Extremely poor performance of JsonParser

Extremely poor performance of JsonParser

[objeck]

Thanks debugging and profiling.

Initial observations:

• Memory usage remains constant even when garbage collection is relaxed
• Most of the time is spent parsing the JSON with low CPU usage
• Considering parallelism, i.e., workers when parsing JSON arrays

[objeck]

The primary issue was memory management thresholds. The VM was tuned to run on low-memory hosts (i.e., RPI4). The initial memory size for programs outside of RPI4 was changed from 4M to 1G. I am looking into setting available VM memory to the 16th of the system memory. For now, things should be faster.

The primary issue was memory management thresholds. The VM was tuned to run on low-memory hosts (i.e., RPI4). The initial memory size for programs outside of RPI4 was changed from 4M to 1G. I am looking into setting available VM memory to the 16th of the system memory. For now, things should be faster.

https://www.baeldung.com/jvm-parameters

[objeck]

The primary issue was memory management thresholds. The VM was tuned to run on low-memory hosts (i.e., RPI4). The initial memory size for programs outside of RPI4 was changed from 4M to 1G. I am looking into setting available VM memory to the 16th of the system memory. For now, things should be faster.

https://www.baeldung.com/jvm-parameters

Yep, I believe the VM has an .ini config file as well. I have tried to avoid VM parameters but will now look into this further.

[objeck]

The primary issue was memory management thresholds. The VM was tuned to run on low-memory hosts (i.e., RPI4). The initial memory size for programs outside of RPI4 was changed from 4M to 1G. I am looking into setting available VM memory to the 16th of the system memory. For now, things should be faster.

Added prepended command line parameter --GC_THRESHOLD= for example 100k, 4m and 1g. The new runtime host memory check is the default.

I believe the VM has an .ini config file as well.

How that file look like? What about a sample .ini config file?

[objeck]

I believe the VM has an .ini config file as well.

How that file look like? What about a sample .ini config file?

Nevermind, it was there. I will add it to the guide tomorrow. The file is config.prop however it only sets system properties.

The file is config.prop however it only sets system properties.

It's in lib\native\misc directory. Seems to be useless to me.

For now, things should be faster.

Was this issue really fixed or not? You have the JSON file, you could test.

[objeck]

@tqo50396 I tested against the JSON file you provided. I then added auto-tuning parameters before adding support for manual configuration. All are based on the sample JSON file you gave me.

@tqo50396 I tested against the JSON file you provided. I then added auto-tuning parameters before adding support for manual configuration. All are based on the sample JSON file you gave me.

But what I want to know is simply if you fixed the issue or not. If you believed that you fixed it, you could close this issue. I can't test since I no longer use Objeck (nor do I want to fire up MSYS2 to compile Objeck from source).

[objeck]

Yes, it is fixed. I will close the issue.

…

On Mon, Dec 4, 2023 at 12:04 PM tqo50396 ***@***.***> wrote: @tqo50396 <https://github.com/tqo50396> I tested against the JSON file you provided. I then added auto-tuning parameters before adding support for manual configuration. All are based on the sample JSON file you gave me. But what I want to know is simply if you fixed the issue or not. If you believed that you fixed it, you could close this issue. I can't test since I no longer use Objeck (nor do I want to fire up MSYS2 to compile Objeck from source). — Reply to this email directly, view it on GitHub <#407 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHRT4MHO5JMLFR7PGFRUODYHYUDNAVCNFSM6AAAAAA77V77J2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZZGM4DQMJVGQ> . You are receiving this because you commented.Message ID: ***@***.***>

Yes, it is fixed. I will close the issue.

The Objeck VM eats more RAM but the performance is still terrible. I'm very disappointed.

Download Adept here: https://github.com/AdeptLanguage/Adept/releases/tag/Nightly

Compile the code with adept test.adept:

pragma compiler_version '2.8'

import basics
import JSON

func main {
    json JSON = JSONFromFile("Web.MsHtml.json")

    if json.kind() == ::UNDEFINED {
        print("Failed to load json")
    } else {
        Constants <JSON> List = json.field("Constants").array().get()
        Constant0 JSON = Constants.get(0)
        print(Constant0.field("Name").string().get())
    }
}

You will see how fast it is compared to your parser.