(no commit message)

[libreriscv.git] / openpower / sv / ldst.mdwn

# SV Load and Store

This section describes how Standard Load/Store Defined Word-instructions are exploited as
Element-level Load/Stores and augmented to create direct equivalents of
Vector Load/Store instructions.  

<!-- hide -->
Links:

* <https://bugs.libre-soc.org/show_bug.cgi?id=561>
* <https://bugs.libre-soc.org/show_bug.cgi?id=572>
* <https://bugs.libre-soc.org/show_bug.cgi?id=571>
* <https://bugs.libre-soc.org/show_bug.cgi?id=940> post autoincrement mode
* <https://bugs.libre-soc.org/show_bug.cgi?id=1047> Data-Dependent Fail-First
* <https://llvm.org/devmtg/2016-11/Slides/Emerson-ScalableVectorizationinLLVMIR.pdf>
* <https://github.com/riscv/riscv-v-spec/blob/master/v-spec.adoc#vector-loads-and-stores>
* [[ldst/discussion]]

## Rationale

All Vector ISAs dating back fifty years have extensive and comprehensive
Load and Store operations that go far beyond the capabilities of Scalar
RISC and most CISC processors, yet at their heart on an individual element
basis may be found to be no different from RISC Scalar equivalents.

The resource savings from Vector LD/ST are significant and stem
from the fact that one single instruction can trigger a dozen (or in
some microarchitectures such as Cray or NEC SX Aurora) hundreds of
element-level Memory accesses.

Additionally, and simply: if the Arithmetic side of an ISA supports
Vector Operations, then in order to keep the ALUs 100% occupied the
Memory infrastructure (and the ISA itself) correspondingly needs Vector
Memory Operations as well.

Vectorized Load and Store also presents an extra dimension (literally)
which creates scenarios unique to Vector applications, that a Scalar (and
even a SIMD) ISA simply never encounters: not even the complex Addressing
Modes of the 68,000 or S/360 resemble Vector Load/Store.
SVP64 endeavours to add the
modes typically found in *all* Scalable Vector ISAs, without changing the
behaviour of the underlying Base (Scalar) v3.0B operations in any way.
(The sole apparent exception is Post-Increment Mode on LD/ST-update
instructions)
<!-- show -->

## Modes overview

Vectorization of Load and Store requires creation, from scalar operations,
a number of different modes:

* **fixed aka "unit" stride** - contiguous sequence with no gaps
* **element strided** - sequential but regularly offset, with gaps
* **vector indexed** - vector of base addresses and vector of offsets
* **Speculative Fault-first** - where it makes sense to do so
* **Data-Dependent Fail-First** - Conditional truncation of Vector Length
* **Structure Packing** - covered in SV by [[sv/remap]] and Pack/Unpack Mode.

*Despite being constructed from Scalar LD/ST none of these Modes exist
or make sense in any Scalar ISA. They **only** exist in Vector ISAs
and are a critical part of its value*.

Also included in SVP64 LD/ST is Element-width overrides and Twin-Predication.

Note also that Indexed [[sv/remap]] mode may be applied to both Scalar
LD/ST Immediate Defined Word-instructions *and* LD/ST Indexed Defined Word-instructions.
LD/ST-Indexed should not be conflated with Indexed REMAP mode:
clarification is provided below.

**Determining the LD/ST Modes**

A minor complication (caused by the retro-fitting of modern Vector
features to a Scalar ISA) is that certain features do not exactly make
sense or are considered a security risk.  Fault-first on Vector Indexed
would allow attackers to probe large numbers of pages from userspace,
where strided Fault-first (by creating contiguous sequential LDs likely
to be in the same Page) does not.

In addition, reduce mode makes no sense.  Realistically we need an
alternative table definition for [[sv/svp64]] `RM.MODE`.  The following
modes make sense:

* simple (no augmentation)
* Fault-first (where Vector Indexed is banned)
* Data-dependent Fail-First (extremely useful for Linked-List pointer-chasing)
* Signed Effective Address computation (Vector Indexed only, on RB)

More than that however it is necessary to fit the usual Vector ISA
capabilities onto both Power ISA LD/ST with immediate and to LD/ST
Indexed. They present subtly different Mode tables, which, due to lack
of space, have the following quirks:

* LD/ST Immediate has no individual control over src/dest zeroing,
  whereas LD/ST Indexed does.
* LD/ST Immediate has saturation but LD/ST Indexed does not.

