Cortex A76 µarch - Backend

Switching to the back-end of the core we have a look at the execution core. 

The integer core contains 6 issue queues and execution ports (4 depicted in the slide plus 2 load/store pipelines). There are 3 integer execution pipelines – two ALUs capable of simple arithmetic operations and a complex pipeline handling also multiplication, division and CRC ops. The three integer pipelines are served by 16 deep issue queues. The same size issue queue can also be found serving the single branch execution port.

Two load/store units are the remaining ports of the integer core and are each served by two 12 deep issue queues. The issue queue stages are 3 cycles deep and while I mentioned that the rename/dispatch is 1 stage deep, the dispatch stage actually overlaps with the first cycle of the issue queues stages.

The ASIMD/floating point core contains two pipelines which are served by two 16-deep issue queues.

When it comes to the backend of a CPU core the two most important metrics are instruction throughput and latency. Where the A76 in particular improves a lot is in terms of instruction latency as it’s able to shave off cycles on very important instructions.

To better overview the improvements I created a table with the most common instruction types. The execution throughput and latencies presented here are for AArch64 instructions and if not otherwise noted represent operations on 64-bit data for integer and 64bit (double precision) FP.

Backend Execution Throughput and Latency
  Cortex-A75 Cortex-A76 Exynos-M3
  Exec Lat Exec Lat Exec Lat
Integer Arithmetic (Add, sub) 2 1 3 1 4 1
Integer Multiply 32b 1 3 1 2 2 3
Integer Multiply 64b 1 3 1 2 1
(2x 0.5)
4
Integer Multiply Accumulate 1 3 1 2 1 3
Integer Division 32b 0.25 12 0.2 < 12 1/12 - 1 < 12
Integer Division 64b 0.25 12 0.2 < 12 1/21 - 1 < 21
Move  (Mov) 2 1 3 1 3 1
Shift ops (Lsl) 2 1 3 1 3 1
Load instructions 2 4 2 4 2 4
Store instructions 2 1 2 1 1 1
FP Arithmetic 2 3 2 2 3 2
FP Multiply 2 3 2 3 3 4
FP Multiply Accumulate 2 5 2 4 3 4
FP Division (S-form) 0.2-0.33 6-10 0.66 7 >0.16
(2x 1+/12)
12
FP Load 2 5 2 5 2 5
FP Store 2 1-N 2 2 2 1
ASIMD Arithmetic 2 3 2 2 3 1
ASIMD Multiply 1 4 1 4 1 3
ASIMD Multiply Accumulate 1 4 1 4 1 3
ASIMD FP Arithmetic 2 3 2 2 3 2
ASIMD FP Multiply 2 3 2 3 1 3
ASIMD FP Chained MAC (VMLA) 2 6 2 5 3 5
ASIMD FP Fused MAC (VFMA) 2 5 2 4 3 4

On the integer operations side the A76 improves the multiplication and multiply accumulate latencies from 3 cycles down to 2 cycles, with the throughput remaining the same when compared to the A75. Obviously because the A76 has 3 integer pipelines simple arithmetic operations see a 50% increase in throughput versus the A75’s 2 pipelines.

The much larger and important improvements can be found in the “VX” (vector execution) pipelines which are in charge of FP and ASIMD operations. Arm calls the new pipeline a “state-of-the-art” design and this is finally the result that’s been hyped up for several years now.

Floating point arithmetic operations have been reduced in latency from 3 cycles down to 2 cycles, and multiply accumulate has also shaved off a cycle from 5 cycles down to 4.

What Arm means by the “Dual 128bit ASIMD” with doubled execution bandwidth is that for the A75 and prior only one of the vector pipelines was capable of 128bit while the other one was still 64-bit. For the A76 both vector pipelines are 128-bit now so quad-precision operations see a doubling of the execution throughput.

Moving onto more details of the data handling side, we see the again the two load/store pipelines which was something first implemented on the A73 and A75. Although depicted as one issue queue in the slide, the LD/S pipelines each have their own queues at 16 entries deep.

The data cache is fixed at 64KB and is 4-way associative. Load latency remains at 4 cycles. The DTLBs run a separate pipeline as tag and data lookup. Arm’s goals here is aiming for maximum MLP/ memory level parallelism to be able to feed the core.

In a perfect machine everything would be already located in the caches, so it’s important to have very robust prefetching capabilities. On the A76 we see a new 4th generation prefetchers introduced to get nearer to this goal of perfect cache-hit operation. In all the A76 has 4 different prefetching engines running in parallel looking at various data patterns and loading data into the caches.

