2014-05-14 1 Mill Computing Patents pending Number eight of a series Drinking from the Firehose Many chips from one – Specification in the Mill™ CPU Architecture.

2014-05-14 1Mill Computing Patents pending

Number eight of a series

Drinking from the Firehose

Many chips from one – Specification in the Mill™ CPU Architecture


The Mill CPU

The Mill is a new general-purpose commercial CPU family.

The Mill has a 10x single-thread power/performance gain over conventional out-of-order superscalar architectures, yet runs the same programs, without rewrite.

This talk will explain:• configurable architecture strategy• attributed specification• operation set specification• component configuration at core/chip/board levels• automatic tool generation


Talks in this series

1. Encoding2. The Belt3. Memory4. Prediction5. Metadata and speculation6. Execution7. Security8. Specification9. …

You are here

Slides and videos of other talks are at:

MillComputing.com/docs


addsx(b2, b5)

The Mill Architecture

Specification and configuration

New with the Mill:

Family members built from specificationsReusable components

Instruction set built by composing attributesFully regular instruction set

Mechanically generated bit-level encodingEntropy-optimal encoding throughout

Configuration-specific generated tool setsAsm, sim, debugger, compiler, …

Generated hardwareVerilog from specification


Caution!

Gross over-simplification!

This talk tries to convey an intuitive understanding to the non-specialist.

The reality is more complicated.


Specification

This talk does not describe the Mill architecture

Unlike other talks in this series…

It describes how the operation set and particular family member micro-architectures are specified.

It describes, and demonstrates, some of the software tools built from the specifications.

It describes how the specification supports manual creation of Mill hardware.


Specification

This talk does not describe the Mill architecture

Unlike other talks in this series…

The specification tools are for internal use in creation of Mill CPUs; the tools are not intended to be products.

By use of these tools, we can create new Mill chip products more quickly and at lower cost than usual.

The intended audience includes tool designers and software developers interested in advanced design.


Specification

abstract Mill CPU architecture

family members Tin

Copper

Silver

Gold

The Mill is a family of member CPUs sharing an abstract operation set and micro-architecture.

specification driven

Members differ in concrete operation set and micro-architecture..

Designers describe a concrete member by writing a specification.


Specification

abstract Mill CPU architecture

family members Tin

Silver

Gold

tools compiler

asmdebugge

rHWgensim

Software automatically creates system software, verification tests, documentation, and a hardware framework for the new member from the specification.

specification driven

data driven

Copper


C++ compiler masquerade

assembly language

The Mill assembler syntax is C++.Suitably disguised.


Two-pass assemblersTraditional assemblers have two passes.

The first pass treats the source as a program in a meta-language, usually a macro language, and interprets that program to produce a different source program in machine language.

The second pass translates the program in machine language to binary and produces the executable file.

source file

macrolanguage

first pass machine

language

second pass

load module

binary


The Mill assembler uses the C++ compiler

The first pass is the C++ compiler, which translates the assembly language source program to an executable.

The second pass is the execution of the C++ program, to emit binary and produce the executable file.

source file

macrolanguage

first pass machine

language

second pass

load module

binary

source file

C++C++

compiler programexecution

load module

binary


C++ is Mill assembly language?

Each assembler operation is a C++ function call.

add b3,b5jump looploop:

add(b3,b5)br("loop")L("loop")

conventional assembler

Mill assembler

add(b0, 3), store(*b5, b7), br("loop“);

A Mill instruction is a C++ statementcomprising operations separated by commas

operations instruction


C++ as meta-language

for(int i = 1; i < 5; ++i) add(b0, i);

Each call of an asm function emits that operation.

gives the same machine code as:

add(b0, 1);add(b0, 2);add(b0, 3);add(b0, 4);

As in a macro assembler, in Mill assembler you can meta-program what your program will be.


Demo: extend a core

The test case contains this code fragment:

con(w(fpemu::st2bin32("3.0")));con(w(fpemu::st2bin32("5.0")));addb(b0, b1);nop();nop();nop();

However:

the Tin CPU does not support native floating-point


Demo: extend a core

Build for Tin – fail


Demo: create new “Demo” member like Tin

Copy code tree Tin -> DemoClear old files from new tree Tell builder tool to use new memberCreate new specification from Tin spec


Demo: update Demo member with FPU

Populate execution pipeline slot with floating point


ISA design by composition

attributesThe semantic pieces of operations


Operation attributes

A Mill operation invocation comprises a core operation and values for some number of attributes.

Specific attribute values are supplied by the operation mnemonic or by an argument to the operation function.

addus(b3, 17)

unsigned integersaturatingb3 – third belt position17 - literal

domainoverflowopand0imm0

attribute value

There are ~50 attributes. Only a handful are meaningful for any particular operation.

core operation add


Attribute values

Most attributes are enumerations:

enum directionCode {leftward, rightward};

enum condSenseCode {allSense, falseSense, trueSense};

enum overflowCode {excepting, modulo, saturating, widening};

Attribute values can be specified individually, or as bitsets with a selection of values of the same attribute.

enum domainCode{binFloat, boolean, decFloat, logical, pointers, signedInt, unsignedInt};


MnemonicsEach opcode and attribute value has a text string nick.

value nick

leftwardrightwardsignedIntunsignedInt

lrsu

Spec software concatenates the nicks of the opcode and attributes to make the assembler mnemonic automatically.

shiftrs operation is shift, right, signed

There are ~120 core ops and ~1000 mnemonics.


