 bits for fractional part. This means that Q31 can only represent numbers in the range [-1,0.9999].
By definition, when converting from float to fixed-point, a multiplication by 2ˇN is done, where N is the fractional part size, in this case 31.
However, I got confused with this code, it doesn't look right, but works:
#define q31_float_to_int(x) ( (int) ( (float)(x)*(float)0x7FFFFFFF ) )
And it seems to work OK. For example:
int a = q31_float_to_int(0.5f);
gives Hex: 0x40000000, which is OK.
Why is the multipication here done with 2ˇ31 - 1, and not just 2ˇ31?

The code above is not a good solution to convert from float to fixed point. I am guessing whoever wrote the code used the scale factor of 0x7FFFFFFF to avoid an overflow when the input is 1.0. The correct scaling factor is 2^31 and not 2^31 - 1. Note that there are also precision issues when converting a float (with 24 bits of precision) to an Q1.31 (with 31 bits of precision). Consider saturating the input data before multiplication:
const float Q31_MAX_F = 0x0.FFFFFFp0F;
const float Q31_MIN_F = -1.0F;
float clamped = fmaxf(fminf(input, Q31_MAX_F), Q31_MIN_F);
The code above will clamp input to the range of [-1.0, 1.0). The constantQ31_MAX_F is approximately 1 - (2 ^ -24), considering 24-bits of precision, and Q31_MIN_F is -1. Then you can multiply clamped by 2^31, or even better, use scalbnf, or ldexpf:
int result = (int) scalbnf(clamped, 31);
And if you want rounding:
int result = (int) roundf(scalbnf(clamped, 31)));

I recently had to use STM32's CORDIC for hardware-accelerated trigonometry, and left unsatisfied with the accepted answer (and everything else I found on the web), I came up with a simpler (but slightly less precise) algorithm for Q31/F32 conversion:
#define Q31_SCALAR (float)M_PI
#define F32_TO_Q31(F) (int32_t)((fmodf((F)+Q31_SCALAR,2.f*Q31_SCALAR) + ((F)<-Q31_SCALAR?Q31_SCALAR:-Q31_SCALAR)) * ((float)(INT32_MAX+1u)/Q31_SCALAR))
#define Q31_TO_F32(Q) ((int32_t)(Q) / (float)(INT32_MAX+1u))
#define CORDIC_COS_SIN(RAD,COS_VAR,SIN_VAR) { hcordic.Instance->WDATA = F32_TO_Q31(RAD); \
(COS_VAR) = Q31_TO_F32(hcordic.Instance->RDATA); (SIN_VAR) = Q31_TO_F32(hcordic.Instance->RDATA); }
This will map floats from [-π, +π] to approximately [INT32_MIN, INT32_MAX[. If the input value is out of range, it will be "wrapped" back into that range (e.g. -5.9π will be treated as 0.1π).
If instead you want to map [-1, +1] as per the original question, simply use the following:
#define Q31_SCALAR 1.f

Related

Computing floating point accuracy (K&R 2-1)