## Format and fields

Fields used in tables below:

* **zz**: both sz and dz are set equal to this flag.
  If predication is enabled will put zeros into the dest
  (or as src in the case of twin pred) when the predicate bit is zero.
  otherwise the element is ignored or skipped, depending on context.
* **inv CR bit** just as in branches (BO) these bits allow testing of
  a CR bit and whether it is set (inv=0) or unset (inv=1)
* **RC1** as if Rc=1, stores CRs *but not the result*
* **SEA** - Signed Effective Address, if enabled performs sign-extension on
  registers that have been reduced due to elwidth overrides
* **PI** - post-increment mode (applies to LD/ST with update only).
  the Effective Address utilised is always just RA, i.e. the computation of
  EA is stored in RA **after** it is actually used.
* **LF** - Load/Store Fail or Fault First: for any reason Load or Store Vectors
  may be truncated to (at least) one element, and VL altered to indicate such.
* **VLi** - Inclusive Data-Dependent Fail-First: the failing element is included
  in the Truncated Vector.
* **els** - Element-strided Mode: the element index (after REMAP)
  is multiplied by the immediate offset (or Scalar RB for Indexed). Restrictions apply.

When VLi=0 on Store Operations the Memory update does **not** take place
on the element that failed.  EA does **not** update into RA on Load/Store
with Update instructions either.

**LD/ST immediate**

The table for [[sv/svp64]] for `immed(RA)` which is `RM.MODE`
(bits 19:23 of `RM`) is:

| 0 | 1 |  2  |  3   4  |  description               |
|---|---| --- |---------|--------------------------- |
|els| 0 | PI  |  zz LF  | post-increment and Fault-First  |
|VLi| 1 | inv | CR-bit  | Data-Dependent  ffirst CR sel   |

The `els` bit is only relevant when `RA.isvec` is clear: this indicates
whether stride is unit or element:

```
    if RA.isvec:
        svctx.ldstmode = indexed
    elif els == 0:
        svctx.ldstmode = unitstride
    elif immediate != 0:
        svctx.ldstmode = elementstride
```

An immediate of zero is a safety-valve to allow `LD-VSPLAT`: in effect
the multiplication of the immediate-offset by zero results in reading from
the exact same memory location, *even with a Vector register*. (Normally
this type of behaviour is reserved for the mapreduce modes)

For `LD-VSPLAT`, on non-cache-inhibited Loads, the read can occur just
the once and be copied, rather than hitting the Data Cache multiple
times with the same memory read at the same location.  The benefit of
Cache-inhibited LD-splats is that it allows for memory-mapped peripherals
to have multiple data values read in quick succession and stored in
sequentially numbered registers (but, see Note below).

For non-cache-inhibited ST from a vector source onto a scalar destination:
with the Vector loop effectively creating multiple memory writes to
the same location, we can deduce that the last of these will be the
"successful" one. Thus, implementations are free and clear to optimise
out the overwriting STs, leaving just the last one as the "winner".
Bear in mind that predicate masks will skip some elements (in source
non-zeroing mode).  Cache-inhibited ST operations on the other hand
**MUST** write out a Vector source multiple successive times to the exact
same Scalar destination. Just like Cache-inhibited LDs, multiple values
may be written out in quick succession to a memory-mapped peripheral
from sequentially-numbered registers.

Note that any memory location may be Cache-inhibited
(Power ISA v3.1, Book III, 1.6.1, p1033)

*Programmer's Note: an immediate also with a Scalar source as a "VSPLAT"
mode is simply not possible: there are not enough Mode bits. One single
Scalar Load operation may be used instead, followed by any arithmetic
operation (including a simple mv) in "Splat" mode.*

**LD/ST Indexed**

The modes for `RA+RB` indexed version are slightly different
but are the same `RM.MODE` bits (19:23 of `RM`):

| 0 | 1 |  2  |  3   4  |  description               |
|---|---| --- |---------|--------------------------- |
|els| 0 | PI  |  zz SEA | post-increment and Fault-First        |
|VLi| 1 | inv | CR-bit  | Data-Dependent ffirst CR sel        |