In terms of the A76 cache hierarchy Arm is said to have made no compromises and got the best of both worlds in terms of bandwidth and latency. The 64KB L1 instruction cache reads up to 32B/cycle and the same bandwidth applies to the L1 data cache in both directions. The L1 is a writeback cache. The L2 cache is configurable in 256 or 512KB sizes and is D-side inclusive with the same 2x 32B/cycle write and read interfaces up to the exclusive L3 cache in the 2nd generation DSU.

Overall the microarchitectural improvements on the core are said to improve memory bandwidth to DRAM by up to 90% in microbenchmarks.

All in all the microarchitecture of the A76 could be summed up in a few focus design points: Maximise memory performance throughout the core by looking at every single cycle. During the design phase the engineers were looking at feature changes with a sensitivity of up to 0.25% in performance or power – if that metric was fulfilled then it was deemed to be a worthwhile change in the core. Small percentages then in turn add up to create significant figures in the end product.

The focus on bandwidth on latency is said to have been extreme, and Arm was very adamant in re-iterating that to be able to take full advantage of the microarchitecture that vendors need to implement an equally capable memory subsystem on the SoC to see full advantages. A figure that was put out there was 0.25% of performance per nanosecond of latency to main memory. As we’ve seen in the Snapdragon 845 one of the reasons the SoC didn’t quite reach Arm’s projected performance metrics was the degraded memory latency figures which might have been introduced by the L4 system cache in the SoC. In the future vendors will need to focus more on providing latency sensitive memory subsystems as otherwise they’ll be letting free performance and power on the table with differences that could amount to basically a generational difference in CPU IP.

Cortex A76 µarch - Frontend Cortex A76 - Performance & Power Projections
Comments Locked

123 Comments

View All Comments

  • darwiniandude - Friday, June 1, 2018 - link

    Lots of Apple hate in these comments. Which is fine, nothing wrong with having your own opinion. Performance is important to me - I edit 4K (not for business purposes) from a few Fuji mirrorless bodies quite happily on iOS - on an iPhone X and iPad Pro. My fastest desktop and notebook machines I own currently are both Core2Duo. They simply cannot do it. I’m not a typical use case. I did have a quad i7, but I sold that machine (MacBook Pro) while I could still get a stupidly high amount of money back for it used. Don’t assume that no one on mobile wants high performance ARM cores - not everyone is just using Facebook messenger taking the occasional selfie all day.
    Also, I remember when AMD smoked Intel at times in the past. People argued, but there was never the “you don’t need that performance” type arguments.
  • leledumbo - Friday, June 1, 2018 - link

    That's what Apple is actually doing: single TDP configurable SoC for both their phones and pads (and tops if the rumor come true).

    The argument is not "you don't need that performance", but "most people don't need that performance". You are one of the few in the performance-needy pool. I know you exist, just not many, and that's what many manufacturers are aware of, so they don't take the Apple route.
  • hlovatt - Friday, June 1, 2018 - link

    But this is Anantech. We want the best. We want to push the envelope. I don’t want to read about ho hum performance at a good price. That’s for Consumer Reports or a myriad of other yawn sites.
  • serendip - Friday, June 1, 2018 - link

    4k editing on an iPad probably won't be using the CPU completely for processing though. There's a lot of stuff that can be passed on to faster and more efficient DSP and IP blocks. I've also run Quicksync encoding on Atom tablets running Windows, it's much faster than using the puny Atom cores directly.
  • darwiniandude - Saturday, June 2, 2018 - link

    https://forums.luma-touch.com/viewtopic.php?t=6493 Some discussion of performance on iPad at 4K. It really does work very well. Must be using the GPU also.
  • tipoo - Tuesday, September 4, 2018 - link

    Afaik, only playback will use dedicated blocks like Quicksync, editing itself, the rendering of new effects, would be heavily assisted by the GPU and partly on the CPU.
  • techconc - Tuesday, June 5, 2018 - link

    The "most people don't need that performance" argument may sound nice to say, but why do you think people buy new phones? They do it when their old phone feels slow,etc. A higher performing phone has a longer effective life span.

    Using phones as an example, Android has about 85% of the market share for devices sold. Yet, when Apple and Google report their active user base, Android barely maintains a 2:1 ratio over iOS devices. Why? The majority of Android devices sold are low end devices that have a much shorter effective life span.
  • Meteor2 - Monday, July 2, 2018 - link

    I kinda want to buy a new phone, but my Nexus 5X simply doesn't feel slow. So I haven't. And it must have less than half the performance of modern high-end phones.
  • Maxiking - Sunday, June 3, 2018 - link

    So another paper dragon, YAY.

    They promise the same every year, so statistically, if they keep repeating the lie every year, they will get there eventually!
  • Herkko - Wednesday, June 6, 2018 - link

    Tell me how much Nintendo Switch power and energy effiency grow if chance old ARM-cortex A57-A53 new ARM-cortex A76 CPU

Log in

Don't have an account? Sign up now