I found Stevens Computing Services – K & R Exercise 2-1 a very thorough answer to K&R 2-1. This slice of the full code computes the maximum value of a float type in the C programming language.
Unluckily my theoretical comprehension of float values is quite limited. I know they are composed of significand (mantissa.. ) and a magnitude which is a power of 2.
#include <stdio.h>
#include <limits.h>
#include <float.h>
main()
{
float flt_a, flt_b, flt_c, flt_r;
/* FLOAT */
printf("\nFLOAT MAX\n");
printf("<limits.h> %E ", FLT_MAX);
flt_a = 2.0;
flt_b = 1.0;
while (flt_a != flt_b) {
flt_m = flt_b; /* MAX POWER OF 2 IN MANTISSA */
flt_a = flt_b = flt_b * 2.0;
flt_a = flt_a + 1.0;
}
flt_m = flt_m + (flt_m - 1); /* MAX VALUE OF MANTISSA */
flt_a = flt_b = flt_c = flt_m;
while (flt_b == flt_c) {
flt_c = flt_a;
flt_a = flt_a * 2.0;
flt_b = flt_a / 2.0;
}
printf("COMPUTED %E\n", flt_c);
}
I understand that the latter part basically checks to which power of 2 it's possible to raise the significand with a three variable algorithm. What about the first part?
I can see that a progression of multiples of 2 should eventually determine the value of the significand, but I tried to trace a few small numbers to check how it should work and it failed to find the right values...
======================================================================
What are the concepts on which this program is based upon and does this program gets more precise as longer and non-integer numbers have to be found?
The first loop determines the number of bits contributing to the significand by finding the least power 2 such that adding 1 to it (using floating-point arithmetic) fails to change its value. If that's the nth power of two, then the significand uses n bits, because with n bits you can express all the integers from 0 through 2^n - 1, but not 2^n. The floating-point representation of 2^n must therefore have an exponent large enough that the (binary) units digit is not significant.
By that same token, having found the first power of 2 whose float representation has worse than unit precision, the maximim float value that does have unit precision is one less. That value is recorded in variable flt_m.
The second loop then tests for the maximum exponent by starting with the maximum unit-precision value, and repeatedly doubling it (thereby increasing the exponent by 1) until it finds that the result cannot be converted back by halving it. The maximum float is the value before that final doubling.
Do note, by the way, that all the above supposes a base-2 floating-point representation. You are unlikely to run into anything different, but C does not actually require any specific representation.
With respect to the second part of your question,
does this program gets more precise as longer and non-integer numbers have to be found?
the program takes care to avoid losing precision. It does assume a binary floating-point representation such as you described, but it will work correctly regardless of the number of bits in the significand or exponent of such a representation. No non-integers are involved, but the program already deals with numbers that have worse than unit precision, and with numbers larger than can be represented with type int.