Vector Indexed Strided Mode is qualified as follows:

```
    if els and !RA.isvec and !RB.isvec:
        svctx.ldstmode = elementstride
```

A summary of the effect of Vectorization of src or dest:

```
    imm(RA)  RT.v   RA.v   no stride allowed
    imm(RA)  RT.s   RA.v   no stride allowed
    imm(RA)  RT.v   RA.s   stride-select allowed
    imm(RA)  RT.s   RA.s   not vectorized
    RA,RB    RT.v  {RA|RB}.v Standard Indexed
    RA,RB    RT.s  {RA|RB}.v Indexed but single LD (no VSPLAT)
    RA,RB    RT.v  {RA&RB}.s VSPLAT possible. stride selectable
    RA,RB    RT.s  {RA&RB}.s not vectorized (scalar identity)
```

Signed Effective Address computation is only relevant for Vector Indexed
Mode, when elwidth overrides are applied.  The source override applies to
RB, and before adding to RA in order to calculate the Effective Address,
if SEA is set then RB is sign-extended from elwidth bits to the full 64 bits.
For other Modes (ffirst), all EA computation with elwidth
overrides is unsigned.  RA is *never* altered (not truncated)
by element-width overrides.

Note that cache-inhibited LD/ST  when VSPLAT is activated will perform
**multiple** LD/ST operations, sequentially.  Even with scalar src
a Cache-inhibited LD will read the same memory location *multiple
times*, storing the result in successive Vector destination registers.
This because the cache-inhibit instructions are typically used to read
and write memory-mapped peripherals.  If a genuine cache-inhibited
LD-VSPLAT is required then a single *scalar* cache-inhibited LD should
be performed, followed by a VSPLAT-augmented mv, copying the one *scalar*
value into multiple register destinations.

Note also that cache-inhibited VSPLAT with Data-Dependent Fail-First is possible.
This allows for example to issue a massive batch of memory-mapped
peripheral reads, stopping at the first NULL-terminated character and
truncating VL to that point. No branch is needed to issue that large
burst of LDs, which may be valuable in Embedded scenarios.

## Vectorization of Scalar Power ISA v3.0B

Scalar Power ISA Load/Store operations may be seen from [[isa/fixedload]]
and [[isa/fixedstore]] pseudocode to be of the form:

```
    lbux RT, RA, RB
    EA <- (RA) + (RB)
    RT <- MEM(EA)
```

and for immediate variants:

```
    lb RT,D(RA)
    EA <- RA + EXTS(D)
    RT <- MEM(EA)
```

Thus in the first example, the source registers may each be independently
marked as scalar or vector, and likewise the destination; in the second
example only the one source and one dest may be marked as scalar or
vector.

Thus we can see that Vector Indexed may be covered, and, as demonstrated
with the pseudocode below, the immediate can be used to give unit
stride or element stride.  With there being no way to tell which from
the Power v3.0B Scalar opcode alone, the choice is provided instead by
the SV Context.

```
    # LD not VLD!  format - ldop RT, immed(RA)
    # op_width: lb=1, lh=2, lw=4, ld=8
    op_load(RT, RA, op_width, immed, svctx, RAupdate):
      ps = get_pred_val(FALSE, RA); # predication on src
      pd = get_pred_val(FALSE, RT); # ... AND on dest
      for (i=0, j=0, u=0; i < VL && j < VL;):
        # skip nonpredicates elements
        if (RA.isvec) while (!(ps & 1<<i)) i++;
        if (RAupdate.isvec) while (!(ps & 1<<u)) u++;
        if (RT.isvec) while (!(pd & 1<<j)) j++;
        if postinc:
            offs = 0; # added afterwards
            if RA.isvec: srcbase = ireg[RA+i]
            else         srcbase = ireg[RA]
        elif svctx.ldstmode == elementstride:
          # element stride mode
          srcbase = ireg[RA]
          offs = i * immed              # j*immed for a ST
        elif svctx.ldstmode == unitstride:
          # unit stride mode
          srcbase = ireg[RA]
          offs = immed + (i * op_width) # j*op_width for ST
        elif RA.isvec:
          # quirky Vector indexed mode but with an immediate
          srcbase = ireg[RA+i]
          offs = immed;
        else
          # standard scalar mode (but predicated)
          # no stride multiplier means VSPLAT mode
          srcbase = ireg[RA]
          offs = immed

        # compute EA
        EA = srcbase + offs
        # load from memory
        ireg[RT+j] <= MEM[EA];
        # check post-increment of EA
        if postinc: EA = srcbase + immed;
        # update RA?
        if RAupdate: ireg[RAupdate+u] = EA;
        if (!RT.isvec)
            break # destination scalar, end now
        if (RA.isvec) i++;
        if (RAupdate.isvec) u++;
        if (RT.isvec) j++;
```

