T0b1/wasmtime
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/master' into no_std

Browse Source
This commit is contained in:
Dan Gohman
2018-03-30 15:14:30 -07:00
parent fc7b0a7e51 b523b69c16
commit 07693048f0
260 changed files with 4509 additions and 5074 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
.travis.yml
View File

@@ -2,6 +2,10 @@ language: rust
rust: rust:
- stable - stable
- beta - beta
- nightly
matrix:
allow_failures:
- rust: nightly
dist: trusty dist: trusty
sudo: false sudo: false
addons: addons:

6
README.rst
View File

@@ -3,7 +3,7 @@ Cretonne Code Generator
======================= =======================
Cretonne is a low-level retargetable code generator. It translates a `target-independent Cretonne is a low-level retargetable code generator. It translates a `target-independent
intermediate language <https://cretonne.readthedocs.io/en/latest/langref.html>`_ into executable intermediate representation <https://cretonne.readthedocs.io/en/latest/langref.html>`_ into executable
machine code. machine code.
*This is a work in progress that is not yet functional.* *This is a work in progress that is not yet functional.*
@@ -16,6 +16,10 @@ machine code.
:target: https://travis-ci.org/Cretonne/cretonne :target: https://travis-ci.org/Cretonne/cretonne
:alt: Build Status :alt: Build Status
.. image:: https://badges.gitter.im/Cretonne/cretonne.png
:target: https://gitter.im/Cretonne/Lobby/~chat
:alt: Gitter chat
For more information, see `the documentation For more information, see `the documentation
<https://cretonne.readthedocs.io/en/latest/?badge=latest>`_. <https://cretonne.readthedocs.io/en/latest/?badge=latest>`_.

10
check-clippy.sh Executable file
View File

@@ -0,0 +1,10 @@
#!/bin/bash
set -euo pipefail
# Usage: check-clippy.sh [--install]
if cargo install --list | tee /dev/null | grep -q "^clippy v0"; then
exit 0
else
exit 1
fi

4
check-rustfmt.sh
View File

@@ -22,11 +22,11 @@ set -euo pipefail
# operation, however that doesn't appear to be possible through "cargo fmt"). # operation, however that doesn't appear to be possible through "cargo fmt").
VERS="0.9.0" VERS="0.9.0"
if cargo install --list | grep -q "^rustfmt v$VERS"; then if cargo install --list | tee /dev/null | grep -q "^rustfmt v$VERS"; then
exit 0 exit 0
fi fi
if [ "$1" != "--install" ]; then if [[ ${1:-""} != "--install" ]]; then
echo "********************************************************************" echo "********************************************************************"
echo "* Please install rustfmt v$VERS to verify formatting. *" echo "* Please install rustfmt v$VERS to verify formatting. *"
echo "* If a newer version of rustfmt is available, update this script. *" echo "* If a newer version of rustfmt is available, update this script. *"

7
clippy-all.sh Executable file
View File

@@ -0,0 +1,7 @@
#!/bin/bash
set -euo pipefail
# Check all sources with clippy.
# In the cton-util crate (root dir) clippy will only work with nightly cargo -
# there is a bug where it will reject the commands passed to it by cargo 0.25.0
cargo +nightly clippy --all

18
cranelift/Cargo.toml
View File

@@ -1,7 +1,7 @@
[package] [package]
name = "cretonne-tools" name = "cretonne-tools"
authors = ["The Cretonne Project Developers"] authors = ["The Cretonne Project Developers"]
version = "0.3.4" version = "0.4.1"
description = "Binaries for testing the Cretonne library" description = "Binaries for testing the Cretonne library"
license = "Apache-2.0" license = "Apache-2.0"
documentation = "https://cretonne.readthedocs.io/" documentation = "https://cretonne.readthedocs.io/"
@@ -13,18 +13,18 @@ name = "cton-util"
path = "src/cton-util.rs" path = "src/cton-util.rs"
[dependencies] [dependencies]
cretonne = { path = "lib/cretonne", version = "0.3.4" } cretonne = { path = "lib/cretonne", version = "0.4.1" }
cretonne-reader = { path = "lib/reader", version = "0.3.4" } cretonne-reader = { path = "lib/reader", version = "0.4.1" }
cretonne-frontend = { path = "lib/frontend", version = "0.3.4" } cretonne-frontend = { path = "lib/frontend", version = "0.4.1" }
cretonne-wasm = { path = "lib/wasm", version = "0.3.4" } cretonne-wasm = { path = "lib/wasm", version = "0.4.1" }
cretonne-native = { path = "lib/native", version = "0.3.4" } cretonne-native = { path = "lib/native", version = "0.4.1" }
filecheck = { path = "lib/filecheck" } cretonne-filetests = { path = "lib/filetests", version = "0.4.1" }
filecheck = "0.2.1"
docopt = "0.8.0" docopt = "0.8.0"
serde = "1.0.8" serde = "1.0.8"
serde_derive = "1.0.8" serde_derive = "1.0.8"
num_cpus = "1.5.1"
tempdir = "0.3.5" tempdir = "0.3.5"
term = "0.5" term = "0.5.1"
[workspace] [workspace]

1
cranelift/clippy.toml Normal file
View File

@@ -0,0 +1 @@
doc-valid-idents = [ "WebAssembly", "NaN", "SetCC" ]

2
cranelift/docs/callex.cton
View File

@@ -1,6 +1,6 @@
test verifier test verifier
function %gcd(i32 uext, i32 uext) -> i32 uext native { function %gcd(i32 uext, i32 uext) -> i32 uext system_v {
fn1 = function %divmod(i32 uext, i32 uext) -> i32 uext, i32 uext fn1 = function %divmod(i32 uext, i32 uext) -> i32 uext, i32 uext
ebb1(v1: i32, v2: i32): ebb1(v1: i32, v2: i32):

20
cranelift/docs/compare-llvm.rst
View File

@@ -16,8 +16,8 @@ highlighting some of the differences and similarities. Both projects:
- Use an ISA-agnostic input language in order to mostly abstract away the - Use an ISA-agnostic input language in order to mostly abstract away the
differences between target instruction set architectures. differences between target instruction set architectures.
- Depend extensively on SSA form. - Depend extensively on SSA form.
- Have both textual and in-memory forms of their primary intermediate language. - Have both textual and in-memory forms of their primary intermediate
(LLVM also has a binary bitcode format; Cretonne doesn't.) representation. (LLVM also has a binary bitcode format; Cretonne doesn't.)
- Can target multiple ISAs. - Can target multiple ISAs.
- Can cross-compile by default without rebuilding the code generator. - Can cross-compile by default without rebuilding the code generator.
@@ -41,8 +41,8 @@ LLVM uses multiple intermediate representations as it translates a program to
binary machine code: binary machine code:
`LLVM IR <https://llvm.org/docs/LangRef.html>`_ `LLVM IR <https://llvm.org/docs/LangRef.html>`_
This is the primary intermediate language which has textual, binary, and This is the primary intermediate representation which has textual, binary, and
in-memory representations. It serves two main purposes: in-memory forms. It serves two main purposes:
- An ISA-agnostic, stable(ish) input language that front ends can generate - An ISA-agnostic, stable(ish) input language that front ends can generate
easily. easily.
@@ -89,9 +89,9 @@ representation. Some target ISAs have a fast instruction selector that can
translate simple code directly to MachineInstrs, bypassing SelectionDAG when translate simple code directly to MachineInstrs, bypassing SelectionDAG when
possible. possible.
:doc:`Cretonne <langref>` uses a single intermediate language to cover these :doc:`Cretonne <langref>` uses a single intermediate representation to cover
levels of abstraction. This is possible in part because of Cretonne's smaller these levels of abstraction. This is possible in part because of Cretonne's
scope. smaller scope.
- Cretonne does not provide assemblers and disassemblers, so it is not - Cretonne does not provide assemblers and disassemblers, so it is not
necessary to be able to represent every weird instruction in an ISA. Only necessary to be able to represent every weird instruction in an ISA. Only
@@ -102,7 +102,7 @@ scope.
- SSA form is preserved throughout. After register allocation, each SSA value - SSA form is preserved throughout. After register allocation, each SSA value
is annotated with an assigned ISA register or stack slot. is annotated with an assigned ISA register or stack slot.
The Cretonne intermediate language is similar to LLVM IR, but at a slightly The Cretonne intermediate representation is similar to LLVM IR, but at a slightly
lower level of abstraction. lower level of abstraction.
Program structure Program structure
@@ -112,12 +112,12 @@ In LLVM IR, the largest representable unit is the *module* which corresponds
more or less to a C translation unit. It is a collection of functions and more or less to a C translation unit. It is a collection of functions and
global variables that may contain references to external symbols too. global variables that may contain references to external symbols too.
In Cretonne IL, the largest representable unit is the *function*. This is so In Cretonne IR, the largest representable unit is the *function*. This is so
that functions can easily be compiled in parallel without worrying about that functions can easily be compiled in parallel without worrying about
references to shared data structures. Cretonne does not have any references to shared data structures. Cretonne does not have any
inter-procedural optimizations like inlining. inter-procedural optimizations like inlining.
An LLVM IR function is a graph of *basic blocks*. A Cretonne IL function is a An LLVM IR function is a graph of *basic blocks*. A Cretonne IR function is a
graph of *extended basic blocks* that may contain internal branch instructions. graph of *extended basic blocks* that may contain internal branch instructions.
The main difference is that an LLVM conditional branch instruction has two The main difference is that an LLVM conditional branch instruction has two
target basic blocks---a true and a false edge. A Cretonne branch instruction target basic blocks---a true and a false edge. A Cretonne branch instruction

16
cranelift/docs/cton_domain.py
View File

@@ -1,6 +1,6 @@
# -*- coding: utf-8 -*- # -*- coding: utf-8 -*-
# #
# Sphinx domain for documenting compiler intermediate languages. # Sphinx domain for documenting compiler intermediate representations.
# #
# This defines a 'cton' Sphinx domain with the following directives and roles: # This defines a 'cton' Sphinx domain with the following directives and roles:
# #
@@ -29,10 +29,10 @@ import sphinx.ext.autodoc
class CtonObject(ObjectDescription): class CtonObject(ObjectDescription):
""" """
Any kind of Cretonne IL object. Any kind of Cretonne IR object.
This is a shared base class for the different kinds of indexable objects This is a shared base class for the different kinds of indexable objects
in the Cretonne IL reference. in the Cretonne IR reference.
""" """
option_spec = { option_spec = {
'noindex': directives.flag, 'noindex': directives.flag,
@@ -98,7 +98,7 @@ def parse_type(name, signode):
class CtonType(CtonObject): class CtonType(CtonObject):
"""A Cretonne IL type description.""" """A Cretonne IR type description."""
def handle_signature(self, sig, signode): def handle_signature(self, sig, signode):
""" """
@@ -112,7 +112,7 @@ class CtonType(CtonObject):
return name return name
def get_index_text(self, name): def get_index_text(self, name):
return name + ' (IL type)' return name + ' (IR type)'
sep_equal = re.compile('\s*=\s*') sep_equal = re.compile('\s*=\s*')
@@ -127,7 +127,7 @@ def parse_params(s, signode):
class CtonInst(CtonObject): class CtonInst(CtonObject):
"""A Cretonne IL instruction.""" """A Cretonne IR instruction."""
doc_field_types = [ doc_field_types = [
TypedField('argument', label=l_('Arguments'), TypedField('argument', label=l_('Arguments'),
@@ -176,11 +176,11 @@ class CtonInst(CtonObject):
class CtonInstGroup(CtonObject): class CtonInstGroup(CtonObject):
"""A Cretonne IL instruction group.""" """A Cretonne IR instruction group."""
class CretonneDomain(Domain): class CretonneDomain(Domain):
"""Cretonne domain for intermediate language objects.""" """Cretonne domain for IR objects."""
name = 'cton' name = 'cton'
label = 'Cretonne' label = 'Cretonne'

2
cranelift/docs/example.cton
View File

@@ -1,6 +1,6 @@
test verifier test verifier
function %average(i32, i32) -> f32 native { function %average(i32, i32) -> f32 system_v {
ss1 = explicit_slot 8 ; Stack slot for ``sum``. ss1 = explicit_slot 8 ; Stack slot for ``sum``.
ebb1(v1: i32, v2: i32): ebb1(v1: i32, v2: i32):

50
cranelift/docs/langref.rst
View File

@@ -5,19 +5,19 @@ Cretonne Language Reference
.. default-domain:: cton .. default-domain:: cton
.. highlight:: cton .. highlight:: cton
The Cretonne intermediate language (:term:`IL`) has two equivalent The Cretonne intermediate representation (:term:`IR`) has two primary forms:
representations: an *in-memory data structure* that the code generator library an *in-memory data structure* that the code generator library is using, and a
is using, and a *text format* which is used for test cases and debug output. *text format* which is used for test cases and debug output.
Files containing Cretonne textual IL have the ``.cton`` filename extension. Files containing Cretonne textual IR have the ``.cton`` filename extension.
This reference uses the text format to describe IL semantics but glosses over This reference uses the text format to describe IR semantics but glosses over
the finer details of the lexical and syntactic structure of the format. the finer details of the lexical and syntactic structure of the format.
Overall structure Overall structure
================= =================
Cretonne compiles functions independently. A ``.cton`` IL file may contain Cretonne compiles functions independently. A ``.cton`` IR file may contain
multiple functions, and the programmatic API can create multiple function multiple functions, and the programmatic API can create multiple function
handles at the same time, but the functions don't share any data or reference handles at the same time, but the functions don't share any data or reference
each other directly. each other directly.
@@ -27,7 +27,7 @@ This is a simple C function that computes the average of an array of floats:
.. literalinclude:: example.c .. literalinclude:: example.c
:language: c :language: c
Here is the same function compiled into Cretonne IL: Here is the same function compiled into Cretonne IR:
.. literalinclude:: example.cton .. literalinclude:: example.cton
:language: cton :language: cton
@@ -77,7 +77,7 @@ variable value for the next iteration.
The `cton_frontend` crate contains utilities for translating from programs The `cton_frontend` crate contains utilities for translating from programs
containing multiple assignments to the same variables into SSA form for containing multiple assignments to the same variables into SSA form for
Cretonne :term:`IL`. Cretonne :term:`IR`.
Such variables can also be presented to Cretonne as :term:`stack slot`\s. Such variables can also be presented to Cretonne as :term:`stack slot`\s.
Stack slots are accessed with the :inst:`stack_store` and :inst:`stack_load` Stack slots are accessed with the :inst:`stack_store` and :inst:`stack_load`
@@ -303,7 +303,7 @@ indicate the different kinds of immediate operands on an instruction.
A floating point condition code. See the :inst:`fcmp` instruction for details. A floating point condition code. See the :inst:`fcmp` instruction for details.
The two IEEE floating point immediate types :type:`ieee32` and :type:`ieee64` The two IEEE floating point immediate types :type:`ieee32` and :type:`ieee64`
are displayed as hexadecimal floating point literals in the textual :term:`IL` are displayed as hexadecimal floating point literals in the textual :term:`IR`
format. Decimal floating point literals are not allowed because some computer format. Decimal floating point literals are not allowed because some computer
systems can round differently when converting to binary. The hexadecimal systems can round differently when converting to binary. The hexadecimal
floating point format is mostly the same as the one used by C99, but extended floating point format is mostly the same as the one used by C99, but extended
@@ -400,11 +400,11 @@ convention:
param : type [paramext] [paramspecial] param : type [paramext] [paramspecial]
paramext : "uext" | "sext" paramext : "uext" | "sext"
paramspecial : "sret" | "link" | "fp" | "csr" | "vmctx" paramspecial : "sret" | "link" | "fp" | "csr" | "vmctx"
callconv : "native" | "spiderwasm" callconv : "system_v" | "spiderwasm"
Parameters and return values have flags whose meaning is mostly target Parameters and return values have flags whose meaning is mostly target
dependent. They make it possible to call native functions on the target dependent. These flags support interfacing with code produced by other
platform. When calling other Cretonne functions, the flags are not necessary. compilers.
Functions that are called directly must be declared in the :term:`function Functions that are called directly must be declared in the :term:`function
preamble`: preamble`:
@@ -563,7 +563,7 @@ runtime data structures.
alignment for storing a pointer. alignment for storing a pointer.
Chains of ``deref`` global variables are possible, but cycles are not Chains of ``deref`` global variables are possible, but cycles are not
allowed. They will be caught by the IL verifier. allowed. They will be caught by the IR verifier.
:arg BaseGV: Global variable containing the base pointer. :arg BaseGV: Global variable containing the base pointer.
:arg Offset: Byte offset from the loaded base pointer to the global :arg Offset: Byte offset from the loaded base pointer to the global
@@ -654,6 +654,11 @@ trap when accessed.
address space reserved for the heap, not including the guard pages. address space reserved for the heap, not including the guard pages.
:arg GuardBytes: Size of the guard pages in bytes. :arg GuardBytes: Size of the guard pages in bytes.
When the base is a global variable, it must be :term:`accessible` and naturally
aligned for a pointer value.
The ``reserved_reg`` option is not yet implemented.
Dynamic heaps Dynamic heaps
~~~~~~~~~~~~~ ~~~~~~~~~~~~~
@@ -672,6 +677,11 @@ is resized. The bound of a dynamic heap is stored in a global variable.
:arg BoundGV: Global variable containing the current heap bound in bytes. :arg BoundGV: Global variable containing the current heap bound in bytes.
:arg GuardBytes: Size of the guard pages in bytes. :arg GuardBytes: Size of the guard pages in bytes.
When the base is a global variable, it must be :term:`accessible` and naturally
aligned for a pointer value.
The ``reserved_reg`` option is not yet implemented.
Heap examples Heap examples
~~~~~~~~~~~~~ ~~~~~~~~~~~~~
@@ -1144,19 +1154,11 @@ Glossary
The extended basic blocks which contain all the executable code in a The extended basic blocks which contain all the executable code in a
function. The function body follows the function preamble. function. The function body follows the function preamble.
intermediate language
IL
The language used to describe functions to Cretonne. This reference
describes the syntax and semantics of the Cretonne IL. The IL has two
forms: Textual and an in-memory intermediate representation
(:term:`IR`).
intermediate representation intermediate representation
IR IR
The in-memory representation of :term:`IL`. The data structures The language used to describe functions to Cretonne. This reference
Cretonne uses to represent a program internally are called the describes the syntax and semantics of Cretonne IR. The IR has two
intermediate representation. Cretonne's IR can be converted to text forms: Textual, and an in-memory data structure.
losslessly.
stack slot stack slot
A fixed size memory allocation in the current function's activation A fixed size memory allocation in the current function's activation

46
cranelift/docs/testing.rst
View File

@@ -89,7 +89,7 @@ easier to provide substantial input functions for the compiler tests.
File tests are :file:`*.cton` files in the :file:`filetests/` directory File tests are :file:`*.cton` files in the :file:`filetests/` directory
hierarchy. Each file has a header describing what to test followed by a number hierarchy. Each file has a header describing what to test followed by a number
of input functions in the :doc:`Cretonne textual intermediate language of input functions in the :doc:`Cretonne textual intermediate representation
<langref>`: <langref>`:
.. productionlist:: .. productionlist::
@@ -136,13 +136,15 @@ This example will run the legalizer test twice. Both runs will have
``opt_level=best``, but they will have different ``is_64bit`` settings. The 32-bit ``opt_level=best``, but they will have different ``is_64bit`` settings. The 32-bit
run will also have the RISC-V specific flag ``supports_m`` disabled. run will also have the RISC-V specific flag ``supports_m`` disabled.
The filetests are run automatically as part of `cargo test`, and they can
also be run manually with the `cton-util test` command.
Filecheck Filecheck
--------- ---------
Many of the test commands described below use *filecheck* to verify their Many of the test commands described below use *filecheck* to verify their
output. Filecheck is a Rust implementation of the LLVM tool of the same name. output. Filecheck is a Rust implementation of the LLVM tool of the same name.
See the :file:`lib/filecheck` `documentation <https://docs.rs/filecheck/>`_ for See the `documentation <https://docs.rs/filecheck/>`_ for details of its syntax.
details of its syntax.
Comments in :file:`.cton` files are associated with the entity they follow. Comments in :file:`.cton` files are associated with the entity they follow.
This typically means an instruction or the whole function. Those tests that This typically means an instruction or the whole function. Those tests that
@@ -164,7 +166,7 @@ Cretonne's tests don't need this.
---------- ----------
This is one of the simplest file tests, used for testing the conversion to and This is one of the simplest file tests, used for testing the conversion to and
from textual IL. The ``test cat`` command simply parses each function and from textual IR. The ``test cat`` command simply parses each function and
converts it back to text again. The text of each function is then matched converts it back to text again. The text of each function is then matched
against the associated filecheck directives. against the associated filecheck directives.
@@ -186,7 +188,7 @@ Example::
`test verifier` `test verifier`
--------------- ---------------
Run each function through the IL verifier and check that it produces the Run each function through the IR verifier and check that it produces the
expected error messages. expected error messages.
Expected error messages are indicated with an ``error:`` directive *on the Expected error messages are indicated with an ``error:`` directive *on the
@@ -324,6 +326,38 @@ Test the simple GVN pass.
The simple GVN pass is run on each function, and then results are run The simple GVN pass is run on each function, and then results are run
through filecheck. through filecheck.
`test licm`
-----------------
Test the LICM pass.
The LICM pass is run on each function, and then results are run
through filecheck.
`test dce`
-----------------
Test the DCE pass.
The DCE pass is run on each function, and then results are run
through filecheck.
`test preopt`
-----------------
Test the preopt pass.
The preopt pass is run on each function, and then results are run
through filecheck.
`test postopt`
-----------------
Test the postopt pass.
The postopt pass is run on each function, and then results are run
through filecheck.
`test compile` `test compile`
-------------- --------------
@@ -333,4 +367,4 @@ Each function is passed through the full ``Context::compile()`` function
which is normally used to compile code. This type of test often depends which is normally used to compile code. This type of test often depends
on assertions or verifier errors, but it is also possible to use on assertions or verifier errors, but it is also possible to use
filecheck directives which will be matched against the final form of the filecheck directives which will be matched against the final form of the
Cretonne IL right before binary machine code emission. Cretonne IR right before binary machine code emission.

46
cranelift/filetests/dce/basic.cton Normal file
View File

@@ -0,0 +1,46 @@
test dce
function %simple() -> i32 {
ebb0:
v2 = iconst.i32 2
v3 = iconst.i32 3
return v3
}
; sameln: function %simple
; nextln: ebb0:
; nextln: v3 = iconst.i32 3
; nextln: return v3
; nextln: }
function %some_branching(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
v3 = iconst.i32 70
v4 = iconst.i32 71
v5 = iconst.i32 72
v8 = iconst.i32 73
brz v0, ebb1
jump ebb2(v8)
ebb1:
v2 = iadd v0, v3
return v0
ebb2(v9: i32):
v6 = iadd v1, v4
v7 = iadd v6, v9
return v7
}
; sameln: function %some_branching
; nextln: ebb0(v0: i32, v1: i32):
; nextln: v4 = iconst.i32 71
; nextln: v8 = iconst.i32 73
; nextln: brz v0, ebb1
; nextln: jump ebb2(v8)
; nextln:
; nextln: ebb1:
; nextln: return v0
; nextln:
; nextln: ebb2(v9: i32):
; nextln: v6 = iadd.i32 v1, v4
; nextln: v7 = iadd v6, v9
; nextln: return v7

2
cranelift/filetests/domtree/loops.cton
View File

@@ -59,7 +59,7 @@ function %test(i32) {
; nextln: ebb5: ; nextln: ebb5:
; nextln: } ; nextln: }
function %loop2(i32) native { function %loop2(i32) system_v {
ebb0(v0: i32): ebb0(v0: i32):
brz v0, ebb1 ; dominates: ebb1 ebb3 ebb4 ebb5 brz v0, ebb1 ; dominates: ebb1 ebb3 ebb4 ebb5
jump ebb2 ; dominates: ebb2 jump ebb2 ; dominates: ebb2

2
cranelift/filetests/domtree/loops2.cton
View File

@@ -43,7 +43,7 @@ function %loop1(i32) {
; nextln: ebb9: ; nextln: ebb9:
; nextln: } ; nextln: }
function %loop2(i32) native { function %loop2(i32) system_v {
ebb0(v0: i32): ebb0(v0: i32):
brz v0, ebb1 ; dominates: ebb1 ebb3 ebb4 ebb5 brz v0, ebb1 ; dominates: ebb1 ebb3 ebb4 ebb5
jump ebb2 ; dominates: ebb2 jump ebb2 ; dominates: ebb2

2
cranelift/filetests/isa/intel/abi-bool.cton
View File

@@ -2,7 +2,7 @@ test compile
set is_64bit=1 set is_64bit=1
isa intel haswell isa intel haswell
function %foo(i64, i64, i64, i32) -> b1 native { function %foo(i64, i64, i64, i32) -> b1 system_v {
ebb3(v0: i64, v1: i64, v2: i64, v3: i32): ebb3(v0: i64, v1: i64, v2: i64, v3: i32):
v5 = icmp ne v2, v2 v5 = icmp ne v2, v2
v8 = iconst.i64 0 v8 = iconst.i64 0

12
cranelift/filetests/isa/intel/abi32.cton
View File

@@ -5,14 +5,14 @@ isa intel
; regex: V=v\d+ ; regex: V=v\d+
function %f() { function %f() {
sig0 = (i32) -> i32 native sig0 = (i32) -> i32 system_v
; check: sig0 = (i32 [0]) -> i32 [%rax] native ; check: sig0 = (i32 [0]) -> i32 [%rax] system_v
sig1 = (i64) -> b1 native sig1 = (i64) -> b1 system_v
; check: sig1 = (i32 [0], i32 [4]) -> b1 [%rax] native ; check: sig1 = (i32 [0], i32 [4]) -> b1 [%rax] system_v
sig2 = (f32, i64) -> f64 native sig2 = (f32, i64) -> f64 system_v
; check: sig2 = (f32 [0], i32 [4], i32 [8]) -> f64 [%xmm0] native ; check: sig2 = (f32 [0], i32 [4], i32 [8]) -> f64 [%xmm0] system_v
ebb0: ebb0:
return return

12
cranelift/filetests/isa/intel/abi64.cton
View File

@@ -6,14 +6,14 @@ isa intel
; regex: V=v\d+ ; regex: V=v\d+
function %f() { function %f() {
sig0 = (i32) -> i32 native sig0 = (i32) -> i32 system_v
; check: sig0 = (i32 [%rdi]) -> i32 [%rax] native ; check: sig0 = (i32 [%rdi]) -> i32 [%rax] system_v
sig1 = (i64) -> b1 native sig1 = (i64) -> b1 system_v
; check: sig1 = (i64 [%rdi]) -> b1 [%rax] native ; check: sig1 = (i64 [%rdi]) -> b1 [%rax] system_v
sig2 = (f32, i64) -> f64 native sig2 = (f32, i64) -> f64 system_v
; check: sig2 = (f32 [%xmm0], i64 [%rdi]) -> f64 [%xmm0] native ; check: sig2 = (f32 [%xmm0], i64 [%rdi]) -> f64 [%xmm0] system_v
ebb0: ebb0:
return return

30
cranelift/filetests/isa/intel/baseline_clz_ctz_popcount.cton
View File

@@ -47,28 +47,23 @@ ebb1(v20: i32):
function %i64_popcount(i64) -> i64 { function %i64_popcount(i64) -> i64 {
ebb0(v30: i64): ebb0(v30: i64):
v31 = popcnt v30; v31 = popcnt v30;
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: iconst.i64 ; check: iconst.i64
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: iadd ; check: iadd
; check: iconst.i64 ; check: iconst.i64
; check: band ; check: band
; check: iconst.i64 ; check: iconst.i64
; check: imul ; check: imul
; check: iconst.i32 ; check: ushr_imm
; check: ushr
return v31; return v31;
} }
@@ -78,27 +73,22 @@ ebb0(v30: i64):
function %i32_popcount(i32) -> i32 { function %i32_popcount(i32) -> i32 {
ebb0(v40: i32): ebb0(v40: i32):
v41 = popcnt v40; v41 = popcnt v40;
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: iconst.i32 ; check: iconst.i32
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: band ; check: band
; check: isub ; check: isub
; check: iconst.i32 ; check: ushr_imm
; check: ushr
; check: iadd ; check: iadd
; check: iconst.i32 ; check: iconst.i32
; check: band ; check: band
; check: iconst.i32 ; check: iconst.i32
; check: imul ; check: imul
; check: iconst.i32 ; check: ushr_imm
; check: ushr
return v41; return v41;
} }

24
cranelift/filetests/isa/intel/baseline_clz_ctz_popcount_encoding.cton
View File

@@ -15,56 +15,56 @@ ebb0:
[-,%r11] v10 = iconst.i64 0x1234 [-,%r11] v10 = iconst.i64 0x1234
; asm: bsfq %r11, %rcx ; asm: bsfq %r11, %rcx
[-,%rcx,%eflags] v11, v12 = x86_bsf v10 ; bin: 49 0f bc cb [-,%rcx,%rflags] v11, v12 = x86_bsf v10 ; bin: 49 0f bc cb
[-,%rdx] v14 = iconst.i64 0x5678 [-,%rdx] v14 = iconst.i64 0x5678
; asm: bsfq %rdx, %r12 ; asm: bsfq %rdx, %r12
[-,%r12,%eflags] v15, v16 = x86_bsf v14 ; bin: 4c 0f bc e2 [-,%r12,%rflags] v15, v16 = x86_bsf v14 ; bin: 4c 0f bc e2
; asm: bsfq %rdx, %rdi ; asm: bsfq %rdx, %rdi
[-,%rdi,%eflags] v17, v18 = x86_bsf v14 ; bin: 48 0f bc fa [-,%rdi,%rflags] v17, v18 = x86_bsf v14 ; bin: 48 0f bc fa
; 32-bit wide bsf ; 32-bit wide bsf
[-,%r11] v20 = iconst.i32 0x1234 [-,%r11] v20 = iconst.i32 0x1234
; asm: bsfl %r11d, %ecx ; asm: bsfl %r11d, %ecx
[-,%rcx,%eflags] v21, v22 = x86_bsf v20 ; bin: 41 0f bc cb [-,%rcx,%rflags] v21, v22 = x86_bsf v20 ; bin: 41 0f bc cb
[-,%rdx] v24 = iconst.i32 0x5678 [-,%rdx] v24 = iconst.i32 0x5678
; asm: bsfl %edx, %r12d ; asm: bsfl %edx, %r12d
[-,%r12,%eflags] v25, v26 = x86_bsf v24 ; bin: 44 0f bc e2 [-,%r12,%rflags] v25, v26 = x86_bsf v24 ; bin: 44 0f bc e2
; asm: bsfl %edx, %esi ; asm: bsfl %edx, %esi
[-,%rsi,%eflags] v27, v28 = x86_bsf v24 ; bin: 0f bc f2 [-,%rsi,%rflags] v27, v28 = x86_bsf v24 ; bin: 0f bc f2
; 64-bit wide bsr ; 64-bit wide bsr
[-,%r11] v30 = iconst.i64 0x1234 [-,%r11] v30 = iconst.i64 0x1234
; asm: bsrq %r11, %rcx ; asm: bsrq %r11, %rcx
[-,%rcx,%eflags] v31, v32 = x86_bsr v30 ; bin: 49 0f bd cb [-,%rcx,%rflags] v31, v32 = x86_bsr v30 ; bin: 49 0f bd cb
[-,%rdx] v34 = iconst.i64 0x5678 [-,%rdx] v34 = iconst.i64 0x5678
; asm: bsrq %rdx, %r12 ; asm: bsrq %rdx, %r12
[-,%r12,%eflags] v35, v36 = x86_bsr v34 ; bin: 4c 0f bd e2 [-,%r12,%rflags] v35, v36 = x86_bsr v34 ; bin: 4c 0f bd e2
; asm: bsrq %rdx, %rdi ; asm: bsrq %rdx, %rdi
[-,%rdi,%eflags] v37, v38 = x86_bsr v34 ; bin: 48 0f bd fa [-,%rdi,%rflags] v37, v38 = x86_bsr v34 ; bin: 48 0f bd fa
; 32-bit wide bsr ; 32-bit wide bsr
[-,%r11] v40 = iconst.i32 0x1234 [-,%r11] v40 = iconst.i32 0x1234
; asm: bsrl %r11d, %ecx ; asm: bsrl %r11d, %ecx
[-,%rcx,%eflags] v41, v42 = x86_bsr v40 ; bin: 41 0f bd cb [-,%rcx,%rflags] v41, v42 = x86_bsr v40 ; bin: 41 0f bd cb
[-,%rdx] v44 = iconst.i32 0x5678 [-,%rdx] v44 = iconst.i32 0x5678
; asm: bsrl %edx, %r12d ; asm: bsrl %edx, %r12d
[-,%r12,%eflags] v45, v46 = x86_bsr v44 ; bin: 44 0f bd e2 [-,%r12,%rflags] v45, v46 = x86_bsr v44 ; bin: 44 0f bd e2
; asm: bsrl %edx, %esi ; asm: bsrl %edx, %esi
[-,%rsi,%eflags] v47, v48 = x86_bsr v44 ; bin: 0f bd f2 [-,%rsi,%rflags] v47, v48 = x86_bsr v44 ; bin: 0f bd f2
; 64-bit wide cmov ; 64-bit wide cmov