Attribute semanticsBesides its type, each attribute has three choices:

how values are expressed in assembler source• by mnemonic, based on the function name• by parameter, based on explicit argument• derived from other attributes, not in source

how values are encoded in target binary code• pinned in a single bit field in all formats• direct in different bit fields in different formats• merged into an opcode super-field• uncoded, for internal use only

how the set of permitted values is determined• universal, same for all slots for all members• by member, same for all slots on a given member• by slot, may vary based on available entropy


A candidate operation

Semantics of the new operation:

shiftaddincrement

Assembler:

Any ALU can do this in one cycle

#define N = 7uint16_t NEWOP (uint16_t a, uint16_t b) {

return (a << N) + b + 1; }

Pick a name: Pick a value for N:


Demo: define a new opcode

Add new opcode – opAttr.hhAdd printname – opAttr.ccAdd traits – attrTraits.cc


Argument signatures

Some attributes get their value from the function arguments in the operation, rather than the mnemonic.

arg kind meaningexuArgimmArgbitArgoffArg

exu-side belt positionsmall immediate constantbit numberload/store offset

Argument nicks are concatenated into signatures.

exuBitSigbaseOffWidthfSigexuExuExuSig

belt position, bit numberaddress base, offset, widththree belt positions

signature arguments, in order

Ops are uniquely identified by their mnemonic and signature


Operation patterns

An operation pattern comprises:• the core operation and its encoding block• the argument list• all meaningful values for all mnemonic attributes

Each pattern defines all the operations that result from the cross-product of attribute values: the models.

There are around a thousand models.

Operations are defined as patterns, not individually.

opPattern(exuBlock, addOp) << floats << roundings << exuArg << exuArg;

This defines 12 models: six different rounding modes for each of binary and decimal floating point


What attributes for our new operation?What domain?

What about overflow?ignore it?mark result as an error?saturate to maximal value?produce a double-width result?

signedInt?unsignedInt?

Where to encode it?

exuBlock?

What arguments?

exuArg, exuArg?


Demo: define a new operation

Add specification – opSpecs.ccAdd sim implementationbuild sim


Say how – or say what?

specificationHardware development made easy.


Abstract Mill-ness


abstract Mill

operation set

micro-architectur

e


Abstract Mill-ness


abstract Mill

operation set

micro-architectur

e


Specifications make concrete from abstract


abstract Mill

operation set

micro-architectur

e concrete Mill chips

Crimson

Monocore

...

specifications


Why specification/configuration?

Creating a CPU by hand is fabulously expensive.

Much of CPU implementation is repetitive, error-prone, tedious and wasteful.

Often the design winds up sub-optimal because it’s too much trouble to change it yet again

The Mill team knew it lacked the resources to implement – and re-implement – a moving target from scratch

So we got the software to do it• can address multiple markets efficiently• fast pivots for new chips• economy for company and customers

Result:


Concrete Mill chips

Each concrete chip is specified as a set of components, including cores, caches, memory controllers, etc.

Crimson

Monocore

...

concrete Mill chips

“Crimson” chip


Concrete Mill chips

Copper core ...

“Crimson” chip

Silver core

caches

“Copper” core

Each concrete chip is specified as a set of components, including cores, caches, memory controllers, etc.


Concrete Mill cores

...

“Copper” core

specRegs

caches

ALUs

Belt

decoders

The component cores in turn specify still more nested components.


Recursive specification

Apologies to Jonathan Swift

Big parts have little parts,within each to excite ’em;and little parts have smaller parts,and so ad infinitum

It’s components, all the way down!


Component parameters

Components have parameters to define their function.

size = 16

belt

bank count = 4line width = 64evict policy = LRUway count = 4 …

cache

exit table size = 2048latency = 2…

predictor

Components of the same kind but different parameter values can be collected in palettes for reuse in designing other Mills.


Behind components

Behind each component kind is hand-written software:

• An emulation function sits in the simulator.It defines what the component does in the machine.

• A generator function sits in the generator.It emits the Verilog starting point for hardware.

The emulation function is definitive; if the hardware doesn’t match the simulator then the simulator is right.


Clock domains

The Mill sim is event-driven at pico-second accuracy.

All components reside in a clock domain. By default sub-components reside in the domain of their parent.

Xtal components create top-level clock domains.

PLL components link different domains. The ratio registers are in MMIO space for program control.

A simulated Mill program can use simulated MMIO to control the simulated hardware and change the simulated clock rate that it itself is running under.


Memory hierarchy

Components that derive from the memLevel type can be hooked together to model the memory hierarchy.

The connections are streams of requests and responses. Each component only deals with the stream. It does not know or care what is on the other end.

The streams use predictive throttling for congestion control, similar to network message methods.

Streams run at full speed, without handshaking delay.


Demo: try it out

run sim - ivan/build/testAsm.sim


Other roads…

There are other architectures that provide operation specification. These differ significantly from the Mill.

purpose:add special-purpose embedded operationsform optimal subsets for family members

encoding:reserved bit patterns, manually selectedautomatically generated optimal-entropy

specification:one-at-a-time manual processpattern-based orthogonal generation


Summary:

The Mill:

Defines members by component lists

Defines operations by composing attributesTool produces cross-product of attributes

Recursive composition – mix and match

Compact notation expresses clock, memorySays what connects to what, tool creates “how”.


For technical info about the Mill CPU architecture:

MillComputing.com/docs

To sign up for future announcements, white papers etc.

MillComputing.com/mailing-list

Shameless plug

2014-05-14 1 Mill Computing Patents pending Number eight of a series Drinking from the Firehose Many chips from one – Specification in the Mill™ CPU Architecture.

Documents

specification slide

computing patents

architecture specification

specification tools

c program

cpu architecture slide

specification abstract

machine language