Indexed LD is:

```
    # format: ldop RT, RA, RB
    function op_ldx(RT, RA, RB, RAupdate=False) # LD not VLD!
      ps = get_pred_val(FALSE, RA); # predication on src
      pd = get_pred_val(FALSE, RT); # ... AND on dest
      for (i=0, j=0, k=0, u=0; i < VL && j < VL && k < VL):
        # skip nonpredicated RA, RB and RT
        if (RA.isvec) while (!(ps & 1<<i)) i++;
        if (RAupdate.isvec) while (!(ps & 1<<u)) u++;
        if (RB.isvec) while (!(ps & 1<<k)) k++;
        if (RT.isvec) while (!(pd & 1<<j)) j++;
        if svctx.ldstmode == elementstride:
            EA = ireg[RA] + ireg[RB]*j   # register-strided
        else
            EA = ireg[RA+i] + ireg[RB+k] # indexed address
        if RAupdate: ireg[RAupdate+u] = EA
        ireg[RT+j] <= MEM[EA];
        if (!RT.isvec)
            break # destination scalar, end immediately
        if (RA.isvec) i++;
        if (RAupdate.isvec) u++;
        if (RB.isvec) k++;
        if (RT.isvec) j++;
```

Note that Element-Strided uses the Destination Step because with both
sources being Scalar as a prerequisite condition of activation of
Element-Stride Mode, the source step (being Scalar) would never advance.

Note in both cases that [[sv/svp64]] allows RA-as-a-dest in "update"
mode (`ldux`) to be effectively a *completely different* register from
RA-as-a-source.  This because there is room in svp64 to extend RA-as-src
as well as RA-as-dest, both independently as scalar or vector *and*
independently extending their range.

*Programmer's note: being able to set RA-as-a-source as separate from
RA-as-a-destination as Scalar is **extremely valuable** once it is
remembered that Simple-V element operations must be in Program Order,
especially in loops, for saving on multiple address computations. Care
does have to be taken however that RA-as-src is not overwritten by
RA-as-dest unless intentionally desired, especially in element-strided
Mode.*

## LD/ST Indexed vs Indexed REMAP

Unfortunately the word "Indexed" is used twice in completely different
contexts, potentially causing confusion.

* There has existed instructions in the Power ISA `ld RT,RA,RB` since
  its creation: these are called "LD/ST Indexed" instructions and their
  name and meaning is well-established.
* There now exists, in Simple-V, a [[sv/remap]] mode called "Indexed"
  Mode that can be applied to *any* instruction **including those
  named LD/ST Indexed**.

Whilst it may be costly in terms of register reads to allow REMAP Indexed
Mode to be applied to any Vectorized LD/ST Indexed operation such as
`sv.ld *RT,RA,*RB`, or even misleadingly labelled  as redundant, firstly
the strict application of the RISC Paradigm that Simple-V follows makes
it awkward to consider *preventing* the application of Indexed REMAP to
such operations, and secondly they are not actually the same at all.

Indexed REMAP, as applied to RB in the instruction `sv.ld *RT,RA,*RB`
effectively performs an *in-place* re-ordering of the offsets, RB.
To achieve the same effect without Indexed REMAP would require taking
a *copy* of the Vector of offsets starting at RB, manually explicitly
reordering them, and finally using the copy of re-ordered offsets in a
non-REMAP'ed `sv.ld`.  Using non-strided LD as an example, pseudocode
showing what actually occurs, where the pseudocode for `indexed_remap`
may be found in [[sv/remap]]:

```
    # sv.ld *RT,RA,*RB with Index REMAP applied to RB
    for i in 0..VL-1:
        if remap.indexed:
            rb_idx = indexed_remap(i) # remap
        else:
            rb_idx = i # use the index as-is
        EA = GPR(RA) + GPR(RB+rb_idx)
        GPR(RT+i) = MEM(EA, 8)
```

Thus it can be seen that the use of Indexed REMAP saves copying
and manual reordering of the Vector of RB offsets.