174
cranelift/filetests/isa/intel/binary32-float.cton
View File

@@ -147,48 +147,48 @@ ebb0:
; Load/Store ; Load/Store
; asm: movd (%ecx), %xmm5 ; asm: movss (%ecx), %xmm5
[-,%xmm5] v100 = load.f32 v0 ; bin: 66 0f 6e 29 [-,%xmm5] v100 = load.f32 v0 ; bin: heap_oob f3 0f 10 29
; asm: movd (%esi), %xmm2 ; asm: movss (%esi), %xmm2
[-,%xmm2] v101 = load.f32 v1 ; bin: 66 0f 6e 16 [-,%xmm2] v101 = load.f32 v1 ; bin: heap_oob f3 0f 10 16
; asm: movd 50(%ecx), %xmm5 ; asm: movss 50(%ecx), %xmm5
[-,%xmm5] v110 = load.f32 v0+50 ; bin: 66 0f 6e 69 32 [-,%xmm5] v110 = load.f32 v0+50 ; bin: heap_oob f3 0f 10 69 32
; asm: movd -50(%esi), %xmm2 ; asm: movss -50(%esi), %xmm2
[-,%xmm2] v111 = load.f32 v1-50 ; bin: 66 0f 6e 56 ce [-,%xmm2] v111 = load.f32 v1-50 ; bin: heap_oob f3 0f 10 56 ce
; asm: movd 10000(%ecx), %xmm5 ; asm: movss 10000(%ecx), %xmm5
[-,%xmm5] v120 = load.f32 v0+10000 ; bin: 66 0f 6e a9 00002710 [-,%xmm5] v120 = load.f32 v0+10000 ; bin: heap_oob f3 0f 10 a9 00002710
; asm: movd -10000(%esi), %xmm2 ; asm: movss -10000(%esi), %xmm2
[-,%xmm2] v121 = load.f32 v1-10000 ; bin: 66 0f 6e 96 ffffd8f0 [-,%xmm2] v121 = load.f32 v1-10000 ; bin: heap_oob f3 0f 10 96 ffffd8f0
; asm: movd %xmm5, (%ecx) ; asm: movss %xmm5, (%ecx)
[-] store.f32 v100, v0 ; bin: 66 0f 7e 29 [-] store.f32 v100, v0 ; bin: heap_oob f3 0f 11 29
; asm: movd %xmm2, (%esi) ; asm: movss %xmm2, (%esi)
[-] store.f32 v101, v1 ; bin: 66 0f 7e 16 [-] store.f32 v101, v1 ; bin: heap_oob f3 0f 11 16
; asm: movd %xmm5, 50(%ecx) ; asm: movss %xmm5, 50(%ecx)
[-] store.f32 v100, v0+50 ; bin: 66 0f 7e 69 32 [-] store.f32 v100, v0+50 ; bin: heap_oob f3 0f 11 69 32
; asm: movd %xmm2, -50(%esi) ; asm: movss %xmm2, -50(%esi)
[-] store.f32 v101, v1-50 ; bin: 66 0f 7e 56 ce [-] store.f32 v101, v1-50 ; bin: heap_oob f3 0f 11 56 ce
; asm: movd %xmm5, 10000(%ecx) ; asm: movss %xmm5, 10000(%ecx)
[-] store.f32 v100, v0+10000 ; bin: 66 0f 7e a9 00002710 [-] store.f32 v100, v0+10000 ; bin: heap_oob f3 0f 11 a9 00002710
; asm: movd %xmm2, -10000(%esi) ; asm: movss %xmm2, -10000(%esi)
[-] store.f32 v101, v1-10000 ; bin: 66 0f 7e 96 ffffd8f0 [-] store.f32 v101, v1-10000 ; bin: heap_oob f3 0f 11 96 ffffd8f0
; Spill / Fill. ; Spill / Fill.
; asm: movd %xmm5, 1032(%esp) ; asm: movss %xmm5, 1032(%esp)
[-,ss1] v200 = spill v100 ; bin: 66 0f 7e ac 24 00000408 [-,ss1] v200 = spill v100 ; bin: f3 0f 11 ac 24 00000408
; asm: movd %xmm2, 1032(%esp) ; asm: movss %xmm2, 1032(%esp)
[-,ss1] v201 = spill v101 ; bin: 66 0f 7e 94 24 00000408 [-,ss1] v201 = spill v101 ; bin: f3 0f 11 94 24 00000408
; asm: movd 1032(%esp), %xmm5 ; asm: movss 1032(%esp), %xmm5
[-,%xmm5] v210 = fill v200 ; bin: 66 0f 6e ac 24 00000408 [-,%xmm5] v210 = fill v200 ; bin: f3 0f 10 ac 24 00000408
; asm: movd 1032(%esp), %xmm2 ; asm: movss 1032(%esp), %xmm2
[-,%xmm2] v211 = fill v201 ; bin: 66 0f 6e 94 24 00000408 [-,%xmm2] v211 = fill v201 ; bin: f3 0f 10 94 24 00000408
; asm: movd %xmm5, 1032(%rsp) ; asm: movss %xmm5, 1032(%rsp)
regspill v100, %xmm5 -> ss1 ; bin: 66 0f 7e ac 24 00000408 regspill v100, %xmm5 -> ss1 ; bin: f3 0f 11 ac 24 00000408
; asm: movd 1032(%rsp), %xmm5 ; asm: movss 1032(%rsp), %xmm5
regfill v100, ss1 -> %xmm5 ; bin: 66 0f 6e ac 24 00000408 regfill v100, ss1 -> %xmm5 ; bin: f3 0f 10 ac 24 00000408
; Comparisons. ; Comparisons.
; ;
@@ -221,11 +221,11 @@ ebb0:
[-,%rdx] v307 = fcmp ule v11, v10 ; bin: 0f 2e d5 0f 96 c2 [-,%rdx] v307 = fcmp ule v11, v10 ; bin: 0f 2e d5 0f 96 c2
; asm: ucomiss %xmm2, %xmm5 ; asm: ucomiss %xmm2, %xmm5
[-,%eflags] v310 = ffcmp v10, v11 ; bin: 0f 2e ea [-,%rflags] v310 = ffcmp v10, v11 ; bin: 0f 2e ea
; asm: ucomiss %xmm2, %xmm5 ; asm: ucomiss %xmm2, %xmm5
[-,%eflags] v311 = ffcmp v11, v10 ; bin: 0f 2e d5 [-,%rflags] v311 = ffcmp v11, v10 ; bin: 0f 2e d5
; asm: ucomiss %xmm5, %xmm5 ; asm: ucomiss %xmm5, %xmm5
[-,%eflags] v312 = ffcmp v10, v10 ; bin: 0f 2e ed [-,%rflags] v312 = ffcmp v10, v10 ; bin: 0f 2e ed
return return
} }
@@ -362,48 +362,48 @@ ebb0:
; Load/Store ; Load/Store
; asm: movq (%ecx), %xmm5 ; asm: movsd (%ecx), %xmm5
[-,%xmm5] v100 = load.f64 v0 ; bin: f3 0f 7e 29 [-,%xmm5] v100 = load.f64 v0 ; bin: heap_oob f2 0f 10 29
; asm: movq (%esi), %xmm2 ; asm: movsd (%esi), %xmm2
[-,%xmm2] v101 = load.f64 v1 ; bin: f3 0f 7e 16 [-,%xmm2] v101 = load.f64 v1 ; bin: heap_oob f2 0f 10 16
; asm: movq 50(%ecx), %xmm5 ; asm: movsd 50(%ecx), %xmm5
[-,%xmm5] v110 = load.f64 v0+50 ; bin: f3 0f 7e 69 32 [-,%xmm5] v110 = load.f64 v0+50 ; bin: heap_oob f2 0f 10 69 32
; asm: movq -50(%esi), %xmm2 ; asm: movsd -50(%esi), %xmm2
[-,%xmm2] v111 = load.f64 v1-50 ; bin: f3 0f 7e 56 ce [-,%xmm2] v111 = load.f64 v1-50 ; bin: heap_oob f2 0f 10 56 ce
; asm: movq 10000(%ecx), %xmm5 ; asm: movsd 10000(%ecx), %xmm5
[-,%xmm5] v120 = load.f64 v0+10000 ; bin: f3 0f 7e a9 00002710 [-,%xmm5] v120 = load.f64 v0+10000 ; bin: heap_oob f2 0f 10 a9 00002710
; asm: movq -10000(%esi), %xmm2 ; asm: movsd -10000(%esi), %xmm2
[-,%xmm2] v121 = load.f64 v1-10000 ; bin: f3 0f 7e 96 ffffd8f0 [-,%xmm2] v121 = load.f64 v1-10000 ; bin: heap_oob f2 0f 10 96 ffffd8f0
; asm: movq %xmm5, (%ecx) ; asm: movsd %xmm5, (%ecx)
[-] store.f64 v100, v0 ; bin: 66 0f d6 29 [-] store.f64 v100, v0 ; bin: heap_oob f2 0f 11 29
; asm: movq %xmm2, (%esi) ; asm: movsd %xmm2, (%esi)
[-] store.f64 v101, v1 ; bin: 66 0f d6 16 [-] store.f64 v101, v1 ; bin: heap_oob f2 0f 11 16
; asm: movq %xmm5, 50(%ecx) ; asm: movsd %xmm5, 50(%ecx)
[-] store.f64 v100, v0+50 ; bin: 66 0f d6 69 32 [-] store.f64 v100, v0+50 ; bin: heap_oob f2 0f 11 69 32
; asm: movq %xmm2, -50(%esi) ; asm: movsd %xmm2, -50(%esi)
[-] store.f64 v101, v1-50 ; bin: 66 0f d6 56 ce [-] store.f64 v101, v1-50 ; bin: heap_oob f2 0f 11 56 ce
; asm: movq %xmm5, 10000(%ecx) ; asm: movsd %xmm5, 10000(%ecx)
[-] store.f64 v100, v0+10000 ; bin: 66 0f d6 a9 00002710 [-] store.f64 v100, v0+10000 ; bin: heap_oob f2 0f 11 a9 00002710
; asm: movq %xmm2, -10000(%esi) ; asm: movsd %xmm2, -10000(%esi)
[-] store.f64 v101, v1-10000 ; bin: 66 0f d6 96 ffffd8f0 [-] store.f64 v101, v1-10000 ; bin: heap_oob f2 0f 11 96 ffffd8f0
; Spill / Fill. ; Spill / Fill.
; asm: movq %xmm5, 1032(%esp) ; asm: movsd %xmm5, 1032(%esp)
[-,ss1] v200 = spill v100 ; bin: 66 0f d6 ac 24 00000408 [-,ss1] v200 = spill v100 ; bin: f2 0f 11 ac 24 00000408
; asm: movq %xmm2, 1032(%esp) ; asm: movsd %xmm2, 1032(%esp)
[-,ss1] v201 = spill v101 ; bin: 66 0f d6 94 24 00000408 [-,ss1] v201 = spill v101 ; bin: f2 0f 11 94 24 00000408
; asm: movq 1032(%esp), %xmm5 ; asm: movsd 1032(%esp), %xmm5
[-,%xmm5] v210 = fill v200 ; bin: f3 0f 7e ac 24 00000408 [-,%xmm5] v210 = fill v200 ; bin: f2 0f 10 ac 24 00000408
; asm: movq 1032(%esp), %xmm2 ; asm: movsd 1032(%esp), %xmm2
[-,%xmm2] v211 = fill v201 ; bin: f3 0f 7e 94 24 00000408 [-,%xmm2] v211 = fill v201 ; bin: f2 0f 10 94 24 00000408
; asm: movq %xmm5, 1032(%rsp) ; asm: movsd %xmm5, 1032(%rsp)
regspill v100, %xmm5 -> ss1 ; bin: 66 0f d6 ac 24 00000408 regspill v100, %xmm5 -> ss1 ; bin: f2 0f 11 ac 24 00000408
; asm: movq 1032(%rsp), %xmm5 ; asm: movsd 1032(%rsp), %xmm5
regfill v100, ss1 -> %xmm5 ; bin: f3 0f 7e ac 24 00000408 regfill v100, ss1 -> %xmm5 ; bin: f2 0f 10 ac 24 00000408
; Comparisons. ; Comparisons.
; ;
@@ -436,11 +436,11 @@ ebb0:
[-,%rdx] v307 = fcmp ule v11, v10 ; bin: 66 0f 2e d5 0f 96 c2 [-,%rdx] v307 = fcmp ule v11, v10 ; bin: 66 0f 2e d5 0f 96 c2
; asm: ucomisd %xmm2, %xmm5 ; asm: ucomisd %xmm2, %xmm5
[-,%eflags] v310 = ffcmp v10, v11 ; bin: 66 0f 2e ea [-,%rflags] v310 = ffcmp v10, v11 ; bin: 66 0f 2e ea
; asm: ucomisd %xmm2, %xmm5 ; asm: ucomisd %xmm2, %xmm5
[-,%eflags] v311 = ffcmp v11, v10 ; bin: 66 0f 2e d5 [-,%rflags] v311 = ffcmp v11, v10 ; bin: 66 0f 2e d5
; asm: ucomisd %xmm5, %xmm5 ; asm: ucomisd %xmm5, %xmm5
[-,%eflags] v312 = ffcmp v10, v10 ; bin: 66 0f 2e ed [-,%rflags] v312 = ffcmp v10, v10 ; bin: 66 0f 2e ed
return return
} }
@@ -448,7 +448,7 @@ ebb0:
function %cpuflags_float(f32 [%xmm0]) { function %cpuflags_float(f32 [%xmm0]) {
ebb0(v0: f32 [%xmm0]): ebb0(v0: f32 [%xmm0]):
; asm: ucomiss %xmm0, %xmm0 ; asm: ucomiss %xmm0, %xmm0
[-,%eflags] v1 = ffcmp v0, v0 ; bin: 0f 2e c0 [-,%rflags] v1 = ffcmp v0, v0 ; bin: 0f 2e c0
jump ebb1 jump ebb1
@@ -471,21 +471,21 @@ ebb1:
brff ule v1, ebb1 ; bin: 76 f0 brff ule v1, ebb1 ; bin: 76 f0
; asm: jp .+4; ud2 ; asm: jp .+4; ud2
trapff ord v1, user0 ; bin: 7a 02 0f 0b trapff ord v1, user0 ; bin: 7a 02 user0 0f 0b
; asm: jnp .+4; ud2 ; asm: jnp .+4; ud2
trapff uno v1, user0 ; bin: 7b 02 0f 0b trapff uno v1, user0 ; bin: 7b 02 user0 0f 0b
; asm: je .+4; ud2 ; asm: je .+4; ud2
trapff one v1, user0 ; bin: 74 02 0f 0b trapff one v1, user0 ; bin: 74 02 user0 0f 0b
; asm: jne .+4; ud2 ; asm: jne .+4; ud2
trapff ueq v1, user0 ; bin: 75 02 0f 0b trapff ueq v1, user0 ; bin: 75 02 user0 0f 0b
; asm: jna .+4; ud2 ; asm: jna .+4; ud2
trapff gt v1, user0 ; bin: 76 02 0f 0b trapff gt v1, user0 ; bin: 76 02 user0 0f 0b
; asm: jnae .+4; ud2 ; asm: jnae .+4; ud2
trapff ge v1, user0 ; bin: 72 02 0f 0b trapff ge v1, user0 ; bin: 72 02 user0 0f 0b
; asm: jnb .+4; ud2 ; asm: jnb .+4; ud2
trapff ult v1, user0 ; bin: 73 02 0f 0b trapff ult v1, user0 ; bin: 73 02 user0 0f 0b
; asm: jnbe .+4; ud2 ; asm: jnbe .+4; ud2
trapff ule v1, user0 ; bin: 77 02 0f 0b trapff ule v1, user0 ; bin: 77 02 user0 0f 0b
; asm: setnp %bl ; asm: setnp %bl
[-,%rbx] v10 = trueff ord v1 ; bin: 0f 9b c3 [-,%rbx] v10 = trueff ord v1 ; bin: 0f 9b c3

152
cranelift/filetests/isa/intel/binary32.cton
View File

@@ -1,4 +1,4 @@
; binary emission of 32-bit code. ; binary emission of x86-32 code.
test binemit test binemit
set is_compressed set is_compressed
isa intel haswell isa intel haswell
@@ -25,6 +25,9 @@ ebb0:
; asm: movl $2, %esi ; asm: movl $2, %esi
[-,%rsi] v2 = iconst.i32 2 ; bin: be 00000002 [-,%rsi] v2 = iconst.i32 2 ; bin: be 00000002
; asm: movb $1, %cl
[-,%rcx] v9007 = bconst.b1 true ; bin: b9 00000001
; Integer Register-Register Operations. ; Integer Register-Register Operations.
; asm: addl %esi, %ecx ; asm: addl %esi, %ecx
@@ -125,13 +128,13 @@ ebb0:
; asm: movl $2, %edx ; asm: movl $2, %edx
[-,%rdx] v53 = iconst.i32 2 ; bin: ba 00000002 [-,%rdx] v53 = iconst.i32 2 ; bin: ba 00000002
; asm: idivl %ecx ; asm: idivl %ecx
[-,%rax,%rdx] v54, v55 = x86_sdivmodx v52, v53, v1 ; bin: f7 f9 [-,%rax,%rdx] v54, v55 = x86_sdivmodx v52, v53, v1 ; bin: int_divz f7 f9
; asm: idivl %esi ; asm: idivl %esi
[-,%rax,%rdx] v56, v57 = x86_sdivmodx v52, v53, v2 ; bin: f7 fe [-,%rax,%rdx] v56, v57 = x86_sdivmodx v52, v53, v2 ; bin: int_divz f7 fe
; asm: divl %ecx ; asm: divl %ecx
[-,%rax,%rdx] v58, v59 = x86_udivmodx v52, v53, v1 ; bin: f7 f1 [-,%rax,%rdx] v58, v59 = x86_udivmodx v52, v53, v1 ; bin: int_divz f7 f1
; asm: divl %esi ; asm: divl %esi
[-,%rax,%rdx] v60, v61 = x86_udivmodx v52, v53, v2 ; bin: f7 f6 [-,%rax,%rdx] v60, v61 = x86_udivmodx v52, v53, v2 ; bin: int_divz f7 f6
; Register copies. ; Register copies.
@@ -152,105 +155,105 @@ ebb0:
; Register indirect addressing with no displacement. ; Register indirect addressing with no displacement.
; asm: movl %ecx, (%esi) ; asm: movl %ecx, (%esi)
store v1, v2 ; bin: 89 0e store v1, v2 ; bin: heap_oob 89 0e
; asm: movl %esi, (%ecx) ; asm: movl %esi, (%ecx)
store v2, v1 ; bin: 89 31 store v2, v1 ; bin: heap_oob 89 31
; asm: movw %cx, (%esi) ; asm: movw %cx, (%esi)
istore16 v1, v2 ; bin: 66 89 0e istore16 v1, v2 ; bin: heap_oob 66 89 0e
; asm: movw %si, (%ecx) ; asm: movw %si, (%ecx)
istore16 v2, v1 ; bin: 66 89 31 istore16 v2, v1 ; bin: heap_oob 66 89 31
; asm: movb %cl, (%esi) ; asm: movb %cl, (%esi)
istore8 v1, v2 ; bin: 88 0e istore8 v1, v2 ; bin: heap_oob 88 0e
; Can't store %sil in 32-bit mode (needs REX prefix). ; Can't store %sil in 32-bit mode (needs REX prefix).
; asm: movl (%ecx), %edi ; asm: movl (%ecx), %edi
[-,%rdi] v100 = load.i32 v1 ; bin: 8b 39 [-,%rdi] v100 = load.i32 v1 ; bin: heap_oob 8b 39
; asm: movl (%esi), %edx ; asm: movl (%esi), %edx
[-,%rdx] v101 = load.i32 v2 ; bin: 8b 16 [-,%rdx] v101 = load.i32 v2 ; bin: heap_oob 8b 16
; asm: movzwl (%ecx), %edi ; asm: movzwl (%ecx), %edi
[-,%rdi] v102 = uload16.i32 v1 ; bin: 0f b7 39 [-,%rdi] v102 = uload16.i32 v1 ; bin: heap_oob 0f b7 39
; asm: movzwl (%esi), %edx ; asm: movzwl (%esi), %edx
[-,%rdx] v103 = uload16.i32 v2 ; bin: 0f b7 16 [-,%rdx] v103 = uload16.i32 v2 ; bin: heap_oob 0f b7 16
; asm: movswl (%ecx), %edi ; asm: movswl (%ecx), %edi
[-,%rdi] v104 = sload16.i32 v1 ; bin: 0f bf 39 [-,%rdi] v104 = sload16.i32 v1 ; bin: heap_oob 0f bf 39
; asm: movswl (%esi), %edx ; asm: movswl (%esi), %edx
[-,%rdx] v105 = sload16.i32 v2 ; bin: 0f bf 16 [-,%rdx] v105 = sload16.i32 v2 ; bin: heap_oob 0f bf 16
; asm: movzbl (%ecx), %edi ; asm: movzbl (%ecx), %edi
[-,%rdi] v106 = uload8.i32 v1 ; bin: 0f b6 39 [-,%rdi] v106 = uload8.i32 v1 ; bin: heap_oob 0f b6 39
; asm: movzbl (%esi), %edx ; asm: movzbl (%esi), %edx
[-,%rdx] v107 = uload8.i32 v2 ; bin: 0f b6 16 [-,%rdx] v107 = uload8.i32 v2 ; bin: heap_oob 0f b6 16
; asm: movsbl (%ecx), %edi ; asm: movsbl (%ecx), %edi
[-,%rdi] v108 = sload8.i32 v1 ; bin: 0f be 39 [-,%rdi] v108 = sload8.i32 v1 ; bin: heap_oob 0f be 39
; asm: movsbl (%esi), %edx ; asm: movsbl (%esi), %edx
[-,%rdx] v109 = sload8.i32 v2 ; bin: 0f be 16 [-,%rdx] v109 = sload8.i32 v2 ; bin: heap_oob 0f be 16
; Register-indirect with 8-bit signed displacement. ; Register-indirect with 8-bit signed displacement.
; asm: movl %ecx, 100(%esi) ; asm: movl %ecx, 100(%esi)
store v1, v2+100 ; bin: 89 4e 64 store v1, v2+100 ; bin: heap_oob 89 4e 64
; asm: movl %esi, -100(%ecx) ; asm: movl %esi, -100(%ecx)
store v2, v1-100 ; bin: 89 71 9c store v2, v1-100 ; bin: heap_oob 89 71 9c
; asm: movw %cx, 100(%esi) ; asm: movw %cx, 100(%esi)
istore16 v1, v2+100 ; bin: 66 89 4e 64 istore16 v1, v2+100 ; bin: heap_oob 66 89 4e 64
; asm: movw %si, -100(%ecx) ; asm: movw %si, -100(%ecx)
istore16 v2, v1-100 ; bin: 66 89 71 9c istore16 v2, v1-100 ; bin: heap_oob 66 89 71 9c
; asm: movb %cl, 100(%esi) ; asm: movb %cl, 100(%esi)
istore8 v1, v2+100 ; bin: 88 4e 64 istore8 v1, v2+100 ; bin: heap_oob 88 4e 64
; asm: movl 50(%ecx), %edi ; asm: movl 50(%ecx), %edi
[-,%rdi] v110 = load.i32 v1+50 ; bin: 8b 79 32 [-,%rdi] v110 = load.i32 v1+50 ; bin: heap_oob 8b 79 32
; asm: movl -50(%esi), %edx ; asm: movl -50(%esi), %edx
[-,%rdx] v111 = load.i32 v2-50 ; bin: 8b 56 ce [-,%rdx] v111 = load.i32 v2-50 ; bin: heap_oob 8b 56 ce
; asm: movzwl 50(%ecx), %edi ; asm: movzwl 50(%ecx), %edi
[-,%rdi] v112 = uload16.i32 v1+50 ; bin: 0f b7 79 32 [-,%rdi] v112 = uload16.i32 v1+50 ; bin: heap_oob 0f b7 79 32
; asm: movzwl -50(%esi), %edx ; asm: movzwl -50(%esi), %edx
[-,%rdx] v113 = uload16.i32 v2-50 ; bin: 0f b7 56 ce [-,%rdx] v113 = uload16.i32 v2-50 ; bin: heap_oob 0f b7 56 ce
; asm: movswl 50(%ecx), %edi ; asm: movswl 50(%ecx), %edi
[-,%rdi] v114 = sload16.i32 v1+50 ; bin: 0f bf 79 32 [-,%rdi] v114 = sload16.i32 v1+50 ; bin: heap_oob 0f bf 79 32
; asm: movswl -50(%esi), %edx ; asm: movswl -50(%esi), %edx
[-,%rdx] v115 = sload16.i32 v2-50 ; bin: 0f bf 56 ce [-,%rdx] v115 = sload16.i32 v2-50 ; bin: heap_oob 0f bf 56 ce
; asm: movzbl 50(%ecx), %edi ; asm: movzbl 50(%ecx), %edi
[-,%rdi] v116 = uload8.i32 v1+50 ; bin: 0f b6 79 32 [-,%rdi] v116 = uload8.i32 v1+50 ; bin: heap_oob 0f b6 79 32
; asm: movzbl -50(%esi), %edx ; asm: movzbl -50(%esi), %edx
[-,%rdx] v117 = uload8.i32 v2-50 ; bin: 0f b6 56 ce [-,%rdx] v117 = uload8.i32 v2-50 ; bin: heap_oob 0f b6 56 ce
; asm: movsbl 50(%ecx), %edi ; asm: movsbl 50(%ecx), %edi
[-,%rdi] v118 = sload8.i32 v1+50 ; bin: 0f be 79 32 [-,%rdi] v118 = sload8.i32 v1+50 ; bin: heap_oob 0f be 79 32
; asm: movsbl -50(%esi), %edx ; asm: movsbl -50(%esi), %edx
[-,%rdx] v119 = sload8.i32 v2-50 ; bin: 0f be 56 ce [-,%rdx] v119 = sload8.i32 v2-50 ; bin: heap_oob 0f be 56 ce
; Register-indirect with 32-bit signed displacement. ; Register-indirect with 32-bit signed displacement.
; asm: movl %ecx, 10000(%esi) ; asm: movl %ecx, 10000(%esi)
store v1, v2+10000 ; bin: 89 8e 00002710 store v1, v2+10000 ; bin: heap_oob 89 8e 00002710
; asm: movl %esi, -10000(%ecx) ; asm: movl %esi, -10000(%ecx)
store v2, v1-10000 ; bin: 89 b1 ffffd8f0 store v2, v1-10000 ; bin: heap_oob 89 b1 ffffd8f0
; asm: movw %cx, 10000(%esi) ; asm: movw %cx, 10000(%esi)
istore16 v1, v2+10000 ; bin: 66 89 8e 00002710 istore16 v1, v2+10000 ; bin: heap_oob 66 89 8e 00002710
; asm: movw %si, -10000(%ecx) ; asm: movw %si, -10000(%ecx)
istore16 v2, v1-10000 ; bin: 66 89 b1 ffffd8f0 istore16 v2, v1-10000 ; bin: heap_oob 66 89 b1 ffffd8f0
; asm: movb %cl, 10000(%esi) ; asm: movb %cl, 10000(%esi)
istore8 v1, v2+10000 ; bin: 88 8e 00002710 istore8 v1, v2+10000 ; bin: heap_oob 88 8e 00002710
; asm: movl 50000(%ecx), %edi ; asm: movl 50000(%ecx), %edi
[-,%rdi] v120 = load.i32 v1+50000 ; bin: 8b b9 0000c350 [-,%rdi] v120 = load.i32 v1+50000 ; bin: heap_oob 8b b9 0000c350
; asm: movl -50000(%esi), %edx ; asm: movl -50000(%esi), %edx
[-,%rdx] v121 = load.i32 v2-50000 ; bin: 8b 96 ffff3cb0 [-,%rdx] v121 = load.i32 v2-50000 ; bin: heap_oob 8b 96 ffff3cb0
; asm: movzwl 50000(%ecx), %edi ; asm: movzwl 50000(%ecx), %edi
[-,%rdi] v122 = uload16.i32 v1+50000 ; bin: 0f b7 b9 0000c350 [-,%rdi] v122 = uload16.i32 v1+50000 ; bin: heap_oob 0f b7 b9 0000c350
; asm: movzwl -50000(%esi), %edx ; asm: movzwl -50000(%esi), %edx
[-,%rdx] v123 = uload16.i32 v2-50000 ; bin: 0f b7 96 ffff3cb0 [-,%rdx] v123 = uload16.i32 v2-50000 ; bin: heap_oob 0f b7 96 ffff3cb0
; asm: movswl 50000(%ecx), %edi ; asm: movswl 50000(%ecx), %edi
[-,%rdi] v124 = sload16.i32 v1+50000 ; bin: 0f bf b9 0000c350 [-,%rdi] v124 = sload16.i32 v1+50000 ; bin: heap_oob 0f bf b9 0000c350
; asm: movswl -50000(%esi), %edx ; asm: movswl -50000(%esi), %edx
[-,%rdx] v125 = sload16.i32 v2-50000 ; bin: 0f bf 96 ffff3cb0 [-,%rdx] v125 = sload16.i32 v2-50000 ; bin: heap_oob 0f bf 96 ffff3cb0
; asm: movzbl 50000(%ecx), %edi ; asm: movzbl 50000(%ecx), %edi
[-,%rdi] v126 = uload8.i32 v1+50000 ; bin: 0f b6 b9 0000c350 [-,%rdi] v126 = uload8.i32 v1+50000 ; bin: heap_oob 0f b6 b9 0000c350
; asm: movzbl -50000(%esi), %edx ; asm: movzbl -50000(%esi), %edx
[-,%rdx] v127 = uload8.i32 v2-50000 ; bin: 0f b6 96 ffff3cb0 [-,%rdx] v127 = uload8.i32 v2-50000 ; bin: heap_oob 0f b6 96 ffff3cb0
; asm: movsbl 50000(%ecx), %edi ; asm: movsbl 50000(%ecx), %edi
[-,%rdi] v128 = sload8.i32 v1+50000 ; bin: 0f be b9 0000c350 [-,%rdi] v128 = sload8.i32 v1+50000 ; bin: heap_oob 0f be b9 0000c350
; asm: movsbl -50000(%esi), %edx ; asm: movsbl -50000(%esi), %edx
[-,%rdx] v129 = sload8.i32 v2-50000 ; bin: 0f be 96 ffff3cb0 [-,%rdx] v129 = sload8.i32 v2-50000 ; bin: heap_oob 0f be 96 ffff3cb0
; Bit-counting instructions. ; Bit-counting instructions.
@@ -403,6 +406,13 @@ ebb0:
; asm: addl $-2147483648, %esp ; asm: addl $-2147483648, %esp
adjust_sp_imm -2147483648 ; bin: 81 c4 80000000 adjust_sp_imm -2147483648 ; bin: 81 c4 80000000
; Shift immediates
; asm: shll $2, %esi
[-,%rsi] v513 = ishl_imm v2, 2 ; bin: c1 e6 02
; asm: sarl $5, %esi
[-,%rsi] v514 = sshr_imm v2, 5 ; bin: c1 fe 05
; asm: shrl $8, %esi
[-,%rsi] v515 = ushr_imm v2, 8 ; bin: c1 ee 08
; asm: testl %ecx, %ecx ; asm: testl %ecx, %ecx
; asm: je ebb1 ; asm: je ebb1
@@ -427,7 +437,7 @@ ebb1:
; asm: ebb2: ; asm: ebb2:
ebb2: ebb2:
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Special branch encodings only for I32 mode. ; Special branch encodings only for I32 mode.
@@ -466,9 +476,9 @@ ebb0:
ebb1: ebb1:
; asm: cmpl %esi, %ecx ; asm: cmpl %esi, %ecx
[-,%eflags] v10 = ifcmp v1, v2 ; bin: 39 f1 [-,%rflags] v10 = ifcmp v1, v2 ; bin: 39 f1
; asm: cmpl %ecx, %esi ; asm: cmpl %ecx, %esi
[-,%eflags] v11 = ifcmp v2, v1 ; bin: 39 ce [-,%rflags] v11 = ifcmp v2, v1 ; bin: 39 ce
; asm: je ebb1 ; asm: je ebb1
brif eq v11, ebb1 ; bin: 74 fa brif eq v11, ebb1 ; bin: 74 fa
@@ -514,41 +524,41 @@ ebb1:
; The trapif instructions are encoded as macros: a conditional jump over a ud2. ; The trapif instructions are encoded as macros: a conditional jump over a ud2.
; asm: jne .+4; ud2 ; asm: jne .+4; ud2
trapif eq v11, user0 ; bin: 75 02 0f 0b trapif eq v11, user0 ; bin: 75 02 user0 0f 0b
; asm: je .+4; ud2 ; asm: je .+4; ud2
trapif ne v11, user0 ; bin: 74 02 0f 0b trapif ne v11, user0 ; bin: 74 02 user0 0f 0b
; asm: jnl .+4; ud2 ; asm: jnl .+4; ud2
trapif slt v11, user0 ; bin: 7d 02 0f 0b trapif slt v11, user0 ; bin: 7d 02 user0 0f 0b
; asm: jnge .+4; ud2 ; asm: jnge .+4; ud2
trapif sge v11, user0 ; bin: 7c 02 0f 0b trapif sge v11, user0 ; bin: 7c 02 user0 0f 0b
; asm: jng .+4; ud2 ; asm: jng .+4; ud2
trapif sgt v11, user0 ; bin: 7e 02 0f 0b trapif sgt v11, user0 ; bin: 7e 02 user0 0f 0b
; asm: jnle .+4; ud2 ; asm: jnle .+4; ud2
trapif sle v11, user0 ; bin: 7f 02 0f 0b trapif sle v11, user0 ; bin: 7f 02 user0 0f 0b
; asm: jnb .+4; ud2 ; asm: jnb .+4; ud2
trapif ult v11, user0 ; bin: 73 02 0f 0b trapif ult v11, user0 ; bin: 73 02 user0 0f 0b
; asm: jnae .+4; ud2 ; asm: jnae .+4; ud2
trapif uge v11, user0 ; bin: 72 02 0f 0b trapif uge v11, user0 ; bin: 72 02 user0 0f 0b
; asm: jna .+4; ud2 ; asm: jna .+4; ud2
trapif ugt v11, user0 ; bin: 76 02 0f 0b trapif ugt v11, user0 ; bin: 76 02 user0 0f 0b
; asm: jnbe .+4; ud2 ; asm: jnbe .+4; ud2
trapif ule v11, user0 ; bin: 77 02 0f 0b trapif ule v11, user0 ; bin: 77 02 user0 0f 0b
; Stack check. ; Stack check.
; asm: cmpl %esp, %ecx ; asm: cmpl %esp, %ecx
[-,%eflags] v40 = ifcmp_sp v1 ; bin: 39 e1 [-,%rflags] v40 = ifcmp_sp v1 ; bin: 39 e1
; asm: cmpl %esp, %esi ; asm: cmpl %esp, %esi
[-,%eflags] v41 = ifcmp_sp v2 ; bin: 39 e6 [-,%rflags] v41 = ifcmp_sp v2 ; bin: 39 e6
; asm: cmpl $-100, %ecx ; asm: cmpl $-100, %ecx
[-,%eflags] v42 = ifcmp_imm v1, -100 ; bin: 83 f9 9c [-,%rflags] v42 = ifcmp_imm v1, -100 ; bin: 83 f9 9c
; asm: cmpl $100, %esi ; asm: cmpl $100, %esi
[-,%eflags] v43 = ifcmp_imm v2, 100 ; bin: 83 fe 64 [-,%rflags] v43 = ifcmp_imm v2, 100 ; bin: 83 fe 64
; asm: cmpl $-10000, %ecx ; asm: cmpl $-10000, %ecx
[-,%eflags] v44 = ifcmp_imm v1, -10000 ; bin: 81 f9 ffffd8f0 [-,%rflags] v44 = ifcmp_imm v1, -10000 ; bin: 81 f9 ffffd8f0
; asm: cmpl $10000, %esi ; asm: cmpl $10000, %esi
[-,%eflags] v45 = ifcmp_imm v2, 10000 ; bin: 81 fe 00002710 [-,%rflags] v45 = ifcmp_imm v2, 10000 ; bin: 81 fe 00002710
return return
} }
@@ -566,7 +576,7 @@ ebb0:
; asm: movzbl %cl, %esi ; asm: movzbl %cl, %esi
[-,%rsi] v30 = uextend.i32 v11 ; bin: 0f b6 f1 [-,%rsi] v30 = uextend.i32 v11 ; bin: 0f b6 f1
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Tests for i32/i16 conversion instructions. ; Tests for i32/i16 conversion instructions.
@@ -582,5 +592,5 @@ ebb0:
; asm: movzwl %cx, %esi ; asm: movzwl %cx, %esi
[-,%rsi] v30 = uextend.i32 v11 ; bin: 0f b7 f1 [-,%rsi] v30 = uextend.i32 v11 ; bin: 0f b7 f1
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }

190
cranelift/filetests/isa/intel/binary64-float.cton
View File

@@ -157,52 +157,52 @@ ebb0:
; Load/Store ; Load/Store
; asm: movd (%r14), %xmm5 ; asm: movss (%r14), %xmm5
[-,%xmm5] v100 = load.f32 v3 ; bin: 66 41 0f 6e 2e [-,%xmm5] v100 = load.f32 v3 ; bin: heap_oob f3 41 0f 10 2e
; asm: movd (%rax), %xmm10 ; asm: movss (%rax), %xmm10
[-,%xmm10] v101 = load.f32 v2 ; bin: 66 44 0f 6e 10 [-,%xmm10] v101 = load.f32 v2 ; bin: heap_oob f3 44 0f 10 10
; asm: movd 50(%r14), %xmm5 ; asm: movss 50(%r14), %xmm5
[-,%xmm5] v110 = load.f32 v3+50 ; bin: 66 41 0f 6e 6e 32 [-,%xmm5] v110 = load.f32 v3+50 ; bin: heap_oob f3 41 0f 10 6e 32
; asm: movd -50(%rax), %xmm10 ; asm: movss -50(%rax), %xmm10
[-,%xmm10] v111 = load.f32 v2-50 ; bin: 66 44 0f 6e 50 ce [-,%xmm10] v111 = load.f32 v2-50 ; bin: heap_oob f3 44 0f 10 50 ce
; asm: movd 10000(%r14), %xmm5 ; asm: movss 10000(%r14), %xmm5
[-,%xmm5] v120 = load.f32 v3+10000 ; bin: 66 41 0f 6e ae 00002710 [-,%xmm5] v120 = load.f32 v3+10000 ; bin: heap_oob f3 41 0f 10 ae 00002710
; asm: movd -10000(%rax), %xmm10 ; asm: movss -10000(%rax), %xmm10
[-,%xmm10] v121 = load.f32 v2-10000 ; bin: 66 44 0f 6e 90 ffffd8f0 [-,%xmm10] v121 = load.f32 v2-10000 ; bin: heap_oob f3 44 0f 10 90 ffffd8f0
; asm: movd %xmm5, (%r14) ; asm: movss %xmm5, (%r14)
[-] store.f32 v100, v3 ; bin: 66 41 0f 7e 2e [-] store.f32 v100, v3 ; bin: heap_oob f3 41 0f 11 2e
; asm: movd %xmm10, (%rax) ; asm: movss %xmm10, (%rax)
[-] store.f32 v101, v2 ; bin: 66 44 0f 7e 10 [-] store.f32 v101, v2 ; bin: heap_oob f3 44 0f 11 10
; asm: movd %xmm5, (%r13) ; asm: movss %xmm5, (%r13)
[-] store.f32 v100, v4 ; bin: 66 41 0f 7e 6d 00 [-] store.f32 v100, v4 ; bin: heap_oob f3 41 0f 11 6d 00
; asm: movd %xmm10, (%r13) ; asm: movss %xmm10, (%r13)
[-] store.f32 v101, v4 ; bin: 66 45 0f 7e 55 00 [-] store.f32 v101, v4 ; bin: heap_oob f3 45 0f 11 55 00
; asm: movd %xmm5, 50(%r14) ; asm: movss %xmm5, 50(%r14)
[-] store.f32 v100, v3+50 ; bin: 66 41 0f 7e 6e 32 [-] store.f32 v100, v3+50 ; bin: heap_oob f3 41 0f 11 6e 32
; asm: movd %xmm10, -50(%rax) ; asm: movss %xmm10, -50(%rax)
[-] store.f32 v101, v2-50 ; bin: 66 44 0f 7e 50 ce [-] store.f32 v101, v2-50 ; bin: heap_oob f3 44 0f 11 50 ce
; asm: movd %xmm5, 10000(%r14) ; asm: movss %xmm5, 10000(%r14)
[-] store.f32 v100, v3+10000 ; bin: 66 41 0f 7e ae 00002710 [-] store.f32 v100, v3+10000 ; bin: heap_oob f3 41 0f 11 ae 00002710
; asm: movd %xmm10, -10000(%rax) ; asm: movss %xmm10, -10000(%rax)
[-] store.f32 v101, v2-10000 ; bin: 66 44 0f 7e 90 ffffd8f0 [-] store.f32 v101, v2-10000 ; bin: heap_oob f3 44 0f 11 90 ffffd8f0
; Spill / Fill. ; Spill / Fill.
; asm: movd %xmm5, 1032(%rsp) ; asm: movss %xmm5, 1032(%rsp)
[-,ss1] v200 = spill v100 ; bin: 66 0f 7e ac 24 00000408 [-,ss1] v200 = spill v100 ; bin: f3 0f 11 ac 24 00000408
; asm: movd %xmm10, 1032(%rsp) ; asm: movss %xmm10, 1032(%rsp)
[-,ss1] v201 = spill v101 ; bin: 66 44 0f 7e 94 24 00000408 [-,ss1] v201 = spill v101 ; bin: f3 44 0f 11 94 24 00000408
; asm: movd 1032(%rsp), %xmm5 ; asm: movss 1032(%rsp), %xmm5
[-,%xmm5] v210 = fill v200 ; bin: 66 0f 6e ac 24 00000408 [-,%xmm5] v210 = fill v200 ; bin: f3 0f 10 ac 24 00000408
; asm: movd 1032(%rsp), %xmm10 ; asm: movss 1032(%rsp), %xmm10
[-,%xmm10] v211 = fill v201 ; bin: 66 44 0f 6e 94 24 00000408 [-,%xmm10] v211 = fill v201 ; bin: f3 44 0f 10 94 24 00000408
; asm: movd %xmm5, 1032(%rsp) ; asm: movss %xmm5, 1032(%rsp)
regspill v100, %xmm5 -> ss1 ; bin: 66 0f 7e ac 24 00000408 regspill v100, %xmm5 -> ss1 ; bin: f3 0f 11 ac 24 00000408
; asm: movd 1032(%rsp), %xmm5 ; asm: movss 1032(%rsp), %xmm5
regfill v100, ss1 -> %xmm5 ; bin: 66 0f 6e ac 24 00000408 regfill v100, ss1 -> %xmm5 ; bin: f3 0f 10 ac 24 00000408
; Comparisons. ; Comparisons.
; ;
@@ -235,11 +235,11 @@ ebb0:
[-,%rdx] v307 = fcmp ule v11, v10 ; bin: 44 0f 2e d5 0f 96 c2 [-,%rdx] v307 = fcmp ule v11, v10 ; bin: 44 0f 2e d5 0f 96 c2
; asm: ucomiss %xmm10, %xmm5 ; asm: ucomiss %xmm10, %xmm5
[-,%eflags] v310 = ffcmp v10, v11 ; bin: 41 0f 2e ea [-,%rflags] v310 = ffcmp v10, v11 ; bin: 41 0f 2e ea
; asm: ucomiss %xmm10, %xmm5 ; asm: ucomiss %xmm10, %xmm5
[-,%eflags] v311 = ffcmp v11, v10 ; bin: 44 0f 2e d5 [-,%rflags] v311 = ffcmp v11, v10 ; bin: 44 0f 2e d5
; asm: ucomiss %xmm5, %xmm5 ; asm: ucomiss %xmm5, %xmm5
[-,%eflags] v312 = ffcmp v10, v10 ; bin: 0f 2e ed [-,%rflags] v312 = ffcmp v10, v10 ; bin: 0f 2e ed
return return
} }
@@ -392,52 +392,52 @@ ebb0:
; Load/Store ; Load/Store
; asm: movq (%r14), %xmm5 ; asm: movsd (%r14), %xmm5
[-,%xmm5] v100 = load.f64 v3 ; bin: f3 41 0f 7e 2e [-,%xmm5] v100 = load.f64 v3 ; bin: heap_oob f2 41 0f 10 2e
; asm: movq (%rax), %xmm10 ; asm: movsd (%rax), %xmm10
[-,%xmm10] v101 = load.f64 v2 ; bin: f3 44 0f 7e 10 [-,%xmm10] v101 = load.f64 v2 ; bin: heap_oob f2 44 0f 10 10
; asm: movq 50(%r14), %xmm5 ; asm: movsd 50(%r14), %xmm5
[-,%xmm5] v110 = load.f64 v3+50 ; bin: f3 41 0f 7e 6e 32 [-,%xmm5] v110 = load.f64 v3+50 ; bin: heap_oob f2 41 0f 10 6e 32
; asm: movq -50(%rax), %xmm10 ; asm: movsd -50(%rax), %xmm10
[-,%xmm10] v111 = load.f64 v2-50 ; bin: f3 44 0f 7e 50 ce [-,%xmm10] v111 = load.f64 v2-50 ; bin: heap_oob f2 44 0f 10 50 ce
; asm: movq 10000(%r14), %xmm5 ; asm: movsd 10000(%r14), %xmm5
[-,%xmm5] v120 = load.f64 v3+10000 ; bin: f3 41 0f 7e ae 00002710 [-,%xmm5] v120 = load.f64 v3+10000 ; bin: heap_oob f2 41 0f 10 ae 00002710
; asm: movq -10000(%rax), %xmm10 ; asm: movsd -10000(%rax), %xmm10
[-,%xmm10] v121 = load.f64 v2-10000 ; bin: f3 44 0f 7e 90 ffffd8f0 [-,%xmm10] v121 = load.f64 v2-10000 ; bin: heap_oob f2 44 0f 10 90 ffffd8f0
; asm: movq %xmm5, (%r14) ; asm: movsd %xmm5, (%r14)
[-] store.f64 v100, v3 ; bin: 66 41 0f d6 2e [-] store.f64 v100, v3 ; bin: heap_oob f2 41 0f 11 2e
; asm: movq %xmm10, (%rax) ; asm: movsd %xmm10, (%rax)
[-] store.f64 v101, v2 ; bin: 66 44 0f d6 10 [-] store.f64 v101, v2 ; bin: heap_oob f2 44 0f 11 10
; asm: movq %xmm5, (%r13) ; asm: movsd %xmm5, (%r13)
[-] store.f64 v100, v4 ; bin: 66 41 0f d6 6d 00 [-] store.f64 v100, v4 ; bin: heap_oob f2 41 0f 11 6d 00
; asm: movq %xmm10, (%r13) ; asm: movsd %xmm10, (%r13)
[-] store.f64 v101, v4 ; bin: 66 45 0f d6 55 00 [-] store.f64 v101, v4 ; bin: heap_oob f2 45 0f 11 55 00
; asm: movq %xmm5, 50(%r14) ; asm: movsd %xmm5, 50(%r14)
[-] store.f64 v100, v3+50 ; bin: 66 41 0f d6 6e 32 [-] store.f64 v100, v3+50 ; bin: heap_oob f2 41 0f 11 6e 32
; asm: movq %xmm10, -50(%rax) ; asm: movsd %xmm10, -50(%rax)
[-] store.f64 v101, v2-50 ; bin: 66 44 0f d6 50 ce [-] store.f64 v101, v2-50 ; bin: heap_oob f2 44 0f 11 50 ce
; asm: movq %xmm5, 10000(%r14) ; asm: movsd %xmm5, 10000(%r14)
[-] store.f64 v100, v3+10000 ; bin: 66 41 0f d6 ae 00002710 [-] store.f64 v100, v3+10000 ; bin: heap_oob f2 41 0f 11 ae 00002710
; asm: movq %xmm10, -10000(%rax) ; asm: movsd %xmm10, -10000(%rax)
[-] store.f64 v101, v2-10000 ; bin: 66 44 0f d6 90 ffffd8f0 [-] store.f64 v101, v2-10000 ; bin: heap_oob f2 44 0f 11 90 ffffd8f0
; Spill / Fill. ; Spill / Fill.
; asm: movq %xmm5, 1032(%rsp) ; asm: movsd %xmm5, 1032(%rsp)
[-,ss1] v200 = spill v100 ; bin: 66 0f d6 ac 24 00000408 [-,ss1] v200 = spill v100 ; bin: f2 0f 11 ac 24 00000408
; asm: movq %xmm10, 1032(%rsp) ; asm: movsd %xmm10, 1032(%rsp)
[-,ss1] v201 = spill v101 ; bin: 66 44 0f d6 94 24 00000408 [-,ss1] v201 = spill v101 ; bin: f2 44 0f 11 94 24 00000408
; asm: movq 1032(%rsp), %xmm5 ; asm: movsd 1032(%rsp), %xmm5
[-,%xmm5] v210 = fill v200 ; bin: f3 0f 7e ac 24 00000408 [-,%xmm5] v210 = fill v200 ; bin: f2 0f 10 ac 24 00000408
; asm: movq 1032(%rsp), %xmm10 ; asm: movsd 1032(%rsp), %xmm10
[-,%xmm10] v211 = fill v201 ; bin: f3 44 0f 7e 94 24 00000408 [-,%xmm10] v211 = fill v201 ; bin: f2 44 0f 10 94 24 00000408
; asm: movq %xmm5, 1032(%rsp) ; asm: movsd %xmm5, 1032(%rsp)
regspill v100, %xmm5 -> ss1 ; bin: 66 0f d6 ac 24 00000408 regspill v100, %xmm5 -> ss1 ; bin: f2 0f 11 ac 24 00000408
; asm: movq 1032(%rsp), %xmm5 ; asm: movsd 1032(%rsp), %xmm5
regfill v100, ss1 -> %xmm5 ; bin: f3 0f 7e ac 24 00000408 regfill v100, ss1 -> %xmm5 ; bin: f2 0f 10 ac 24 00000408
; Comparisons. ; Comparisons.
; ;
@@ -470,11 +470,11 @@ ebb0:
[-,%rdx] v307 = fcmp ule v11, v10 ; bin: 66 44 0f 2e d5 0f 96 c2 [-,%rdx] v307 = fcmp ule v11, v10 ; bin: 66 44 0f 2e d5 0f 96 c2
; asm: ucomisd %xmm10, %xmm5 ; asm: ucomisd %xmm10, %xmm5
[-,%eflags] v310 = ffcmp v10, v11 ; bin: 66 41 0f 2e ea [-,%rflags] v310 = ffcmp v10, v11 ; bin: 66 41 0f 2e ea
; asm: ucomisd %xmm10, %xmm5 ; asm: ucomisd %xmm10, %xmm5
[-,%eflags] v311 = ffcmp v11, v10 ; bin: 66 44 0f 2e d5 [-,%rflags] v311 = ffcmp v11, v10 ; bin: 66 44 0f 2e d5
; asm: ucomisd %xmm5, %xmm5 ; asm: ucomisd %xmm5, %xmm5
[-,%eflags] v312 = ffcmp v10, v10 ; bin: 66 0f 2e ed [-,%rflags] v312 = ffcmp v10, v10 ; bin: 66 0f 2e ed
return return
} }
@@ -482,7 +482,7 @@ ebb0:
function %cpuflags_float(f32 [%xmm0]) { function %cpuflags_float(f32 [%xmm0]) {
ebb0(v0: f32 [%xmm0]): ebb0(v0: f32 [%xmm0]):
; asm: ucomiss %xmm0, %xmm0 ; asm: ucomiss %xmm0, %xmm0
[-,%eflags] v1 = ffcmp v0, v0 ; bin: 0f 2e c0 [-,%rflags] v1 = ffcmp v0, v0 ; bin: 0f 2e c0
jump ebb1 jump ebb1
@@ -505,21 +505,21 @@ ebb1:
brff ule v1, ebb1 ; bin: 76 f0 brff ule v1, ebb1 ; bin: 76 f0
; asm: jp .+4; ud2 ; asm: jp .+4; ud2
trapff ord v1, user0 ; bin: 7a 02 0f 0b trapff ord v1, user0 ; bin: 7a 02 user0 0f 0b
; asm: jnp .+4; ud2 ; asm: jnp .+4; ud2
trapff uno v1, user0 ; bin: 7b 02 0f 0b trapff uno v1, user0 ; bin: 7b 02 user0 0f 0b
; asm: je .+4; ud2 ; asm: je .+4; ud2
trapff one v1, user0 ; bin: 74 02 0f 0b trapff one v1, user0 ; bin: 74 02 user0 0f 0b
; asm: jne .+4; ud2 ; asm: jne .+4; ud2
trapff ueq v1, user0 ; bin: 75 02 0f 0b trapff ueq v1, user0 ; bin: 75 02 user0 0f 0b
; asm: jna .+4; ud2 ; asm: jna .+4; ud2
trapff gt v1, user0 ; bin: 76 02 0f 0b trapff gt v1, user0 ; bin: 76 02 user0 0f 0b
; asm: jnae .+4; ud2 ; asm: jnae .+4; ud2
trapff ge v1, user0 ; bin: 72 02 0f 0b trapff ge v1, user0 ; bin: 72 02 user0 0f 0b
; asm: jnb .+4; ud2 ; asm: jnb .+4; ud2
trapff ult v1, user0 ; bin: 73 02 0f 0b trapff ult v1, user0 ; bin: 73 02 user0 0f 0b
; asm: jnbe .+4; ud2 ; asm: jnbe .+4; ud2
trapff ule v1, user0 ; bin: 77 02 0f 0b trapff ule v1, user0 ; bin: 77 02 user0 0f 0b
; asm: setnp %bl ; asm: setnp %bl
[-,%rbx] v10 = trueff ord v1 ; bin: 0f 9b c3 [-,%rbx] v10 = trueff ord v1 ; bin: 0f 9b c3

