Custom Code to Understand a Custom Core

Section by Anand Shimpi

All Computer Engineers at NCSU had to take mandatory programming courses. Given that my dad is a Computer Science professor, I always had exposure to programming, but I never considered it my strong suit - perhaps me gravitating towards hardware was some passive rebellious thing. Either way I knew that in order to really understand Swift, I'd have to do some coding on my own. The only problem? I have zero experience writing Objective-C code for iOS, and not enough time to go through a crash course.

I had code that I wanted to time/execute in C, but I needed it ported to a format that I could easily run/monitor on an iPhone. I enlisted the help of a talented developer friend who graduated around the same time I did from NCSU, Nirdhar Khazanie. Nirdhar has been working on mobile development for years now, and he quickly made the garbled C code I wanted to run into something that executed beautifully on the iPhone. He gave me a framework where I could vary instructions as well as data set sizes, which made this next set of experiments possible. It's always helpful to know a good programmer.

So what did Nirdhar's app let me do? Let's start at the beginning. ARM's Cortex A9 has two independent integer ALUs, does Swift have more? To test this theory I created a loop of independent integer adds. The variables are all independent of one another, which should allow for some great instruction level parallelism. The code loops many times, which should make for some easily predictable branches. My code is hardly optimal but I did keep track of how many millions of adds were executed per second. I also reported how long each iteration of the loop took, on average.

Integer Add Code
  Apple A5 (2 x Cortex A9 @ 800MHz Apple A5 Scaled (2 x Cortex A9 @ 1300MHz Apple A6 (2 x Swift @ 1300MHz Swift / A9 Perf Advantage @ 1300MHz
Integer Add Test 207 MIPS 336 MIPS 369 MIPS 9.8%
Integer Add Latency in Clocks 23 clocks   21 clocks  

The code here should be fairly bound by the integer execution path. We're showing a 9.8% increase in performance. Average latency is improved slightly by 2 clocks, but we're not seeing the sort of ILP increase that would come from having a third ALU that can easily be populated. The slight improvement in performance here could be due to a number of things. A quick look at some of Apple's own documentation confirms what we've seen here: Swift has two integer ALUs and can issue 3 operations per cycle (implying a 3-wide decoder as well). I don't know if the third decoder is responsible for the slight gains in performance here or not.

What about floating point performance? ARM's Cortex A9 only has a single issue port for FP operations which seriously hampers FP performance. Here I modified the code from earlier to do a bunch of single and double precision FP multiplies:

FP Add Code
  Apple A5 (2 x Cortex A9 @ 800MHz Apple A5 Scaled (2 x Cortex A9 @ 1300MHz Apple A6 (2 x Swift @ 1300MHz Swift / A9 Perf Advantage @ 1300MHz
FP Mul Test (single precision) 94 MFLOPS 153 MFLOPS 143 MFLOPS -7%
FP Mul Test (double precision) 87 MFLOPS 141 MFLOPS 315 MFLOPS 123%

There's actually a slight regression in performance if we look at single precision FP multiply performance, likely due to the fact that performance wouldn't scale perfectly linearly from 800MHz to 1.3GHz. Notice what happens when we double up the size of our FP multiplies though, performance goes up on Swift but remains unchanged on the Cortex A9. Given the support for ARM's VFPv4 extensions, Apple likely has a second FP unit in Swift that can help with FMAs or to improve double precision FP performance. It's also possible that Swift is a 128-bit wide NEON machine and my DP test compiles down to NEON code which enjoys the benefits of a wider engine. I ran the same test with FP adds and didn't notice any changes to the data above.

Sanity Check with Linpack & Passmark

Section by Anand Shimpi

Not completely trusting my own code, I wanted some additional data points to help understand the Swift architecture. I first turned to the iOS port of Linpack and graphed FP performance vs. problem size:

Even though I ran the benchmark for hundreds of iterations at each data point, the curves didn't come out as smooth as I would've liked them to. Regardless there's a clear trend. Swift maintains a huge performance advantage, even at small problem sizes which supports the theory of having two ports to dedicated FP hardware. There's also a much smaller relative drop in performance when going out to main memory. If you do the math on the original unscaled 4S scores you get the following data:

Linpack Throughput: Cycles per Operation
  Apple Swift @ 1300MHz (iPhone 5) ARM Cortex A9 @ 800MHz (iPhone 4S)
~300KB Problem Size 1.45 cycles 3.55 cycles
~8MB Problem Size 2.08 cycles 6.75 cycles
Increase 43% 90%

Swift is simply able to hide memory latency better than the Cortex A9. Concurrent FP/memory operations seem to do very well on Swift...

As the last sanity check I used Passmark, another general purpose iOS microbenchmark.

Passmark CPU Performance
  Apple A5 (2 x Cortex A9 @ 800MHz Apple A5 Scaled (2 x Cortex A9 @ 1300MHz Apple A6 (2 x Swift @ 1300MHz Swift / A9 Perf Advantage @ 1300MHz
Integer 257 418 614 47.0%
FP 230 374 813 118%
Primality 54 87 183 109%
String qsort 1065 1730 2126 22.8%
Encryption 38.1 61.9 93.5 51.0%
Compression 1.18 1.92 2.26 17.9%

The integer math test uses a large dataset and performs a number of add, subtract, multiply and divide operations on the values. The dataset measures 240KB per core, which is enough to stress the L2 cache of these processors. Note the 47% increase in performance over a scaled Cortex A9.