## LD/ST ffirst (Fault-First)

LD/ST ffirst treats the first LD/ST in a vector (element 0 if REMAP
is not active and predication is not applied)
as an ordinary one, with all behaviour with respect to
Interrupts Exceptions Page Faults Memory Management being identical
in every regard to Scalar v3.0 Power ISA LD/ST. However for elements
1 and above, if an exception would occur, then VL is **truncated**
to the previous element: the exception is **not** then raised because
the LD/ST that would otherwise have caused an exception is *required*
to be cancelled. Additionally an implementor may choose to truncate VL
for any arbitrary reason *except for the very first*.

ffirst LD/ST to multiple pages via a Vectorized Index base is
considered a security risk due to the abuse of probing multiple
pages in rapid succession and getting speculative feedback on which
pages would fail.  Therefore Vector Indexed LD/ST is prohibited
entirely, and the Mode bit instead used for element-strided LD/ST.
<!-- hide -->
See <https://bugs.libre-soc.org/show_bug.cgi?id=561>
<!-- show -->

```
    for(i = 0; i < VL; i++)
        reg[rt + i] = mem[reg[ra] + i * reg[rb]];
```

High security implementations where any kind of speculative probing of
memory pages is considered a risk should take advantage of the fact
that implementations may truncate VL at any point, without requiring
software to be rewritten and made non-portable. Such implementations may
choose to *always* set VL=1 which will have the effect of terminating
any speculative probing (and also adversely affect performance), but
will at least not require applications to be rewritten.

Low-performance simpler hardware implementations may also choose (always)
to also set VL=1 as the bare minimum compliant implementation of LD/ST
Fail-First. It is however critically important to remember that the first
element LD/ST **MUST** be treated as an ordinary LD/ST, i.e.  **MUST**
raise exceptions exactly like an ordinary LD/ST.

For ffirst LD/STs, VL may be truncated arbitrarily to a nonzero value
for any implementation-specific reason. For example: it is perfectly
reasonable for implementations to alter VL when ffirst LD or ST operations
are initiated on a nonaligned boundary, such that within a loop the
subsequent iteration of that loop begins the following ffirst LD/ST
operations on an aligned boundary such as the beginning of a cache line,
or beginning of a Virtual Memory page. Likewise, to reduce workloads or
balance resources.

When Predication is used, the "first" element is considered to be the first
non-predicated element rather than specifically `srcstep=0`.

Vertical-First Mode is slightly strange in that only one element at a time
is ever executed anyway.  Given that programmers may legitimately choose
to alter srcstep and dststep in non-sequential order as part of explicit
loops, it is neither possible nor safe to make speculative assumptions
about future LD/STs.  Therefore, Fail-First LD/ST in Vertical-First is
`UNDEFINED`.  This is very different from Arithmetic (Data-dependent)
FFirst where Vertical-First Mode is fully deterministic, not speculative.

## Data-Dependent Fail-First (not Fail/Fault-First)

Not to be confused with Fail/Fault First, Data-Fail-First performs an
additional check on the data, and if the test
fails then VL is truncated and further looping terminates.
This is precisely the same as Arithmetic Data-Dependent Fail-First,
the only difference being that the result comes from the LD/ST
rather than from an Arithmetic operation.

Important to note is that reduce mode is implied by Data-Dependent Fail-First.
In other words where normally if the destination is Scalar, the looping
terminates at the first load/store, Data-Dependent Fail-First *continues*
just as it does in reduce mode.

Also a crucial difference between Arithmetic and LD/ST Data-Dependent Fail-First:
except for Store-Conditional a 4-bit Condition Register Field test is created
for testing purposes
*but not stored* (thus there is no RC1 Mode as there is in Arithmetic).
The reason why a CR Field is not stored is because Load/Store, particularly
the Update instructions, is already expensive in register terms,
and adding an extra Vector write would be too costly in hardware.

