(Draft Status)

Letter regarding ISAMUX / NS

This is a quick overview of the way that we would like to add changes that we are proposing to the PowerPC instruction set. It is based on a Open Standardisation of the way that existing "mode switches", already found in the POWER instruction set, are added:

  • FPSCR's "NI" bit, setting non-IEEE754 FP mode
  • MSR's "LE" bit (and associated HILE bit), setting little-endian mode
  • MSR's "SF" bit, setting either 32-bit or 64-bit mode

All of these are set by one instruction, that, once set, radically changes the entire behaviour and characteristics of subsequent instructions.

With these (and other) long-established precedents already in POWER, there is therefore essentially conceptually nothing new about what we propose: we simply seek that the process by which such "switching" is added is formalised and standardised, such that we (and others) have a clear, standards-non-disruptive, atomic and non-intrusive path to extend the POWER ISA.

Summary of Libre-SOC Project

TODO brief summary of Libre-SOC project (hybrid CPU-GPU-VPU), thereby helping explain exactly why we need extensive augmentation of POWER ISA.

Basically it's because it is not a separate GPU-VPU, it's an actual CPU-GPU-VPU. No separate GPU, because the CPU is the GPU. No separate VPU, because the CPU is the GPU. There is not even a separate "pipeline": the CPU pipelines are the GPU and VPU pipelines.

Closest equivalents include the ARC core (which has VPU extensions and 3D extensions in the form of Broadcom's VideoCore IV) and the ICubeCorp IC3128. Both are considered "hybrid" CPU-GPU-VPU processors.

With the project being Libre - not proprietary and secretive and never to be published, ever - it is no good having the extensions as "custom" because "custom" is specifically for the cases where the augmented toolchain is never, under any circumstances, published and made public by the proprietary company. For business commercial reasons, Libre-SOC is the total opposite of this proprietary, secretive approach.

Overview

The PowerPC Instruction Set Architecture (ISA) is an abstract model of a computer. This is what programmers use when they write programs for the machine, even if indirectly via a compiler for a high level language. We must be conservative in how we add to the ISA to:

  • not break existing programs
  • be mindful as to how others may wish to add to the ISA in the future

This document describes our strategy.

ISA modes and escape sequences

New chips usually need to be able to run older (legacy) software that is incompatible with the latest and greatest ISA. Eg: 64 bit chip must be able to run older 16 bit and 32 bit software.

To enable backwards compatability the CPU will be set into 'legacy' mode. This is done with an ISA Mode switch, also known as ISA Muxing or ISA Namespaces.

The operating system is able to quickly change between 'modern' ISA mode and various legacy modes.

Another technique is an ISA escape-sequence. This is a type of mode that is only operational for a short time, unlike 32 or 64 bit which would be for the entire run of a program.

What are we adding to the ISA

When high quality graphical display were developed the CPUs at the time were shown to not be able to run the display fast enough. The solution was the use of Graphics cards, these are specialised computers that are good at rendering pixels; often by doing the same thing in different parts of the screen at the same time (in parallel). These specialised computers are called Graphical Processing Units (GPUs).

The parallelism of some GPUs is thousands. This has led to GPUs being used to solve non graphical problems where high parallelism is useful.

break

Letter regarding ISAMUX / NS

Hardware-level dynamic ISA Muxing (also known as ISA Namespaces and ISA escape-sequencing) is commonly used in instruction sets, in an arbitrary and ad-hoc fashion, added often on an on-demand basis. Examples include:

  • Setting a SPR to switch the meaning of certain opcodes for Little-Endian / Big-Endian behaviour (present in POWER and SPARC)
  • Setting a SPR to provide "backwards-compatibility" for features from older versions of an ISA (such as changing to new ratified versions of the IEEE754 standard)

(These we term "ISA Muxing" because, ultimately, they are extra bits (or change existing bits) in the actual instruction decoder phase, which involves "MUXes" to switch them on and off).

The Libre-SOC team, developing a hybrid CPU-VPU-GPU, needs to add significantly and strategically to the POWER ISA to support, for example, Khronos Vulkan IEEE754 Conformance, whilst at the same time being able to run full POWER9 compliant instructions.

There is absolutely no way that we are going to duplicate the entire FP opcode set as a custom extension to POWER, just to add a literally-identical suite of FP opcodes that are compliant with the Khronos Conformance Suites: this would be a significant and irresponsible use of opcode space.

In addition, as this processor is likely to be used for parallel compute purposes in high-efficiency environments, we also need to add FP16 support. Again: there is no way that we are going to add triple duplicated opcodes to POWER, given that the opcodes needed are absolutely identical to those that already exist, apart from the FP bitwidth (32 / 64).

There are several other strategically critical uses to which we would like to put such a scheme (related to power consumption and reducing throughput bottlenecks needed for heavy-computation workloads in GPU and VPU scenarios).

In addition, the scheme has several other key advantages over other ISA "extending" ideas (such as extending the general ISA POWER space to 64 bit) in that, unlike 64 bit opcodes, its judicious and careful use does not require large increases in I-Cache size because all opcodes, ultimately, remain 32-bit. The scheme also allows future official POWER extensions to the ISA - managed by the OpenPOWER Foundation - to be strategically managed in a controlled, long-term, non-damaging way to the reputation and stability of OpenPOWER.

Therefore we advocate being able to set "ISAMUX/NS" mode-switching bits that, like the existing LE/BE mode-switching bits, change the behaviour of existing opcodes to an alternative "meaning" (followed by another mode-switch that returns them to their original meaning. Note: to reduce binary code-size, alternative schemes include setting a countdown which, when it expires, automatically disables the requested mode-switch)

Note also that to ensure that kernels and hypervisors are not impacted by userspace ISAMUX/NS mode-switching, it is critical that Supervisor and Hypervisor modes have their own completely separate ISAMUX/NS SPRs (imagine a userspace application setting the LE/BE bit on a global basis, or setting a global IEEE754 FP Standards compatibility flag).

Further, that Supervisor / Hypervisor modes have access to and control over userspace ISAMUX/NS SPRs (without themselves being affected by setting of userspace ISAMUX/NS SPRs), in order to be able to correctly context-switch userspace applications to their correct (former) running state.

Given the number of mode-switch bits that we anticipate using, we advocate that such a scheme be formalised, and that the OpenPOWER Foundation be the "atomic arbiter" similar to IANA and JEDEC in the formal allocation of mode-switch bits to OpenPOWER implementors.

We envisage that some of these bits will be unary, some will be binary, some will be allocated for exclusive use by the OpenPOWER Foundation, some allocated to OpenPOWER Members (by the OpenPOWER Foundation), and some reserved for "custom and experimentation usage".

(This latter - custom experimentation - to be explicitly documented that upstream compiler and toolchain support will never, under any circumstances be accepted by the OpenPOWER Foundation, and that this be enforced through the EULA and through Trademark law).

However as we are quite new to POWER 3.0B (1300+ page PDF), we do appreciate that such a formal scheme may already be present in POWER9 3.0B, that we have simply overlooked.