accuracy of sqrt of integers

I have a loop like this:
for(uint64_t i=0; i*i<n; i++) {
This requires doing a multiplication every iteration. If I could calculate the sqrt before the loop then I could avoid this.
unsigned cut = sqrt(n)
for(uint64_t i=0; i<cut; i++) {
In my case it's okay if the sqrt function rounds up to the next integer but it's not okay if it rounds down.
My question is: is the sqrt function accurate enough to do this for all cases?
Edit: Let me list some cases. If n is a perfect square so that n = y^2 my question would be - is cut=sqrt(n)>=y for all n? If cut=y-1 then there is a problem. E.g. if n = 120 and cut = 10 it's okay but if n=121 (11^2) and cut is still 10 then it won't work.
My first concern was the fractional part of float only has 23 bits and double 52 so they can't store all the digits of some 32-bit or 64-bit integers. However, I don't think this is a problem. Let's assume we want the sqrt of some number y but we can't store all the digits of y. If we let the fraction of y we can store be x we can write y = x + dx then we want to make sure that whatever dx we choose does not move us to the next integer.
sqrt(x+dx) < sqrt(x) + 1 //solve
dx < 2*sqrt(x) + 1
// e.g for x = 100 dx < 21
// sqrt(100+20) < sqrt(100) + 1
Float can store 23 bits so we let y = 2^23 + 2^9. This is more than sufficient since 2^9 < 2*sqrt(2^23) + 1. It's easy to show this for double as well with 64-bit integers. So although they can't store all the digits as long as the sqrt of what they can store is accurate then the sqrt(fraction) should be sufficient. Now let's look at what happens for integers close to INT_MAX and the sqrt:
unsigned xi = -1-1;
printf("%u %u\n", xi, (unsigned)(float)xi); //4294967294 4294967295
printf("%u %u\n", (unsigned)sqrt(xi), (unsigned)sqrtf(xi)); //65535 65536
Since float can't store all the digits of 2^31-2 and double can they get different results for the sqrt. But the float version of the sqrt is one integer larger. This is what I want. For 64-bit integers as long as the sqrt of the double always rounds up it's okay.
First, integer multiplication is really quite cheap. So long as you have more than a few cycles of work per loop iteration and one spare execute slot, it should be entirely hidden by reorder on most non-tiny processors.
If you did have a processor with dramatically slow integer multiply, a truly clever compiler might transform your loop to:
for (uint64_t i = 0, j = 0; j < cut; j += 2*i+1, i++)
replacing the multiply with an lea or a shift and two adds.
Those notes aside, let’s look at your question as stated. No, you can’t just use i < sqrt(n). Counter-example: n = 0x20000000000000. Assuming adherence to IEEE-754, you will have cut = 0x5a82799, and cut*cut is 0x1ffffff8eff971.
However, a basic floating-point error analysis shows that the error in computing sqrt(n) (before conversion to integer) is bounded by 3/4 of an ULP. So you can safely use:
uint32_t cut = sqrt(n) + 1;
and you’ll perform at most one extra loop iteration, which is probably acceptable. If you want to be totally precise, instead use:
uint32_t cut = sqrt(n);
cut += (uint64_t)cut*cut < n;
Edit: z boson clarifies that for his purposes, this only matters when n is an exact square (otherwise, getting a value of cut that is “too small by one” is acceptable). In that case, there is no need for the adjustment and on can safely just use:
uint32_t cut = sqrt(n);
Why is this true? It’s pretty simple to see, actually. Converting n to double introduces a perturbation:
double_n = n*(1 + e)
which satisfies |e| < 2^-53. The mathematical square root of this value can be expanded as follows:
square_root(double_n) = square_root(n)*square_root(1+e)
Now, since n is assumed to be a perfect square with at most 64 bits, square_root(n) is an exact integer with at most 32 bits, and is the mathematically precise value that we hope to compute. To analyze the square_root(1+e) term, use a taylor series about 1:
square_root(1+e) = 1 + e/2 + O(e^2)
= 1 + d with |d| <~ 2^-54
Thus, the mathematically exact value square_root(double_n) is less than half an ULP away from[1] the desired exact answer, and necessarily rounds to that value.
[1] I’m being fast and loose here in my abuse of relative error estimates, where the relative size of an ULP actually varies across a binade — I’m trying to give a bit of the flavor of the proof without getting too bogged down in details. This can all be made perfectly rigorous, it just gets to be a bit wordy for Stack Overflow.
All my answer is useless if you have access to IEEE 754 double precision floating point, since Stephen Canon demonstrated both
a simple way to avoid imul in loop
a simple way to compute the ceiling sqrt
Otherwise, if for some reason you have a non IEEE 754 compliant platform, or only single precision, you could get the integer part of square root with a simple Newton-Raphson loop. For example in Squeak Smalltalk we have this method in Integer:
sqrtFloor
"Return the integer part of the square root of self"
| guess delta |
guess := 1 bitShift: (self highBit + 1) // 2.
[
delta := (guess squared - self) // (guess + guess).
delta = 0 ] whileFalse: [
guess := guess - delta ].
^guess - 1
Where // is operator for quotient of integer division.
Final guard guess*guess <= self ifTrue: [^guess]. can be avoided if initial guess is fed in excess of exact solution as is the case here.
Initializing with approximate float sqrt was not an option because integers are arbitrarily large and might overflow
But here, you could seed the initial guess with floating point sqrt approximation, and my bet is that the exact solution will be found in very few loops. In C that would be:
uint32_t sqrtFloor(uint64_t n)
{
int64_t diff;
int64_t delta;
uint64_t guess=sqrt(n); /* implicit conversions here... */
while( (delta = (diff=guess*guess-n) / (guess+guess)) != 0 )
guess -= delta;
return guess-(diff>0);
}
That's a few integer multiplications and divisions, but outside the main loop.
What you are looking for is a way to calculate a rational upper bound of the square root of a natural number. Continued fraction is what you need see wikipedia.
For x>0, there is
.
To make the notation more compact, rewriting the above formula as
Truncate the continued fraction by removing the tail term (x-1)/2's at each recursion depth, one gets a sequence of approximations of sqrt(x) as below:
Upper bounds appear at lines with odd line numbers, and gets tighter. When distance between an upper bound and its neighboring lower bound is less than 1, that approximation is what you need. Using that value as the value of cut, here cut must be a float number, solves the problem.
For very large number, rational number should be used, so no precision is lost during conversion between integer and floating point number.