*Programmer's note: Programmers
may use Data-Dependent Load with a test to truncate VL, and may then
follow up with a `sv.cmpi` or other operation. The important aspect is
that the Vector Load truncated on finding a NULL pointer, for example.*

*Programmer's note: Load-with-Update may be used to update
the register used in Effective Address computation of th
next element.  This may be used to perform single-linked-list
walking, where Data-Dependent Fail-First terminates and
truncates the Vector at the first NULL.*

**Load/Store Data-Dependent Fail-First, VLi=0**

In the case of Store operations there is a quirk when VLi (VL inclusive
is "Valid") is clear. Bear in mind the criteria is that the truncated
Vector of results, when VLi is clear, must all pass the "test", but when
VLi is set the *current failed test* is permitted to be included.  Thus,
the actual update (store) to Memory is **not permitted to take place**
should the test fail.

Additionally in any Load/Store with Update instruction,
when VLi=0 and a test fails then RA does **not** receive a
copy of the Effective Address.  Hardware implementations with Out-of-Order
Micro-Architectures should use speculative Shadow-Hold and Cancellation
(or other Transactional Rollback mechanism) when the test fails.

* **Load, VLi=0**: perform the Memory Load, do not put the result into the regfile yet (or EA into RA). Test the Loaded data: if fail do not store the Load in the register file (or EA into RA). Otherwise proceed with updating regfiles. VL is truncated to "only elements that passed the test"
* **Store, VLi=0**: even before the Store takes place, perform the test on the data to *be* stored.  If fail do not proceed with the Store at all. VL is truncated to "only elements that passed the test"

**Load/Store Data-Dependent Fail-First, VLi=1**

By contrast if VLi=1 and the test fails, the Store may proceed *and then*
looping terminates.  In this way, when Inclusive the Vector of Truncated results
contains the first-failed data (including RA on Updates)

* **Load, VLi=1**: perform the Memory Load, complete it in full (including EA into RA). Test the Loaded data: if fail then VL is truncated to "elements tested".
* **Store, VLi=0**: same as Load. Perform the Store in full and after-the-fact carry out the test of the original data requested to be stored. If fail then VL is truncated to "elements tested".

Below is an example of loading the starting addresses of Linked-List
nodes.  If VLi=1 it will load the NULL pointer into the Vector of results.
If however VLi=0 it will *exclude* the NULL pointer by truncating VL to
one Element earlier (only loading non-NULL data into registers).

*Programmer's Note: by also setting the RC1 qualifier as well as setting
VLi=1 it is possible to establish a Predicate Mask such that the first
zero in the predicate will be the NULL pointer*

```
   RT=1 # vec - deliberately overlaps by one with RA
   RA=0 # vec - first one is valid, contains ptr
   imm = 8 # offset_of(ptr->next)
   for i in range(VL):
       # this part is the Scalar Defined Word-instruction (standard scalar ld operation)
       EA = GPR(RA+i) + imm          # ptr + offset(next)
       data = MEM(EA, 8)             # 64-bit address of ptr->next
       # was a normal vector-ld up to this point. now the Data-Fail-First
       cr_test = conditions(data)
       if Rc=1 or RC1: CR.field(i) = cr_test # only store if Rc=1/RC1
       action_load = True
       if cr_test.EQ == testbit:             # check if zero
           if VLI then
              VL = i+1            # update VL, inclusive
           else
              VL = i              # update VL, exclusive current
              action_load = False # current load excluded
           stop = True            # stop looping
       if action_load:
          GPR(RT+i) = data        # happens to be read on next loop!
       if stop: break
```

**Data-Dependent Fail-First on Store-Conditional (Rc=1)**

There are very few instructions that allow Rc=1 for Load/Store:
one of those is the `stdcx.` and other Atomic Store-Conditional
instructions.  With Simple-V being a loop around Scalar instructions
strictly obeying Scalar Program Order a Horizontal-First Fail-First loop
on an Atomic Store-Conditional will always fail the second and all other
Store-Conditional instructions because
Load-Reservation and Store-Conditional are required to be executed
in pairs.

By contrast, in Vertical-First Mode it is in fact possible to issue
the pairs, and consequently allowing Vectorized Data-Dependent Fail-First is
useful.