327
cranelift/filetests/isa/intel/binary64.cton
View File

@@ -1,4 +1,4 @@
; binary emission of 64-bit code. ; binary emission of x86-64 code.
test binemit test binemit
set is_64bit set is_64bit
set is_compressed set is_compressed
@@ -38,6 +38,11 @@ ebb0:
; asm: movq $0xffffffff88001122, %r14 # 32-bit sign-extended constant. ; asm: movq $0xffffffff88001122, %r14 # 32-bit sign-extended constant.
[-,%r14] v5 = iconst.i64 0xffff_ffff_8800_1122 ; bin: 49 c7 c6 88001122 [-,%r14] v5 = iconst.i64 0xffff_ffff_8800_1122 ; bin: 49 c7 c6 88001122
; asm: movb $1, %cl
[-,%rcx] v9007 = bconst.b1 true ; bin: b9 00000001
; asm: movb $1, %sil
[-,%r10] v9008 = bconst.b1 true ; bin: 41 ba 00000001
; Integer Register-Register Operations. ; Integer Register-Register Operations.
; asm: addq %rsi, %rcx ; asm: addq %rsi, %rcx
@@ -170,146 +175,146 @@ ebb0:
; Register indirect addressing with no displacement. ; Register indirect addressing with no displacement.
; asm: movq %rcx, (%r10) ; asm: movq %rcx, (%r10)
store v1, v3 ; bin: 49 89 0a store v1, v3 ; bin: heap_oob 49 89 0a
; asm: movq %r10, (%rcx) ; asm: movq %r10, (%rcx)
store v3, v1 ; bin: 4c 89 11 store v3, v1 ; bin: heap_oob 4c 89 11
; asm: movl %ecx, (%r10) ; asm: movl %ecx, (%r10)
istore32 v1, v3 ; bin: 41 89 0a istore32 v1, v3 ; bin: heap_oob 41 89 0a
; asm: movl %r10d, (%rcx) ; asm: movl %r10d, (%rcx)
istore32 v3, v1 ; bin: 44 89 11 istore32 v3, v1 ; bin: heap_oob 44 89 11
; asm: movw %cx, (%r10) ; asm: movw %cx, (%r10)
istore16 v1, v3 ; bin: 66 41 89 0a istore16 v1, v3 ; bin: heap_oob 66 41 89 0a
; asm: movw %r10w, (%rcx) ; asm: movw %r10w, (%rcx)
istore16 v3, v1 ; bin: 66 44 89 11 istore16 v3, v1 ; bin: heap_oob 66 44 89 11
; asm: movb %cl, (%r10) ; asm: movb %cl, (%r10)
istore8 v1, v3 ; bin: 41 88 0a istore8 v1, v3 ; bin: heap_oob 41 88 0a
; asm: movb %r10b, (%rcx) ; asm: movb %r10b, (%rcx)
istore8 v3, v1 ; bin: 44 88 11 istore8 v3, v1 ; bin: heap_oob 44 88 11
; asm: movq (%rcx), %r14 ; asm: movq (%rcx), %r14
[-,%r14] v120 = load.i64 v1 ; bin: 4c 8b 31 [-,%r14] v120 = load.i64 v1 ; bin: heap_oob 4c 8b 31
; asm: movq (%r10), %rdx ; asm: movq (%r10), %rdx
[-,%rdx] v121 = load.i64 v3 ; bin: 49 8b 12 [-,%rdx] v121 = load.i64 v3 ; bin: heap_oob 49 8b 12
; asm: movl (%rcx), %r14d ; asm: movl (%rcx), %r14d
[-,%r14] v122 = uload32.i64 v1 ; bin: 44 8b 31 [-,%r14] v122 = uload32.i64 v1 ; bin: heap_oob 44 8b 31
; asm: movl (%r10), %edx ; asm: movl (%r10), %edx
[-,%rdx] v123 = uload32.i64 v3 ; bin: 41 8b 12 [-,%rdx] v123 = uload32.i64 v3 ; bin: heap_oob 41 8b 12
; asm: movslq (%rcx), %r14 ; asm: movslq (%rcx), %r14
[-,%r14] v124 = sload32.i64 v1 ; bin: 4c 63 31 [-,%r14] v124 = sload32.i64 v1 ; bin: heap_oob 4c 63 31
; asm: movslq (%r10), %rdx ; asm: movslq (%r10), %rdx
[-,%rdx] v125 = sload32.i64 v3 ; bin: 49 63 12 [-,%rdx] v125 = sload32.i64 v3 ; bin: heap_oob 49 63 12
; asm: movzwq (%rcx), %r14 ; asm: movzwq (%rcx), %r14
[-,%r14] v126 = uload16.i64 v1 ; bin: 4c 0f b7 31 [-,%r14] v126 = uload16.i64 v1 ; bin: heap_oob 4c 0f b7 31
; asm: movzwq (%r10), %rdx ; asm: movzwq (%r10), %rdx
[-,%rdx] v127 = uload16.i64 v3 ; bin: 49 0f b7 12 [-,%rdx] v127 = uload16.i64 v3 ; bin: heap_oob 49 0f b7 12
; asm: movswq (%rcx), %r14 ; asm: movswq (%rcx), %r14
[-,%r14] v128 = sload16.i64 v1 ; bin: 4c 0f bf 31 [-,%r14] v128 = sload16.i64 v1 ; bin: heap_oob 4c 0f bf 31
; asm: movswq (%r10), %rdx ; asm: movswq (%r10), %rdx
[-,%rdx] v129 = sload16.i64 v3 ; bin: 49 0f bf 12 [-,%rdx] v129 = sload16.i64 v3 ; bin: heap_oob 49 0f bf 12
; asm: movzbq (%rcx), %r14 ; asm: movzbq (%rcx), %r14
[-,%r14] v130 = uload8.i64 v1 ; bin: 4c 0f b6 31 [-,%r14] v130 = uload8.i64 v1 ; bin: heap_oob 4c 0f b6 31
; asm: movzbq (%r10), %rdx ; asm: movzbq (%r10), %rdx
[-,%rdx] v131 = uload8.i64 v3 ; bin: 49 0f b6 12 [-,%rdx] v131 = uload8.i64 v3 ; bin: heap_oob 49 0f b6 12
; asm: movsbq (%rcx), %r14 ; asm: movsbq (%rcx), %r14
[-,%r14] v132 = sload8.i64 v1 ; bin: 4c 0f be 31 [-,%r14] v132 = sload8.i64 v1 ; bin: heap_oob 4c 0f be 31
; asm: movsbq (%r10), %rdx ; asm: movsbq (%r10), %rdx
[-,%rdx] v133 = sload8.i64 v3 ; bin: 49 0f be 12 [-,%rdx] v133 = sload8.i64 v3 ; bin: heap_oob 49 0f be 12
; Register-indirect with 8-bit signed displacement. ; Register-indirect with 8-bit signed displacement.
; asm: movq %rcx, 100(%r10) ; asm: movq %rcx, 100(%r10)
store v1, v3+100 ; bin: 49 89 4a 64 store v1, v3+100 ; bin: heap_oob 49 89 4a 64
; asm: movq %r10, -100(%rcx) ; asm: movq %r10, -100(%rcx)
store v3, v1-100 ; bin: 4c 89 51 9c store v3, v1-100 ; bin: heap_oob 4c 89 51 9c
; asm: movl %ecx, 100(%r10) ; asm: movl %ecx, 100(%r10)
istore32 v1, v3+100 ; bin: 41 89 4a 64 istore32 v1, v3+100 ; bin: heap_oob 41 89 4a 64
; asm: movl %r10d, -100(%rcx) ; asm: movl %r10d, -100(%rcx)
istore32 v3, v1-100 ; bin: 44 89 51 9c istore32 v3, v1-100 ; bin: heap_oob 44 89 51 9c
; asm: movw %cx, 100(%r10) ; asm: movw %cx, 100(%r10)
istore16 v1, v3+100 ; bin: 66 41 89 4a 64 istore16 v1, v3+100 ; bin: heap_oob 66 41 89 4a 64
; asm: movw %r10w, -100(%rcx) ; asm: movw %r10w, -100(%rcx)
istore16 v3, v1-100 ; bin: 66 44 89 51 9c istore16 v3, v1-100 ; bin: heap_oob 66 44 89 51 9c
; asm: movb %cl, 100(%r10) ; asm: movb %cl, 100(%r10)
istore8 v1, v3+100 ; bin: 41 88 4a 64 istore8 v1, v3+100 ; bin: heap_oob 41 88 4a 64
; asm: movb %r10b, 100(%rcx) ; asm: movb %r10b, 100(%rcx)
istore8 v3, v1+100 ; bin: 44 88 51 64 istore8 v3, v1+100 ; bin: heap_oob 44 88 51 64
; asm: movq 50(%rcx), %r10 ; asm: movq 50(%rcx), %r10
[-,%r10] v140 = load.i64 v1+50 ; bin: 4c 8b 51 32 [-,%r10] v140 = load.i64 v1+50 ; bin: heap_oob 4c 8b 51 32
; asm: movq -50(%r10), %rdx ; asm: movq -50(%r10), %rdx
[-,%rdx] v141 = load.i64 v3-50 ; bin: 49 8b 52 ce [-,%rdx] v141 = load.i64 v3-50 ; bin: heap_oob 49 8b 52 ce
; asm: movl 50(%rcx), %edi ; asm: movl 50(%rcx), %edi
[-,%rdi] v142 = uload32.i64 v1+50 ; bin: 8b 79 32 [-,%rdi] v142 = uload32.i64 v1+50 ; bin: heap_oob 8b 79 32
; asm: movl -50(%rsi), %edx ; asm: movl -50(%rsi), %edx
[-,%rdx] v143 = uload32.i64 v2-50 ; bin: 8b 56 ce [-,%rdx] v143 = uload32.i64 v2-50 ; bin: heap_oob 8b 56 ce
; asm: movslq 50(%rcx), %rdi ; asm: movslq 50(%rcx), %rdi
[-,%rdi] v144 = sload32.i64 v1+50 ; bin: 48 63 79 32 [-,%rdi] v144 = sload32.i64 v1+50 ; bin: heap_oob 48 63 79 32
; asm: movslq -50(%rsi), %rdx ; asm: movslq -50(%rsi), %rdx
[-,%rdx] v145 = sload32.i64 v2-50 ; bin: 48 63 56 ce [-,%rdx] v145 = sload32.i64 v2-50 ; bin: heap_oob 48 63 56 ce
; asm: movzwq 50(%rcx), %rdi ; asm: movzwq 50(%rcx), %rdi
[-,%rdi] v146 = uload16.i64 v1+50 ; bin: 48 0f b7 79 32 [-,%rdi] v146 = uload16.i64 v1+50 ; bin: heap_oob 48 0f b7 79 32
; asm: movzwq -50(%rsi), %rdx ; asm: movzwq -50(%rsi), %rdx
[-,%rdx] v147 = uload16.i64 v2-50 ; bin: 48 0f b7 56 ce [-,%rdx] v147 = uload16.i64 v2-50 ; bin: heap_oob 48 0f b7 56 ce
; asm: movswq 50(%rcx), %rdi ; asm: movswq 50(%rcx), %rdi
[-,%rdi] v148 = sload16.i64 v1+50 ; bin: 48 0f bf 79 32 [-,%rdi] v148 = sload16.i64 v1+50 ; bin: heap_oob 48 0f bf 79 32
; asm: movswq -50(%rsi), %rdx ; asm: movswq -50(%rsi), %rdx
[-,%rdx] v149 = sload16.i64 v2-50 ; bin: 48 0f bf 56 ce [-,%rdx] v149 = sload16.i64 v2-50 ; bin: heap_oob 48 0f bf 56 ce
; asm: movzbq 50(%rcx), %rdi ; asm: movzbq 50(%rcx), %rdi
[-,%rdi] v150 = uload8.i64 v1+50 ; bin: 48 0f b6 79 32 [-,%rdi] v150 = uload8.i64 v1+50 ; bin: heap_oob 48 0f b6 79 32
; asm: movzbq -50(%rsi), %rdx ; asm: movzbq -50(%rsi), %rdx
[-,%rdx] v151 = uload8.i64 v2-50 ; bin: 48 0f b6 56 ce [-,%rdx] v151 = uload8.i64 v2-50 ; bin: heap_oob 48 0f b6 56 ce
; asm: movsbq 50(%rcx), %rdi ; asm: movsbq 50(%rcx), %rdi
[-,%rdi] v152 = sload8.i64 v1+50 ; bin: 48 0f be 79 32 [-,%rdi] v152 = sload8.i64 v1+50 ; bin: heap_oob 48 0f be 79 32
; asm: movsbq -50(%rsi), %rdx ; asm: movsbq -50(%rsi), %rdx
[-,%rdx] v153 = sload8.i64 v2-50 ; bin: 48 0f be 56 ce [-,%rdx] v153 = sload8.i64 v2-50 ; bin: heap_oob 48 0f be 56 ce
; Register-indirect with 32-bit signed displacement. ; Register-indirect with 32-bit signed displacement.
; asm: movq %rcx, 10000(%r10) ; asm: movq %rcx, 10000(%r10)
store v1, v3+10000 ; bin: 49 89 8a 00002710 store v1, v3+10000 ; bin: heap_oob 49 89 8a 00002710
; asm: movq %r10, -10000(%rcx) ; asm: movq %r10, -10000(%rcx)
store v3, v1-10000 ; bin: 4c 89 91 ffffd8f0 store v3, v1-10000 ; bin: heap_oob 4c 89 91 ffffd8f0
; asm: movl %ecx, 10000(%rsi) ; asm: movl %ecx, 10000(%rsi)
istore32 v1, v2+10000 ; bin: 89 8e 00002710 istore32 v1, v2+10000 ; bin: heap_oob 89 8e 00002710
; asm: movl %esi, -10000(%rcx) ; asm: movl %esi, -10000(%rcx)
istore32 v2, v1-10000 ; bin: 89 b1 ffffd8f0 istore32 v2, v1-10000 ; bin: heap_oob 89 b1 ffffd8f0
; asm: movw %cx, 10000(%rsi) ; asm: movw %cx, 10000(%rsi)
istore16 v1, v2+10000 ; bin: 66 89 8e 00002710 istore16 v1, v2+10000 ; bin: heap_oob 66 89 8e 00002710
; asm: movw %si, -10000(%rcx) ; asm: movw %si, -10000(%rcx)
istore16 v2, v1-10000 ; bin: 66 89 b1 ffffd8f0 istore16 v2, v1-10000 ; bin: heap_oob 66 89 b1 ffffd8f0
; asm: movb %cl, 10000(%rsi) ; asm: movb %cl, 10000(%rsi)
istore8 v1, v2+10000 ; bin: 88 8e 00002710 istore8 v1, v2+10000 ; bin: heap_oob 88 8e 00002710
; asm: movb %sil, 10000(%rcx) ; asm: movb %sil, 10000(%rcx)
istore8 v2, v1+10000 ; bin: 40 88 b1 00002710 istore8 v2, v1+10000 ; bin: heap_oob 40 88 b1 00002710
; asm: movq 50000(%rcx), %r10 ; asm: movq 50000(%rcx), %r10
[-,%r10] v160 = load.i64 v1+50000 ; bin: 4c 8b 91 0000c350 [-,%r10] v160 = load.i64 v1+50000 ; bin: heap_oob 4c 8b 91 0000c350
; asm: movq -50000(%r10), %rdx ; asm: movq -50000(%r10), %rdx
[-,%rdx] v161 = load.i64 v3-50000 ; bin: 49 8b 92 ffff3cb0 [-,%rdx] v161 = load.i64 v3-50000 ; bin: heap_oob 49 8b 92 ffff3cb0
; asm: movl 50000(%rcx), %edi ; asm: movl 50000(%rcx), %edi
[-,%rdi] v162 = uload32.i64 v1+50000 ; bin: 8b b9 0000c350 [-,%rdi] v162 = uload32.i64 v1+50000 ; bin: heap_oob 8b b9 0000c350
; asm: movl -50000(%rsi), %edx ; asm: movl -50000(%rsi), %edx
[-,%rdx] v163 = uload32.i64 v2-50000 ; bin: 8b 96 ffff3cb0 [-,%rdx] v163 = uload32.i64 v2-50000 ; bin: heap_oob 8b 96 ffff3cb0
; asm: movslq 50000(%rcx), %rdi ; asm: movslq 50000(%rcx), %rdi
[-,%rdi] v164 = sload32.i64 v1+50000 ; bin: 48 63 b9 0000c350 [-,%rdi] v164 = sload32.i64 v1+50000 ; bin: heap_oob 48 63 b9 0000c350
; asm: movslq -50000(%rsi), %rdx ; asm: movslq -50000(%rsi), %rdx
[-,%rdx] v165 = sload32.i64 v2-50000 ; bin: 48 63 96 ffff3cb0 [-,%rdx] v165 = sload32.i64 v2-50000 ; bin: heap_oob 48 63 96 ffff3cb0
; asm: movzwq 50000(%rcx), %rdi ; asm: movzwq 50000(%rcx), %rdi
[-,%rdi] v166 = uload16.i64 v1+50000 ; bin: 48 0f b7 b9 0000c350 [-,%rdi] v166 = uload16.i64 v1+50000 ; bin: heap_oob 48 0f b7 b9 0000c350
; asm: movzwq -50000(%rsi), %rdx ; asm: movzwq -50000(%rsi), %rdx
[-,%rdx] v167 = uload16.i64 v2-50000 ; bin: 48 0f b7 96 ffff3cb0 [-,%rdx] v167 = uload16.i64 v2-50000 ; bin: heap_oob 48 0f b7 96 ffff3cb0
; asm: movswq 50000(%rcx), %rdi ; asm: movswq 50000(%rcx), %rdi
[-,%rdi] v168 = sload16.i64 v1+50000 ; bin: 48 0f bf b9 0000c350 [-,%rdi] v168 = sload16.i64 v1+50000 ; bin: heap_oob 48 0f bf b9 0000c350
; asm: movswq -50000(%rsi), %rdx ; asm: movswq -50000(%rsi), %rdx
[-,%rdx] v169 = sload16.i64 v2-50000 ; bin: 48 0f bf 96 ffff3cb0 [-,%rdx] v169 = sload16.i64 v2-50000 ; bin: heap_oob 48 0f bf 96 ffff3cb0
; asm: movzbq 50000(%rcx), %rdi ; asm: movzbq 50000(%rcx), %rdi
[-,%rdi] v170 = uload8.i64 v1+50000 ; bin: 48 0f b6 b9 0000c350 [-,%rdi] v170 = uload8.i64 v1+50000 ; bin: heap_oob 48 0f b6 b9 0000c350
; asm: movzbq -50000(%rsi), %rdx ; asm: movzbq -50000(%rsi), %rdx
[-,%rdx] v171 = uload8.i64 v2-50000 ; bin: 48 0f b6 96 ffff3cb0 [-,%rdx] v171 = uload8.i64 v2-50000 ; bin: heap_oob 48 0f b6 96 ffff3cb0
; asm: movsbq 50000(%rcx), %rdi ; asm: movsbq 50000(%rcx), %rdi
[-,%rdi] v172 = sload8.i64 v1+50000 ; bin: 48 0f be b9 0000c350 [-,%rdi] v172 = sload8.i64 v1+50000 ; bin: heap_oob 48 0f be b9 0000c350
; asm: movsbq -50000(%rsi), %rdx ; asm: movsbq -50000(%rsi), %rdx
[-,%rdx] v173 = sload8.i64 v2-50000 ; bin: 48 0f be 96 ffff3cb0 [-,%rdx] v173 = sload8.i64 v2-50000 ; bin: heap_oob 48 0f be 96 ffff3cb0
; More arithmetic. ; More arithmetic.
@@ -324,17 +329,17 @@ ebb0:
[-,%rax] v190 = iconst.i64 1 [-,%rax] v190 = iconst.i64 1
[-,%rdx] v191 = iconst.i64 2 [-,%rdx] v191 = iconst.i64 2
; asm: idivq %rcx ; asm: idivq %rcx
[-,%rax,%rdx] v192, v193 = x86_sdivmodx v190, v191, v1 ; bin: 48 f7 f9 [-,%rax,%rdx] v192, v193 = x86_sdivmodx v190, v191, v1 ; bin: int_divz 48 f7 f9
; asm: idivq %rsi ; asm: idivq %rsi
[-,%rax,%rdx] v194, v195 = x86_sdivmodx v190, v191, v2 ; bin: 48 f7 fe [-,%rax,%rdx] v194, v195 = x86_sdivmodx v190, v191, v2 ; bin: int_divz 48 f7 fe
; asm: idivq %r10 ; asm: idivq %r10
[-,%rax,%rdx] v196, v197 = x86_sdivmodx v190, v191, v3 ; bin: 49 f7 fa [-,%rax,%rdx] v196, v197 = x86_sdivmodx v190, v191, v3 ; bin: int_divz 49 f7 fa
; asm: divq %rcx ; asm: divq %rcx
[-,%rax,%rdx] v198, v199 = x86_udivmodx v190, v191, v1 ; bin: 48 f7 f1 [-,%rax,%rdx] v198, v199 = x86_udivmodx v190, v191, v1 ; bin: int_divz 48 f7 f1
; asm: divq %rsi ; asm: divq %rsi
[-,%rax,%rdx] v200, v201 = x86_udivmodx v190, v191, v2 ; bin: 48 f7 f6 [-,%rax,%rdx] v200, v201 = x86_udivmodx v190, v191, v2 ; bin: int_divz 48 f7 f6
; asm: divq %r10 ; asm: divq %r10
[-,%rax,%rdx] v202, v203 = x86_udivmodx v190, v191, v3 ; bin: 49 f7 f2 [-,%rax,%rdx] v202, v203 = x86_udivmodx v190, v191, v3 ; bin: int_divz 49 f7 f2
; double-length multiply instructions, 64 bit ; double-length multiply instructions, 64 bit
[-,%rax] v1001 = iconst.i64 1 [-,%rax] v1001 = iconst.i64 1
@@ -453,6 +458,14 @@ ebb0:
; asm: setbe %dl ; asm: setbe %dl
[-,%rdx] v319 = icmp ule v2, v3 ; bin: 4c 39 d6 0f 96 c2 [-,%rdx] v319 = icmp ule v2, v3 ; bin: 4c 39 d6 0f 96 c2
; asm: cmpq $37, %rcx
; asm: setl %bl
[-,%rbx] v320 = icmp_imm slt v1, 37 ; bin: 48 83 f9 25 0f 9c c3
; asm: cmpq $100000, %rcx
; asm: setl %bl
[-,%rbx] v321 = icmp_imm slt v1, 100000 ; bin: 48 81 f9 000186a0 0f 9c c3
; Bool-to-int conversions. ; Bool-to-int conversions.
; asm: movzbq %bl, %rcx ; asm: movzbq %bl, %rcx
@@ -529,6 +542,21 @@ ebb0:
; asm: addq $-2147483648, %rsp ; asm: addq $-2147483648, %rsp
adjust_sp_imm -2147483648 ; bin: 48 81 c4 80000000 adjust_sp_imm -2147483648 ; bin: 48 81 c4 80000000
; Shift immediates
; asm: shlq $12, %rsi
[-,%rsi] v515 = ishl_imm v2, 12 ; bin: 48 c1 e6 0c
; asm: shlq $13, %r8
[-,%r8] v516 = ishl_imm v4, 13 ; bin: 49 c1 e0 0d
; asm: sarq $32, %rsi
[-,%rsi] v517 = sshr_imm v2, 32 ; bin: 48 c1 fe 20
; asm: sarq $33, %r8
[-,%r8] v518 = sshr_imm v4, 33 ; bin: 49 c1 f8 21
; asm: shrl $62, %rsi
[-,%rsi] v519 = ushr_imm v2, 62 ; bin: 48 c1 ee 3e
; asm: shrl $63, %r8
[-,%r8] v520 = ushr_imm v4, 63 ; bin: 49 c1 e8 3f
; asm: testq %rcx, %rcx ; asm: testq %rcx, %rcx
; asm: je ebb1 ; asm: je ebb1
brz v1, ebb1 ; bin: 48 85 c9 74 1b brz v1, ebb1 ; bin: 48 85 c9 74 1b
@@ -569,9 +597,9 @@ ebb0:
ebb1: ebb1:
; asm: cmpq %r10, %rcx ; asm: cmpq %r10, %rcx
[-,%eflags] v10 = ifcmp v1, v2 ; bin: 4c 39 d1 [-,%rflags] v10 = ifcmp v1, v2 ; bin: 4c 39 d1
; asm: cmpq %rcx, %r10 ; asm: cmpq %rcx, %r10
[-,%eflags] v11 = ifcmp v2, v1 ; bin: 49 39 ca [-,%rflags] v11 = ifcmp v2, v1 ; bin: 49 39 ca
; asm: je ebb1 ; asm: je ebb1
brif eq v11, ebb1 ; bin: 74 f8 brif eq v11, ebb1 ; bin: 74 f8
@@ -617,41 +645,42 @@ ebb1:
; The trapif instructions are encoded as macros: a conditional jump over a ud2. ; The trapif instructions are encoded as macros: a conditional jump over a ud2.
; asm: jne .+4; ud2 ; asm: jne .+4; ud2
trapif eq v11, user0 ; bin: 75 02 0f 0b trapif eq v11, user0 ; bin: 75 02 user0 0f 0b
; asm: je .+4; ud2 ; asm: je .+4; ud2
trapif ne v11, user0 ; bin: 74 02 0f 0b trapif ne v11, user0 ; bin: 74 02 user0 0f 0b
; asm: jnl .+4; ud2 ; asm: jnl .+4; ud2
trapif slt v11, user0 ; bin: 7d 02 0f 0b trapif slt v11, user0 ; bin: 7d 02 user0 0f 0b
; asm: jnge .+4; ud2 ; asm: jnge .+4; ud2
trapif sge v11, user0 ; bin: 7c 02 0f 0b trapif sge v11, user0 ; bin: 7c 02 user0 0f 0b
; asm: jng .+4; ud2 ; asm: jng .+4; ud2
trapif sgt v11, user0 ; bin: 7e 02 0f 0b trapif sgt v11, user0 ; bin: 7e 02 user0 0f 0b
; asm: jnle .+4; ud2 ; asm: jnle .+4; ud2
trapif sle v11, user0 ; bin: 7f 02 0f 0b trapif sle v11, user0 ; bin: 7f 02 user0 0f 0b
; asm: jnb .+4; ud2 ; asm: jnb .+4; ud2
trapif ult v11, user0 ; bin: 73 02 0f 0b trapif ult v11, user0 ; bin: 73 02 user0 0f 0b
; asm: jnae .+4; ud2 ; asm: jnae .+4; ud2
trapif uge v11, user0 ; bin: 72 02 0f 0b trapif uge v11, user0 ; bin: 72 02 user0 0f 0b
; asm: jna .+4; ud2 ; asm: jna .+4; ud2
trapif ugt v11, user0 ; bin: 76 02 0f 0b trapif ugt v11, user0 ; bin: 76 02 user0 0f 0b
; asm: jnbe .+4; ud2 ; asm: jnbe .+4; ud2
trapif ule v11, user0 ; bin: 77 02 0f 0b trapif ule v11, user0 ; bin: 77 02 user0 0f 0b
; Stack check. ; Stack check.
; asm: cmpq %rsp, %rcx ; asm: cmpq %rsp, %rcx
[-,%eflags] v40 = ifcmp_sp v1 ; bin: 48 39 e1 [-,%rflags] v40 = ifcmp_sp v1 ; bin: 48 39 e1
; asm: cmpq %rsp, %r10 ; asm: cmpq %rsp, %r10
[-,%eflags] v41 = ifcmp_sp v2 ; bin: 49 39 e2 [-,%rflags] v41 = ifcmp_sp v2 ; bin: 49 39 e2
; asm: cmpq $-100, %rcx ; asm: cmpq $-100, %rcx
[-,%eflags] v522 = ifcmp_imm v1, -100 ; bin: 48 83 f9 9c [-,%rflags] v522 = ifcmp_imm v1, -100 ; bin: 48 83 f9 9c
; asm: cmpq $100, %r10 ; asm: cmpq $100, %r10
[-,%eflags] v523 = ifcmp_imm v2, 100 ; bin: 49 83 fa 64 [-,%rflags] v523 = ifcmp_imm v2, 100 ; bin: 49 83 fa 64
; asm: cmpq $-10000, %rcx ; asm: cmpq $-10000, %rcx
[-,%eflags] v524 = ifcmp_imm v1, -10000 ; bin: 48 81 f9 ffffd8f0 [-,%rflags] v524 = ifcmp_imm v1, -10000 ; bin: 48 81 f9 ffffd8f0
; asm: cmpq $10000, %r10 ; asm: cmpq $10000, %r10
[-,%eflags] v525 = ifcmp_imm v2, 10000 ; bin: 49 81 fa 00002710 [-,%rflags] v525 = ifcmp_imm v2, 10000 ; bin: 49 81 fa 00002710
return return
} }
@@ -708,71 +737,71 @@ ebb0:
; Register indirect addressing with no displacement. ; Register indirect addressing with no displacement.
; asm: movl (%rcx), %edi ; asm: movl (%rcx), %edi
[-,%rdi] v10 = load.i32 v1 ; bin: 8b 39 [-,%rdi] v10 = load.i32 v1 ; bin: heap_oob 8b 39
; asm: movl (%rsi), %edx ; asm: movl (%rsi), %edx
[-,%rdx] v11 = load.i32 v2 ; bin: 8b 16 [-,%rdx] v11 = load.i32 v2 ; bin: heap_oob 8b 16
; asm: movzwl (%rcx), %edi ; asm: movzwl (%rcx), %edi
[-,%rdi] v12 = uload16.i32 v1 ; bin: 0f b7 39 [-,%rdi] v12 = uload16.i32 v1 ; bin: heap_oob 0f b7 39
; asm: movzwl (%rsi), %edx ; asm: movzwl (%rsi), %edx
[-,%rdx] v13 = uload16.i32 v2 ; bin: 0f b7 16 [-,%rdx] v13 = uload16.i32 v2 ; bin: heap_oob 0f b7 16
; asm: movswl (%rcx), %edi ; asm: movswl (%rcx), %edi
[-,%rdi] v14 = sload16.i32 v1 ; bin: 0f bf 39 [-,%rdi] v14 = sload16.i32 v1 ; bin: heap_oob 0f bf 39
; asm: movswl (%rsi), %edx ; asm: movswl (%rsi), %edx
[-,%rdx] v15 = sload16.i32 v2 ; bin: 0f bf 16 [-,%rdx] v15 = sload16.i32 v2 ; bin: heap_oob 0f bf 16
; asm: movzbl (%rcx), %edi ; asm: movzbl (%rcx), %edi
[-,%rdi] v16 = uload8.i32 v1 ; bin: 0f b6 39 [-,%rdi] v16 = uload8.i32 v1 ; bin: heap_oob 0f b6 39
; asm: movzbl (%rsi), %edx ; asm: movzbl (%rsi), %edx
[-,%rdx] v17 = uload8.i32 v2 ; bin: 0f b6 16 [-,%rdx] v17 = uload8.i32 v2 ; bin: heap_oob 0f b6 16
; asm: movsbl (%rcx), %edi ; asm: movsbl (%rcx), %edi
[-,%rdi] v18 = sload8.i32 v1 ; bin: 0f be 39 [-,%rdi] v18 = sload8.i32 v1 ; bin: heap_oob 0f be 39
; asm: movsbl (%rsi), %edx ; asm: movsbl (%rsi), %edx
[-,%rdx] v19 = sload8.i32 v2 ; bin: 0f be 16 [-,%rdx] v19 = sload8.i32 v2 ; bin: heap_oob 0f be 16
; Register-indirect with 8-bit signed displacement. ; Register-indirect with 8-bit signed displacement.
; asm: movl 50(%rcx), %edi ; asm: movl 50(%rcx), %edi
[-,%rdi] v20 = load.i32 v1+50 ; bin: 8b 79 32 [-,%rdi] v20 = load.i32 v1+50 ; bin: heap_oob 8b 79 32
; asm: movl -50(%rsi), %edx ; asm: movl -50(%rsi), %edx
[-,%rdx] v21 = load.i32 v2-50 ; bin: 8b 56 ce [-,%rdx] v21 = load.i32 v2-50 ; bin: heap_oob 8b 56 ce
; asm: movzwl 50(%rcx), %edi ; asm: movzwl 50(%rcx), %edi
[-,%rdi] v22 = uload16.i32 v1+50 ; bin: 0f b7 79 32 [-,%rdi] v22 = uload16.i32 v1+50 ; bin: heap_oob 0f b7 79 32
; asm: movzwl -50(%rsi), %edx ; asm: movzwl -50(%rsi), %edx
[-,%rdx] v23 = uload16.i32 v2-50 ; bin: 0f b7 56 ce [-,%rdx] v23 = uload16.i32 v2-50 ; bin: heap_oob 0f b7 56 ce
; asm: movswl 50(%rcx), %edi ; asm: movswl 50(%rcx), %edi
[-,%rdi] v24 = sload16.i32 v1+50 ; bin: 0f bf 79 32 [-,%rdi] v24 = sload16.i32 v1+50 ; bin: heap_oob 0f bf 79 32
; asm: movswl -50(%rsi), %edx ; asm: movswl -50(%rsi), %edx
[-,%rdx] v25 = sload16.i32 v2-50 ; bin: 0f bf 56 ce [-,%rdx] v25 = sload16.i32 v2-50 ; bin: heap_oob 0f bf 56 ce
; asm: movzbl 50(%rcx), %edi ; asm: movzbl 50(%rcx), %edi
[-,%rdi] v26 = uload8.i32 v1+50 ; bin: 0f b6 79 32 [-,%rdi] v26 = uload8.i32 v1+50 ; bin: heap_oob 0f b6 79 32
; asm: movzbl -50(%rsi), %edx ; asm: movzbl -50(%rsi), %edx
[-,%rdx] v27 = uload8.i32 v2-50 ; bin: 0f b6 56 ce [-,%rdx] v27 = uload8.i32 v2-50 ; bin: heap_oob 0f b6 56 ce
; asm: movsbl 50(%rcx), %edi ; asm: movsbl 50(%rcx), %edi
[-,%rdi] v28 = sload8.i32 v1+50 ; bin: 0f be 79 32 [-,%rdi] v28 = sload8.i32 v1+50 ; bin: heap_oob 0f be 79 32
; asm: movsbl -50(%rsi), %edx ; asm: movsbl -50(%rsi), %edx
[-,%rdx] v29 = sload8.i32 v2-50 ; bin: 0f be 56 ce [-,%rdx] v29 = sload8.i32 v2-50 ; bin: heap_oob 0f be 56 ce
; Register-indirect with 32-bit signed displacement. ; Register-indirect with 32-bit signed displacement.
; asm: movl 50000(%rcx), %edi ; asm: movl 50000(%rcx), %edi
[-,%rdi] v30 = load.i32 v1+50000 ; bin: 8b b9 0000c350 [-,%rdi] v30 = load.i32 v1+50000 ; bin: heap_oob 8b b9 0000c350
; asm: movl -50000(%rsi), %edx ; asm: movl -50000(%rsi), %edx
[-,%rdx] v31 = load.i32 v2-50000 ; bin: 8b 96 ffff3cb0 [-,%rdx] v31 = load.i32 v2-50000 ; bin: heap_oob 8b 96 ffff3cb0
; asm: movzwl 50000(%rcx), %edi ; asm: movzwl 50000(%rcx), %edi
[-,%rdi] v32 = uload16.i32 v1+50000 ; bin: 0f b7 b9 0000c350 [-,%rdi] v32 = uload16.i32 v1+50000 ; bin: heap_oob 0f b7 b9 0000c350
; asm: movzwl -50000(%rsi), %edx ; asm: movzwl -50000(%rsi), %edx
[-,%rdx] v33 = uload16.i32 v2-50000 ; bin: 0f b7 96 ffff3cb0 [-,%rdx] v33 = uload16.i32 v2-50000 ; bin: heap_oob 0f b7 96 ffff3cb0
; asm: movswl 50000(%rcx), %edi ; asm: movswl 50000(%rcx), %edi
[-,%rdi] v34 = sload16.i32 v1+50000 ; bin: 0f bf b9 0000c350 [-,%rdi] v34 = sload16.i32 v1+50000 ; bin: heap_oob 0f bf b9 0000c350
; asm: movswl -50000(%rsi), %edx ; asm: movswl -50000(%rsi), %edx
[-,%rdx] v35 = sload16.i32 v2-50000 ; bin: 0f bf 96 ffff3cb0 [-,%rdx] v35 = sload16.i32 v2-50000 ; bin: heap_oob 0f bf 96 ffff3cb0
; asm: movzbl 50000(%rcx), %edi ; asm: movzbl 50000(%rcx), %edi
[-,%rdi] v36 = uload8.i32 v1+50000 ; bin: 0f b6 b9 0000c350 [-,%rdi] v36 = uload8.i32 v1+50000 ; bin: heap_oob 0f b6 b9 0000c350
; asm: movzbl -50000(%rsi), %edx ; asm: movzbl -50000(%rsi), %edx
[-,%rdx] v37 = uload8.i32 v2-50000 ; bin: 0f b6 96 ffff3cb0 [-,%rdx] v37 = uload8.i32 v2-50000 ; bin: heap_oob 0f b6 96 ffff3cb0
; asm: movsbl 50000(%rcx), %edi ; asm: movsbl 50000(%rcx), %edi
[-,%rdi] v38 = sload8.i32 v1+50000 ; bin: 0f be b9 0000c350 [-,%rdi] v38 = sload8.i32 v1+50000 ; bin: heap_oob 0f be b9 0000c350
; asm: movsbl -50000(%rsi), %edx ; asm: movsbl -50000(%rsi), %edx
[-,%rdx] v39 = sload8.i32 v2-50000 ; bin: 0f be 96 ffff3cb0 [-,%rdx] v39 = sload8.i32 v2-50000 ; bin: heap_oob 0f be 96 ffff3cb0
; Integer Register-Register Operations. ; Integer Register-Register Operations.
@@ -903,17 +932,17 @@ ebb0:
[-,%rax] v160 = iconst.i32 1 [-,%rax] v160 = iconst.i32 1
[-,%rdx] v161 = iconst.i32 2 [-,%rdx] v161 = iconst.i32 2
; asm: idivl %ecx ; asm: idivl %ecx
[-,%rax,%rdx] v162, v163 = x86_sdivmodx v160, v161, v1 ; bin: f7 f9 [-,%rax,%rdx] v162, v163 = x86_sdivmodx v160, v161, v1 ; bin: int_divz f7 f9
; asm: idivl %esi ; asm: idivl %esi
[-,%rax,%rdx] v164, v165 = x86_sdivmodx v160, v161, v2 ; bin: f7 fe [-,%rax,%rdx] v164, v165 = x86_sdivmodx v160, v161, v2 ; bin: int_divz f7 fe
; asm: idivl %r10d ; asm: idivl %r10d
[-,%rax,%rdx] v166, v167 = x86_sdivmodx v160, v161, v3 ; bin: 41 f7 fa [-,%rax,%rdx] v166, v167 = x86_sdivmodx v160, v161, v3 ; bin: int_divz 41 f7 fa
; asm: divl %ecx ; asm: divl %ecx
[-,%rax,%rdx] v168, v169 = x86_udivmodx v160, v161, v1 ; bin: f7 f1 [-,%rax,%rdx] v168, v169 = x86_udivmodx v160, v161, v1 ; bin: int_divz f7 f1
; asm: divl %esi ; asm: divl %esi
[-,%rax,%rdx] v170, v171 = x86_udivmodx v160, v161, v2 ; bin: f7 f6 [-,%rax,%rdx] v170, v171 = x86_udivmodx v160, v161, v2 ; bin: int_divz f7 f6
; asm: divl %r10d ; asm: divl %r10d
[-,%rax,%rdx] v172, v173 = x86_udivmodx v160, v161, v3 ; bin: 41 f7 f2 [-,%rax,%rdx] v172, v173 = x86_udivmodx v160, v161, v3 ; bin: int_divz 41 f7 f2
; Bit-counting instructions. ; Bit-counting instructions.
@@ -1010,6 +1039,14 @@ ebb0:
; asm: setbe %dl ; asm: setbe %dl
[-,%rdx] v319 = icmp ule v2, v3 ; bin: 44 39 d6 0f 96 c2 [-,%rdx] v319 = icmp ule v2, v3 ; bin: 44 39 d6 0f 96 c2
; asm: cmpl $37, %ecx
; asm: setl %bl
[-,%rbx] v320 = icmp_imm slt v1, 37 ; bin: 83 f9 25 0f 9c c3
; asm: cmpq $100000, %ecx
; asm: setl %bl
[-,%rbx] v321 = icmp_imm slt v1, 100000 ; bin: 81 f9 000186a0 0f 9c c3
; Bool-to-int conversions. ; Bool-to-int conversions.
; asm: movzbl %bl, %ecx ; asm: movzbl %bl, %ecx
@@ -1039,19 +1076,32 @@ ebb0:
regfill v1, ss1 -> %rcx ; bin: 8b 8c 24 00000408 regfill v1, ss1 -> %rcx ; bin: 8b 8c 24 00000408
; asm: cmpl %esi, %ecx ; asm: cmpl %esi, %ecx
[-,%eflags] v520 = ifcmp v1, v2 ; bin: 39 f1 [-,%rflags] v520 = ifcmp v1, v2 ; bin: 39 f1
; asm: cmpl %r10d, %esi ; asm: cmpl %r10d, %esi
[-,%eflags] v521 = ifcmp v2, v3 ; bin: 44 39 d6 [-,%rflags] v521 = ifcmp v2, v3 ; bin: 44 39 d6
; asm: cmpl $-100, %ecx ; asm: cmpl $-100, %ecx
[-,%eflags] v522 = ifcmp_imm v1, -100 ; bin: 83 f9 9c [-,%rflags] v522 = ifcmp_imm v1, -100 ; bin: 83 f9 9c
; asm: cmpl $100, %r10d ; asm: cmpl $100, %r10d
[-,%eflags] v523 = ifcmp_imm v3, 100 ; bin: 41 83 fa 64 [-,%rflags] v523 = ifcmp_imm v3, 100 ; bin: 41 83 fa 64
; asm: cmpl $-10000, %ecx ; asm: cmpl $-10000, %ecx
[-,%eflags] v524 = ifcmp_imm v1, -10000 ; bin: 81 f9 ffffd8f0 [-,%rflags] v524 = ifcmp_imm v1, -10000 ; bin: 81 f9 ffffd8f0
; asm: cmpl $10000, %r10d ; asm: cmpl $10000, %r10d
[-,%eflags] v525 = ifcmp_imm v3, 10000 ; bin: 41 81 fa 00002710 [-,%rflags] v525 = ifcmp_imm v3, 10000 ; bin: 41 81 fa 00002710
; asm: shll $2, %esi
[-,%rsi] v526 = ishl_imm v2, 2 ; bin: c1 e6 02
; asm: shll $12, %r10d
[-,%r10] v527 = ishl_imm v3, 12 ; bin: 41 c1 e2 0c
; asm: sarl $5, %esi
[-,%rsi] v529 = sshr_imm v2, 5 ; bin: c1 fe 05
; asm: sarl $32, %r10d
[-,%r10] v530 = sshr_imm v3, 32 ; bin: 41 c1 fa 20
; asm: shrl $8, %esi
[-,%rsi] v532 = ushr_imm v2, 8 ; bin: c1 ee 08
; asm: shrl $31, %r10d
[-,%r10] v533 = ushr_imm v3, 31 ; bin: 41 c1 ea 1f
; asm: testl %ecx, %ecx ; asm: testl %ecx, %ecx
; asm: je ebb1x ; asm: je ebb1x
@@ -1082,6 +1132,7 @@ ebb1:
; asm: ebb2x: ; asm: ebb2x:
ebb2: ebb2:
jump ebb1 ; bin: eb fd jump ebb1 ; bin: eb fd
} }
; Tests for i32/i8 conversion instructions. ; Tests for i32/i8 conversion instructions.
@@ -1109,7 +1160,7 @@ ebb0:
; asm: movzbl %r10b, %ecx ; asm: movzbl %r10b, %ecx
[-,%rcx] v32 = uextend.i32 v13 ; bin: 41 0f b6 ca [-,%rcx] v32 = uextend.i32 v13 ; bin: 41 0f b6 ca
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Tests for i32/i16 conversion instructions. ; Tests for i32/i16 conversion instructions.
@@ -1137,7 +1188,7 @@ ebb0:
; asm: movzwl %r10w, %ecx ; asm: movzwl %r10w, %ecx
[-,%rcx] v32 = uextend.i32 v13 ; bin: 41 0f b7 ca [-,%rcx] v32 = uextend.i32 v13 ; bin: 41 0f b7 ca
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Tests for i64/i8 conversion instructions. ; Tests for i64/i8 conversion instructions.
@@ -1165,7 +1216,7 @@ ebb0:
; asm: movzbl %r10b, %ecx ; asm: movzbl %r10b, %ecx
[-,%rcx] v32 = uextend.i64 v13 ; bin: 41 0f b6 ca [-,%rcx] v32 = uextend.i64 v13 ; bin: 41 0f b6 ca
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Tests for i64/i16 conversion instructions. ; Tests for i64/i16 conversion instructions.
@@ -1193,7 +1244,7 @@ ebb0:
; asm: movzwl %r10w, %ecx ; asm: movzwl %r10w, %ecx
[-,%rcx] v32 = uextend.i64 v13 ; bin: 41 0f b7 ca [-,%rcx] v32 = uextend.i64 v13 ; bin: 41 0f b7 ca
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }
; Tests for i64/i32 conversion instructions. ; Tests for i64/i32 conversion instructions.
@@ -1221,5 +1272,5 @@ ebb0:
; asm: movl %r10d, %ecx ; asm: movl %r10d, %ecx
[-,%rcx] v32 = uextend.i64 v13 ; bin: 44 89 d1 [-,%rcx] v32 = uextend.i64 v13 ; bin: 44 89 d1
trap user0 ; bin: 0f 0b trap user0 ; bin: user0 0f 0b
} }

4
cranelift/filetests/isa/intel/legalize-div-traps.cton
View File

@@ -40,7 +40,7 @@ ebb0(v0: i64, v1: i64):
; nextln: brif eq $fm1, $(m1=$EBB) ; nextln: brif eq $fm1, $(m1=$EBB)
; nextln: $(fz=$V) = ifcmp_imm v1, 0 ; nextln: $(fz=$V) = ifcmp_imm v1, 0
; nextln: trapif eq $fz, int_divz ; nextln: trapif eq $fz, int_divz
; check: $(hi=$V) = sshr ; check: $(hi=$V) = sshr_imm
; nextln: $(q=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1 ; nextln: $(q=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1
; nextln: jump $(done=$EBB)($q) ; nextln: jump $(done=$EBB)($q)
; check: $m1: ; check: $m1:
@@ -60,7 +60,7 @@ ebb0(v0: i64, v1: i64):
v2 = srem v0, v1 v2 = srem v0, v1
; nextln: $(fm1=$V) = ifcmp_imm v1, -1 ; nextln: $(fm1=$V) = ifcmp_imm v1, -1
; nextln: brif eq $fm1, $(m1=$EBB) ; nextln: brif eq $fm1, $(m1=$EBB)
; check: $(hi=$V) = sshr ; check: $(hi=$V) = sshr_imm
; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1 ; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1
; nextln: jump $(done=$EBB)($r) ; nextln: jump $(done=$EBB)($r)
; check: $m1: ; check: $m1:

4
cranelift/filetests/isa/intel/legalize-div.cton
View File

@@ -32,7 +32,7 @@ function %sdiv(i64, i64) -> i64 {
ebb0(v0: i64, v1: i64): ebb0(v0: i64, v1: i64):
; check: ebb0( ; check: ebb0(
v2 = sdiv v0, v1 v2 = sdiv v0, v1
; check: $(hi=$V) = sshr ; check: $(hi=$V) = sshr_imm
; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1 ; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1
return v2 return v2
; nextln: return $d ; nextln: return $d
@@ -46,7 +46,7 @@ ebb0(v0: i64, v1: i64):
v2 = srem v0, v1 v2 = srem v0, v1
; nextln: $(fm1=$V) = ifcmp_imm v1, -1 ; nextln: $(fm1=$V) = ifcmp_imm v1, -1
; nextln: brif eq $fm1, $(m1=$EBB) ; nextln: brif eq $fm1, $(m1=$EBB)
; check: $(hi=$V) = sshr ; check: $(hi=$V) = sshr_imm
; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1 ; nextln: $(d=$V), $(r=$V) = x86_sdivmodx v0, $hi, v1
; nextln: jump $(done=$EBB)($r) ; nextln: jump $(done=$EBB)($r)
; check: $m1: ; check: $m1:

4
cranelift/filetests/isa/intel/legalize-libcall.cton
View File

@@ -9,7 +9,7 @@ ebb0(v0: f32):
v1 = floor v0 v1 = floor v0
return v1 return v1
} }
; check: function %floor(f32 [%xmm0]) -> f32 [%xmm0] native { ; check: function %floor(f32 [%xmm0]) -> f32 [%xmm0] system_v {
; check: sig0 = (f32) -> f32 native ; check: sig0 = (f32) -> f32 system_v
; check: fn0 = sig0 %FloorF32 ; check: fn0 = sig0 %FloorF32
; check: v1 = call fn0(v0) ; check: v1 = call fn0(v0)

6
cranelift/filetests/isa/intel/legalize-memory.cton
View File

@@ -23,7 +23,7 @@ function %deref(i64 vmctx) -> i64 {
ebb1(v1: i64): ebb1(v1: i64):
v2 = global_addr.i64 gv2 v2 = global_addr.i64 gv2
; check: $(a1=$V) = iadd_imm v1, -16 ; check: $(a1=$V) = iadd_imm v1, -16
; check: $(p1=$V) = load.i64 $a1 ; check: $(p1=$V) = load.i64 notrap aligned $a1
; check: v2 = iadd_imm $p1, 32 ; check: v2 = iadd_imm $p1, 32
return v2 return v2
; check: return v2 ; check: return v2
@@ -55,7 +55,7 @@ ebb0(v0: i32, v999: i64):
; Checks here are assuming that no pipehole opts fold the load offsets. ; Checks here are assuming that no pipehole opts fold the load offsets.
; nextln: $(xoff=$V) = uextend.i64 v0 ; nextln: $(xoff=$V) = uextend.i64 v0
; nextln: $(haddr=$V) = iadd_imm v999, 64 ; nextln: $(haddr=$V) = iadd_imm v999, 64
; nextln: $(hbase=$V) = load.i64 $haddr ; nextln: $(hbase=$V) = load.i64 notrap aligned $haddr
; nextln: v1 = iadd $hbase, $xoff ; nextln: v1 = iadd $hbase, $xoff
v2 = load.f32 v1+16 v2 = load.f32 v1+16
; nextln: v2 = load.f32 v1+16 ; nextln: v2 = load.f32 v1+16
@@ -103,7 +103,7 @@ ebb0(v0: i32, v999: i64):
; Checks here are assuming that no pipehole opts fold the load offsets. ; Checks here are assuming that no pipehole opts fold the load offsets.
; nextln: $(xoff=$V) = uextend.i64 v0 ; nextln: $(xoff=$V) = uextend.i64 v0
; nextln: $(haddr=$V) = iadd_imm.i64 v999, 64 ; nextln: $(haddr=$V) = iadd_imm.i64 v999, 64
; nextln: $(hbase=$V) = load.i64 $haddr ; nextln: $(hbase=$V) = load.i64 notrap aligned $haddr
; nextln: v1 = iadd $hbase, $xoff ; nextln: v1 = iadd $hbase, $xoff
v2 = load.f32 v1+0x7fff_ffff v2 = load.f32 v1+0x7fff_ffff
; nextln: v2 = load.f32 v1+0x7fff_ffff ; nextln: v2 = load.f32 v1+0x7fff_ffff