The FP test is identical to the integer test (including size) but it works on 32 and 64-bit floating point values. The performance increase here despite facing the same workload lends credibility to the theory that there are multiple FP pipelines in Swift.

The Primality benchmark is branch heavy and features a lot of FP math and compares. Once again we see huge scaling compared to the Cortex A9.

The qsort test features integer math and is very branch heavy. The memory footprint of the test is around 5MB, but the gains here aren't as large as we've seen elsewhere. It's possible that Swift features a much larger branch mispredict penalty than the A9.

The Encryption test works on a very small dataset that can easily fit in the L1 cache but is very heavy on the math. Performance scales very well here, almost mirroring the integer benchmark results.

Finally the compression test shows us the smallest gains once you take into account Swift's higher operating frequency. There's not much more to conclude here other than we won't always see greater than generational scaling from Swift over the previous Cortex A9.

Decoding Swift Apple's Swift: Visualized
POST A COMMENT

276 Comments

View All Comments

  • medi01 - Wednesday, October 17, 2012 - link

    1) Compare ipad2's gamut, cough
    2) Check values on toms
    http://media.bestofmicro.com/3/4/331888/original/g...
    http://www.tomshardware.com/reviews/ipad-3-benchma...

    Unlike anand, toms was beyond primitive contrast/brightness benchmarking for quite a while.
    Reply
  • thunng8 - Thursday, October 18, 2012 - link

    Not sure if I should trust Tom's figures compared to Anands's.

    In any case, both show the ipad3 has higher gamut, especially in sRGB.
    Reply
  • steven75 - Wednesday, October 17, 2012 - link

    I think what you meant to say is that AMOLEDs win on black levels and that's about it. LCDs still win in accuracy and most importantly ability to see them in outdoor settings. Reply
  • KoolAidMan1 - Tuesday, October 16, 2012 - link

    Not even close. Even the better Android displays like the Galaxy S3 has a PenTile display. Despite having more "pixels" it actually has fewer subpixels than the iPhone does. Unless you have bad eyesight the S3 display looks really bad in comparison, and this is before we get to even worse smartphone displays out there by HTC, etc. Reply
  • Sufo - Tuesday, October 16, 2012 - link

    Old pentile displays were visibly jaggy on vertical lines - even my old lumia 800 exhibited this to some extent. On the GS3 tho, it is not noticeable and it has nothing to do with eyesight.

    Your comment makes it sound (to someone who has seen many different smartphone displays in person) as though you haven't spent much time with the GS3 (read: many smartphones) at all. Simply mentioning that is uses pentile subpix config, from you, sounds like regurgitated information. Not only that, but you seem to gloss over the many benefits that amoled panels bring. It's arguable that these benefits are more important than an accurate colourspace on (specifically) a mobile phone - although it is ofc entirely subjective.

    This brings me to the last tell of ignorance I noted; your mention of HTC. Have you used a One X? For those who do not like amoled panels, the display on the one x is perhaps nicer than both the gs3 and the ip5. Ofc you may say Android is not your cup of tea, and that's a perfectly justifiable stance, however it has nothing to do with display tech.

    tl;dr You sound like you don't know what you're talking about
    Reply
  • KoolAidMan1 - Tuesday, October 16, 2012 - link

    I do know what I'm talking about given that I've seen many smartphones, and I've calibrated my share of desktop displays to sRGB.

    Differences in display tech aside, Android phones have never gotten color profiles right, EVER. They're almost always oversaturated, have too much contrast, and are inaccurate. Anand even posted a difference in color accuracy between several devices, and the profile for the S3 is totally what I expected.

    The S3 really doesn't look good, period, but then again there are people who argue that TN panels are just fine against IPS. I'm used to hearing nonsense on forums when it comes to display from people who don't know what to look for.
    Reply
  • KoolAidMan1 - Tuesday, October 16, 2012 - link

    BTW, apologies if that came out harsh, but the difference in color and contrast accuracy between something like the S3 and a properly calibrated device is a night and day difference to me. I'm pretty sensitive to display quality though; my main desktop display at home is still an NEC and my plasma is a Pioneer Elite (RIP) Reply
  • rocketbuddha - Tuesday, October 16, 2012 - link

    For Android you have the following 720p HD Displays

    SLCD - HTC Rezound (2011 tech)
    SLCD 2 - HTC One X, Sony HD
    HD SAMOLED Pentile - GS3, Galaxy Nexus, Moto Razr HD
    HD SAMOLED RGB - Galaxy Note II
    True IPS LCD - LG Optimus 4X, Optimus G
    Super IPS LCD -Asus Padphone, Sharp phones etc

    So you have big set of choices. If dark contrasts are important then SAMOLED is the way to go. SAMOLED RGB over SAMOLED Pentile.
    If overall color and whites are important go with SLCD2.
    IPS LCDs are the closest to the Retina Display and u have a choices there too. You can pick and choose what is good for you and have alternatives.
    Reply
  • Spunjji - Thursday, October 18, 2012 - link

    The HTC One X has what is hailed to be one of the best LCD smartphone displays out there. Your claim is invalid.

    Similarly, the Galaxy Note 2 has an AMOLED display without PenTile. Sure, it's lower density, but one does not hold a 5.5" screen so close to one's face.
    Reply
  • medi01 - Wednesday, October 17, 2012 - link

    ""The iPhone 5 display is better than any current Android display.""
    Why don't you go hit your dumb head with something heavy, ipad would do?
    Reply

Log in

Don't have an account? Sign up now