Save CSRs in interrupt functions

Function attribute for standard vector calling convention variant: riscv_vector_cc

Document constraint "s"

We have a list of macros, that should be extended for each new
extension. Let's rephrase the section such, that the wording
targets all extensions that are supported by compilers and
remove the incomplete list.

Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>

Standard vector calling convention variant will only enabled when function has
vector argument or returing value by default, however user may also want to
invoke function without that during a vectorized loop at some
situation, but it will cause a huge performance penalty due to vector
register store/restore.

So user can declare function with this riscv_vector_cc attribute like below,
that could enforce function will use standard vector calling convention
variant.

```c
void foo() __attribute__((riscv_vector_cc));
```

"s" can be used to create an artificial reference for linker garbage
collection, or define sections to hold symbol addresses. "S" does not
work for a preemptible symbol in GCC while "s" does.

Add missing `__riscv_q` predefine to riscv-c-api.md

The Q extension and its `__riscv_q` predefined macro are missing from the list, but compilers do actually define this alongside `__riscv_f` and `__riscv_d`.


Signed-off-by: Roland McGrath <mcgrathr@google.com>

Replace 'complier' with 'compiler'.

Fix typo: replace errof with error

This fixes a simple copy-paste mistake in the orc_b_32 emulation comment
for rv64.

Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>

This patch defines the intrinsic function API for the Zimop extension.

Co-authored-by: lyut <lyudovikos.nersisyan@ispras.ru>
Co-authored-by: ln8-8 <73429801+ln8-8@users.noreply.github.com>
Signed-off-by: ln8-8 <73429801+ln8-8@users.noreply.github.com>

Fix the __riscv_clmul* intrinsics.

Locality for prefetch built-in proposal

These should be  __riscv_clmul*(uintxlen_t rs1, uintxlen_t rs2). There
are only 1 argument.

These should be __riscv_sha512sum0r and __riscv_sha512sum1r. There
are only 2 instructions not 4.

`__riscv_64e` is defined analogously to `__riscv_32e`.

Changing the meaning of `__riscv_32e` should not cause any problems
since if we are just testing for existence of the `E` extension,
there is `__riscv_e` preprocessor macro.

Add zicsr and zifencei to extension test macros

Adds the __riscv_zicsr and __riscv_zifencei extension test macros to the
existing table.

Signed-off-by: Daniel Mangum <georgedanielmangum@gmail.com>

This is a proposal for scalar crypto intrinsics for the Zk* extensions.

I've removed the recommendation to use 'long' for XLEN. There's nothing
guaranteeing that 'long' is XLEN. For example, if we were to support
ILP32 on RV64, 'long' would become 32 bits while XLEN would still be 64 bits.

Other design decisions are spelled out in the document.

Some text has been copied from #23.

One open question is whether we should emulate 64-bit intrinsics on RV32
where it is feasible.  This could be useful for the sha512 intrinsics
and some others.

Adding target attribute for function.

- Because we don't has any spec to define how to resolve, also it's
  confusing among target, target_clones and target_version.

- Also same behavior for target attribute specfied twice.

- Add link for mapping symobl.
- File scope -> translation unit scope

- Drop some sentence not belong this spec

Document operand modifier

Zawrs: Define extension availability macro

- Drop `disable  a set of features or extensions`, we don't have syntax
  to drop now.

- Tweak description for `arch=`: describe two different cases of `arch=`
  clearly.

- Drop last sentence, that will happened on multi-version funciton but
  not happen in `target` attribute.

command line option.

Co-authored-by: Luís Marques <luis@luismarques.eu>
Signed-off-by: Kito Cheng <kito.cheng@gmail.com>

- Fix the syntax to allow more than two ATTR.
- Speicfy the interactions among `arch`, `tune`, and `cpu`.
- Describe the behavior if same ATTR appear more than once.
- Mention compiler should always generate the function even the
  attribute compiler will generate same code without attribute.