2
cranelift/filetests/isa/intel/prologue-epilogue.cton
View File

@@ -9,7 +9,7 @@ ebb0:
return return
} }
; check: function %foo(i64 fp [%rbp], i64 csr [%rbx], i64 csr [%r12], i64 csr [%r13], i64 csr [%r14], i64 csr [%r15]) -> i64 fp [%rbp], i64 csr [%rbx], i64 csr [%r12], i64 csr [%r13], i64 csr [%r14], i64 csr [%r15] native { ; check: function %foo(i64 fp [%rbp], i64 csr [%rbx], i64 csr [%r12], i64 csr [%r13], i64 csr [%r14], i64 csr [%r15]) -> i64 fp [%rbp], i64 csr [%rbx], i64 csr [%r12], i64 csr [%r13], i64 csr [%r14], i64 csr [%r15] system_v {
; nextln: ss0 = explicit_slot 168, offset -224 ; nextln: ss0 = explicit_slot 168, offset -224
; nextln: ss1 = incoming_arg 56, offset -56 ; nextln: ss1 = incoming_arg 56, offset -56
; check: ebb0(v0: i64 [%rbp], v1: i64 [%rbx], v2: i64 [%r12], v3: i64 [%r13], v4: i64 [%r14], v5: i64 [%r15]): ; check: ebb0(v0: i64 [%rbp], v1: i64 [%rbx], v2: i64 [%r12], v3: i64 [%r13], v4: i64 [%r14], v5: i64 [%r15]):

4
cranelift/filetests/isa/riscv/abi-e.cton
View File

@@ -7,8 +7,8 @@ isa riscv enable_e
function %f() { function %f() {
; Spilling into the stack args after %x15 since %16 and up are not ; Spilling into the stack args after %x15 since %16 and up are not
; available in RV32E. ; available in RV32E.
sig0 = (i64, i64, i64, i64) -> i64 native sig0 = (i64, i64, i64, i64) -> i64 system_v
; check: sig0 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13], i32 [%x14], i32 [%x15], i32 [0], i32 [4]) -> i32 [%x10], i32 [%x11] native ; check: sig0 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13], i32 [%x14], i32 [%x15], i32 [0], i32 [4]) -> i32 [%x10], i32 [%x11] system_v
ebb0: ebb0:
return return
} }

24
cranelift/filetests/isa/riscv/abi.cton
View File

@@ -5,27 +5,27 @@ isa riscv
; regex: V=v\d+ ; regex: V=v\d+
function %f() { function %f() {
sig0 = (i32) -> i32 native sig0 = (i32) -> i32 system_v
; check: sig0 = (i32 [%x10]) -> i32 [%x10] native ; check: sig0 = (i32 [%x10]) -> i32 [%x10] system_v
sig1 = (i64) -> b1 native sig1 = (i64) -> b1 system_v
; check: sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] native ; check: sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] system_v
; The i64 argument must go in an even-odd register pair. ; The i64 argument must go in an even-odd register pair.
sig2 = (f32, i64) -> f64 native sig2 = (f32, i64) -> f64 system_v
; check: sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] native ; check: sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] system_v
; Spilling into the stack args. ; Spilling into the stack args.
sig3 = (f64, f64, f64, f64, f64, f64, f64, i64) -> f64 native sig3 = (f64, f64, f64, f64, f64, f64, f64, i64) -> f64 system_v
; check: sig3 = (f64 [%f10], f64 [%f11], f64 [%f12], f64 [%f13], f64 [%f14], f64 [%f15], f64 [%f16], i32 [0], i32 [4]) -> f64 [%f10] native ; check: sig3 = (f64 [%f10], f64 [%f11], f64 [%f12], f64 [%f13], f64 [%f14], f64 [%f15], f64 [%f16], i32 [0], i32 [4]) -> f64 [%f10] system_v
; Splitting vectors. ; Splitting vectors.
sig4 = (i32x4) native sig4 = (i32x4) system_v
; check: sig4 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13]) native ; check: sig4 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13]) system_v
; Splitting vectors, then splitting ints. ; Splitting vectors, then splitting ints.
sig5 = (i64x4) native sig5 = (i64x4) system_v
; check: sig5 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13], i32 [%x14], i32 [%x15], i32 [%x16], i32 [%x17]) native ; check: sig5 = (i32 [%x10], i32 [%x11], i32 [%x12], i32 [%x13], i32 [%x14], i32 [%x15], i32 [%x16], i32 [%x17]) system_v
ebb0: ebb0:
return return

4
cranelift/filetests/isa/riscv/legalize-abi.cton
View File

@@ -106,7 +106,7 @@ ebb0(v0: i64x4):
} }
function %indirect(i32) { function %indirect(i32) {
sig1 = () native sig1 = () system_v
ebb0(v0: i32): ebb0(v0: i32):
call_indirect sig1, v0() call_indirect sig1, v0()
return return
@@ -114,7 +114,7 @@ ebb0(v0: i32):
; The first argument to call_indirect doesn't get altered. ; The first argument to call_indirect doesn't get altered.
function %indirect_arg(i32, f32x2) { function %indirect_arg(i32, f32x2) {
sig1 = (f32x2) native sig1 = (f32x2) system_v
ebb0(v0: i32, v1: f32x2): ebb0(v0: i32, v1: f32x2):
call_indirect sig1, v0(v1) call_indirect sig1, v0(v1)
; check: call_indirect sig1, v0($V, $V) ; check: call_indirect sig1, v0($V, $V)

30
cranelift/filetests/isa/riscv/parse-encoding.cton
View File

@@ -3,32 +3,32 @@ test legalizer
isa riscv isa riscv
function %parse_encoding(i32 [%x5]) -> i32 [%x10] { function %parse_encoding(i32 [%x5]) -> i32 [%x10] {
; check: function %parse_encoding(i32 [%x5], i32 link [%x1]) -> i32 [%x10], i32 link [%x1] native { ; check: function %parse_encoding(i32 [%x5], i32 link [%x1]) -> i32 [%x10], i32 link [%x1] system_v {
sig0 = (i32 [%x10]) -> i32 [%x10] native sig0 = (i32 [%x10]) -> i32 [%x10] system_v
; check: sig0 = (i32 [%x10]) -> i32 [%x10] native ; check: sig0 = (i32 [%x10]) -> i32 [%x10] system_v
sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] native sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] system_v
; check: sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] native ; check: sig1 = (i32 [%x10], i32 [%x11]) -> b1 [%x10] system_v
sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] native sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] system_v
; check: sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] native ; check: sig2 = (f32 [%f10], i32 [%x12], i32 [%x13]) -> f64 [%f10] system_v
; Arguments on stack where not necessary ; Arguments on stack where not necessary
sig3 = (f64 [%f10], i32 [0], i32 [4]) -> f64 [%f10] native sig3 = (f64 [%f10], i32 [0], i32 [4]) -> f64 [%f10] system_v
; check: sig3 = (f64 [%f10], i32 [0], i32 [4]) -> f64 [%f10] native ; check: sig3 = (f64 [%f10], i32 [0], i32 [4]) -> f64 [%f10] system_v
; Stack argument before register argument ; Stack argument before register argument
sig4 = (f32 [72], i32 [%x10]) native sig4 = (f32 [72], i32 [%x10]) system_v
; check: sig4 = (f32 [72], i32 [%x10]) native ; check: sig4 = (f32 [72], i32 [%x10]) system_v
; Return value on stack ; Return value on stack
sig5 = () -> f32 [0] native sig5 = () -> f32 [0] system_v
; check: sig5 = () -> f32 [0] native ; check: sig5 = () -> f32 [0] system_v
; function + signature ; function + signature
fn0 = function %bar(i32 [%x10]) -> b1 [%x10] native fn0 = function %bar(i32 [%x10]) -> b1 [%x10] system_v
; check: sig6 = (i32 [%x10]) -> b1 [%x10] native ; check: sig6 = (i32 [%x10]) -> b1 [%x10] system_v
; nextln: fn0 = sig6 %bar ; nextln: fn0 = sig6 %bar
ebb0(v0: i32): ebb0(v0: i32):

2
cranelift/filetests/licm/complex.cton
View File

@@ -1,6 +1,6 @@
test licm test licm
function %complex(i32) -> i32 native { function %complex(i32) -> i32 system_v {
ebb0(v0: i32): ebb0(v0: i32):
jump ebb1(v0) jump ebb1(v0)

81
cranelift/filetests/licm/reject.cton Normal file
View File

@@ -0,0 +1,81 @@
test licm
function %other_side_effects(i32) -> i32 {
ebb0(v0: i32):
jump ebb1(v0)
ebb1(v1: i32):
regmove.i32 v0, %10 -> %20
; check: ebb1(v1: i32):
; check: regmove.i32 v0, %10 -> %20
v2 = iconst.i32 1
brz v1, ebb2(v1)
v5 = isub v1, v2
jump ebb1(v5)
ebb2(v6: i32):
return v6
}
function %cpu_flags(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
jump ebb1(v0, v1)
ebb1(v2: i32, v3: i32):
v4 = ifcmp.i32 v0, v1
v5 = selectif.i32 eq v4, v2, v3
; check: ebb1(v2: i32, v3: i32):
; check: ifcmp.i32 v0, v1
; check: v5 = selectif.i32 eq v4, v2, v3
v8 = iconst.i32 1
brz v1, ebb2(v1)
v9 = isub v1, v8
v10 = iadd v1, v8
jump ebb1(v9, v10)
ebb2(v6: i32):
return v6
}
function %spill(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
v2 = spill.i32 v0
jump ebb1(v0, v1)
ebb1(v3: i32, v4: i32):
v5 = spill.i32 v1
v6 = fill.i32 v2
v7 = fill.i32 v5
; check: ebb1(v3: i32, v4: i32):
; check: v5 = spill.i32 v1
; check: v6 = fill.i32 v2
; check: v7 = fill v5
brz v1, ebb2(v1)
v9 = isub v1, v4
jump ebb1(v9, v3)
ebb2(v10: i32):
return v10
}
function %non_invariant_aliases(i32) -> i32 {
ebb0(v0: i32):
jump ebb1(v0)
ebb1(v1: i32):
v8 -> v1
v9 -> v1
v2 = iadd v8, v9
; check: ebb1(v1: i32):
; check: v2 = iadd v8, v9
brz v1, ebb2(v1)
v5 = isub v1, v2
jump ebb1(v5)
ebb2(v6: i32):
return v6
}

12
cranelift/filetests/parser/branch.cton
View File

@@ -9,7 +9,7 @@ ebb0:
ebb1: ebb1:
jump ebb0() jump ebb0()
} }
; sameln: function %minimal() native { ; sameln: function %minimal() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: jump ebb1 ; nextln: jump ebb1
; nextln: ; nextln:
@@ -25,7 +25,7 @@ ebb0(v90: i32):
ebb1(v91: i32): ebb1(v91: i32):
jump ebb0(v91) jump ebb0(v91)
} }
; sameln: function %onearg(i32) native { ; sameln: function %onearg(i32) system_v {
; nextln: ebb0(v90: i32): ; nextln: ebb0(v90: i32):
; nextln: jump ebb1(v90) ; nextln: jump ebb1(v90)
; nextln: ; nextln:
@@ -41,7 +41,7 @@ ebb0(v90: i32, v91: f32):
ebb1(v92: i32, v93: f32): ebb1(v92: i32, v93: f32):
jump ebb0(v92, v93) jump ebb0(v92, v93)
} }
; sameln: function %twoargs(i32, f32) native { ; sameln: function %twoargs(i32, f32) system_v {
; nextln: ebb0(v90: i32, v91: f32): ; nextln: ebb0(v90: i32, v91: f32):
; nextln: jump ebb1(v90, v91) ; nextln: jump ebb1(v90, v91)
; nextln: ; nextln:
@@ -57,7 +57,7 @@ ebb0(v90: i32):
ebb1: ebb1:
brnz v90, ebb1() brnz v90, ebb1()
} }
; sameln: function %minimal(i32) native { ; sameln: function %minimal(i32) system_v {
; nextln: ebb0(v90: i32): ; nextln: ebb0(v90: i32):
; nextln: brz v90, ebb1 ; nextln: brz v90, ebb1
; nextln: ; nextln:
@@ -72,7 +72,7 @@ ebb0(v90: i32, v91: f32):
ebb1(v92: i32, v93: f32): ebb1(v92: i32, v93: f32):
brnz v90, ebb0(v92, v93) brnz v90, ebb0(v92, v93)
} }
; sameln: function %twoargs(i32, f32) native { ; sameln: function %twoargs(i32, f32) system_v {
; nextln: ebb0(v90: i32, v91: f32): ; nextln: ebb0(v90: i32, v91: f32):
; nextln: brz v90, ebb1(v90, v91) ; nextln: brz v90, ebb1(v90, v91)
; nextln: ; nextln:
@@ -94,7 +94,7 @@ ebb30:
ebb40: ebb40:
trap user4 trap user4
} }
; sameln: function %jumptable(i32) native { ; sameln: function %jumptable(i32) system_v {
; check: jt2 = jump_table 0, 0, ebb10, ebb40, ebb20, ebb30 ; check: jt2 = jump_table 0, 0, ebb10, ebb40, ebb20, ebb30
; check: jt200 = jump_table 0 ; check: jt200 = jump_table 0
; check: ebb10(v3: i32): ; check: ebb10(v3: i32):

10
cranelift/filetests/parser/call.cton
View File

@@ -5,7 +5,7 @@ function %mini() {
ebb1: ebb1:
return return
} }
; sameln: function %mini() native { ; sameln: function %mini() system_v {
; nextln: ebb1: ; nextln: ebb1:
; nextln: return ; nextln: return
; nextln: } ; nextln: }
@@ -29,10 +29,10 @@ function %signatures() {
fn5 = sig11 %foo fn5 = sig11 %foo
fn8 = function %bar(i32) -> b1 fn8 = function %bar(i32) -> b1
} }
; sameln: function %signatures() native { ; sameln: function %signatures() system_v {
; check: sig10 = () native ; check: sig10 = () system_v
; check: sig11 = (i32, f64) -> i32, b1 spiderwasm ; check: sig11 = (i32, f64) -> i32, b1 spiderwasm
; check: sig12 = (i32) -> b1 native ; check: sig12 = (i32) -> b1 system_v
; not: fn0 ; not: fn0
; check: fn5 = sig11 %foo ; check: fn5 = sig11 %foo
; check: fn8 = sig12 %bar ; check: fn8 = sig12 %bar
@@ -88,7 +88,7 @@ function %special1(i32 sret, i32 fp, i32 csr, i32 link) -> i32 link, i32 fp, i32
ebb0(v1: i32, v2: i32, v3: i32, v4: i32): ebb0(v1: i32, v2: i32, v3: i32, v4: i32):
return v4, v2, v3, v1 return v4, v2, v3, v1
} }
; check: function %special1(i32 sret, i32 fp, i32 csr, i32 link) -> i32 link, i32 fp, i32 csr, i32 sret native { ; check: function %special1(i32 sret, i32 fp, i32 csr, i32 link) -> i32 link, i32 fp, i32 csr, i32 sret system_v {
; check: ebb0(v1: i32, v2: i32, v3: i32, v4: i32): ; check: ebb0(v1: i32, v2: i32, v3: i32, v4: i32):
; check: return v4, v2, v3, v1 ; check: return v4, v2, v3, v1
; check: } ; check: }

2
cranelift/filetests/parser/instruction_encoding.cton
View File

@@ -13,7 +13,7 @@ ebb1(v0: i32 [%x8], v1: i32):
@55 v9 = iadd v8, v7 @55 v9 = iadd v8, v7
@a5 [Iret#5] return v0, v8 @a5 [Iret#5] return v0, v8
} }
; sameln: function %foo(i32, i32) native { ; sameln: function %foo(i32, i32) system_v {
; nextln: ebb1(v0: i32 [%x8], v1: i32): ; nextln: ebb1(v0: i32 [%x8], v1: i32):
; nextln: [-,-]$WS v2 = iadd v0, v1 ; nextln: [-,-]$WS v2 = iadd v0, v1
; nextln: [-]$WS trap heap_oob ; nextln: [-]$WS trap heap_oob

2
cranelift/filetests/parser/keywords.cton
View File

@@ -2,4 +2,4 @@ test cat
; 'function' is not a keyword, and can be used as the name of a function too. ; 'function' is not a keyword, and can be used as the name of a function too.
function %function() {} function %function() {}
; check: function %function() native ; check: function %function() system_v

4
cranelift/filetests/parser/rewrite.cton
View File

@@ -9,7 +9,7 @@ ebb100(v20: i32):
v9200 = f64const 0x4.0p0 v9200 = f64const 0x4.0p0
trap user4 trap user4
} }
; sameln: function %defs() native { ; sameln: function %defs() system_v {
; nextln: ebb100(v20: i32): ; nextln: ebb100(v20: i32):
; nextln: v1000 = iconst.i32x8 5 ; nextln: v1000 = iconst.i32x8 5
; nextln: v9200 = f64const 0x1.0000000000000p2 ; nextln: v9200 = f64const 0x1.0000000000000p2
@@ -23,7 +23,7 @@ ebb100(v20: i32):
v200 = iadd v20, v1000 v200 = iadd v20, v1000
jump ebb100(v1000) jump ebb100(v1000)
} }
; sameln: function %use_value() native { ; sameln: function %use_value() system_v {
; nextln: ebb100(v20: i32): ; nextln: ebb100(v20: i32):
; nextln: v1000 = iadd_imm v20, 5 ; nextln: v1000 = iadd_imm v20, 5
; nextln: v200 = iadd v20, v1000 ; nextln: v200 = iadd v20, v1000

28
cranelift/filetests/parser/tiny.cton
View File

@@ -5,7 +5,7 @@ function %minimal() {
ebb0: ebb0:
trap user0 trap user0
} }
; sameln: function %minimal() native { ; sameln: function %minimal() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: trap user0 ; nextln: trap user0
; nextln: } ; nextln: }
@@ -18,7 +18,7 @@ ebb0:
v1 = iconst.i8 6 v1 = iconst.i8 6
v2 = ishl v0, v1 v2 = ishl v0, v1
} }
; sameln: function %ivalues() native { ; sameln: function %ivalues() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: v0 = iconst.i32 2 ; nextln: v0 = iconst.i32 2
; nextln: v1 = iconst.i8 6 ; nextln: v1 = iconst.i8 6
@@ -34,7 +34,7 @@ ebb0:
v2 = bextend.b32 v1 v2 = bextend.b32 v1
v3 = bxor v0, v2 v3 = bxor v0, v2
} }
; sameln: function %bvalues() native { ; sameln: function %bvalues() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: v0 = bconst.b32 true ; nextln: v0 = bconst.b32 true
; nextln: v1 = bconst.b8 false ; nextln: v1 = bconst.b8 false
@@ -47,17 +47,17 @@ function %select() {
ebb0(v90: i32, v91: i32, v92: b1): ebb0(v90: i32, v91: i32, v92: b1):
v0 = select v92, v90, v91 v0 = select v92, v90, v91
} }
; sameln: function %select() native { ; sameln: function %select() system_v {
; nextln: ebb0(v90: i32, v91: i32, v92: b1): ; nextln: ebb0(v90: i32, v91: i32, v92: b1):
; nextln: v0 = select v92, v90, v91 ; nextln: v0 = select v92, v90, v91
; nextln: } ; nextln: }
; Polymorphic instruction controlled by third operand. ; Polymorphic instruction controlled by third operand.
function %selectif() native { function %selectif() system_v {
ebb0(v95: i32, v96: i32, v97: b1): ebb0(v95: i32, v96: i32, v97: b1):
v98 = selectif.i32 eq v97, v95, v96 v98 = selectif.i32 eq v97, v95, v96
} }
; sameln: function %selectif() native { ; sameln: function %selectif() system_v {
; nextln: ebb0(v95: i32, v96: i32, v97: b1): ; nextln: ebb0(v95: i32, v96: i32, v97: b1):
; nextln: v98 = selectif.i32 eq v97, v95, v96 ; nextln: v98 = selectif.i32 eq v97, v95, v96
; nextln: } ; nextln: }
@@ -69,7 +69,7 @@ ebb0:
v1 = extractlane v0, 3 v1 = extractlane v0, 3
v2 = insertlane v0, 1, v1 v2 = insertlane v0, 1, v1
} }
; sameln: function %lanes() native { ; sameln: function %lanes() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: v0 = iconst.i32x4 2 ; nextln: v0 = iconst.i32x4 2
; nextln: v1 = extractlane v0, 3 ; nextln: v1 = extractlane v0, 3
@@ -85,7 +85,7 @@ ebb0(v90: i32, v91: i32):
v3 = irsub_imm v91, 45 v3 = irsub_imm v91, 45
br_icmp eq v90, v91, ebb0(v91, v90) br_icmp eq v90, v91, ebb0(v91, v90)
} }
; sameln: function %icmp(i32, i32) native { ; sameln: function %icmp(i32, i32) system_v {
; nextln: ebb0(v90: i32, v91: i32): ; nextln: ebb0(v90: i32, v91: i32):
; nextln: v0 = icmp eq v90, v91 ; nextln: v0 = icmp eq v90, v91
; nextln: v1 = icmp ult v90, v91 ; nextln: v1 = icmp ult v90, v91
@@ -101,7 +101,7 @@ ebb0(v90: f32, v91: f32):
v1 = fcmp uno v90, v91 v1 = fcmp uno v90, v91
v2 = fcmp lt v90, v91 v2 = fcmp lt v90, v91
} }
; sameln: function %fcmp(f32, f32) native { ; sameln: function %fcmp(f32, f32) system_v {
; nextln: ebb0(v90: f32, v91: f32): ; nextln: ebb0(v90: f32, v91: f32):
; nextln: v0 = fcmp eq v90, v91 ; nextln: v0 = fcmp eq v90, v91
; nextln: v1 = fcmp uno v90, v91 ; nextln: v1 = fcmp uno v90, v91
@@ -115,7 +115,7 @@ ebb0(v90: i32, v91: f32):
v0 = bitcast.i8x4 v90 v0 = bitcast.i8x4 v90
v1 = bitcast.i32 v91 v1 = bitcast.i32 v91
} }
; sameln: function %bitcast(i32, f32) native { ; sameln: function %bitcast(i32, f32) system_v {
; nextln: ebb0(v90: i32, v91: f32): ; nextln: ebb0(v90: i32, v91: f32):
; nextln: v0 = bitcast.i8x4 v90 ; nextln: v0 = bitcast.i8x4 v90
; nextln: v1 = bitcast.i32 v91 ; nextln: v1 = bitcast.i32 v91
@@ -135,7 +135,7 @@ ebb0:
stack_store v1, ss10+2 stack_store v1, ss10+2
stack_store v2, ss2 stack_store v2, ss2
} }
; sameln: function %stack() native { ; sameln: function %stack() system_v {
; check: ss2 = explicit_slot 4 ; check: ss2 = explicit_slot 4
; check: ss3 = incoming_arg 4, offset 8 ; check: ss3 = incoming_arg 4, offset 8
; check: ss4 = outgoing_arg 4 ; check: ss4 = outgoing_arg 4
@@ -162,7 +162,7 @@ ebb0(v1: i32):
store aligned v3, v1+12 store aligned v3, v1+12
store notrap aligned v3, v1-12 store notrap aligned v3, v1-12
} }
; sameln: function %memory(i32) native { ; sameln: function %memory(i32) system_v {
; nextln: ebb0(v1: i32): ; nextln: ebb0(v1: i32):
; nextln: v2 = load.i64 v1 ; nextln: v2 = load.i64 v1
; nextln: v3 = load.i64 aligned v1 ; nextln: v3 = load.i64 aligned v1
@@ -187,7 +187,7 @@ ebb0(v1: i32):
regfill v1, ss0 -> %10 regfill v1, ss0 -> %10
return return
} }
; sameln: function %diversion(i32) native { ; sameln: function %diversion(i32) system_v {
; nextln: ss0 = spill_slot 4 ; nextln: ss0 = spill_slot 4
; check: ebb0(v1: i32): ; check: ebb0(v1: i32):
; nextln: regmove v1, %10 -> %20 ; nextln: regmove v1, %10 -> %20
@@ -204,7 +204,7 @@ ebb0:
copy_special %20 -> %10 copy_special %20 -> %10
return return
} }
; sameln: function %copy_special() native { ; sameln: function %copy_special() system_v {
; nextln: ebb0: ; nextln: ebb0:
; nextln: copy_special %10 -> %20 ; nextln: copy_special %10 -> %20
; nextln: copy_special %20 -> %10 ; nextln: copy_special %20 -> %10

100
cranelift/filetests/postopt/basic.cton Normal file
View File

@@ -0,0 +1,100 @@
test postopt
isa intel
; Test that compare+branch sequences are folded effectively on x86.
function %br_icmp(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
[Op1icscc#39,%rdx] v2 = icmp slt v0, v1
[Op1t8jccd_long#85] brnz v2, ebb1
[Op1ret#c3] return v1
ebb1:
[Op1puid#b8,%rax] v8 = iconst.i32 3
[Op1ret#c3] return v8
}
; sameln: function %br_icmp
; nextln: ebb0(v0: i32, v1: i32):
; nextln: v9 = ifcmp v0, v1
; nextln: v2 = trueif slt v9
; nextln: brif slt v9, ebb1
; nextln: return v1
; nextln:
; nextln: ebb1:
; nextln: v8 = iconst.i32 3
; nextln: return v8
; nextln: }
; Use brz instead of brnz, so the condition is inverted.
function %br_icmp_inverse(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
[Op1icscc#39,%rdx] v2 = icmp slt v0, v1
[Op1t8jccd_long#84] brz v2, ebb1
[Op1ret#c3] return v1
ebb1:
[Op1puid#b8,%rax] v8 = iconst.i32 3
[Op1ret#c3] return v8
}
; sameln: function %br_icmp_inverse
; nextln: ebb0(v0: i32, v1: i32):
; nextln: v9 = ifcmp v0, v1
; nextln: v2 = trueif slt v9
; nextln: brif sge v9, ebb1
; nextln: return v1
; nextln:
; nextln: ebb1:
; nextln: v8 = iconst.i32 3
; nextln: return v8
; nextln: }
; Use icmp_imm instead of icmp.
function %br_icmp_imm(i32, i32) -> i32 {
ebb0(v0: i32, v1: i32):
[Op1icsccib#7083] v2 = icmp_imm slt v0, 2
[Op1t8jccd_long#84] brz v2, ebb1
[Op1ret#c3] return v1
ebb1:
[Op1puid#b8,%rax] v8 = iconst.i32 3
[Op1ret#c3] return v8
}
; sameln: function %br_icmp_imm
; nextln: ebb0(v0: i32, v1: i32):
; nextln: v9 = ifcmp_imm v0, 2
; nextln: v2 = trueif slt v9
; nextln: brif sge v9, ebb1
; nextln: return v1
; nextln:
; nextln: ebb1:
; nextln: v8 = iconst.i32 3
; nextln: return v8
; nextln: }
; Use fcmp instead of icmp.
function %br_fcmp(f32, f32) -> f32 {
ebb0(v0: f32, v1: f32):
[Op2fcscc#42e,%rdx] v2 = fcmp gt v0, v1
[Op1t8jccd_long#84] brz v2, ebb1
[Op1ret#c3] return v1
ebb1:
[Op1puid#b8,%rax] v18 = iconst.i32 0x40a8_0000
[Mp2frurm#56e,%xmm0] v8 = bitcast.f32 v18
[Op1ret#c3] return v8
}
; sameln: function %br_fcmp
; nextln: ebb0(v0: f32, v1: f32):
; nextln: v19 = ffcmp v0, v1
; nextln: v2 = trueff gt v19
; nextln: brff ule v19, ebb1
; nextln: return v1
; nextln:
; nextln: ebb1:
; nextln: v18 = iconst.i32 0x40a8_0000
; nextln: v8 = bitcast.f32 v18
; nextln: return v8
; nextln: }

80
cranelift/filetests/preopt/simplify.cton Normal file
View File

@@ -0,0 +1,80 @@
test preopt
isa intel
function %iadd_imm(i32) -> i32 {
ebb0(v0: i32):
v1 = iconst.i32 2
v2 = iadd v0, v1
return v2
}
; sameln: function %iadd_imm
; nextln: ebb0(v0: i32):
; nextln: v1 = iconst.i32 2
; nextln: v2 = iadd_imm v0, 2
; nextln: return v2
; nextln: }
function %isub_imm(i32) -> i32 {
ebb0(v0: i32):
v1 = iconst.i32 2
v2 = isub v0, v1
return v2
}
; sameln: function %isub_imm
; nextln: ebb0(v0: i32):
; nextln: v1 = iconst.i32 2
; nextln: v2 = iadd_imm v0, -2
; nextln: return v2
; nextln: }
function %icmp_imm(i32) -> i32 {
ebb0(v0: i32):
v1 = iconst.i32 2
v2 = icmp slt v0, v1
v3 = bint.i32 v2
return v3
}
; sameln: function %icmp_imm
; nextln: ebb0(v0: i32):
; nextln: v1 = iconst.i32 2
; nextln: v2 = icmp_imm slt v0, 2
; nextln: v3 = bint.i32 v2
; nextln: return v3
; nextln: }
function %brz_bint(i32) {
ebb0(v0: i32):
v3 = icmp_imm slt v0, 0
v1 = bint.i32 v3
v2 = select v1, v1, v1
trapz v1, user0
brz v1, ebb1
jump ebb2
ebb1:
return
ebb2:
return
}
; sameln: function %brz_bint
; nextln: (v0: i32):
; nextln: v3 = icmp_imm slt v0, 0
; nextln: v1 = bint.i32 v3
; nextln: v2 = select v3, v1, v1
; nextln: trapz v3, user0
; nextln: brz v3, ebb1
; nextln: jump ebb2
function %irsub_imm(i32) -> i32 {
ebb0(v0: i32):
v1 = iconst.i32 2
v2 = isub v1, v0
return v2
}
; sameln: function %irsub_imm
; nextln: ebb0(v0: i32):
; nextln: v1 = iconst.i32 2
; nextln: v2 = irsub_imm v1, 2
; nextln: return v2
; nextln: }

2
cranelift/filetests/regalloc/coalesce.cton
View File

@@ -109,7 +109,7 @@ ebb1(v10: i32):
return v11 return v11
} }
function %gvn_unremovable_phi(i32) native { function %gvn_unremovable_phi(i32) system_v {
ebb0(v0: i32): ebb0(v0: i32):
v2 = iconst.i32 0 v2 = iconst.i32 0
jump ebb2(v2, v0) jump ebb2(v2, v0)

12
cranelift/filetests/regalloc/coalescing-207.cton
View File

@@ -5,12 +5,12 @@ isa intel haswell
; Reported as https://github.com/Cretonne/cretonne/issues/207 ; Reported as https://github.com/Cretonne/cretonne/issues/207
; ;
; The coalescer creates a virtual register with two interfering values. ; The coalescer creates a virtual register with two interfering values.
function %pr207(i64 vmctx, i32, i32) -> i32 native { function %pr207(i64 vmctx, i32, i32) -> i32 system_v {
gv0 = vmctx-8 gv0 = vmctx-8
heap0 = static gv0, min 0, bound 0x5000, guard 0x0040_0000 heap0 = static gv0, min 0, bound 0x5000, guard 0x0040_0000
sig0 = (i64 vmctx, i32, i32) -> i32 native sig0 = (i64 vmctx, i32, i32) -> i32 system_v
sig1 = (i64 vmctx, i32, i32, i32) -> i32 native sig1 = (i64 vmctx, i32, i32, i32) -> i32 system_v
sig2 = (i64 vmctx, i32, i32, i32) -> i32 native sig2 = (i64 vmctx, i32, i32, i32) -> i32 system_v
fn0 = sig0 u0:2 fn0 = sig0 u0:2
fn1 = sig1 u0:0 fn1 = sig1 u0:0
fn2 = sig2 u0:1 fn2 = sig2 u0:1
@@ -1034,10 +1034,10 @@ ebb92(v767: i32):
} }
; Same problem from musl.wasm. ; Same problem from musl.wasm.
function %musl(f64 [%xmm0], i64 vmctx [%rdi]) -> f64 [%xmm0] native { function %musl(f64 [%xmm0], i64 vmctx [%rdi]) -> f64 [%xmm0] system_v {
gv0 = vmctx gv0 = vmctx
heap0 = static gv0, min 0, bound 0x0001_0000_0000, guard 0x8000_0000 heap0 = static gv0, min 0, bound 0x0001_0000_0000, guard 0x8000_0000
sig0 = (f64 [%xmm0], i32 [%rdi], i64 vmctx [%rsi]) -> f64 [%xmm0] native sig0 = (f64 [%xmm0], i32 [%rdi], i64 vmctx [%rsi]) -> f64 [%xmm0] system_v
fn0 = sig0 u0:517 fn0 = sig0 u0:517
ebb0(v0: f64, v1: i64): ebb0(v0: f64, v1: i64):

2
cranelift/filetests/regalloc/coalescing-216.cton
View File

@@ -5,7 +5,7 @@ isa intel haswell
; Reported as https://github.com/Cretonne/cretonne/issues/216 from the Binaryen fuzzer. ; Reported as https://github.com/Cretonne/cretonne/issues/216 from the Binaryen fuzzer.
; ;
; The (old) coalescer creates a virtual register with two identical values. ; The (old) coalescer creates a virtual register with two identical values.
function %pr216(i32 [%rdi], i64 vmctx [%rsi]) -> i64 [%rax] native { function %pr216(i32 [%rdi], i64 vmctx [%rsi]) -> i64 [%rax] system_v {
ebb0(v0: i32, v1: i64): ebb0(v0: i32, v1: i64):
v3 = iconst.i64 0 v3 = iconst.i64 0
v5 = iconst.i32 0 v5 = iconst.i32 0

4
cranelift/filetests/regalloc/coloring-227.cton
View File

@@ -2,7 +2,7 @@ test regalloc
set is_64bit set is_64bit
isa intel haswell isa intel haswell
function %pr227(i32 [%rdi], i32 [%rsi], i32 [%rdx], i32 [%rcx], i64 vmctx [%r8]) native { function %pr227(i32 [%rdi], i32 [%rsi], i32 [%rdx], i32 [%rcx], i64 vmctx [%r8]) system_v {
gv0 = vmctx gv0 = vmctx
heap0 = static gv0, min 0, bound 0x0001_0000_0000, guard 0x8000_0000 heap0 = static gv0, min 0, bound 0x0001_0000_0000, guard 0x8000_0000
@@ -21,7 +21,7 @@ function %pr227(i32 [%rdi], i32 [%rsi], i32 [%rdx], i32 [%rcx], i64 vmctx [%r8])
@0011 [RexOp1puid#b8] v9 = iconst.i32 0 @0011 [RexOp1puid#b8] v9 = iconst.i32 0
@0015 [RexOp1puid#b8] v11 = iconst.i32 0 @0015 [RexOp1puid#b8] v11 = iconst.i32 0
@0017 [RexOp1icscc#39] v12 = icmp.i32 eq v15, v11 @0017 [RexOp1icscc#39] v12 = icmp.i32 eq v15, v11
@0017 [RexOp2urm#4b6] v13 = bint.i32 v12 @0017 [RexOp2urm_noflags#4b6] v13 = bint.i32 v12
@001a [RexOp1rr#21] v14 = band v9, v13 @001a [RexOp1rr#21] v14 = band v9, v13
@001b [RexOp1tjccb#75] brnz v14, ebb6 @001b [RexOp1tjccb#75] brnz v14, ebb6
@001d [RexOp1jmpb#eb] jump ebb7 @001d [RexOp1jmpb#eb] jump ebb7

2
cranelift/filetests/regalloc/ghost-param.cton
View File

@@ -9,7 +9,7 @@ isa intel haswell
; ;
; Test case by binaryen fuzzer! ; Test case by binaryen fuzzer!
function %pr215(i64 vmctx [%rdi]) native { function %pr215(i64 vmctx [%rdi]) system_v {
ebb0(v0: i64): ebb0(v0: i64):
v10 = iconst.i64 0 v10 = iconst.i64 0
v1 = bitcast.f64 v10 v1 = bitcast.f64 v10

2
cranelift/filetests/regalloc/global-fixed.cton
View File

@@ -2,7 +2,7 @@ test regalloc
set is_64bit=1 set is_64bit=1
isa intel haswell isa intel haswell
function %foo() native { function %foo() system_v {
ebb4: ebb4:
v3 = iconst.i32 0 v3 = iconst.i32 0
jump ebb3 jump ebb3

2
cranelift/filetests/regalloc/intel-regres.cton
View File

@@ -11,7 +11,7 @@ isa intel
; This ended up confusong the constraint solver which had not made a record of ; This ended up confusong the constraint solver which had not made a record of
; the fixed register assignment for v9 since it was already in the correct ; the fixed register assignment for v9 since it was already in the correct
; register. ; register.
function %pr147(i32) -> i32 native { function %pr147(i32) -> i32 system_v {
ebb0(v0: i32): ebb0(v0: i32):
v1 = iconst.i32 0 v1 = iconst.i32 0
v2 = iconst.i32 1 v2 = iconst.i32 1

2
cranelift/filetests/regalloc/output-interference.cton
View File

@@ -2,7 +2,7 @@ test regalloc
set is_64bit=1 set is_64bit=1
isa intel haswell isa intel haswell
function %test(i64) -> i64 native { function %test(i64) -> i64 system_v {
ebb0(v0: i64): ebb0(v0: i64):
v2 = iconst.i64 12 v2 = iconst.i64 12
; This division clobbers two of its fixed input registers on Intel. ; This division clobbers two of its fixed input registers on Intel.

