|
|
Post by AnthroHeart on May 9, 2021 21:18:01 GMT
I updated from v3.3 to v3.4 for the Intention Repeater MAX.
v3.4 is pretty exact on the timer now. Though it is about 5X slower than v3.3 due to not being able to use the benchmarking loop anymore. It still hits over 10PHz with "I am calm" even when I'm doing transcoding of videos to mp3 in the background.
So take your pick:
v3.3 - Faster but inexact timer. v3.4 - Slower but exact timer.
I am on Linux, so I will need to see if someone can compile v3.4 for Windows.
You will need a 64-bit version of g++.
Compile command for Windows (this includes the redistributable libraries):
g++ .\intention_repeater_max_3.4.cpp -O3 -o .\intention_repeater_max_3.4.exe -static-libgcc -static-libstdc++ -Wl,-Bstatic -lstdc++ -lpthread -Wl,-Bdynamic
I will later make a new version to allow selection of either or if you want faster with inexact timer or slower with exact timer.
I will need to do this to CUDA as well.
|
|
|
|
Post by AnthroHeart on May 9, 2021 21:23:42 GMT
Also, can you suggest if I should use maybe 3.3A instead of 3.4 for the exact time version?
I will need to combine this functionality so I don't have to manage two different versions.
Sometime in the next few days.
|
|
|
|
Post by reden on May 9, 2021 21:59:52 GMT
The old timer wasn't that inexact, was it? I never had any issues with it and it still worked to tell me how much time had passed.
|
|
|
|
Post by AnthroHeart on May 9, 2021 22:24:12 GMT
I'm not sure, but when recording a video of Repeater MAX running, and playing an mp3 on VLC, after about 30 mins the timer in MAX would differ by 5 minutes or so.
The 3.3 coding wasn't that great using the benchmark, because it may run benchmark # of iterations one too many times, for each second, or sometimes the right amount.
It only used a benchmark loop to make it much faster.
I'll add a --timer flag to specify inexact or exact timer.
Exact timer is accurate to within 0.046 milliseconds every second.
I didn't measure the inexact one.
|
|
|
|
Post by AnthroHeart on May 9, 2021 23:11:13 GMT
For version 3.3, the average variance over 10 iterations is 7.67ms for each second. For version 3.4, the average variance over 10 iterations is 0.043ms for each second.
C++ lets you measure in nanoseconds, so it was a pretty good measure.
Version 3.3 is 178.4X more inaccurate for timing for each second than v3.4. After one hour, v3.3 would be off from true time an average of 27.6 seconds. After one hour, v3.4 would be off from true time an average of 0.15 seconds. This is the most accurate I can get with my hardware. This is using 10GB of RAM for "I am calm." But v3.3 is 6.3X faster than v3.4 for # of iterations/second. If doing screen recording and playing mp3 music while running the Repeater, the time seems to vary more wildly. |
|
|
|
Post by Forester on May 10, 2021 16:02:51 GMT
I updated from v3.3 to v3.4 for the Intention Repeater MAX.
v3.4 is pretty exact on the timer now. Though it is about 5X slower than v3.3 due to not being able to use the benchmarking loop anymore. It still hits over 10PHz with "I am calm" even when I'm doing transcoding of videos to mp3 in the background.
So take your pick:
v3.3 - Faster but inexact timer. v3.4 - Slower but exact timer.
In addition to time measurement, are there other benefits, such as faster execution speed? |
|
|
|
Post by AnthroHeart on May 10, 2021 16:07:01 GMT
The latest version is actually slower execution speed.
Nothing else was changed except doing the most accurate repetitions possible.
In one second, v3.4 can actually take like 1,000,053,108ns.
Though not sure about exact precision, C++ can measure elapsed time in nanoseconds.
|
|
|
|
Post by Forester on May 10, 2021 16:23:24 GMT
The latest version is actually slower execution speed. Nothing else was changed except doing the most accurate repetitions possible. In one second, v3.4 can actually take like 1,000,053,108ns. Though not sure about exact precision, C++ can measure elapsed time in nanoseconds. So it's slower but more accurate in manifestation, right? |
|
|
|
Post by reden on May 10, 2021 17:13:11 GMT
The latest version is actually slower execution speed. Nothing else was changed except doing the most accurate repetitions possible. In one second, v3.4 can actually take like 1,000,053,108ns. Though not sure about exact precision, C++ can measure elapsed time in nanoseconds. So it's slower but more accurate in manifestation, right? It's supposed to be only the program timer. For instance if you play a 30 minute MP3 alongside the Repeater, the Repeater's timer will be off by like +5 minutes after the MP3 ends. If it somehow affects manifestation precision it will be very interesting. But I don't think it will do that. |
|
|
|
Post by Forester on May 10, 2021 17:38:37 GMT
Ok then I keep the current version. Thanks for the explanation
|
|
|
|
Post by AnthroHeart on May 10, 2021 17:46:40 GMT
So it's slower but more accurate in manifestation, right? It's supposed to be only the program timer. For instance if you play a 30 minute MP3 alongside the Repeater, the Repeater's timer will be off by like +5 minutes after the MP3 ends. If it somehow affects manifestation precision it will be very interesting. But I don't think it will do that. At this point I am not sure if there's much of a difference between a 10PHz and a 60PHz repetition rate.
It still rocks the energy.
I am still blown away at how simple the concept is of repeated intention. Holo-Stones are even easier to use.
I like the term "stupidly simple".
I also love "automagically".
Heck, even the word from my childhood "snagglefraggle" is quite fun.
|
|
|
|
Post by AnthroHeart on May 10, 2021 17:51:52 GMT
So it's slower but more accurate in manifestation, right? It's supposed to be only the program timer. For instance if you play a 30 minute MP3 alongside the Repeater, the Repeater's timer will be off by like +5 minutes after the MP3 ends. If it somehow affects manifestation precision it will be very interesting. But I don't think it will do that. It would be off by 5 mins if also doing screen recording using OBS Studio while playing an MP3. Just playing an mp3 might not make a big difference because it's not processor intensive.
BTW I was able to make the update myself as the developer I tried to get (who wanted $170) wasn't paying attention and just kept asking questions about what I had already covered.
Took me 20 minutes to make the update.
|
|
|
|
Post by AnthroHeart on May 10, 2021 17:56:00 GMT
When I update to v3.5, I will combine both features (faster speed w/more inexact timer, and slower speed with more exact timer). I will have a flag like: --timer With options "exact" and "inexact".
What should the default be (if the timer flag isn't set)?
1) Faster (by 6X) with 180X more inaccurate timer 2) Slower (by 6X) with accurate timer
|
|
|
|
Post by AnthroHeart on May 10, 2021 18:02:33 GMT
The latest version is actually slower execution speed. Nothing else was changed except doing the most accurate repetitions possible. In one second, v3.4 can actually take like 1,000,053,108ns. Though not sure about exact precision, C++ can measure elapsed time in nanoseconds. So it's slower but more accurate in manifestation, right? Well, manifestation is always accurate, but may manifest in a different way than we expect.
Not so much accurate in manifestation, but more intensity of manifestation.
The higher the reps, the more power that goes into it it seems.
Also, higher reps means higher frequency, and from what I understand higher frequency not only has more power, but can manifest better.
|
|
|
|
Post by Forester on May 10, 2021 21:17:59 GMT
The higher the reps, the more power that goes into it it seems.
Also, higher reps means higher frequency, and from what I understand higher frequency not only has more power, but can manifest better.
so is this what version 3.4 does? |
|
