|
My little MCU can't do math.
It runs at 600MHz.
It's running the same code as something that runs at 240MHz, but it's going like 10 times as fast on the 600MHz CPU. They're both 32-bit CPUs.
10 times faster, by my rough estimation.
That doesn't add up. I'm not really complaining per se, but it is a bit like putting something together, it working, and yet having parts left over.
Edit: after actually benchmarking it came out 9 times as fast. pretty close for a wild guess on my part.
To err is human. Fortune favors the monsters.
modified 17-Aug-22 19:59pm.
|
|
|
|
|
Quote: My little MCU can't do math.
I know, it`s music based probably. I guess the processor has groups of cords arranged in levels, the sounds travel between cords and levels which results in logic being produced. It looks like this: <internet link="">alu_schematic.jpg
|
|
|
|
|
In some cases math can be optimised by allowing [a great deal of] loss in precision, and baking your own, eg cos(n) with a precomputed table. Dunno what your little MCU wants to achieive.
"If we don't change direction, we'll end up where we're going"
|
|
|
|
|
In this, this ARM has a fantastic floating point processor. Floats are as fast as ints according to the documentation. Doubles, like two ints.
I think I'll continue to lean on it.
To err is human. Fortune favors the monsters.
|
|
|
|
|
That's not surprising. It could be that the execution clocks required per instruction are less on your 600 MHz MCU.
There used to be a tool in Norton Utilities that attempted to calculate the CPU's clock speed by timing the runtime of a specific task. As long as you were using the same CPU family (an 8026, IIRC), it gave accurate results. It gave too low results on an 8086, and too high results on an 80386+, all due to the changes in required clocks per instruction on the different CPUs.
EDIT: fixed typos
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
modified 18-Aug-22 4:36am.
|
|
|
|
|
That's probably part of it, I suspect the faster chip also having more pipelines inside the cpu core is also a factor.
Did you ever see history portrayed as an old man with a wise brow and pulseless heart, weighing all things in the balance of reason?
Is not rather the genius of history like an eternal, imploring maiden, full of fire, with a burning heart and flaming soul, humanly warm and humanly beautiful?
--Zachris Topelius
|
|
|
|
|
The superscalar nature of the processor, according to the documentation, runs typical C code at 2 instructions per clock cycle about 40% of the time.
I am inclined to think it's the great floating point processor and branch prediction that's winning the day here.
To err is human. Fortune favors the monsters.
|
|
|
|
|
Have you looked at the processor differences? Pre-fetch, caching, instruction cycles, etc.?
|
|
|
|
|
Some of them.
To err is human. Fortune favors the monsters.
|
|
|
|
|
Clock speed is only part of the equation. For instance, if this processor can do an operation in half the number of clock cycles then it will show a 50% improvement even at the same clock speed (at least for that instruction).
|
|
|
|
|
I'm well aware of that.
Now name a processor architecture that one 32-bit processor will have that another will not, that - clock speeds considered equal, runs over 3 times faster than the next one.
Especially given both processors are superscalar.
There's a profound difference here, and it can't be readily explained by clock speed or CPU architecture.
To err is human. Fortune favors the monsters.
modified 19-Aug-22 3:27am.
|
|
|
|
|
What is the memory bus width on both systems?
Back in the old days, some of the chip families had similar CPU, but 8, 16, or 32 bit memory bus. 2 fetches to grab a 16bit instruction on an 8 bit bus.
With the newer CPUs, maybe your floating point library is 100% cached on the 600MHz
|
|
|
|
|
I think part of the reasoning for the 8 bit bus, less pins means cheaper, easier boards
|
|
|
|
|
On this issue part of me wants to take a page from Mostly Harmless wherein Arthur Dent is talking about Perfectly Normal Beasts. Nobody has any idea where they come from. Some fear if they search for the answer, the question might be taken away, so to speak. This kind of feels like that to me. 
To err is human. Fortune favors the monsters.
|
|
|
|
|
I wanna say since they're probably using similar SRAM chips it would be similar, but I can't be sure how any of it is configured to that detail without poring over datasheets.
To err is human. Fortune favors the monsters.
|
|
|
|
|
I've been wanting to switch over to ARM Cortex processors for awhile - I figure coding against them opens a lot more doors for me than just using boutique Tensilica CPUs for example.
The Teensy is Arduino compatible, and the most capable (by far) of the Arduino compatible MCUs on the market. The ARM it uses is a superscalar 32-bit 600mhz processor, with branch prediction, and 64-bit floating point coprocessing. It also has two programmable USB ports, SPDIF, I2S, I2C, SPI, a zillion PWM channels, SDMMC and stuff I don't even remember right now.
So for those of you that have been dabbling in IoT, or want to, I can recommend the Teensy, particularly version 4.1
They are costlier than most other kits, but well worth it. You'll understand where the money went.
I'll probably still use ESP32s sometimes, but the Teensy blows them away.
To err is human. Fortune favors the monsters.
|
|
|
|
|
honey the codewitch wrote: The ARM it uses is a superscalar 32-bit 600mhz processor
600 miliHerz? That's going to take ages...   
M.D.V. 
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
i'm a descriptivist.
*hides*
To err is human. Fortune favors the monsters.
|
|
|
|
|
we all knew what you meant. Case sensitive units of measure begs for human error.
"A little time, a little trouble, your better day"
Badfinger
|
|
|
|
|
Wordle 425 3/6
🟩⬛🟩⬛🟨
🟩⬛🟩🟩⬛
🟩🟩🟩🟩🟩
|
|
|
|
|
Wordle 425 3/6
⬜🟨⬜🟩⬜
🟩⬜🟩🟩🟩
🟩🟩🟩🟩🟩
So close, so close
"If we don't change direction, we'll end up where we're going"
|
|
|
|
|
Wordle 425 4/6
🟨⬜⬜⬜⬜
🟨⬜⬜🟨⬜
⬜⬜⬜⬜⬜
🟩🟩🟩🟩🟩
Software rusts. Simon Stephenson, ca 1994. So does this signature. me, 2012
|
|
|
|
|
Wordle 425 3/6
⬛⬛🟩🟨⬛
🟨⬛🟩🟩🟨
🟩🟩🟩🟩🟩
|
|
|
|
|
Wordle 425 5/6
🟨⬜⬜⬜⬜
⬜🟨⬜⬜⬜
⬜🟨🟩⬜⬜
🟩⬜🟩🟩⬜
🟩🟩🟩🟩🟩
|
|
|
|
|
Wordle 425 4/6
⬛⬛🟩🟩⬛
⬛⬛🟩🟩🟨
🟩⬛🟩🟩⬛
🟩🟩🟩🟩🟩
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
