 |
|
|
RENESAS TOOL NEWS on May 1, 2004: RSO-M3T-CC32R_1-040501D
A Note on Using C-Compiler Package
M3T-CC32R
--On the Power Functions in the Standard Library--
|
|---|
Please take note of the following problem in using the M3T-CC32R C-compiler package for the M32R family MCUs:
- On the power functions (powf5, powf, and pow) in the standard library
- Versions Concerned
M3T-CC32R V.4.20 Release 1 and 1A
- Description
Using the powf5, powf, and pow functions may generate incorrect results of operations.
- 2.1 Conditions
- This problem occurs if the following two conditions are satisfied:
| (1) |
Any of the following functions are called:
| (a) | a powf5 function |
| (b) | a powf function with the -m32re5 compile option selected (*1) |
| (c) | a pow function with the -float_only and -m32re5 compile options selected (*1) |
|
| (2) |
When x and y represent the first and second input arguments of the function in (1),
none of the following cases apply to them:
| (a) |
Either of x and y is 0.0. |
| (b) |
The value of x is negative (including -0.0) and y is not an integer (*2). |
| (c) |
Either of x and y is infinity (∞) or not a number (NaN) (*2 and *3). |
NOTES:
| *1. |
Selecting these options replaces the calling of the powf or pow function with that of the powf5. |
| *2. |
If these cases are applied, no power operation is performed;
macro name "EDOM" is assigned to external variable "errno", which holds an error number. |
| *3. |
Infinity (∞) and not a number (NaN) are the special symbols
returned if illegal operations are performed concerning floating-point operations (for example, division by 0.0).
So don't use them in usual operations. |
|
- 2.2 Example
Source file: sample.c
--------------------------------------------------------------------
#include <math.h>
#include <mathf.h>
float ans1, ans2, ans3, ans4;
void func(void)
{
/* Conditions (1)-(a) and (2) */
ans1 = powf5(4.0, 0.5);
/* Conditions (1)-(b)(-m32re5 selected) and (2) */
ans2 = powf(10000.0, 0.25);
/* Conditions (1)-(c)(-float_only and -m32re5 selected) and (2) */
ans3 = pow(27.0, 1.0/3.0);
/* No condition is met */
ans4 = powf5(0.0, 1.0);
}
--------------------------------------------------------------------
- Workaround
This problem can be circumvented in either of the following ways:
| (1) |
Immediately after the include statements of mathf.h and mathf.h,
put two lines for replacing powf5 with powf.
Modification of the source file sample.c
--------------------------------------------------------------------
#include <math.h>
#include <mathf.h>
#undef powf /* Added */
#define powf5 powf /* Added */
float ans1, ans2, ans3, ans4;
void func(void)
{
ans1 = powf5(4.0, 0.5);
ans2 = powf(10000.0, 0.25); /* No problem arises
even if -m32re5 selected */
ans3 = pow(27.0, 1.0/3.0); /* No problem arises
even if -float_only and -m32re5 selected */
ans4 = powf5(0.0, 1.0);
}
--------------------------------------------------------------------
|
| (2) |
Use the fixed_powf5 function.
Download the file named fixed_powf5.zip (3.01KB) of the fixed_powf5 function
(the modification of powf5) and use it following the instructions of
"How to Use the fixed_powf5 Function." described below.
The fixed_powf5.lzh file consists of the following files:
| small/fixed_powf5.mo | ... | the fixed_powf5 function for the small memory model |
| small/fixed_powf5.stk | ... | the file for calculating stack size for the small memory model |
| medium/fixed_powf5.mo | ... | the fixed_powf5 function for the medium memory model |
| medium/fixed_powf5.stk | ... | the file for calculating stack size for the medium memory model |
| large/fixed_powf5.mo | ... | the fixed_powf5 function for the large memory model |
| large/fixed_powf5.stk | ... | the file for calculating stack size for the large memory model |
How to Use the fixed_powf5 Function
- Search for all the functions that satisfy Condition (1) above.
- Replace every function involved with the fixed_powf5 one.
This explicitly indicates that the replacement has been made to circumvent the problem.
- Be sure to add the declaration shown below immediately after
the include statements of math.h and mathf.h.
---------------------------------------
extern float fixed_powf5(float,float);
---------------------------------------
- To create a load module, link to it the fixed_powf5.mo file (*) included in the downloaded fixed_powf5.lzh file.
NOTE:
| * |
Select the file appropriate to the memory model you use out of the three. |
|
- Schedule of Fixing the Problem
We plan to fix this problem in our next release of the product.
|
|
