-
|
I have a bot program that uses rust-skia to generate images. However, over time, I've noticed that the program's memory usage gradually increases from around 70MB(with some static resources) at the beginning to about 200MB after a few days. I've only been learning Rust for less than two months, and I'd like to know if this is normal. Here is a code example: let mut surface = surfaces::raster_n32_premul(ISize::new(1080, img_height)).expect("Cannot create surface");
let canvas = surface.canvas();
let mut paint = Paint::default();
paint.set_anti_alias(true);
// Some drawing code, such as:
let img = Image::from_encoded(Data::new_copy(&bytes));
canvas.draw_image(&img, Point::new(100.0, 100.0), Some(&paint));
paint.reset();
let image = surface.image_snapshot();
return match image.encode(None, EncodedImageFormat::JPEG, 50) {
Some(data) => Some(data.as_bytes().to_vec()),
None => {
error!("Image generation failed");
None
}
};
// repeat this almost every 10 minutesTranslated by ChatGPT because my English is really bad😂 |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 5 replies
-
|
The symptom you’re describing doesn’t look normal to me. Do you have a standalone sample that reproduces the issue? |
Beta Was this translation helpful? Give feedback.
-
|
No, I don't have a standalone sample because I don't know which part is wrong😭, but I can give you part of my original code. It is kind of messy, sorry about that |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the sample. I looked a bit through it but it's not obvious to me where it could leak :(. Are you sure that it's a leak without bounds or could it be that there is a ceiling? Otherwise I suggest to run this with a leak detector like heaptrack. |
Beta Was this translation helpful? Give feedback.
-
|
I tried using heaptrack in my project, but I think it was interfered with by other parts, so I plan to set up a separate testing environment.<(。_。)> In addition, I noticed that the examples of rust-skia usually only create a surface once. In my project, because the size of the images is uncertain each time, a new surface is created each time. Do you think the problem might be here? |
Beta Was this translation helpful? Give feedback.
-
|
I don’t think that might be the problem. |
Beta Was this translation helpful? Give feedback.
-
|
I tried conducting a comparative experiment and found that after several hours, the memory usage of sample that draws images is 60mb higher than sample that doesn't draw images. I think that the issue might lie in this. |
Beta Was this translation helpful? Give feedback.
-
|
@tronical This sample seems to have the same problem use skia_safe::*;
use sysinfo::{System, SystemExt, ProcessExt};
use walkdir::WalkDir;
fn main() {
const MB: f32 = 1048576.0;
let mut sys = System::new();
let pid = sysinfo::get_current_pid().unwrap();
{
sys.refresh_system();
sys.refresh_process(pid);
let process = sys.process(pid.clone()).unwrap();
let prog = process.memory() as f32;
println!("Initial memory: {:.2}MB", prog / MB);
for entry in WalkDir::new("./imgs") { // example imgset
let entry = entry.unwrap();
if (&entry).file_type().is_dir() { continue; }
let bytes = std::fs::read(&entry.path()).unwrap();
println!("{:?}", entry);
let img = Image::from_encoded(Data::new_copy(&bytes)).unwrap();
let mut surface =
surfaces::raster_n32_premul(ISize::new(1080, img.height())).expect("Cannot create surface");
let canvas = surface.canvas();
canvas.draw_image(
&img,
Point::new(0.0, 0.0),
None,
);
let _output = surface.image_snapshot();
}
}
sys.refresh_process(pid);
let process = sys.process(pid.clone()).unwrap();
let prog = process.memory() as f32;
println!("Current memory: {:.2}MB", prog / MB);
}It outputs |
Beta Was this translation helpful? Give feedback.
-
Images are cached. I didn't see this. 😂 |
Beta Was this translation helpful? Give feedback.
from_encoded_with_alpha_type
Images are cached. I didn't see this. 😂