6
cranelift/filetests/regalloc/reload-208.cton
View File

@@ -11,11 +11,11 @@ isa intel haswell
; ;
; The problem was the reload pass rewriting EBB arguments on "brnz v9, ebb3(v9)" ; The problem was the reload pass rewriting EBB arguments on "brnz v9, ebb3(v9)"
function %pr208(i64 vmctx [%rdi]) native { function %pr208(i64 vmctx [%rdi]) system_v {
gv0 = vmctx-8 gv0 = vmctx-8
heap0 = static gv0, min 0, bound 0x5000, guard 0x0040_0000 heap0 = static gv0, min 0, bound 0x5000, guard 0x0040_0000
sig0 = (i64 vmctx [%rdi]) -> i32 [%rax] native sig0 = (i64 vmctx [%rdi]) -> i32 [%rax] system_v
sig1 = (i64 vmctx [%rdi], i32 [%rsi]) native sig1 = (i64 vmctx [%rdi], i32 [%rsi]) system_v
fn0 = sig0 u0:1 fn0 = sig0 u0:1
fn1 = sig1 u0:3 fn1 = sig1 u0:3

2
cranelift/filetests/regalloc/reload.cton
View File

@@ -5,7 +5,7 @@ isa riscv enable_e
; Check that we can handle a function return value that got spilled. ; Check that we can handle a function return value that got spilled.
function %spill_return() -> i32 { function %spill_return() -> i32 {
fn0 = function %foo() -> i32 native fn0 = function %foo() -> i32 system_v
ebb0: ebb0:
v0 = call fn0() v0 = call fn0()

4
cranelift/filetests/regalloc/schedule-moves.cton
View File

@@ -1,7 +1,7 @@
test regalloc test regalloc
isa intel haswell isa intel haswell
function %pr165() native { function %pr165() system_v {
ebb0: ebb0:
v0 = iconst.i32 0x0102_0304 v0 = iconst.i32 0x0102_0304
v1 = iconst.i32 0x1102_0304 v1 = iconst.i32 0x1102_0304
@@ -19,7 +19,7 @@ ebb0:
; Same as above, but use so many registers that spilling is required. ; Same as above, but use so many registers that spilling is required.
; Note: This is also a candidate for using xchg instructions. ; Note: This is also a candidate for using xchg instructions.
function %emergency_spill() native { function %emergency_spill() system_v {
ebb0: ebb0:
v0 = iconst.i32 0x0102_0304 v0 = iconst.i32 0x0102_0304
v1 = iconst.i32 0x1102_0304 v1 = iconst.i32 0x1102_0304

2
cranelift/filetests/regalloc/spill-noregs.cton
View File

@@ -13,7 +13,7 @@ isa intel
; ;
; The spiller was not releasing register pressure for dead EBB parameters. ; The spiller was not releasing register pressure for dead EBB parameters.
function %pr223(i32 [%rdi], i64 vmctx [%rsi]) -> i64 [%rax] native { function %pr223(i32 [%rdi], i64 vmctx [%rsi]) -> i64 [%rax] system_v {
ebb0(v0: i32, v1: i64): ebb0(v0: i32, v1: i64):
v2 = iconst.i32 0 v2 = iconst.i32 0
v3 = iconst.i64 0 v3 = iconst.i64 0

4
cranelift/filetests/regalloc/spill.cton
View File

@@ -93,7 +93,7 @@ ebb0(v0: i32):
; The same value used as indirect callee and argument. ; The same value used as indirect callee and argument.
function %doubleuse_icall1(i32) { function %doubleuse_icall1(i32) {
sig0 = (i32) native sig0 = (i32) system_v
ebb0(v0: i32): ebb0(v0: i32):
; not:copy ; not:copy
call_indirect sig0, v0(v0) call_indirect sig0, v0(v0)
@@ -102,7 +102,7 @@ ebb0(v0: i32):
; The same value used as indirect callee and two arguments. ; The same value used as indirect callee and two arguments.
function %doubleuse_icall2(i32) { function %doubleuse_icall2(i32) {
sig0 = (i32, i32) native sig0 = (i32, i32) system_v
ebb0(v0: i32): ebb0(v0: i32):
; check: $(c=$V) = copy v0 ; check: $(c=$V) = copy v0
call_indirect sig0, v0(v0, v0) call_indirect sig0, v0(v0, v0)

16
cranelift/filetests/verifier/defs_dominates_uses.cton Normal file
View File

@@ -0,0 +1,16 @@
test verifier
; Test verification that uses properly dominate defs.
function %non_dominating(i32) -> i32 system_v {
ebb0(v0: i32):
v1 = iadd.i32 v2, v0 ; error: uses value from non-dominating
v2 = iadd.i32 v1, v0
return v2
}
function %inst_uses_its_own_values(i32) -> i32 system_v {
ebb0(v0: i32):
v1 = iadd.i32 v1, v0 ; error: uses value from itself
return v1
}

12
cranelift/filetests/verifier/flags.cton
View File

@@ -6,7 +6,7 @@ function %simple(i32) -> i32 {
ebb0(v0: i32): ebb0(v0: i32):
[Op1rcmp#39] v1 = ifcmp v0, v0 [Op1rcmp#39] v1 = ifcmp v0, v0
[Op2seti_abcd#490] v2 = trueif ugt v1 [Op2seti_abcd#490] v2 = trueif ugt v1
[Op2urm_abcd#4b6] v3 = bint.i32 v2 [Op2urm_noflags_abcd#4b6] v3 = bint.i32 v2
[Op1ret#c3] return v3 [Op1ret#c3] return v3
} }
@@ -18,7 +18,7 @@ function %overlap(i32, f32) -> i32 {
[Op2setf_abcd#490] v4 = trueff gt v3 ; error: conflicting live CPU flags: v2 and v3 [Op2setf_abcd#490] v4 = trueff gt v3 ; error: conflicting live CPU flags: v2 and v3
[Op2seti_abcd#490] v5 = trueif ugt v2 [Op2seti_abcd#490] v5 = trueif ugt v2
[Op1rr#21] v6 = band v4, v5 [Op1rr#21] v6 = band v4, v5
[Op2urm_abcd#4b6] v7 = bint.i32 v6 [Op2urm_noflags_abcd#4b6] v7 = bint.i32 v6
[Op1ret#c3] return v7 [Op1ret#c3] return v7
} }
@@ -28,7 +28,7 @@ function %clobbered(i32) -> i32 {
[Op1rcmp#39] v1 = ifcmp v0, v0 [Op1rcmp#39] v1 = ifcmp v0, v0
[Op1rr#01] v2 = iadd v0, v0 ; error: encoding clobbers live CPU flags in v1 [Op1rr#01] v2 = iadd v0, v0 ; error: encoding clobbers live CPU flags in v1
[Op2seti_abcd#490] v3 = trueif ugt v1 [Op2seti_abcd#490] v3 = trueif ugt v1
[Op2urm_abcd#4b6] v4 = bint.i32 v3 [Op2urm_noflags_abcd#4b6] v4 = bint.i32 v3
[Op1ret#c3] return v4 [Op1ret#c3] return v4
} }
@@ -38,7 +38,7 @@ function %live_across_load(i32) -> i32 {
[Op1rcmp#39] v1 = ifcmp v0, v0 [Op1rcmp#39] v1 = ifcmp v0, v0
[Op1ld#8b] v2 = load.i32 v0 [Op1ld#8b] v2 = load.i32 v0
[Op2seti_abcd#490] v3 = trueif ugt v1 [Op2seti_abcd#490] v3 = trueif ugt v1
[Op2urm_abcd#4b6] v4 = bint.i32 v3 [Op2urm_noflags_abcd#4b6] v4 = bint.i32 v3
[Op1ret#c3] return v4 [Op1ret#c3] return v4
} }
@@ -49,7 +49,7 @@ function %live_across_ebb(i32) -> i32 {
[Op1jmpb#eb] jump ebb1 [Op1jmpb#eb] jump ebb1
ebb1: ebb1:
[Op2seti_abcd#490] v2 = trueif ugt v1 [Op2seti_abcd#490] v2 = trueif ugt v1
[Op2urm_abcd#4b6] v3 = bint.i32 v2 [Op2urm_noflags_abcd#4b6] v3 = bint.i32 v2
[Op1ret#c3] return v3 [Op1ret#c3] return v3
} }
@@ -58,7 +58,7 @@ function %live_across_ebb_backwards(i32) -> i32 {
[Op1jmpb#eb] jump ebb2 [Op1jmpb#eb] jump ebb2
ebb1: ebb1:
[Op2seti_abcd#490] v2 = trueif ugt v1 [Op2seti_abcd#490] v2 = trueif ugt v1
[Op2urm_abcd#4b6] v3 = bint.i32 v2 [Op2urm_noflags_abcd#4b6] v3 = bint.i32 v2
[Op1ret#c3] return v3 [Op1ret#c3] return v3
ebb2: ebb2:
[Op1rcmp#39] v1 = ifcmp v0, v0 [Op1rcmp#39] v1 = ifcmp v0, v0

8
cranelift/publish-all.sh
View File

@@ -4,17 +4,13 @@ cd $(dirname "$0")
topdir="$(pwd)" topdir="$(pwd)"
# All the cretonne-* crates have the same version number # All the cretonne-* crates have the same version number
# The filecheck crate version is managed independently. version="0.4.1"
version="0.3.4"
# Update all of the Cargo.toml files. # Update all of the Cargo.toml files.
# #
# The main Cargo.toml in the top-level directory is the cretonne-tools crate which we don't publish. # The main Cargo.toml in the top-level directory is the cretonne-tools crate which we don't publish.
echo "Updating crate versions to $version" echo "Updating crate versions to $version"
for crate in . lib/*; do for crate in . lib/*; do
if [ "$crate" = "lib/filecheck" ]; then
continue
fi
# Update the version number of this crate to $version. # Update the version number of this crate to $version.
sed -i.bk -e "s/^version = .*/version = \"$version\"/" "$crate/Cargo.toml" sed -i.bk -e "s/^version = .*/version = \"$version\"/" "$crate/Cargo.toml"
# Update the required version number of any cretonne* dependencies. # Update the required version number of any cretonne* dependencies.
@@ -31,7 +27,7 @@ cargo update
echo git commit -a -m "\"Bump version to $version"\" echo git commit -a -m "\"Bump version to $version"\"
echo git push echo git push
for crate in filecheck cretonne frontend native reader wasm; do for crate in cretonne frontend native reader wasm; do
echo cargo publish --manifest-path "lib/$crate/Cargo.toml" echo cargo publish --manifest-path "lib/$crate/Cargo.toml"
done done
echo echo

42
cranelift/src/cat.rs
View File

@@ -1,16 +1,13 @@
//! The `cat` sub-command. //! The `cat` sub-command.
//! //!
//! Read a sequence of Cretonne IL files and print them again to stdout. This has the effect of //! Read a sequence of Cretonne IR files and print them again to stdout. This has the effect of
//! normalizing formatting and removing comments. //! normalizing formatting and removing comments.
use std::borrow::Cow;
use cretonne::ir::Function;
use cton_reader::{parse_functions, TestCommand};
use CommandResult; use CommandResult;
use cton_reader::parse_functions;
use utils::read_to_string; use utils::read_to_string;
use filetest::subtest::{self, SubTest, Context, Result as STResult};
pub fn run(files: Vec<String>) -> CommandResult { pub fn run(files: &[String]) -> CommandResult {
for (i, f) in files.into_iter().enumerate() { for (i, f) in files.into_iter().enumerate() {
if i != 0 { if i != 0 {
println!(); println!();
@@ -20,7 +17,7 @@ pub fn run(files: Vec<String>) -> CommandResult {
Ok(()) Ok(())
} }
fn cat_one(filename: String) -> CommandResult { fn cat_one(filename: &str) -> CommandResult {
let buffer = read_to_string(&filename).map_err( let buffer = read_to_string(&filename).map_err(
|e| format!("{}: {}", filename, e), |e| format!("{}: {}", filename, e),
)?; )?;
@@ -37,34 +34,3 @@ fn cat_one(filename: String) -> CommandResult {
Ok(()) Ok(())
} }
/// Object implementing the `test cat` sub-test.
///
/// This command is used for testing the parser and function printer. It simply parses a function
/// and prints it out again.
///
/// The result is verified by filecheck.
struct TestCat;
pub fn subtest(parsed: &TestCommand) -> STResult<Box<SubTest>> {
assert_eq!(parsed.command, "cat");
if !parsed.options.is_empty() {
Err(format!("No options allowed on {}", parsed))
} else {
Ok(Box::new(TestCat))
}
}
impl SubTest for TestCat {
fn name(&self) -> Cow<str> {
Cow::from("cat")
}
fn needs_verifier(&self) -> bool {
false
}
fn run(&self, func: Cow<Function>, context: &Context) -> STResult<()> {
subtest::run_filecheck(&func.display(context.isa).to_string(), context)
}
}

36
cranelift/src/compile.rs
View File

@@ -1,14 +1,13 @@
//! CLI tool to compile cretonne IL into native code. //! CLI tool to read Cretonne IR files and compile them into native code.
//!
//! Reads IR files into Cretonne IL and compiles it.
use cton_reader::parse_test;
use std::path::PathBuf;
use cretonne::Context; use cretonne::Context;
use cretonne::print_errors::pretty_error;
use cretonne::settings::FlagsOrIsa; use cretonne::settings::FlagsOrIsa;
use cretonne::{binemit, ir}; use cretonne::{binemit, ir};
use cton_reader::parse_test;
use std::path::Path; use std::path::Path;
use utils::{pretty_error, read_to_string, parse_sets_and_isa}; use std::path::PathBuf;
use utils::{parse_sets_and_isa, read_to_string};
struct PrintRelocs { struct PrintRelocs {
flag_print: bool, flag_print: bool,
@@ -45,26 +44,38 @@ impl binemit::RelocSink for PrintRelocs {
} }
} }
struct PrintTraps {
flag_print: bool,
}
impl binemit::TrapSink for PrintTraps {
fn trap(&mut self, offset: binemit::CodeOffset, _srcloc: ir::SourceLoc, code: ir::TrapCode) {
if self.flag_print {
println!("trap: {} at {}", code, offset);
}
}
}
pub fn run( pub fn run(
files: Vec<String>, files: Vec<String>,
flag_print: bool, flag_print: bool,
flag_set: Vec<String>, flag_set: &[String],
flag_isa: String, flag_isa: &str,
) -> Result<(), String> { ) -> Result<(), String> {
let parsed = parse_sets_and_isa(flag_set, flag_isa)?; let parsed = parse_sets_and_isa(flag_set, flag_isa)?;
for filename in files { for filename in files {
let path = Path::new(&filename); let path = Path::new(&filename);
let name = String::from(path.as_os_str().to_string_lossy()); let name = String::from(path.as_os_str().to_string_lossy());
handle_module(flag_print, path.to_path_buf(), name, parsed.as_fisa())?; handle_module(flag_print, &path.to_path_buf(), &name, parsed.as_fisa())?;
} }
Ok(()) Ok(())
} }
fn handle_module( fn handle_module(
flag_print: bool, flag_print: bool,
path: PathBuf, path: &PathBuf,
name: String, name: &str,
fisa: FlagsOrIsa, fisa: FlagsOrIsa,
) -> Result<(), String> { ) -> Result<(), String> {
let buffer = read_to_string(&path).map_err( let buffer = read_to_string(&path).map_err(
@@ -95,8 +106,9 @@ fn handle_module(
// Encode the result as machine code. // Encode the result as machine code.
let mut mem = Vec::new(); let mut mem = Vec::new();
let mut relocs = PrintRelocs { flag_print }; let mut relocs = PrintRelocs { flag_print };
let mut traps = PrintTraps { flag_print };
mem.resize(size as usize, 0); mem.resize(size as usize, 0);
context.emit_to_memory(mem.as_mut_ptr(), &mut relocs, &*isa); context.emit_to_memory(mem.as_mut_ptr(), &mut relocs, &mut traps, &*isa);
if flag_print { if flag_print {
print!(".byte "); print!(".byte ");

37
cranelift/src/cton-util.rs
View File

@@ -1,27 +1,25 @@
#[macro_use(dbg)]
extern crate cretonne; extern crate cretonne;
extern crate cton_filetests;
extern crate cton_reader; extern crate cton_reader;
extern crate cton_wasm; extern crate cton_wasm;
extern crate docopt; extern crate docopt;
extern crate filecheck;
#[macro_use] #[macro_use]
extern crate serde_derive; extern crate serde_derive;
extern crate filecheck;
extern crate num_cpus;
extern crate tempdir; extern crate tempdir;
extern crate term; extern crate term;
use cretonne::{VERSION, timing}; use cretonne::{timing, VERSION};
use docopt::Docopt; use docopt::Docopt;
use std::io::{self, Write}; use std::io::{self, Write};
use std::process; use std::process;
mod utils;
mod filetest;
mod cat; mod cat;
mod compile;
mod print_cfg; mod print_cfg;
mod rsfilecheck; mod rsfilecheck;
mod utils;
mod wasm; mod wasm;
mod compile;
const USAGE: &str = " const USAGE: &str = "
Cretonne code generator utility Cretonne code generator utility
@@ -40,12 +38,12 @@ Options:
-T, --time-passes -T, --time-passes
print pass timing report print pass timing report
-t, --just-decode -t, --just-decode
just decode WebAssembly to Cretonne IL just decode WebAssembly to Cretonne IR
-s, --print-size -s, --print-size
prints generated code size prints generated code size
-c, --check-translation -c, --check-translation
just checks the correctness of Cretonne IL translated from WebAssembly just checks the correctness of Cretonne IR translated from WebAssembly
-p, --print print the resulting Cretonne IL -p, --print print the resulting Cretonne IR
-h, --help print this help message -h, --help print this help message
--set=<set> configure Cretonne settings --set=<set> configure Cretonne settings
--isa=<isa> specify the Cretonne ISA --isa=<isa> specify the Cretonne ISA
@@ -88,15 +86,20 @@ fn cton_util() -> CommandResult {
// Find the sub-command to execute. // Find the sub-command to execute.
let result = if args.cmd_test { let result = if args.cmd_test {
filetest::run(args.flag_verbose, args.arg_file) cton_filetests::run(args.flag_verbose, &args.arg_file).map(|_time| ())
} else if args.cmd_cat { } else if args.cmd_cat {
cat::run(args.arg_file) cat::run(&args.arg_file)
} else if args.cmd_filecheck { } else if args.cmd_filecheck {
rsfilecheck::run(args.arg_file, args.flag_verbose) rsfilecheck::run(&args.arg_file, args.flag_verbose)
} else if args.cmd_print_cfg { } else if args.cmd_print_cfg {
print_cfg::run(args.arg_file) print_cfg::run(&args.arg_file)
} else if args.cmd_compile { } else if args.cmd_compile {
compile::run(args.arg_file, args.flag_print, args.flag_set, args.flag_isa) compile::run(
args.arg_file,
args.flag_print,
&args.flag_set,
&args.flag_isa,
)
} else if args.cmd_wasm { } else if args.cmd_wasm {
wasm::run( wasm::run(
args.arg_file, args.arg_file,
@@ -104,8 +107,8 @@ fn cton_util() -> CommandResult {
args.flag_just_decode, args.flag_just_decode,
args.flag_check_translation, args.flag_check_translation,
args.flag_print, args.flag_print,
args.flag_set, &args.flag_set,
args.flag_isa, &args.flag_isa,
args.flag_print_size, args.flag_print_size,
) )
} else { } else {

75
cranelift/src/filetest/mod.rs
View File

@@ -1,75 +0,0 @@
//! File tests.
//!
//! This module contains the main driver for `cton-util test` as well as implementations of the
//! available test commands.
use std::path::Path;
use std::time;
use cton_reader::TestCommand;
use CommandResult;
use cat;
use print_cfg;
use filetest::runner::TestRunner;
pub mod subtest;
mod binemit;
mod compile;
mod concurrent;
mod domtree;
mod legalizer;
mod licm;
mod preopt;
mod regalloc;
mod runner;
mod runone;
mod simple_gvn;
mod verifier;
/// The result of running the test in a file.
pub type TestResult = Result<time::Duration, String>;
/// Main entry point for `cton-util test`.
///
/// Take a list of filenames which can be either `.cton` files or directories.
///
/// Files are interpreted as test cases and executed immediately.
///
/// Directories are scanned recursively for test cases ending in `.cton`. These test cases are
/// executed on background threads.
///
pub fn run(verbose: bool, files: Vec<String>) -> CommandResult {
let mut runner = TestRunner::new(verbose);
for path in files.iter().map(Path::new) {
if path.is_file() {
runner.push_test(path);
} else {
runner.push_dir(path);
}
}
runner.start_threads();
runner.run()
}
/// Create a new subcommand trait object to match `parsed.command`.
///
/// This function knows how to create all of the possible `test <foo>` commands that can appear in
/// a `.cton` test file.
fn new_subtest(parsed: &TestCommand) -> subtest::Result<Box<subtest::SubTest>> {
match parsed.command {
"binemit" => binemit::subtest(parsed),
"cat" => cat::subtest(parsed),
"compile" => compile::subtest(parsed),
"domtree" => domtree::subtest(parsed),
"legalizer" => legalizer::subtest(parsed),
"licm" => licm::subtest(parsed),
"preopt" => preopt::subtest(parsed),
"print-cfg" => print_cfg::subtest(parsed),
"regalloc" => regalloc::subtest(parsed),
"simple-gvn" => simple_gvn::subtest(parsed),
"verifier" => verifier::subtest(parsed),
_ => Err(format!("unknown test command '{}'", parsed.command)),
}
}

110
cranelift/src/print_cfg.rs
View File

@@ -1,20 +1,14 @@
//! The `print-cfg` sub-command. //! The `print-cfg` sub-command.
//! //!
//! Read a series of Cretonne IL files and print their control flow graphs //! Read a series of Cretonne IR files and print their control flow graphs
//! in graphviz format. //! in graphviz format.
use std::borrow::Cow;
use std::fmt::{Result, Write, Display, Formatter};
use CommandResult; use CommandResult;
use cretonne::flowgraph::ControlFlowGraph; use cretonne::cfg_printer::CFGPrinter;
use cretonne::ir::Function; use cton_reader::parse_functions;
use cretonne::ir::instructions::BranchInfo;
use cton_reader::{parse_functions, TestCommand};
use filetest::subtest::{self, SubTest, Context, Result as STResult};
use utils::read_to_string; use utils::read_to_string;
pub fn run(files: Vec<String>) -> CommandResult { pub fn run(files: &[String]) -> CommandResult {
for (i, f) in files.into_iter().enumerate() { for (i, f) in files.into_iter().enumerate() {
if i != 0 { if i != 0 {
println!(); println!();
@@ -24,74 +18,8 @@ pub fn run(files: Vec<String>) -> CommandResult {
Ok(()) Ok(())
} }
struct CFGPrinter<'a> { fn print_cfg(filename: &str) -> CommandResult {
func: &'a Function, let buffer = read_to_string(filename).map_err(
cfg: ControlFlowGraph,
}
impl<'a> CFGPrinter<'a> {
pub fn new(func: &'a Function) -> CFGPrinter<'a> {
CFGPrinter {
func,
cfg: ControlFlowGraph::with_function(func),
}
}
/// Write the CFG for this function to `w`.
pub fn write(&self, w: &mut Write) -> Result {
self.header(w)?;
self.ebb_nodes(w)?;
self.cfg_connections(w)?;
writeln!(w, "}}")
}
fn header(&self, w: &mut Write) -> Result {
writeln!(w, "digraph \"{}\" {{", self.func.name)?;
if let Some(entry) = self.func.layout.entry_block() {
writeln!(w, " {{rank=min; {}}}", entry)?;
}
Ok(())
}
fn ebb_nodes(&self, w: &mut Write) -> Result {
for ebb in &self.func.layout {
write!(w, " {} [shape=record, label=\"{{{}", ebb, ebb)?;
// Add all outgoing branch instructions to the label.
for inst in self.func.layout.ebb_insts(ebb) {
let idata = &self.func.dfg[inst];
match idata.analyze_branch(&self.func.dfg.value_lists) {
BranchInfo::SingleDest(dest, _) => {
write!(w, " | <{}>{} {}", inst, idata.opcode(), dest)?
}
BranchInfo::Table(table) => {
write!(w, " | <{}>{} {}", inst, idata.opcode(), table)?
}
BranchInfo::NotABranch => {}
}
}
writeln!(w, "}}\"]")?
}
Ok(())
}
fn cfg_connections(&self, w: &mut Write) -> Result {
for ebb in &self.func.layout {
for (parent, inst) in self.cfg.pred_iter(ebb) {
writeln!(w, " {}:{} -> {}", parent, inst, ebb)?;
}
}
Ok(())
}
}
impl<'a> Display for CFGPrinter<'a> {
fn fmt(&self, f: &mut Formatter) -> Result {
self.write(f)
}
}
fn print_cfg(filename: String) -> CommandResult {
let buffer = read_to_string(&filename).map_err(
|e| format!("{}: {}", filename, e), |e| format!("{}: {}", filename, e),
)?; )?;
let items = parse_functions(&buffer).map_err( let items = parse_functions(&buffer).map_err(
@@ -107,29 +35,3 @@ fn print_cfg(filename: String) -> CommandResult {
Ok(()) Ok(())
} }
/// Object implementing the `test print-cfg` sub-test.
struct TestPrintCfg;
pub fn subtest(parsed: &TestCommand) -> STResult<Box<SubTest>> {
assert_eq!(parsed.command, "print-cfg");
if !parsed.options.is_empty() {
Err(format!("No options allowed on {}", parsed))
} else {
Ok(Box::new(TestPrintCfg))
}
}
impl SubTest for TestPrintCfg {
fn name(&self) -> Cow<str> {
Cow::from("print-cfg")
}
fn needs_verifier(&self) -> bool {
false
}
fn run(&self, func: Cow<Function>, context: &Context) -> STResult<()> {
subtest::run_filecheck(&CFGPrinter::new(&func).to_string(), context)
}
}

14
cranelift/src/rsfilecheck.rs
View File

@@ -1,9 +1,13 @@
use CommandResult; //! The `filecheck` sub-command.
use utils::read_to_string; //!
use filecheck::{CheckerBuilder, Checker, NO_VARIABLES}; //! This file is named to avoid a name collision with the filecheck crate.
use std::io::{self, Read};
pub fn run(files: Vec<String>, verbose: bool) -> CommandResult { use CommandResult;
use filecheck::{Checker, CheckerBuilder, NO_VARIABLES};
use std::io::{self, Read};
use utils::read_to_string;
pub fn run(files: &[String], verbose: bool) -> CommandResult {
if files.is_empty() { if files.is_empty() {
return Err("No check files".to_string()); return Err("No check files".to_string());
} }

65
cranelift/src/utils.rs
View File

@@ -1,13 +1,9 @@
//! Utility functions. //! Utility functions.
use cretonne::ir::entities::AnyEntity; use cretonne::isa;
use cretonne::{ir, verifier};
use cretonne::result::CtonError;
use cretonne::isa::TargetIsa; use cretonne::isa::TargetIsa;
use cretonne::settings::{self, FlagsOrIsa}; use cretonne::settings::{self, FlagsOrIsa};
use cretonne::isa;
use cton_reader::{parse_options, Location}; use cton_reader::{parse_options, Location};
use std::fmt::Write;
use std::fs::File; use std::fs::File;
use std::io::{self, Read}; use std::io::{self, Read};
use std::path::Path; use std::path::Path;
@@ -28,51 +24,6 @@ pub fn read_to_end<P: AsRef<Path>>(path: P) -> io::Result<Vec<u8>> {
Ok(buffer) Ok(buffer)
} }
/// Look for a directive in a comment string.
/// The directive is of the form "foo:" and should follow the leading `;` in the comment:
///
/// ; dominates: ebb3 ebb4
///
/// Return the comment text following the directive.
pub fn match_directive<'a>(comment: &'a str, directive: &str) -> Option<&'a str> {
assert!(
directive.ends_with(':'),
"Directive must include trailing colon"
);
let text = comment.trim_left_matches(';').trim_left();
if text.starts_with(directive) {
Some(text[directive.len()..].trim())
} else {
None
}
}
/// Pretty-print a verifier error.
pub fn pretty_verifier_error(
func: &ir::Function,
isa: Option<&TargetIsa>,
err: verifier::Error,
) -> String {
let mut msg = err.to_string();
match err.location {
AnyEntity::Inst(inst) => {
write!(msg, "\n{}: {}\n\n", inst, func.dfg.display_inst(inst, isa)).unwrap()
}
_ => msg.push('\n'),
}
write!(msg, "{}", func.display(isa)).unwrap();
msg
}
/// Pretty-print a Cretonne error.
pub fn pretty_error(func: &ir::Function, isa: Option<&TargetIsa>, err: CtonError) -> String {
if let CtonError::Verifier(e) = err {
pretty_verifier_error(func, isa, e)
} else {
err.to_string()
}
}
/// Like `FlagsOrIsa`, but holds ownership. /// Like `FlagsOrIsa`, but holds ownership.
pub enum OwnedFlagsOrIsa { pub enum OwnedFlagsOrIsa {
Flags(settings::Flags), Flags(settings::Flags),
@@ -90,10 +41,7 @@ impl OwnedFlagsOrIsa {
} }
/// Parse "set" and "isa" commands. /// Parse "set" and "isa" commands.
pub fn parse_sets_and_isa( pub fn parse_sets_and_isa(flag_set: &[String], flag_isa: &str) -> Result<OwnedFlagsOrIsa, String> {
flag_set: Vec<String>,
flag_isa: String,
) -> Result<OwnedFlagsOrIsa, String> {
let mut flag_builder = settings::builder(); let mut flag_builder = settings::builder();
parse_options( parse_options(
flag_set.iter().map(|x| x.as_str()), flag_set.iter().map(|x| x.as_str()),
@@ -119,12 +67,3 @@ pub fn parse_sets_and_isa(
Ok(OwnedFlagsOrIsa::Flags(settings::Flags::new(&flag_builder))) Ok(OwnedFlagsOrIsa::Flags(settings::Flags::new(&flag_builder)))
} }
} }
#[test]
fn test_match_directive() {
assert_eq!(match_directive("; foo: bar ", "foo:"), Some("bar"));
assert_eq!(match_directive(" foo:bar", "foo:"), Some("bar"));
assert_eq!(match_directive("foo:bar", "foo:"), Some("bar"));
assert_eq!(match_directive(";x foo: bar", "foo:"), None);
assert_eq!(match_directive(";;; foo: bar", "foo:"), Some("bar"));
}