Rounding up integer without using float, double, or division

Its an embedded platform thats why such restrictions.
original equation: 0.02035*c*c - 2.4038*c
Did this:
int32_t val = 112; // this value is arbitrary
int32_t result = (val*((val * 0x535A8) - 0x2675F70));
result = result>>24;
The precision is still poor. When we multiply val*0x535A8 Is there a way we can further improve the precision by rounding up, but without using any float, double, or division.
The problem is not precision. You're using plenty of bits.
I suspect the problem is that you're comparing two different methods of converting to int. The first is a cast of a double, the second is a truncation by right-shifting.
Converting floating point to integer simply drops the fractional part, leading to a round towards zero; right-shifting does a round down or floor. For positive numbers there's no difference, but for negative numbers the two methods will be 1 off from each other. See an example at http://ideone.com/rkckuy and some background reading at Wikipedia.
Your original code is easy to fix:
int32_t result = (val*((val * 0x535A8) - 0x2675F70));
if (result < 0)
result += 0xffffff;
result = result>>24;
See the results at http://ideone.com/D0pNPF
You might also just decide that the right shift result is OK as is. The conversion error isn't greater than it is for the other method, just different.
Edit: If you want to do rounding instead of truncation the answer is even easier.
int32_t result = (val*((val * 0x535A8) - 0x2675F70));
result = (result + (1L << 23)) >> 24;
I'm going to join in with some of the others in suggesting that you use a constant expression to replace those magic constants with something that documents how they were derived.
static const int32_t a = (int32_t)(0.02035 * (1L << 24) + 0.5);
static const int32_t b = (int32_t)(2.4038 * (1L << 24) + 0.5);
int32_t result = (val*((val * a) - b));
How about just scaling your constants by 10000. The maximum number you then get is 2035*120*120 - 24038*120 = 26419440, which is far below the 2^31 limit. So maybe there is no need to do real bit-tweaking here.
As noted by Joe Hass, your problem is that you shift your precision bits into the dustbin.
Whether shifting your decimals by 2 or by 10 to the left does actually not matter. Just pretend your decimal point is not behind the last bit but at the shifted position. If you keep computing with the result, shifting by 2 is likely easier to handle. If you just want to output the result, shift by powers of ten as proposed above, convert the digits and insert the decimal point 5 characters from the right.
Givens:
Lets assume 1 <= c <= 120,
original equation: 0.02035*c*c - 2.4038*c
then -70.98586 < f(c) < 4.585
--> -71 <= result <= 5
rounding f(c) to nearest int32_t.
Arguments A = 0.02035 and B = 2.4038
A & B may change a bit with subsequent compiles, but not at run-time.
Allow coder to input values like 0.02035 & 2.4038. The key components shown here and by others it to scale the factors like 0.02035 to by some power-of-2, do the equation (simplified into the form (A*c - B)*c) and the scale the result back.
Important features:
1 When determining A and B, insure the compile time floating point multiplication and final conversion occurs via a round and not a truncation. With positive values, the + 0.5 achieves that. Without a rounded answer UD_A*UD_Scaling could end up just under a whole number and truncate away 0.999999 when converting to the int32_t
2 Instead of doing expensive division at run-time, we do >> (right shift). By adding half the divisor (as suggested by #Joe Hass), before the division, we get a nicely rounded answer. It is important not to code in / here as some_signed_int / 4 and some_signed_int >> 2 do not round the same way. With 2's complement, >> truncates toward INT_MIN whereas / truncates toward 0.
#define UD_A (0.02035)
#define UD_B (2.4038)
#define UD_Shift (24)
#define UD_Scaling ((int32_t) 1 << UD_Shift)
#define UD_ScA ((int32_t) (UD_A*UD_Scaling + 0.5))
#define UD_ScB ((int32_t) (UD_B*UD_Scaling + 0.5))
for (int32_t val = 1; val <= 120; val++) {
int32_t result = ((UD_A*val - UD_B)*val + UD_Scaling/2) >> UD_Shift;
printf("%" PRId32 "%" PRId32 "\n", val, result);
}
Example differences:
val, OP equation, OP code, This code
1, -2.38345, -3, -2
54, -70.46460, -71, -70
120, 4.58400, 4, 5
This is a new answer. My old +1 answer deleted.
If you r input uses max 7 bits and you have 32 bit available then your best bet is to shift everything by as many bits as possible and work with that:
int32_t result;
result = (val * (int32_t)(0.02035 * 0x1000000)) - (int32_t)(2.4038 * 0x1000000);
result >>= 8; // make room for another 7 bit multiplication
result *= val;
result >>= 16;
Constant conversion will be done by an optimising compiler at compile time.