Programmer's note: Care should be taken when VL is truncated in
Vertical-First Mode.

**Future potential**

Although Rc=1 on LD/ST is a rare occurrence at present, future versions
of Power ISA *might* conceivably have Rc=1 LD/ST Scalar instructions, and
with the SVP64 Vectorization Prefixing being itself a RISC-paradigm that
is itself fully-independent of the Scalar Suffix Defined Word-instructions, prohibiting
the possibility of Rc=1 Data-Dependent Mode on future potential LD/ST
operations is not strategically sound.

## LOAD/STORE Elwidths <a name="elwidth"></a>

Loads and Stores are almost unique in that the Power Scalar ISA
provides a width for the operation (lb, lh, lw, ld).  Only `extsb` and
others like it provide an explicit operation width.  There are therefore
*three* widths involved:

* operation width (lb=8, lh=16, lw=32, ld=64)
* src element width override (8/16/32/default)
* destination element width override (8/16/32/default)

Some care is therefore needed to express and make clear the transformations,
which are expressly in this order:

* Calculate the Effective Address from RA at full width
  but (on Indexed Load) allow srcwidth overrides on RB
* Load at the operation width (lb/lh/lw/ld) as usual
* byte-reversal as usual
* zero-extension or truncation from operation width to dest elwidth
* place result in destination at dest elwidth

In order to respect Power v3.0B Scalar behaviour the memory side
is treated effectively as completely separate and distinct from SV
augmentation.  This is primarily down to quirks surrounding LE/BE and
byte-reversal.

It is rather unfortunately possible to request an elwidth override on
the memory side which does not mesh with the overridden operation width:
these result in `UNDEFINED` behaviour.  The reason is that the effect
of attempting a 64-bit `sv.ld` operation with a source elwidth override
of 8/16/32 would result in overlapping memory requests, particularly
on unit and element strided operations.  Thus it is `UNDEFINED` when
the elwidth is smaller than the memory operation width. Examples include
`sv.lw/sw=16/els` which requests (overlapping) 4-byte memory reads offset
from each other at 2-byte intervals.  Store likewise is also `UNDEFINED`
where the dest elwidth override is less than the operation width.

Note the following regarding the pseudocode to follow:

* `scalar identity behaviour` SV Context parameter conditions turn this
  into a straight absolute fully-compliant Scalar v3.0B LD operation
* `brev` selects whether the operation is the byte-reversed variant (`ldbrx`
  rather than `ld`)
* `op_width` specifies the operation width (`lb`, `lh`, `lw`, `ld`) as
  a "normal" part of Scalar v3.0B LD
* `imm_offs` specifies the immediate offset `ld r3, imm_offs(r5)`, again
  as a "normal" part of Scalar v3.0B LD
* `svctx` specifies the SV Context and includes VL as well as
  source and destination elwidth overrides.

Below is the pseudocode for Unit-Strided LD (which includes Vector
capability). Observe in particular that RA, as the base address in both
Immediate and Indexed LD/ST, does not have element-width overriding
applied to it.

Note that predication, predication-zeroing, and other modes
have all been removed, for clarity and simplicity:

```
    # LD not VLD!
    # this covers unit stride mode and a type of vector offset
    function op_ld(RT, RA, op_width, imm_offs, svctx)
      for (int i = 0, int j = 0; i < svctx.VL && j < svctx.VL):
        if not svctx.unit/el-strided:
            # strange vector mode, compute 64 bit address which is
            # not polymorphic! elwidth hardcoded to 64 here
            srcbase = get_polymorphed_reg(RA, 64, i)
        else:
            # unit / element stride mode, compute 64 bit address
            srcbase = get_polymorphed_reg(RA, 64, 0)
            # adjust for unit/el-stride
            srcbase += .... uses op_width here

        # read the underlying memory
        memread <= MEM(srcbase + imm_offs, op_width)

        # truncate/extend to over-ridden dest width.
        memread = adjust_wid(memread, op_width, svctx.elwidth)

        # takes care of inserting memory-read (now correctly byteswapped)
        # into regfile underlying LE-defined order, into the right place
        # using Element-Packing starting at register RT, respecting destination
        # element bitwidth, and the element index (j)
        set_polymorphed_reg(RT, svctx.elwidth, j, memread)

        # increments both src and dest element indices (no predication here)
        i++;
        j++;
```

Note above that the source elwidth is *not used at all* in LD-immediate: RA
never has elwidth overrides, leaving the elwidth free for truncation/extension
of the result.