37
cranelift/src/wasm.rs
View File

@@ -1,19 +1,21 @@
//! CLI tool to use the functions provided by the [cretonne-wasm](../cton_wasm/index.html) crate. //! CLI tool to use the functions provided by the [cretonne-wasm](../cton_wasm/index.html) crate.
//! //!
//! Reads Wasm binary files, translates the functions' code to Cretonne IL. //! Reads Wasm binary files, translates the functions' code to Cretonne IR.
#![cfg_attr(feature = "cargo-clippy", allow(too_many_arguments, cyclomatic_complexity))]
use cton_wasm::{translate_module, DummyEnvironment, ModuleEnvironment};
use std::path::PathBuf;
use cretonne::Context; use cretonne::Context;
use cretonne::print_errors::{pretty_error, pretty_verifier_error};
use cretonne::settings::FlagsOrIsa; use cretonne::settings::FlagsOrIsa;
use std::fs::File; use cton_wasm::{translate_module, DummyEnvironment, ModuleEnvironment};
use std::error::Error; use std::error::Error;
use std::fs::File;
use std::io; use std::io;
use std::path::Path; use std::path::Path;
use std::path::PathBuf;
use std::process::Command; use std::process::Command;
use tempdir::TempDir; use tempdir::TempDir;
use term; use term;
use utils::{pretty_verifier_error, pretty_error, parse_sets_and_isa, read_to_end}; use utils::{parse_sets_and_isa, read_to_end};
macro_rules! vprintln { macro_rules! vprintln {
($x: expr, $($tts:tt)*) => { ($x: expr, $($tts:tt)*) => {
@@ -37,8 +39,8 @@ pub fn run(
flag_just_decode: bool, flag_just_decode: bool,
flag_check_translation: bool, flag_check_translation: bool,
flag_print: bool, flag_print: bool,
flag_set: Vec<String>, flag_set: &[String],
flag_isa: String, flag_isa: &str,
flag_print_size: bool, flag_print_size: bool,
) -> Result<(), String> { ) -> Result<(), String> {
let parsed = parse_sets_and_isa(flag_set, flag_isa)?; let parsed = parse_sets_and_isa(flag_set, flag_isa)?;
@@ -52,8 +54,8 @@ pub fn run(
flag_check_translation, flag_check_translation,
flag_print, flag_print,
flag_print_size, flag_print_size,
path.to_path_buf(), &path.to_path_buf(),
name, &name,
parsed.as_fisa(), parsed.as_fisa(),
)?; )?;
} }
@@ -66,8 +68,8 @@ fn handle_module(
flag_check_translation: bool, flag_check_translation: bool,
flag_print: bool, flag_print: bool,
flag_print_size: bool, flag_print_size: bool,
path: PathBuf, path: &PathBuf,
name: String, name: &str,
fisa: FlagsOrIsa, fisa: FlagsOrIsa,
) -> Result<(), String> { ) -> Result<(), String> {
let mut terminal = term::stdout().unwrap(); let mut terminal = term::stdout().unwrap();
@@ -152,10 +154,9 @@ fn handle_module(
context.func = func.clone(); context.func = func.clone();
if flag_check_translation { if flag_check_translation {
context.verify(fisa).map_err(|err| { context.verify(fisa).map_err(|err| {
pretty_verifier_error(&context.func, fisa.isa, err) pretty_verifier_error(&context.func, fisa.isa, &err)
})?; })?;
} else { } else if let Some(isa) = fisa.isa {
if let Some(isa) = fisa.isa {
let compiled_size = context.compile(isa).map_err(|err| { let compiled_size = context.compile(isa).map_err(|err| {
pretty_error(&context.func, fisa.isa, err) pretty_error(&context.func, fisa.isa, err)
})?; })?;
@@ -169,13 +170,12 @@ fn handle_module(
println!( println!(
"Function #{} bytecode size: {} bytes", "Function #{} bytecode size: {} bytes",
func_index, func_index,
dummy_environ.func_bytecode_sizes[func_index] dummy_environ.func_bytecode_sizes[def_index]
); );
} }
} else { } else {
return Err(String::from("compilation requires a target isa")); return Err(String::from("compilation requires a target isa"));
} }
}
if flag_print { if flag_print {
vprintln!(flag_verbose, ""); vprintln!(flag_verbose, "");
if let Some(start_func) = dummy_environ.info.start_func { if let Some(start_func) = dummy_environ.info.start_func {
@@ -193,10 +193,7 @@ fn handle_module(
if !flag_check_translation && flag_print_size { if !flag_check_translation && flag_print_size {
println!("Total module code size: {} bytes", total_module_code_size); println!("Total module code size: {} bytes", total_module_code_size);
let total_bytecode_size = dummy_environ.func_bytecode_sizes.iter().fold( let total_bytecode_size: usize = dummy_environ.func_bytecode_sizes.iter().sum();
0,
|sum, x| sum + x,
);
println!("Total module bytecode size: {} bytes", total_bytecode_size); println!("Total module bytecode size: {} bytes", total_bytecode_size);
} }

41
cranelift/test-all.sh
View File

@@ -3,11 +3,10 @@ set -euo pipefail
# This is the top-level test script: # This is the top-level test script:
# #
# - Build documentation for Rust code in 'src/tools/target/doc'. # - Make a debug build.
# - Run unit tests for all Rust crates. # - Make a release build.
# - Make a debug build of all crates. # - Run unit tests for all Rust crates (including the filetests)
# - Make a release build of cton-util. # - Build API documentation.
# - Run file-level tests with the release build of cton-util.
# #
# All tests run by this script should be passing at all times. # All tests run by this script should be passing at all times.
@@ -42,22 +41,26 @@ if [ -n "$needcheck" ]; then
touch $tsfile || echo no target directory touch $tsfile || echo no target directory
fi fi
cd "$topdir" # Make sure the code builds in debug mode.
banner "Rust unit tests" banner "Rust debug build"
cargo test --all cargo build
# Build cton-util for parser testing. # Make sure the code builds in release mode, and run the unit tests. We run
cd "$topdir" # these in release mode for speed, but note that the top-level Cargo.toml file
banner "Rust documentation" # does enable debug assertions in release builds.
echo "open $topdir/target/doc/cretonne/index.html" banner "Rust release build and unit tests"
cargo test --all --release
# Make sure the documentation builds.
banner "Rust documentation: $topdir/target/doc/cretonne/index.html"
cargo doc cargo doc
banner "Rust release build"
cargo build --release
export CTONUTIL="$topdir/target/release/cton-util" # Run clippy if we have it.
banner "Rust linter"
cd "$topdir" if $topdir/check-clippy.sh; then
banner "File tests" $topdir/clippy-all.sh --write-mode=diff
"$CTONUTIL" test filetests docs else
echo "\`cargo +nightly install clippy\` for optional rust linting"
fi
banner "OK" banner "OK"

7
cranelift/tests/filetests.rs Normal file
View File

@@ -0,0 +1,7 @@
extern crate cton_filetests;
#[test]
fn filetests() {
// Run all the filetests in the following directories.
cton_filetests::run(false, &["filetests".into(), "docs".into()]).expect("test harness");
}

6
lib/cretonne/Cargo.toml
View File

@@ -1,7 +1,7 @@
[package] [package]
authors = ["The Cretonne Project Developers"] authors = ["The Cretonne Project Developers"]
name = "cretonne" name = "cretonne"
version = "0.3.4" version = "0.4.1"
description = "Low-level code generator library" description = "Low-level code generator library"
license = "Apache-2.0" license = "Apache-2.0"
documentation = "https://cretonne.readthedocs.io/" documentation = "https://cretonne.readthedocs.io/"
@@ -32,3 +32,7 @@ optional = true
default = ["std"] default = ["std"]
std = [] std = []
core = ["hashmap_core"] core = ["hashmap_core"]
[badges]
maintenance = { status = "experimental" }
travis-ci = { repository = "Cretonne/cretonne" }

2
lib/cretonne/README.md
View File

@@ -1,2 +1,2 @@
This crate contains the core Cretonne code generator. It translates code from an This crate contains the core Cretonne code generator. It translates code from an
intermediate language into executable machine code. intermediate representation into executable machine code.

1
lib/cretonne/build.rs
View File

@@ -18,7 +18,6 @@
// The build script expects to be run from the directory where this build.rs file lives. The // The build script expects to be run from the directory where this build.rs file lives. The
// current directory is used to find the sources. // current directory is used to find the sources.
use std::env; use std::env;
use std::process; use std::process;

2
lib/cretonne/meta/base/formats.py
View File

@@ -2,7 +2,7 @@
The cretonne.formats defines all instruction formats. The cretonne.formats defines all instruction formats.
Every instruction format has a corresponding `InstructionData` variant in the Every instruction format has a corresponding `InstructionData` variant in the
Rust representation of cretonne IL, so all instruction formats must be defined Rust representation of Cretonne IR, so all instruction formats must be defined
in this module. in this module.
""" """
from __future__ import absolute_import from __future__ import absolute_import

3
lib/cretonne/meta/base/instructions.py
View File

@@ -588,6 +588,9 @@ stack_check = Instruction(
Read the stack limit from ``GV`` and compare it to the stack pointer. If Read the stack limit from ``GV`` and compare it to the stack pointer. If
the stack pointer has reached or exceeded the limit, generate a trap with a the stack pointer has reached or exceeded the limit, generate a trap with a
``stk_ovf`` code. ``stk_ovf`` code.
The global variable must be accessible and naturally aligned for a
pointer-sized value.
""", """,
ins=GV, can_trap=True) ins=GV, can_trap=True)

2
lib/cretonne/meta/base/legalize.py
View File

@@ -41,6 +41,8 @@ widen = XFormGroup('widen', """
The transformations in the 'widen' group work by expressing The transformations in the 'widen' group work by expressing
instructions in terms of larger types. instructions in terms of larger types.
This group is not yet implemented.
""") """)
expand = XFormGroup('expand', """ expand = XFormGroup('expand', """

4
lib/cretonne/meta/base/settings.py
View File

@@ -20,10 +20,10 @@ opt_level = EnumSetting(
enable_verifier = BoolSetting( enable_verifier = BoolSetting(
""" """
Run the Cretonne IL verifier at strategic times during compilation. Run the Cretonne IR verifier at strategic times during compilation.
This makes compilation slower but catches many bugs. The verifier is This makes compilation slower but catches many bugs. The verifier is
disabled by default, except when reading Cretonne IL from a text file. disabled by default, except when reading Cretonne IR from a text file.
""", """,
default=True) default=True)

10
lib/cretonne/meta/build.py
View File

@@ -14,7 +14,11 @@ import gen_legalizer
import gen_registers import gen_registers
import gen_binemit import gen_binemit
parser = argparse.ArgumentParser(description='Generate sources for Cretonne.')
def main():
# type: () -> None
parser = argparse.ArgumentParser(
description='Generate sources for Cretonne.')
parser.add_argument('--out-dir', help='set output directory') parser.add_argument('--out-dir', help='set output directory')
args = parser.parse_args() args = parser.parse_args()
@@ -30,3 +34,7 @@ gen_legalizer.generate(isas, out_dir)
gen_registers.generate(isas, out_dir) gen_registers.generate(isas, out_dir)
gen_binemit.generate(isas, out_dir) gen_binemit.generate(isas, out_dir)
gen_build_deps.generate() gen_build_deps.generate()
if __name__ == "__main__":
main()

2
lib/cretonne/meta/cdsl/ast.py
View File

@@ -559,7 +559,7 @@ class Enumerator(Literal):
is an AST leaf node representing one of the values. is an AST leaf node representing one of the values.
:param kind: The enumerated `ImmediateKind` containing the value. :param kind: The enumerated `ImmediateKind` containing the value.
:param value: The textual IL representation of the value. :param value: The textual IR representation of the value.
`Enumerator` nodes are not usually created directly. They are created by `Enumerator` nodes are not usually created directly. They are created by
using the dot syntax on immediate kinds: `intcc.ult`. using the dot syntax on immediate kinds: `intcc.ult`.

7
lib/cretonne/meta/cdsl/isa.py
View File

@@ -12,7 +12,7 @@ from .instructions import InstructionGroup
try: try:
from typing import Tuple, Union, Any, Iterable, Sequence, List, Set, Dict, TYPE_CHECKING # noqa from typing import Tuple, Union, Any, Iterable, Sequence, List, Set, Dict, TYPE_CHECKING # noqa
if TYPE_CHECKING: if TYPE_CHECKING:
from .instructions import MaybeBoundInst, InstructionGroup, InstructionFormat # noqa from .instructions import MaybeBoundInst, InstructionFormat # noqa
from .predicates import PredNode, PredKey # noqa from .predicates import PredNode, PredKey # noqa
from .settings import SettingGroup # noqa from .settings import SettingGroup # noqa
from .registers import RegBank # noqa from .registers import RegBank # noqa
@@ -172,8 +172,7 @@ class TargetISA(object):
""" """
for cpumode in self.cpumodes: for cpumode in self.cpumodes:
self.legalize_code(cpumode.default_legalize) self.legalize_code(cpumode.default_legalize)
for x in sorted(cpumode.type_legalize.values(), for x in cpumode.type_legalize.values():
key=lambda x: x.name):
self.legalize_code(x) self.legalize_code(x)
def legalize_code(self, xgrp): def legalize_code(self, xgrp):
@@ -232,7 +231,7 @@ class CPUMode(object):
# Tables for configuring legalization actions when no valid encoding # Tables for configuring legalization actions when no valid encoding
# exists for an instruction. # exists for an instruction.
self.default_legalize = None # type: XFormGroup self.default_legalize = None # type: XFormGroup
self.type_legalize = dict() # type: Dict[ValueType, XFormGroup] self.type_legalize = OrderedDict() # type: OrderedDict[ValueType, XFormGroup] # noqa
def __str__(self): def __str__(self):
# type: () -> str # type: () -> str

2
lib/cretonne/meta/check.sh
View File

@@ -2,7 +2,7 @@
set -euo pipefail set -euo pipefail
cd $(dirname "$0") cd $(dirname "$0")
runif() { function runif() {
if command -v "$1" > /dev/null; then if command -v "$1" > /dev/null; then
echo " === $1 ===" echo " === $1 ==="
"$@" "$@"

2
lib/cretonne/meta/gen_binemit.py
View File

@@ -152,7 +152,7 @@ def gen_isa(isa, fmt):
fmt.line('let bits = encoding.bits();') fmt.line('let bits = encoding.bits();')
with fmt.indented('match func.encodings[inst].recipe() {', '}'): with fmt.indented('match func.encodings[inst].recipe() {', '}'):
for i, recipe in enumerate(isa.all_recipes): for i, recipe in enumerate(isa.all_recipes):
fmt.comment(recipe.name) fmt.comment('Recipe {}'.format(recipe.name))
with fmt.indented('{} => {{'.format(i), '}'): with fmt.indented('{} => {{'.format(i), '}'):
gen_recipe(recipe, fmt) gen_recipe(recipe, fmt)
fmt.line('_ => {},') fmt.line('_ => {},')

8
lib/cretonne/meta/gen_encoding.py
View File

@@ -600,8 +600,8 @@ def make_tables(cpumode):
table[ty][inst].encodings.append(enc) table[ty][inst].encodings.append(enc)
# Ensure there are level 1 table entries for all types with a custom # Ensure there are level 1 table entries for all types with a custom
# legalize action. Try to be stable relative to dict ordering. # legalize action.
for ty in sorted(cpumode.type_legalize.keys(), key=str): for ty in cpumode.type_legalize.keys():
table[ty] table[ty]
return table return table
@@ -756,7 +756,7 @@ def emit_recipe_constraints(isa, fmt):
'static RECIPE_CONSTRAINTS: [RecipeConstraints; {}] = [' 'static RECIPE_CONSTRAINTS: [RecipeConstraints; {}] = ['
.format(len(isa.all_recipes)), '];'): .format(len(isa.all_recipes)), '];'):
for r in isa.all_recipes: for r in isa.all_recipes:
fmt.comment(r.name) fmt.comment('Constraints for recipe {}:'.format(r.name))
tied_i2o, tied_o2i = r.ties() tied_i2o, tied_o2i = r.ties()
fixed_ins, fixed_outs = r.fixed_ops() fixed_ins, fixed_outs = r.fixed_ops()
with fmt.indented('RecipeConstraints {', '},'): with fmt.indented('RecipeConstraints {', '},'):
@@ -830,7 +830,7 @@ def emit_recipe_sizing(isa, fmt):
'static RECIPE_SIZING: [RecipeSizing; {}] = [' 'static RECIPE_SIZING: [RecipeSizing; {}] = ['
.format(len(isa.all_recipes)), '];'): .format(len(isa.all_recipes)), '];'):
for r in isa.all_recipes: for r in isa.all_recipes:
fmt.comment(r.name) fmt.comment('Code size information for recipe {}:'.format(r.name))
with fmt.indented('RecipeSizing {', '},'): with fmt.indented('RecipeSizing {', '},'):
fmt.format('bytes: {},', r.size) fmt.format('bytes: {},', r.size)
if r.branch_range: if r.branch_range:

120
lib/cretonne/meta/gen_instr.py
View File

@@ -49,11 +49,11 @@ def gen_formats(fmt):
with fmt.indented( with fmt.indented(
"fn from(inst: &'a InstructionData) -> InstructionFormat {", "fn from(inst: &'a InstructionData) -> InstructionFormat {",
'}'): '}'):
with fmt.indented('match *inst {', '}'): m = srcgen.Match('*inst')
for f in InstructionFormat.all_formats: for f in InstructionFormat.all_formats:
fmt.line(('InstructionData::{} {{ .. }} => ' + m.arm('InstructionData::' + f.name, ['..'],
'InstructionFormat::{},') 'InstructionFormat::' + f.name)
.format(f.name, f.name)) fmt.match(m)
fmt.line() fmt.line()
@@ -74,7 +74,7 @@ def gen_arguments_method(fmt, is_mut):
'pool: &\'a {m}ir::ValueListPool) -> ' 'pool: &\'a {m}ir::ValueListPool) -> '
'&{m}[Value] {{' '&{m}[Value] {{'
.format(f=method, m=mut), '}'): .format(f=method, m=mut), '}'):
with fmt.indented('match *self {', '}'): m = srcgen.Match('*self')
for f in InstructionFormat.all_formats: for f in InstructionFormat.all_formats:
n = 'InstructionData::' + f.name n = 'InstructionData::' + f.name
@@ -82,25 +82,56 @@ def gen_arguments_method(fmt, is_mut):
# list. We don't split them up, just return it all as variable # list. We don't split them up, just return it all as variable
# arguments. (I expect the distinction to go away). # arguments. (I expect the distinction to go away).
if f.has_value_list: if f.has_value_list:
arg = ''.format(mut) m.arm(n, ['ref {}args'.format(mut), '..'],
fmt.line( 'args.{}(pool)'.format(as_slice))
'{} {{ ref {}args, .. }} => args.{}(pool),'
.format(n, mut, as_slice))
continue continue
# Fixed args. # Fixed args.
fields = []
if f.num_value_operands == 0: if f.num_value_operands == 0:
arg = '&{}[]'.format(mut) arg = '&{}[]'.format(mut)
capture = ''
elif f.num_value_operands == 1: elif f.num_value_operands == 1:
capture = 'ref {}arg, '.format(mut) fields.append('ref {}arg'.format(mut))
arg = '{}(arg)'.format(rslice) arg = '{}(arg)'.format(rslice)
else: else:
capture = 'ref {}args, '.format(mut) args = 'args_arity{}'.format(f.num_value_operands)
arg = 'args' fields.append('args: ref {}{}'.format(mut, args))
arg = args
fields.append('..')
m.arm(n, fields, arg)
fmt.match(m)
def gen_instruction_data(fmt):
# type: (srcgen.Formatter) -> None
"""
Generate the InstructionData enum.
Every variant must contain `opcode` and `ty` fields. An instruction that
doesn't produce a value should have its `ty` field set to `VOID`. The size
of `InstructionData` should be kept at 16 bytes on 64-bit architectures. If
more space is needed to represent an instruction, use a `Box<AuxData>` to
store the additional information out of line.
"""
fmt.line('#[derive(Clone, Debug, Hash, PartialEq, Eq)]')
fmt.line('#[allow(missing_docs)]')
with fmt.indented('pub enum InstructionData {', '}'):
for f in InstructionFormat.all_formats:
with fmt.indented('{} {{'.format(f.name), '},'):
fmt.line('opcode: Opcode,')
if f.typevar_operand is None:
pass
elif f.has_value_list:
fmt.line('args: ValueList,')
elif f.num_value_operands == 1:
fmt.line('arg: Value,')
else:
fmt.line('args: [Value; {}],'.format(f.num_value_operands))
for field in f.imm_fields:
fmt.line( fmt.line(
'{} {{ {}.. }} => {},' '{}: {},'
.format(n, capture, arg)) .format(field.member, field.kind.rust_type))
def gen_instruction_data_impl(fmt): def gen_instruction_data_impl(fmt):
@@ -123,39 +154,37 @@ def gen_instruction_data_impl(fmt):
with fmt.indented('impl InstructionData {', '}'): with fmt.indented('impl InstructionData {', '}'):
fmt.doc_comment('Get the opcode of this instruction.') fmt.doc_comment('Get the opcode of this instruction.')
with fmt.indented('pub fn opcode(&self) -> Opcode {', '}'): with fmt.indented('pub fn opcode(&self) -> Opcode {', '}'):
with fmt.indented('match *self {', '}'): m = srcgen.Match('*self')
for f in InstructionFormat.all_formats: for f in InstructionFormat.all_formats:
fmt.line( m.arm('InstructionData::' + f.name, ['opcode', '..'],
'InstructionData::{} {{ opcode, .. }} => opcode,' 'opcode')
.format(f.name)) fmt.match(m)
fmt.line() fmt.line()
fmt.doc_comment('Get the controlling type variable operand.') fmt.doc_comment('Get the controlling type variable operand.')
with fmt.indented( with fmt.indented(
'pub fn typevar_operand(&self, pool: &ir::ValueListPool) -> ' 'pub fn typevar_operand(&self, pool: &ir::ValueListPool) -> '
'Option<Value> {', '}'): 'Option<Value> {', '}'):
with fmt.indented('match *self {', '}'): m = srcgen.Match('*self')
for f in InstructionFormat.all_formats: for f in InstructionFormat.all_formats:
n = 'InstructionData::' + f.name n = 'InstructionData::' + f.name
if f.typevar_operand is None: if f.typevar_operand is None:
fmt.line(n + ' { .. } => None,') m.arm(n, ['..'], 'None')
elif f.has_value_list: elif f.has_value_list:
# We keep all arguments in a value list. # We keep all arguments in a value list.
i = f.typevar_operand i = f.typevar_operand
fmt.line( m.arm(n, ['ref args', '..'],
'{} {{ ref args, .. }} => ' 'args.get({}, pool)'.format(i))
'args.get({}, pool),'.format(n, i))
elif f.num_value_operands == 1: elif f.num_value_operands == 1:
# We have a single value operand called 'arg'. # We have a single value operand called 'arg'.
fmt.line(n + ' { arg, .. } => Some(arg),') m.arm(n, ['arg', '..'], 'Some(arg)')
else: else:
# We have multiple value operands and an array `args`. # We have multiple value operands and an array `args`.
# Which `args` index to use? # Which `args` index to use?
i = f.typevar_operand args = 'args_arity{}'.format(f.num_value_operands)
fmt.line( m.arm(n, ['args: ref {}'.format(args), '..'],
n + 'Some({}[{}])'.format(args, f.typevar_operand))
' {{ ref args, .. }} => Some(args[{}]),' fmt.match(m)
.format(i))
fmt.line() fmt.line()
fmt.doc_comment( fmt.doc_comment(
@@ -184,13 +213,13 @@ def gen_instruction_data_impl(fmt):
with fmt.indented( with fmt.indented(
'pub fn take_value_list(&mut self) -> Option<ir::ValueList> {', 'pub fn take_value_list(&mut self) -> Option<ir::ValueList> {',
'}'): '}'):
with fmt.indented('match *self {', '}'): m = srcgen.Match('*self')
for f in InstructionFormat.all_formats: for f in InstructionFormat.all_formats:
n = 'InstructionData::' + f.name n = 'InstructionData::' + f.name
if f.has_value_list: if f.has_value_list:
fmt.line( m.arm(n, ['ref mut args', '..'], 'Some(args.take())')
n + ' { ref mut args, .. } => Some(args.take()),') m.arm('_', [], 'None')
fmt.line('_ => None,') fmt.match(m)
fmt.line() fmt.line()
fmt.doc_comment( fmt.doc_comment(
@@ -275,14 +304,12 @@ def gen_opcodes(groups, fmt):
fmt.doc_comment(Instruction.ATTRIBS[attr]) fmt.doc_comment(Instruction.ATTRIBS[attr])
with fmt.indented('pub fn {}(self) -> bool {{' with fmt.indented('pub fn {}(self) -> bool {{'
.format(attr), '}'): .format(attr), '}'):
with fmt.indented('match self {', '}'): m = srcgen.Match('self')
for i in instrs: for i in instrs:
if getattr(i, attr): if getattr(i, attr):
fmt.format( m.arm('Opcode::' + i.camel_name, [], 'true')
'Opcode::{} => true,', m.arm('_', [], 'false')
i.camel_name, i.name) fmt.match(m)
fmt.line('_ => false,')
fmt.line() fmt.line()
fmt.line() fmt.line()
@@ -299,9 +326,10 @@ def gen_opcodes(groups, fmt):
# Generate a private opcode_name function. # Generate a private opcode_name function.
with fmt.indented('fn opcode_name(opc: Opcode) -> &\'static str {', '}'): with fmt.indented('fn opcode_name(opc: Opcode) -> &\'static str {', '}'):
with fmt.indented('match opc {', '}'): m = srcgen.Match('opc')
for i in instrs: for i in instrs:
fmt.format('Opcode::{} => "{}",', i.camel_name, i.name) m.arm('Opcode::' + i.camel_name, [], '"{}"'.format(i.name))
fmt.match(m)
fmt.line() fmt.line()
# Generate an opcode hash table for looking up opcodes by name. # Generate an opcode hash table for looking up opcodes by name.
@@ -655,7 +683,7 @@ def gen_builder(insts, fmt):
fmt.doc_comment(""" fmt.doc_comment("""
Convenience methods for building instructions. Convenience methods for building instructions.
The `InstrBuilder` trait has one method per instruction opcode for The `InstBuilder` trait has one method per instruction opcode for
conveniently constructing the instruction with minimum arguments. conveniently constructing the instruction with minimum arguments.
Polymorphic instructions infer their result types from the input Polymorphic instructions infer their result types from the input
arguments when possible. In some cases, an explicit `ctrl_typevar` arguments when possible. In some cases, an explicit `ctrl_typevar`
@@ -682,13 +710,15 @@ def generate(isas, out_dir):
# opcodes.rs # opcodes.rs
fmt = srcgen.Formatter() fmt = srcgen.Formatter()
gen_formats(fmt) gen_formats(fmt)
gen_instruction_data(fmt)
fmt.line()
gen_instruction_data_impl(fmt) gen_instruction_data_impl(fmt)
fmt.line() fmt.line()
instrs = gen_opcodes(groups, fmt) instrs = gen_opcodes(groups, fmt)
gen_type_constraints(fmt, instrs) gen_type_constraints(fmt, instrs)
fmt.update_file('opcodes.rs', out_dir) fmt.update_file('opcodes.rs', out_dir)
# builder.rs # inst_builder.rs
fmt = srcgen.Formatter() fmt = srcgen.Formatter()
gen_builder(instrs, fmt) gen_builder(instrs, fmt)
fmt.update_file('builder.rs', out_dir) fmt.update_file('inst_builder.rs', out_dir)

13
lib/cretonne/meta/gen_legalizer.py
View File

@@ -103,18 +103,19 @@ def emit_runtime_typecheck(check, fmt, type_sets):
base_exp = build_derived_expr(tv.base) base_exp = build_derived_expr(tv.base)
if (tv.derived_func == TypeVar.LANEOF): if (tv.derived_func == TypeVar.LANEOF):
return "{}.map(|t: Type| -> t.lane_type())".format(base_exp) return "{}.map(|t: ir::Type| t.lane_type())".format(base_exp)
elif (tv.derived_func == TypeVar.ASBOOL): elif (tv.derived_func == TypeVar.ASBOOL):
return "{}.map(|t: Type| -> t.as_bool())".format(base_exp) return "{}.map(|t: ir::Type| t.as_bool())".format(base_exp)
elif (tv.derived_func == TypeVar.HALFWIDTH): elif (tv.derived_func == TypeVar.HALFWIDTH):
return "{}.and_then(|t: Type| -> t.half_width())".format(base_exp) return "{}.and_then(|t: ir::Type| t.half_width())".format(base_exp)
elif (tv.derived_func == TypeVar.DOUBLEWIDTH): elif (tv.derived_func == TypeVar.DOUBLEWIDTH):
return "{}.and_then(|t: Type| -> t.double_width())"\ return "{}.and_then(|t: ir::Type| t.double_width())"\
.format(base_exp) .format(base_exp)
elif (tv.derived_func == TypeVar.HALFVECTOR): elif (tv.derived_func == TypeVar.HALFVECTOR):
return "{}.and_then(|t: Type| -> t.half_vector())".format(base_exp) return "{}.and_then(|t: ir::Type| t.half_vector())"\
.format(base_exp)
elif (tv.derived_func == TypeVar.DOUBLEVECTOR): elif (tv.derived_func == TypeVar.DOUBLEVECTOR):
return "{}.and_then(|t: Type| -> t.by(2))".format(base_exp) return "{}.and_then(|t: ir::Type| t.by(2))".format(base_exp)
else: else:
assert False, "Unknown derived function {}".format(tv.derived_func) assert False, "Unknown derived function {}".format(tv.derived_func)

11
lib/cretonne/meta/gen_settings.py
View File

@@ -28,7 +28,7 @@ def gen_enum_types(sgrp, fmt):
if not isinstance(setting, EnumSetting): if not isinstance(setting, EnumSetting):
continue continue
ty = camel_case(setting.name) ty = camel_case(setting.name)
fmt.doc_comment('Values for {}.'.format(setting)) fmt.doc_comment('Values for `{}`.'.format(setting))
fmt.line('#[derive(Debug, PartialEq, Eq)]') fmt.line('#[derive(Debug, PartialEq, Eq)]')
with fmt.indented('pub enum {} {{'.format(ty), '}'): with fmt.indented('pub enum {} {{'.format(ty), '}'):
for v in setting.values: for v in setting.values:
@@ -57,12 +57,11 @@ def gen_getter(setting, sgrp, fmt):
ty = camel_case(setting.name) ty = camel_case(setting.name)
proto = 'pub fn {}(&self) -> {}'.format(setting.name, ty) proto = 'pub fn {}(&self) -> {}'.format(setting.name, ty)
with fmt.indented(proto + ' {', '}'): with fmt.indented(proto + ' {', '}'):
with fmt.indented( m = srcgen.Match('self.bytes[{}]'.format(setting.byte_offset))
'match self.bytes[{}] {{'
.format(setting.byte_offset), '}'):
for i, v in enumerate(setting.values): for i, v in enumerate(setting.values):
fmt.line('{} => {}::{},'.format(i, ty, camel_case(v))) m.arm(str(i), [], '{}::{}'.format(ty, camel_case(v)))
fmt.line('_ => panic!("Invalid enum value"),') m.arm('_', [], 'panic!("Invalid enum value")')
fmt.match(m)
else: else:
raise AssertionError("Unknown setting kind") raise AssertionError("Unknown setting kind")

4
lib/cretonne/meta/isa/intel/defs.py
View File

@@ -12,8 +12,8 @@ from base.immediates import floatcc
ISA = TargetISA('intel', [base.instructions.GROUP, x86.GROUP]) ISA = TargetISA('intel', [base.instructions.GROUP, x86.GROUP])
# CPU modes for 32-bit and 64-bit operation. # CPU modes for 32-bit and 64-bit operation.
I64 = CPUMode('I64', ISA) X86_64 = CPUMode('I64', ISA)
I32 = CPUMode('I32', ISA) X86_32 = CPUMode('I32', ISA)
# The set of floating point condition codes that are directly supported. # The set of floating point condition codes that are directly supported.
# Other condition codes need to be reversed or expressed as two tests. # Other condition codes need to be reversed or expressed as two tests.

