royfengsss opened a new pull request, #10626:
URL: https://github.com/apache/nuttx/pull/10626
## Summary
When we are using newlib as libm, we can consider newlib-esp32 for esp32
hardware. An it contains some optimizations for esp32 hardware. Besides this
change, two small issues has been fixed.
1. add the tar ball to .gitignore
2. Fix the some build warnings
## Impact
Using a seperated config option to switch and default is n. So no impact to
normal case.
For enable newlib-esp32 case, it has a little performance improvement for
float calculation.
## Testing
Write a piece of simple test code and update `example/hello` app as below:
```
int main(int argc, FAR char *argv[])
{
struct timespec ts0;
struct timespec ts1;
uint64_t elapsed;
char *endp;
long n = 10000;
double number = 0.0;
double sum = 0.0;
double result;
printf("Hello, World!!\n");
if (argc == 2)
{
n = (int)strtol(argv[1], &endp, 10);
if (argv[1] == endp)
{
printf("invalid parameter!\n");
exit(-1);
}
}
srand(time(0));
clock_gettime(CLOCK_REALTIME, &ts0);
for (int i = 1; i <= n; i++)
{
number = (double)(rand() % 9000001 + 1000000) + (double)(rand() % 10001)
/ 10000.0;
sum += 1.0 / sqrt(number);
}
result = sum / (double)n;
clock_gettime(CLOCK_REALTIME, &ts1);
elapsed = (((uint64_t)ts1.tv_sec * NSEC_PER_SEC) + ts1.tv_nsec);
elapsed -= (((uint64_t)ts0.tv_sec * NSEC_PER_SEC) + ts0.tv_nsec);
elapsed /= NSEC_PER_MSEC; /* msec */
printf("The %ld round result is %.7f\n", n, result);
printf("it took %" PRIu64 " msec\n", elapsed);
return 0;
}
```
Use esp32-devkitc:nsh config and enable below options
```
CONFIG_LIBM_NEWLIB=y
CONFIG_LIBM_USE_NEWLIB_ESP32=y
```
Here is the test result:
For the case of using default newlib (sourceware.org)
```
nsh> hello
Hello, World!!
The 10000 round result is 0.0004855
it took 180 msec
nsh> hello 100000
Hello, World!!
The 100000 round result is 0.0004858
it took 1820 msec
nsh> hello 1000000
Hello, World!!
The 1000000 round result is 0.0005035
it took 18190 msec
nsh> hello
Hello, World!!
The 10000 round result is 0.0004856
it took 180 msec
nsh> hello 100000
Hello, World!!
The 100000 round result is 0.0004858
it took 1820 msec
nsh> hello 1000000
Hello, World!!
The 1000000 round result is 0.0004858
it took 18190 msec
nsh>
```
For the case of using newlib-esp32:
```
nsh> hello
Hello, World!!
The 10000 round result is 0.0004854
it took 170 msec
nsh> hello 100000
Hello, World!!
The 100000 round result is 0.0004858
it took 1750 msec
nsh> hello 1000000
Hello, World!!
The 1000000 round result is 0.0004858
it took 17470 msec
nsh> hello
Hello, World!!
The 10000 round result is 0.0004848
it took 180 msec
nsh> hello 100000
Hello, World!!
The 100000 round result is 0.0005035
it took 1750 msec
nsh> hello 1000000
Hello, World!!
The 1000000 round result is 0.0004858
it took 17470 msec
nsh>
```
We can see the newlib-esp32 case is a little better in performance than the
default.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