Need Floating Point Precision Using Unsigned Int

I'm working with a microchip that doesn't have room for floating point precision, however. I need to account for fractional values during some equations. So far I've had good luck using the old *100 -> /100 method like so:
increment = (short int)(((value1 - value2)*100 / totalSteps));
// later in the code I loop through the number of totolSteps
// adding back the increment to arrive at the total I want at the precise time
// time I need it.
newValue = oldValue + (increment / 100);
This works great for values from 0-255 divided by a totalSteps of up to 300. After 300, the fractional values to the right of the decimal place, become important, because they add up over time of course.
I'm curious if anyone has a better way to save decimal accuracy within an integer paradigm? I tried using *1000 /1000, but that didn't work at all.
Thank you in advance.
Fractions with integers is called fixed point math.
Try Googling "fixed point".
Fixed point tips and tricks are out of the scope of SO answer...
Example: 5 tap FIR filter
// C is the filter coefficients using 2.8 fixed precision.
// 2 MSB (of 10) is for integer part and 8 LSB (of 10) is the fraction part.
// Actual fraction precision here is 1/256.
int FIR_5(int* in, // input samples
int inPrec, // sample fraction precision
int* c, // filter coefficients
int cPrec) // coefficients fraction precision
{
const int coefHalf = (cPrec > 0) ? 1 << (cPrec - 1) : 0; // value of 0.5 using cPrec
int sum = 0;
for ( int i = 0; i < 5; ++i )
{
sum += in[i] * c[i];
}
// sum's precision is X.N. where N = inPrec + cPrec;
// return to original precision (inPrec)
sum = (sum + coefHalf) >> cPrec; // adding coefHalf for rounding
return sum;
}
int main()
{
const int filterPrec = 8;
int C[5] = { 8, 16, 208, 16, 8 }; // 1.0 == 256 in 2.8 fixed point. Filter value are 8/256, 16/256, 208/256, etc.
int W[5] = { 10, 203, 40, 50, 72}; // A sampling window (example)
int res = FIR_5(W, 0, C, filterPrec);
return 0;
}
Notes:
In the above example:
the samples are integers (no fraction)
the coefs have fractions of 8 bit.
8 bit fractions mean that each change of 1 is treated as 1/256. 1 << 8 == 256.
Useful notation is Y.Xu or Y.Xs. where Y is how many bits are allocated for the integer part and X for he fraction. u/s denote signed/unsigned.
when multiplying 2 fixed point numbers, their precision (size of fraction bits) are added to each other.
Example A is 0.8u, B is 0.2U. C=A*B. C is 0.10u
when dividing, use a shift operation to lower the result precision. Amount of shifting is up to you. Before lowering precision it's better to add a half to lower the error.
Example: A=129 in 0.8u which is a little over 0.5 (129/256). We want the integer part so we right shift it by 8. Before that we want to add a half which is 128 (1<<7). So A = (A + 128) >> 8 --> 1.
Without adding a half you'll get a larger error in the final result.
Don't use this approach.
New paradigm: Do not accumulate using FP math or fixed point math. Do your accumulation and other equations with integer math. Anytime you need to get some scaled value, divide by your scale factor (100), but do the "add up" part with the raw, unscaled values.
Here's a quick attempt at a precise rational (Bresenham-esque) version of the interpolation if you truly cannot afford to directly interpolate at each step.
div_t frac_step = div(target - source, num_steps);
if(frac_step.rem < 0) {
// Annoying special case to deal with rounding towards zero.
// Alternatively check for the error term slipping to < -num_steps as well
frac_step.rem = -frac_step.rem;
--frac_step.quot;
}
unsigned int error = 0;
do {
// Add the integer term plus an accumulated fraction
error += frac_step.rem;
if(error >= num_steps) {
// Time to carry
error -= num_steps;
++source;
}
source += frac_step.quot;
} while(--num_steps);
A major drawback compared to the fixed-point solution is that the fractional term gets rounded off between iterations if you are using the function to continually walk towards a moving target at differing step lengths.
Oh, and for the record your original code does not seem to be properly accumulating the fractions when stepping, e.g. a 1/100 increment will always be truncated to 0 in the addition no matter how many times the step is taken. Instead you really want to add the increment to a higher-precision fixed-point accumulator and then divide it by 100 (or preferably right shift to divide by a power-of-two) each iteration in order to compute the integer "position".
Do take care with the different integer types and ranges required in your calculations. A multiplication by 1000 will overflow a 16-bit integer unless one term is a long. Go through you calculations and keep track of input ranges and the headroom at each step, then select your integer types to match.
Maybe you can simulate floating point behaviour by saving
it using the IEEE 754 specification
So you save mantisse, exponent, and sign as unsigned int values.
For calculation you use then bitwise addition of mantisse and exponent and so on.
Multiplication and Division you can replace by bitwise addition operations.
I think it is a lot of programming staff to emulate that but it should work.
Your choice of type is the problem: short int is likely to be 16 bits wide. That's why large multipliers don't work - you're limited to +/-32767. Use a 32 bit long int, assuming that your compiler supports it. What chip is it, by the way, and what compiler?