For LD/Indexed, the key is that in the calculation of the Effective Address,
RA has no elwidth override but RB does.  Pseudocode below is simplified
for clarity: predication and all modes are removed:

```
    # LD not VLD! ld*rx if brev else ld*
    function op_ld(RT, RA, RB, op_width, svctx, brev)
      for (int i = 0, int j = 0; i < svctx.VL && j < svctx.VL):
        if not svctx.el-strided:
            # RA not polymorphic! elwidth hardcoded to 64 here
            srcbase = get_polymorphed_reg(RA, 64, i)
        else:
            # element stride mode, again RA not polymorphic
            srcbase = get_polymorphed_reg(RA, 64, 0)
        # RB *is* polymorphic
        offs = get_polymorphed_reg(RB, svctx.src_elwidth, i)
        # sign-extend
        if svctx.SEA: offs = sext(offs, svctx.src_elwidth, 64)

        # takes care of (merges) processor LE/BE and ld/ldbrx
        bytereverse = brev XNOR MSR.LE

        # read the underlying memory
        memread <= MEM(srcbase + offs, op_width)

        # optionally performs byteswap at op width
        if (bytereverse):
            memread = byteswap(memread, op_width)

        # truncate/extend to over-ridden dest width.
        dest_width = op_width if RT.isvec else 64
        memread = adjust_wid(memread, op_width, dest_width)

        # takes care of inserting memory-read (now correctly byteswapped)
        # into regfile underlying LE-defined order, into the right place
        # within the NEON-like register, respecting destination element
        # bitwidth, and the element index (j)
        set_polymorphed_reg(RT, destwidth, j, memread)

        # increments both src and dest element indices (no predication here)
        i++;
        j++;
```

*Programmer's note: with no destination elwidth override the destination
width must be implicitly ascertained.  The assumption is that if the destination
is a Scalar that the entire 64-bit register must be written, thus the width is
extended to 64-bit.  If however the destination is a Vector then it is deemed
appropriate to use the LD/ST width and to perform contiguous register element
packing at that width.  The justification for doing so is that if further
sign-extension or saturation is required after a LD, these may be performed by a
follow-up instruction that uses a source elwidth override matching the exact width
of the LD operation.  Correspondingly for a ST a destination elwidth override
on a prior instruction may match the exact width of the ST instruction.*

## Remapped LD/ST

In the [[sv/remap]] page the concept of "Remapping" is described.  Whilst
it is expensive to set up (2 64-bit opcodes minimum) it provides a way to
arbitrarily perform 1D, 2D and 3D "remapping" of up to 64 elements worth
of LDs or STs.  The usual interest in such re-mapping is for example in
separating out 24-bit RGB channel data into separate contiguous registers.
NEON covers this as shown in the diagram below:

![Load/Store remap](/openpower/sv/load-store.svg)

REMAP easily covers this capability, and with dest elwidth overrides
and saturation may do so with built-in conversion that would normally
require additional width-extension, sign-extension and min/max Vectorized
instructions as post-processing stages.

Thus we do not need to provide specialist LD/ST "Structure Packed" opcodes
because the generic abstracted concept of "Remapping", when applied to
LD/ST, will give that same capability, with far more flexibility.

It is worth noting that Pack/Unpack Modes of SVSTATE, which may be
established through `svstep`, are also an easy way to perform regular
Structure Packing, at the vec2/vec3/vec4 granularity level.  Beyond that,
REMAP will need to be used.

**Parallel Reduction REMAP**

No REMAP Schedule is prohibited in SVP64 because the RISC-paradigm Prefix
is completely separate from the RISC-paradigm Scalar Defined Word-instructions.  Although
obscure there does exist the outside possibility that a potential use for
Parallel Reduction Schedules on LD/ST would find a use in Computer Science.
Readers are invited to contact the authors of this document if one is ever
found.

--------

[[!tag standards]]

\newpage{}