Fixed-point scaling and accuracy in multiplication

I need to perform a multiplication operation on a fixed-point variable x (unsigned 16-bit integer [U16] type with binary point 6 [BP6]) with a coefficient A, which I know will always be between 0 and 1. Code is being written in C for a 32-bit embedded platform.
I know that if I were to also make this coefficient a U16 BP6, then I would end up with a U32 BP12 from the multiplication. I want to rescale this result back down to U16 BP6, so I just lop off the first 10 bits and the last 6.
However, since the coefficient is limited in precision by the number of fractional bits, and I do not necessarily need the full 10 bits of integer, I was thinking that I could just make the coefficient variable A a U16 BP15 to yield a more precise result.
I have worked out the following example (bear with me):
Let's say that x = 172.0 (decimal) and I want to use a coefficient A = 0.82 (decimal). The ideal decimal result would be 172.0 * 0.82 = 141.04.
In binary, x = 0010101100.000000.
If I am using BP6 for A, the binary representation will be either
A_1 = 0000000000.110100 = 0.8125 or
A_2 = 0000000000.110101 = 0.828125
(depending on whether value is based on floor or ceiling).
Performing the binary multiplication between x and either value of A yields (leaving out leading zeroes):
A_1 * x = 10001011.110000000000 = 139.75
A_2 * x = 10001110.011100000000 = 142.4375
In both cases, triming down the last 6 bits would not affect the result.
Now, if I expanded A to have BP15, then
A_3 = 0.110100011110110 = 0.82000732421875
and the resulting multiplication yields
A_3 * x = 10001101.000010101001000000000 = 141.041259765625
When trimming the extra 15 fractional bits, the result is
A_3 * x = 10001101.000010 = 141.03125
So it's pretty clear here that by expanding the coefficient to have more fractional bits yields a more precise result (at least in my example). Is this something which will hold true in general? Is this good/bad to use in practice? Am I missing or misunderstanding something?
EDIT: I should have said "accuracy" in place of "precision" here. I am looking for a result which is closer to my expected value rather than a result which contains more fractional bits.
Having done similar code, I'd say you what you are doing will hold true in general with the following concerns.
It is very easy to get unexpected overflow when shifting around your binary point. Rigorous testing/analysis and/or code detect is recommended. Notable failure: Ariane_5
You want precision, thus I disagree with "lop off ... last 6". Instead I recommend rounding your results as processing time allows. Use the MSBit to be lopped off to possibly adjust the result.

Resources