331
lib/cretonne/meta/isa/intel/encodings.py
View File

@@ -5,7 +5,7 @@ from __future__ import absolute_import
from cdsl.predicates import IsUnsignedInt, Not, And from cdsl.predicates import IsUnsignedInt, Not, And
from base import instructions as base from base import instructions as base
from base.formats import UnaryImm from base.formats import UnaryImm
from .defs import I32, I64 from .defs import X86_64, X86_32
from . import recipes as r from . import recipes as r
from . import settings as cfg from . import settings as cfg
from . import instructions as x86 from . import instructions as x86
@@ -22,16 +22,16 @@ except ImportError:
pass pass
I32.legalize_monomorphic(expand_flags) X86_32.legalize_monomorphic(expand_flags)
I32.legalize_type( X86_32.legalize_type(
default=narrow, default=narrow,
b1=expand_flags, b1=expand_flags,
i32=intel_expand, i32=intel_expand,
f32=intel_expand, f32=intel_expand,
f64=intel_expand) f64=intel_expand)
I64.legalize_monomorphic(expand_flags) X86_64.legalize_monomorphic(expand_flags)
I64.legalize_type( X86_64.legalize_type(
default=narrow, default=narrow,
b1=expand_flags, b1=expand_flags,
i32=intel_expand, i32=intel_expand,
@@ -44,61 +44,61 @@ I64.legalize_type(
# Helper functions for generating encodings. # Helper functions for generating encodings.
# #
def enc_i64(inst, recipe, *args, **kwargs): def enc_x86_64(inst, recipe, *args, **kwargs):
# type: (MaybeBoundInst, r.TailRecipe, *int, **int) -> None # type: (MaybeBoundInst, r.TailRecipe, *int, **int) -> None
""" """
Add encodings for `inst` to I64 with and without a REX prefix. Add encodings for `inst` to X86_64 with and without a REX prefix.
""" """
I64.enc(inst, *recipe.rex(*args, **kwargs)) X86_64.enc(inst, *recipe.rex(*args, **kwargs))
I64.enc(inst, *recipe(*args, **kwargs)) X86_64.enc(inst, *recipe(*args, **kwargs))
def enc_both(inst, recipe, *args, **kwargs): def enc_both(inst, recipe, *args, **kwargs):
# type: (MaybeBoundInst, r.TailRecipe, *int, **Any) -> None # type: (MaybeBoundInst, r.TailRecipe, *int, **Any) -> None
""" """
Add encodings for `inst` to both I32 and I64. Add encodings for `inst` to both X86_32 and X86_64.
""" """
I32.enc(inst, *recipe(*args, **kwargs)) X86_32.enc(inst, *recipe(*args, **kwargs))
enc_i64(inst, recipe, *args, **kwargs) enc_x86_64(inst, recipe, *args, **kwargs)
def enc_i32_i64(inst, recipe, *args, **kwargs): def enc_i32_i64(inst, recipe, *args, **kwargs):
# type: (MaybeBoundInst, r.TailRecipe, *int, **int) -> None # type: (MaybeBoundInst, r.TailRecipe, *int, **int) -> None
""" """
Add encodings for `inst.i32` to I32. Add encodings for `inst.i32` to X86_32.
Add encodings for `inst.i32` to I64 with and without REX. Add encodings for `inst.i32` to X86_64 with and without REX.
Add encodings for `inst.i64` to I64 with a REX.W prefix. Add encodings for `inst.i64` to X86_64 with a REX.W prefix.
""" """
I32.enc(inst.i32, *recipe(*args, **kwargs)) X86_32.enc(inst.i32, *recipe(*args, **kwargs))
# REX-less encoding must come after REX encoding so we don't use it by # REX-less encoding must come after REX encoding so we don't use it by
# default. Otherwise reg-alloc would never use r8 and up. # default. Otherwise reg-alloc would never use r8 and up.
I64.enc(inst.i32, *recipe.rex(*args, **kwargs)) X86_64.enc(inst.i32, *recipe.rex(*args, **kwargs))
I64.enc(inst.i32, *recipe(*args, **kwargs)) X86_64.enc(inst.i32, *recipe(*args, **kwargs))
I64.enc(inst.i64, *recipe.rex(*args, w=1, **kwargs)) X86_64.enc(inst.i64, *recipe.rex(*args, w=1, **kwargs))
def enc_i32_i64_ld_st(inst, w_bit, recipe, *args, **kwargs): def enc_i32_i64_ld_st(inst, w_bit, recipe, *args, **kwargs):
# type: (MaybeBoundInst, bool, r.TailRecipe, *int, **int) -> None # type: (MaybeBoundInst, bool, r.TailRecipe, *int, **int) -> None
""" """
Add encodings for `inst.i32` to I32. Add encodings for `inst.i32` to X86_32.
Add encodings for `inst.i32` to I64 with and without REX. Add encodings for `inst.i32` to X86_64 with and without REX.
Add encodings for `inst.i64` to I64 with a REX prefix, using the `w_bit` Add encodings for `inst.i64` to X86_64 with a REX prefix, using the `w_bit`
argument to determine whether or not to set the REX.W bit. argument to determine whether or not to set the REX.W bit.
""" """
I32.enc(inst.i32.any, *recipe(*args, **kwargs)) X86_32.enc(inst.i32.any, *recipe(*args, **kwargs))
# REX-less encoding must come after REX encoding so we don't use it by # REX-less encoding must come after REX encoding so we don't use it by
# default. Otherwise reg-alloc would never use r8 and up. # default. Otherwise reg-alloc would never use r8 and up.
I64.enc(inst.i32.any, *recipe.rex(*args, **kwargs)) X86_64.enc(inst.i32.any, *recipe.rex(*args, **kwargs))
I64.enc(inst.i32.any, *recipe(*args, **kwargs)) X86_64.enc(inst.i32.any, *recipe(*args, **kwargs))
if w_bit: if w_bit:
I64.enc(inst.i64.any, *recipe.rex(*args, w=1, **kwargs)) X86_64.enc(inst.i64.any, *recipe.rex(*args, w=1, **kwargs))
else: else:
I64.enc(inst.i64.any, *recipe.rex(*args, **kwargs)) X86_64.enc(inst.i64.any, *recipe.rex(*args, **kwargs))
I64.enc(inst.i64.any, *recipe(*args, **kwargs)) X86_64.enc(inst.i64.any, *recipe(*args, **kwargs))
for inst, opc in [ for inst, opc in [
@@ -141,19 +141,22 @@ for inst, rrr in [
# band_imm.i32. Can even use the single-byte immediate for 0xffff_ffXX masks. # band_imm.i32. Can even use the single-byte immediate for 0xffff_ffXX masks.
# Immediate constants. # Immediate constants.
I32.enc(base.iconst.i32, *r.puid(0xb8)) X86_32.enc(base.iconst.i32, *r.puid(0xb8))
I64.enc(base.iconst.i32, *r.puid.rex(0xb8)) X86_64.enc(base.iconst.i32, *r.puid.rex(0xb8))
I64.enc(base.iconst.i32, *r.puid(0xb8)) X86_64.enc(base.iconst.i32, *r.puid(0xb8))
# The 32-bit immediate movl also zero-extends to 64 bits. # The 32-bit immediate movl also zero-extends to 64 bits.
I64.enc(base.iconst.i64, *r.puid.rex(0xb8), X86_64.enc(base.iconst.i64, *r.puid.rex(0xb8),
instp=IsUnsignedInt(UnaryImm.imm, 32)) instp=IsUnsignedInt(UnaryImm.imm, 32))
I64.enc(base.iconst.i64, *r.puid(0xb8), X86_64.enc(base.iconst.i64, *r.puid(0xb8),
instp=IsUnsignedInt(UnaryImm.imm, 32)) instp=IsUnsignedInt(UnaryImm.imm, 32))
# Sign-extended 32-bit immediate. # Sign-extended 32-bit immediate.
I64.enc(base.iconst.i64, *r.uid.rex(0xc7, rrr=0, w=1)) X86_64.enc(base.iconst.i64, *r.uid.rex(0xc7, rrr=0, w=1))
# Finally, the 0xb8 opcode takes an 8-byte immediate with a REX.W prefix. # Finally, the 0xb8 opcode takes an 8-byte immediate with a REX.W prefix.
I64.enc(base.iconst.i64, *r.puiq.rex(0xb8, w=1)) X86_64.enc(base.iconst.i64, *r.puiq.rex(0xb8, w=1))
# bool constants.
enc_both(base.bconst.b1, r.puid_bool, 0xb8)
# Shifts and rotates. # Shifts and rotates.
# Note that the dynamic shift amount is only masked by 5 or 6 bits; the 8-bit # Note that the dynamic shift amount is only masked by 5 or 6 bits; the 8-bit
@@ -164,38 +167,46 @@ for inst, rrr in [
(base.ishl, 4), (base.ishl, 4),
(base.ushr, 5), (base.ushr, 5),
(base.sshr, 7)]: (base.sshr, 7)]:
I32.enc(inst.i32.any, *r.rc(0xd3, rrr=rrr)) # Cannot use enc_i32_i64 for this pattern because instructions require
I64.enc(inst.i64.any, *r.rc.rex(0xd3, rrr=rrr, w=1)) # .any suffix.
I64.enc(inst.i32.any, *r.rc.rex(0xd3, rrr=rrr)) X86_32.enc(inst.i32.any, *r.rc(0xd3, rrr=rrr))
I64.enc(inst.i32.any, *r.rc(0xd3, rrr=rrr)) X86_64.enc(inst.i64.any, *r.rc.rex(0xd3, rrr=rrr, w=1))
X86_64.enc(inst.i32.any, *r.rc.rex(0xd3, rrr=rrr))
X86_64.enc(inst.i32.any, *r.rc(0xd3, rrr=rrr))
for inst, rrr in [
(base.ishl_imm, 4),
(base.ushr_imm, 5),
(base.sshr_imm, 7)]:
enc_i32_i64(inst, r.rib, 0xc1, rrr=rrr)
# Population count. # Population count.
I32.enc(base.popcnt.i32, *r.urm(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt) X86_32.enc(base.popcnt.i32, *r.urm(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt)
I64.enc(base.popcnt.i64, *r.urm.rex(0xf3, 0x0f, 0xb8, w=1), X86_64.enc(base.popcnt.i64, *r.urm.rex(0xf3, 0x0f, 0xb8, w=1),
isap=cfg.use_popcnt) isap=cfg.use_popcnt)
I64.enc(base.popcnt.i32, *r.urm.rex(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt) X86_64.enc(base.popcnt.i32, *r.urm.rex(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt)
I64.enc(base.popcnt.i32, *r.urm(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt) X86_64.enc(base.popcnt.i32, *r.urm(0xf3, 0x0f, 0xb8), isap=cfg.use_popcnt)
# Count leading zero bits. # Count leading zero bits.
I32.enc(base.clz.i32, *r.urm(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt) X86_32.enc(base.clz.i32, *r.urm(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt)
I64.enc(base.clz.i64, *r.urm.rex(0xf3, 0x0f, 0xbd, w=1), X86_64.enc(base.clz.i64, *r.urm.rex(0xf3, 0x0f, 0xbd, w=1),
isap=cfg.use_lzcnt) isap=cfg.use_lzcnt)
I64.enc(base.clz.i32, *r.urm.rex(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt) X86_64.enc(base.clz.i32, *r.urm.rex(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt)
I64.enc(base.clz.i32, *r.urm(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt) X86_64.enc(base.clz.i32, *r.urm(0xf3, 0x0f, 0xbd), isap=cfg.use_lzcnt)
# Count trailing zero bits. # Count trailing zero bits.
I32.enc(base.ctz.i32, *r.urm(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1) X86_32.enc(base.ctz.i32, *r.urm(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1)
I64.enc(base.ctz.i64, *r.urm.rex(0xf3, 0x0f, 0xbc, w=1), X86_64.enc(base.ctz.i64, *r.urm.rex(0xf3, 0x0f, 0xbc, w=1),
isap=cfg.use_bmi1) isap=cfg.use_bmi1)
I64.enc(base.ctz.i32, *r.urm.rex(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1) X86_64.enc(base.ctz.i32, *r.urm.rex(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1)
I64.enc(base.ctz.i32, *r.urm(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1) X86_64.enc(base.ctz.i32, *r.urm(0xf3, 0x0f, 0xbc), isap=cfg.use_bmi1)
# #
# Loads and stores. # Loads and stores.
# #
for recipe in [r.st, r.stDisp8, r.stDisp32]: for recipe in [r.st, r.stDisp8, r.stDisp32]:
enc_i32_i64_ld_st(base.store, True, recipe, 0x89) enc_i32_i64_ld_st(base.store, True, recipe, 0x89)
enc_i64(base.istore32.i64.any, recipe, 0x89) enc_x86_64(base.istore32.i64.any, recipe, 0x89)
enc_i32_i64_ld_st(base.istore16, False, recipe, 0x66, 0x89) enc_i32_i64_ld_st(base.istore16, False, recipe, 0x66, 0x89)
# Byte stores are more complicated because the registers they can address # Byte stores are more complicated because the registers they can address
@@ -203,121 +214,121 @@ for recipe in [r.st, r.stDisp8, r.stDisp32]:
# the corresponding st* recipes when a REX prefix is applied. # the corresponding st* recipes when a REX prefix is applied.
for recipe in [r.st_abcd, r.stDisp8_abcd, r.stDisp32_abcd]: for recipe in [r.st_abcd, r.stDisp8_abcd, r.stDisp32_abcd]:
enc_both(base.istore8.i32.any, recipe, 0x88) enc_both(base.istore8.i32.any, recipe, 0x88)
enc_i64(base.istore8.i64.any, recipe, 0x88) enc_x86_64(base.istore8.i64.any, recipe, 0x88)
enc_i32_i64(base.spill, r.spSib32, 0x89) enc_i32_i64(base.spill, r.spillSib32, 0x89)
enc_i32_i64(base.regspill, r.rsp32, 0x89) enc_i32_i64(base.regspill, r.regspill32, 0x89)
# Use a 32-bit write for spilling `b1` to avoid constraining the permitted # Use a 32-bit write for spilling `b1` to avoid constraining the permitted
# registers. # registers.
# See MIN_SPILL_SLOT_SIZE which makes this safe. # See MIN_SPILL_SLOT_SIZE which makes this safe.
enc_both(base.spill.b1, r.spSib32, 0x89) enc_both(base.spill.b1, r.spillSib32, 0x89)
enc_both(base.regspill.b1, r.rsp32, 0x89) enc_both(base.regspill.b1, r.regspill32, 0x89)
for recipe in [r.ld, r.ldDisp8, r.ldDisp32]: for recipe in [r.ld, r.ldDisp8, r.ldDisp32]:
enc_i32_i64_ld_st(base.load, True, recipe, 0x8b) enc_i32_i64_ld_st(base.load, True, recipe, 0x8b)
enc_i64(base.uload32.i64, recipe, 0x8b) enc_x86_64(base.uload32.i64, recipe, 0x8b)
I64.enc(base.sload32.i64, *recipe.rex(0x63, w=1)) X86_64.enc(base.sload32.i64, *recipe.rex(0x63, w=1))
enc_i32_i64_ld_st(base.uload16, True, recipe, 0x0f, 0xb7) enc_i32_i64_ld_st(base.uload16, True, recipe, 0x0f, 0xb7)
enc_i32_i64_ld_st(base.sload16, True, recipe, 0x0f, 0xbf) enc_i32_i64_ld_st(base.sload16, True, recipe, 0x0f, 0xbf)
enc_i32_i64_ld_st(base.uload8, True, recipe, 0x0f, 0xb6) enc_i32_i64_ld_st(base.uload8, True, recipe, 0x0f, 0xb6)
enc_i32_i64_ld_st(base.sload8, True, recipe, 0x0f, 0xbe) enc_i32_i64_ld_st(base.sload8, True, recipe, 0x0f, 0xbe)
enc_i32_i64(base.fill, r.fiSib32, 0x8b) enc_i32_i64(base.fill, r.fillSib32, 0x8b)
enc_i32_i64(base.regfill, r.rfi32, 0x8b) enc_i32_i64(base.regfill, r.regfill32, 0x8b)
# Load 32 bits from `b1` spill slots. See `spill.b1` above. # Load 32 bits from `b1` spill slots. See `spill.b1` above.
enc_both(base.fill.b1, r.fiSib32, 0x8b) enc_both(base.fill.b1, r.fillSib32, 0x8b)
enc_both(base.regfill.b1, r.rfi32, 0x8b) enc_both(base.regfill.b1, r.regfill32, 0x8b)
# Push and Pop # Push and Pop
I32.enc(x86.push.i32, *r.pushq(0x50)) X86_32.enc(x86.push.i32, *r.pushq(0x50))
enc_i64(x86.push.i64, r.pushq, 0x50) enc_x86_64(x86.push.i64, r.pushq, 0x50)
I32.enc(x86.pop.i32, *r.popq(0x58)) X86_32.enc(x86.pop.i32, *r.popq(0x58))
enc_i64(x86.pop.i64, r.popq, 0x58) enc_x86_64(x86.pop.i64, r.popq, 0x58)
# Copy Special # Copy Special
I64.enc(base.copy_special, *r.copysp.rex(0x89, w=1)) X86_64.enc(base.copy_special, *r.copysp.rex(0x89, w=1))
I32.enc(base.copy_special, *r.copysp(0x89)) X86_32.enc(base.copy_special, *r.copysp(0x89))
# Adjust SP Imm # Adjust SP Imm
I32.enc(base.adjust_sp_imm, *r.adjustsp8(0x83)) X86_32.enc(base.adjust_sp_imm, *r.adjustsp8(0x83))
I32.enc(base.adjust_sp_imm, *r.adjustsp32(0x81)) X86_32.enc(base.adjust_sp_imm, *r.adjustsp32(0x81))
I64.enc(base.adjust_sp_imm, *r.adjustsp8.rex(0x83, w=1)) X86_64.enc(base.adjust_sp_imm, *r.adjustsp8.rex(0x83, w=1))
I64.enc(base.adjust_sp_imm, *r.adjustsp32.rex(0x81, w=1)) X86_64.enc(base.adjust_sp_imm, *r.adjustsp32.rex(0x81, w=1))
# #
# Float loads and stores. # Float loads and stores.
# #
enc_both(base.load.f32.any, r.fld, 0x66, 0x0f, 0x6e) enc_both(base.load.f32.any, r.fld, 0xf3, 0x0f, 0x10)
enc_both(base.load.f32.any, r.fldDisp8, 0x66, 0x0f, 0x6e) enc_both(base.load.f32.any, r.fldDisp8, 0xf3, 0x0f, 0x10)
enc_both(base.load.f32.any, r.fldDisp32, 0x66, 0x0f, 0x6e) enc_both(base.load.f32.any, r.fldDisp32, 0xf3, 0x0f, 0x10)
enc_both(base.load.f64.any, r.fld, 0xf3, 0x0f, 0x7e) enc_both(base.load.f64.any, r.fld, 0xf2, 0x0f, 0x10)
enc_both(base.load.f64.any, r.fldDisp8, 0xf3, 0x0f, 0x7e) enc_both(base.load.f64.any, r.fldDisp8, 0xf2, 0x0f, 0x10)
enc_both(base.load.f64.any, r.fldDisp32, 0xf3, 0x0f, 0x7e) enc_both(base.load.f64.any, r.fldDisp32, 0xf2, 0x0f, 0x10)
enc_both(base.store.f32.any, r.fst, 0x66, 0x0f, 0x7e) enc_both(base.store.f32.any, r.fst, 0xf3, 0x0f, 0x11)
enc_both(base.store.f32.any, r.fstDisp8, 0x66, 0x0f, 0x7e) enc_both(base.store.f32.any, r.fstDisp8, 0xf3, 0x0f, 0x11)
enc_both(base.store.f32.any, r.fstDisp32, 0x66, 0x0f, 0x7e) enc_both(base.store.f32.any, r.fstDisp32, 0xf3, 0x0f, 0x11)
enc_both(base.store.f64.any, r.fst, 0x66, 0x0f, 0xd6) enc_both(base.store.f64.any, r.fst, 0xf2, 0x0f, 0x11)
enc_both(base.store.f64.any, r.fstDisp8, 0x66, 0x0f, 0xd6) enc_both(base.store.f64.any, r.fstDisp8, 0xf2, 0x0f, 0x11)
enc_both(base.store.f64.any, r.fstDisp32, 0x66, 0x0f, 0xd6) enc_both(base.store.f64.any, r.fstDisp32, 0xf2, 0x0f, 0x11)
enc_both(base.fill.f32, r.ffiSib32, 0x66, 0x0f, 0x6e) enc_both(base.fill.f32, r.ffillSib32, 0xf3, 0x0f, 0x10)
enc_both(base.regfill.f32, r.frfi32, 0x66, 0x0f, 0x6e) enc_both(base.regfill.f32, r.fregfill32, 0xf3, 0x0f, 0x10)
enc_both(base.fill.f64, r.ffiSib32, 0xf3, 0x0f, 0x7e) enc_both(base.fill.f64, r.ffillSib32, 0xf2, 0x0f, 0x10)
enc_both(base.regfill.f64, r.frfi32, 0xf3, 0x0f, 0x7e) enc_both(base.regfill.f64, r.fregfill32, 0xf2, 0x0f, 0x10)
enc_both(base.spill.f32, r.fspSib32, 0x66, 0x0f, 0x7e) enc_both(base.spill.f32, r.fspillSib32, 0xf3, 0x0f, 0x11)
enc_both(base.regspill.f32, r.frsp32, 0x66, 0x0f, 0x7e) enc_both(base.regspill.f32, r.fregspill32, 0xf3, 0x0f, 0x11)
enc_both(base.spill.f64, r.fspSib32, 0x66, 0x0f, 0xd6) enc_both(base.spill.f64, r.fspillSib32, 0xf2, 0x0f, 0x11)
enc_both(base.regspill.f64, r.frsp32, 0x66, 0x0f, 0xd6) enc_both(base.regspill.f64, r.fregspill32, 0xf2, 0x0f, 0x11)
# #
# Function addresses. # Function addresses.
# #
I32.enc(base.func_addr.i32, *r.fnaddr4(0xb8), X86_32.enc(base.func_addr.i32, *r.fnaddr4(0xb8),
isap=Not(allones_funcaddrs)) isap=Not(allones_funcaddrs))
I64.enc(base.func_addr.i64, *r.fnaddr8.rex(0xb8, w=1), X86_64.enc(base.func_addr.i64, *r.fnaddr8.rex(0xb8, w=1),
isap=And(Not(allones_funcaddrs), Not(is_pic))) isap=And(Not(allones_funcaddrs), Not(is_pic)))
I32.enc(base.func_addr.i32, *r.allones_fnaddr4(0xb8), X86_32.enc(base.func_addr.i32, *r.allones_fnaddr4(0xb8),
isap=allones_funcaddrs) isap=allones_funcaddrs)
I64.enc(base.func_addr.i64, *r.allones_fnaddr8.rex(0xb8, w=1), X86_64.enc(base.func_addr.i64, *r.allones_fnaddr8.rex(0xb8, w=1),
isap=And(allones_funcaddrs, Not(is_pic))) isap=And(allones_funcaddrs, Not(is_pic)))
I64.enc(base.func_addr.i64, *r.got_fnaddr8.rex(0x8b, w=1), X86_64.enc(base.func_addr.i64, *r.got_fnaddr8.rex(0x8b, w=1),
isap=is_pic) isap=is_pic)
# #
# Global addresses. # Global addresses.
# #
I32.enc(base.globalsym_addr.i32, *r.gvaddr4(0xb8)) X86_32.enc(base.globalsym_addr.i32, *r.gvaddr4(0xb8))
I64.enc(base.globalsym_addr.i64, *r.gvaddr8.rex(0xb8, w=1), X86_64.enc(base.globalsym_addr.i64, *r.gvaddr8.rex(0xb8, w=1),
isap=Not(is_pic)) isap=Not(is_pic))
I64.enc(base.globalsym_addr.i64, *r.got_gvaddr8.rex(0x8b, w=1), X86_64.enc(base.globalsym_addr.i64, *r.got_gvaddr8.rex(0x8b, w=1),
isap=is_pic) isap=is_pic)
# #
# Call/return # Call/return
# #
I32.enc(base.call, *r.call_id(0xe8)) X86_32.enc(base.call, *r.call_id(0xe8))
I64.enc(base.call, *r.call_id(0xe8), isap=Not(is_pic)) X86_64.enc(base.call, *r.call_id(0xe8), isap=Not(is_pic))
I64.enc(base.call, *r.call_plt_id(0xe8), isap=is_pic) X86_64.enc(base.call, *r.call_plt_id(0xe8), isap=is_pic)
I32.enc(base.call_indirect.i32, *r.call_r(0xff, rrr=2)) X86_32.enc(base.call_indirect.i32, *r.call_r(0xff, rrr=2))
I64.enc(base.call_indirect.i64, *r.call_r.rex(0xff, rrr=2)) X86_64.enc(base.call_indirect.i64, *r.call_r.rex(0xff, rrr=2))
I64.enc(base.call_indirect.i64, *r.call_r(0xff, rrr=2)) X86_64.enc(base.call_indirect.i64, *r.call_r(0xff, rrr=2))
I32.enc(base.x_return, *r.ret(0xc3)) X86_32.enc(base.x_return, *r.ret(0xc3))
I64.enc(base.x_return, *r.ret(0xc3)) X86_64.enc(base.x_return, *r.ret(0xc3))
# #
# Branches # Branches
@@ -341,10 +352,10 @@ enc_i32_i64(base.brnz, r.tjccd, 0x85)
# Branch on a b1 value in a register only looks at the low 8 bits. See also # Branch on a b1 value in a register only looks at the low 8 bits. See also
# bint encodings below. # bint encodings below.
# #
# Start with the worst-case encoding for I32 only. The register allocator can't # Start with the worst-case encoding for X86_32 only. The register allocator
# handle a branch with an ABCD-constrained operand. # can't handle a branch with an ABCD-constrained operand.
I32.enc(base.brz.b1, *r.t8jccd_long(0x84)) X86_32.enc(base.brz.b1, *r.t8jccd_long(0x84))
I32.enc(base.brnz.b1, *r.t8jccd_long(0x85)) X86_32.enc(base.brnz.b1, *r.t8jccd_long(0x85))
enc_both(base.brz.b1, r.t8jccb_abcd, 0x74) enc_both(base.brz.b1, r.t8jccb_abcd, 0x74)
enc_both(base.brz.b1, r.t8jccd_abcd, 0x84) enc_both(base.brz.b1, r.t8jccd_abcd, 0x84)
@@ -354,26 +365,28 @@ enc_both(base.brnz.b1, r.t8jccd_abcd, 0x85)
# #
# Trap as ud2 # Trap as ud2
# #
I32.enc(base.trap, *r.trap(0x0f, 0x0b)) X86_32.enc(base.trap, *r.trap(0x0f, 0x0b))
I64.enc(base.trap, *r.trap(0x0f, 0x0b)) X86_64.enc(base.trap, *r.trap(0x0f, 0x0b))
# Using a standard EncRecipe, not the TailRecipe. # Using a standard EncRecipe, not the TailRecipe.
I32.enc(base.trapif, r.trapif, 0) X86_32.enc(base.trapif, r.trapif, 0)
I64.enc(base.trapif, r.trapif, 0) X86_64.enc(base.trapif, r.trapif, 0)
I32.enc(base.trapff, r.trapff, 0) X86_32.enc(base.trapff, r.trapff, 0)
I64.enc(base.trapff, r.trapff, 0) X86_64.enc(base.trapff, r.trapff, 0)
# #
# Comparisons # Comparisons
# #
enc_i32_i64(base.icmp, r.icscc, 0x39) enc_i32_i64(base.icmp, r.icscc, 0x39)
enc_i32_i64(base.icmp_imm, r.icsccib, 0x83, rrr=7)
enc_i32_i64(base.icmp_imm, r.icsccid, 0x81, rrr=7)
enc_i32_i64(base.ifcmp, r.rcmp, 0x39) enc_i32_i64(base.ifcmp, r.rcmp, 0x39)
enc_i32_i64(base.ifcmp_imm, r.rcmpib, 0x83, rrr=7) enc_i32_i64(base.ifcmp_imm, r.rcmpib, 0x83, rrr=7)
enc_i32_i64(base.ifcmp_imm, r.rcmpid, 0x81, rrr=7) enc_i32_i64(base.ifcmp_imm, r.rcmpid, 0x81, rrr=7)
# TODO: We could special-case ifcmp_imm(x, 0) to TEST(x, x). # TODO: We could special-case ifcmp_imm(x, 0) to TEST(x, x).
I32.enc(base.ifcmp_sp.i32, *r.rcmp_sp(0x39)) X86_32.enc(base.ifcmp_sp.i32, *r.rcmp_sp(0x39))
I64.enc(base.ifcmp_sp.i64, *r.rcmp_sp.rex(0x39, w=1)) X86_64.enc(base.ifcmp_sp.i64, *r.rcmp_sp.rex(0x39, w=1))
# #
# Convert flags to bool. # Convert flags to bool.
@@ -398,66 +411,68 @@ enc_i32_i64(x86.bsr, r.bsf_and_bsr, 0x0F, 0xBD)
# #
# This assumes that b1 is represented as an 8-bit low register with the value 0 # This assumes that b1 is represented as an 8-bit low register with the value 0
# or 1. # or 1.
I32.enc(base.bint.i32.b1, *r.urm_abcd(0x0f, 0xb6)) #
I64.enc(base.bint.i64.b1, *r.urm.rex(0x0f, 0xb6)) # zext to i64 implicit. # Encode movzbq as movzbl, because it's equivalent and shorter.
I64.enc(base.bint.i64.b1, *r.urm_abcd(0x0f, 0xb6)) # zext to i64 implicit. X86_32.enc(base.bint.i32.b1, *r.urm_noflags_abcd(0x0f, 0xb6))
I64.enc(base.bint.i32.b1, *r.urm.rex(0x0f, 0xb6)) X86_64.enc(base.bint.i64.b1, *r.urm_noflags.rex(0x0f, 0xb6))
I64.enc(base.bint.i32.b1, *r.urm_abcd(0x0f, 0xb6)) X86_64.enc(base.bint.i64.b1, *r.urm_noflags_abcd(0x0f, 0xb6))
X86_64.enc(base.bint.i32.b1, *r.urm_noflags.rex(0x0f, 0xb6))
X86_64.enc(base.bint.i32.b1, *r.urm_noflags_abcd(0x0f, 0xb6))
# Numerical conversions. # Numerical conversions.
# Reducing an integer is a no-op. # Reducing an integer is a no-op.
I32.enc(base.ireduce.i8.i32, r.null, 0) X86_32.enc(base.ireduce.i8.i32, r.null, 0)
I32.enc(base.ireduce.i16.i32, r.null, 0) X86_32.enc(base.ireduce.i16.i32, r.null, 0)
I64.enc(base.ireduce.i8.i32, r.null, 0) X86_64.enc(base.ireduce.i8.i32, r.null, 0)
I64.enc(base.ireduce.i16.i32, r.null, 0) X86_64.enc(base.ireduce.i16.i32, r.null, 0)
I64.enc(base.ireduce.i8.i64, r.null, 0) X86_64.enc(base.ireduce.i8.i64, r.null, 0)
I64.enc(base.ireduce.i16.i64, r.null, 0) X86_64.enc(base.ireduce.i16.i64, r.null, 0)
I64.enc(base.ireduce.i32.i64, r.null, 0) X86_64.enc(base.ireduce.i32.i64, r.null, 0)
# TODO: Add encodings for cbw, cwde, cdqe, which are sign-extending # TODO: Add encodings for cbw, cwde, cdqe, which are sign-extending
# instructions for %al/%ax/%eax to %ax/%eax/%rax. # instructions for %al/%ax/%eax to %ax/%eax/%rax.
# movsbl # movsbl
I32.enc(base.sextend.i32.i8, *r.urm(0x0f, 0xbe)) X86_32.enc(base.sextend.i32.i8, *r.urm_noflags(0x0f, 0xbe))
I64.enc(base.sextend.i32.i8, *r.urm.rex(0x0f, 0xbe)) X86_64.enc(base.sextend.i32.i8, *r.urm_noflags.rex(0x0f, 0xbe))
I64.enc(base.sextend.i32.i8, *r.urm(0x0f, 0xbe)) X86_64.enc(base.sextend.i32.i8, *r.urm_noflags(0x0f, 0xbe))
# movswl # movswl
I32.enc(base.sextend.i32.i16, *r.urm(0x0f, 0xbf)) X86_32.enc(base.sextend.i32.i16, *r.urm_noflags(0x0f, 0xbf))
I64.enc(base.sextend.i32.i16, *r.urm.rex(0x0f, 0xbf)) X86_64.enc(base.sextend.i32.i16, *r.urm_noflags.rex(0x0f, 0xbf))
I64.enc(base.sextend.i32.i16, *r.urm(0x0f, 0xbf)) X86_64.enc(base.sextend.i32.i16, *r.urm_noflags(0x0f, 0xbf))
# movsbq # movsbq
I64.enc(base.sextend.i64.i8, *r.urm.rex(0x0f, 0xbe, w=1)) X86_64.enc(base.sextend.i64.i8, *r.urm_noflags.rex(0x0f, 0xbe, w=1))
# movswq # movswq
I64.enc(base.sextend.i64.i16, *r.urm.rex(0x0f, 0xbf, w=1)) X86_64.enc(base.sextend.i64.i16, *r.urm_noflags.rex(0x0f, 0xbf, w=1))
# movslq # movslq
I64.enc(base.sextend.i64.i32, *r.urm.rex(0x63, w=1)) X86_64.enc(base.sextend.i64.i32, *r.urm_noflags.rex(0x63, w=1))
# movzbl # movzbl
I32.enc(base.uextend.i32.i8, *r.urm(0x0f, 0xb6)) X86_32.enc(base.uextend.i32.i8, *r.urm_noflags(0x0f, 0xb6))
I64.enc(base.uextend.i32.i8, *r.urm.rex(0x0f, 0xb6)) X86_64.enc(base.uextend.i32.i8, *r.urm_noflags.rex(0x0f, 0xb6))
I64.enc(base.uextend.i32.i8, *r.urm(0x0f, 0xb6)) X86_64.enc(base.uextend.i32.i8, *r.urm_noflags(0x0f, 0xb6))
# movzwl # movzwl
I32.enc(base.uextend.i32.i16, *r.urm(0x0f, 0xb7)) X86_32.enc(base.uextend.i32.i16, *r.urm_noflags(0x0f, 0xb7))
I64.enc(base.uextend.i32.i16, *r.urm.rex(0x0f, 0xb7)) X86_64.enc(base.uextend.i32.i16, *r.urm_noflags.rex(0x0f, 0xb7))
I64.enc(base.uextend.i32.i16, *r.urm(0x0f, 0xb7)) X86_64.enc(base.uextend.i32.i16, *r.urm_noflags(0x0f, 0xb7))
# movzbq, encoded as movzbl because it's equivalent and shorter # movzbq, encoded as movzbl because it's equivalent and shorter
I64.enc(base.uextend.i64.i8, *r.urm.rex(0x0f, 0xb6)) X86_64.enc(base.uextend.i64.i8, *r.urm_noflags.rex(0x0f, 0xb6))
I64.enc(base.uextend.i64.i8, *r.urm(0x0f, 0xb6)) X86_64.enc(base.uextend.i64.i8, *r.urm_noflags(0x0f, 0xb6))
# movzwq, encoded as movzwl because it's equivalent and shorter # movzwq, encoded as movzwl because it's equivalent and shorter
I64.enc(base.uextend.i64.i16, *r.urm.rex(0x0f, 0xb7)) X86_64.enc(base.uextend.i64.i16, *r.urm_noflags.rex(0x0f, 0xb7))
I64.enc(base.uextend.i64.i16, *r.urm(0x0f, 0xb7)) X86_64.enc(base.uextend.i64.i16, *r.urm_noflags(0x0f, 0xb7))
# A 32-bit register copy clears the high 32 bits. # A 32-bit register copy clears the high 32 bits.
I64.enc(base.uextend.i64.i32, *r.umr.rex(0x89)) X86_64.enc(base.uextend.i64.i32, *r.umr.rex(0x89))
I64.enc(base.uextend.i64.i32, *r.umr(0x89)) X86_64.enc(base.uextend.i64.i32, *r.umr(0x89))
# #
@@ -469,8 +484,8 @@ enc_both(base.bitcast.f32.i32, r.frurm, 0x66, 0x0f, 0x6e)
enc_both(base.bitcast.i32.f32, r.rfumr, 0x66, 0x0f, 0x7e) enc_both(base.bitcast.i32.f32, r.rfumr, 0x66, 0x0f, 0x7e)
# movq # movq
I64.enc(base.bitcast.f64.i64, *r.frurm.rex(0x66, 0x0f, 0x6e, w=1)) X86_64.enc(base.bitcast.f64.i64, *r.frurm.rex(0x66, 0x0f, 0x6e, w=1))
I64.enc(base.bitcast.i64.f64, *r.rfumr.rex(0x66, 0x0f, 0x7e, w=1)) X86_64.enc(base.bitcast.i64.f64, *r.rfumr.rex(0x66, 0x0f, 0x7e, w=1))
# movaps # movaps
enc_both(base.copy.f32, r.furm, 0x0f, 0x28) enc_both(base.copy.f32, r.furm, 0x0f, 0x28)
@@ -492,11 +507,11 @@ enc_both(base.fdemote.f32.f64, r.furm, 0xf2, 0x0f, 0x5a)
# cvttss2si # cvttss2si
enc_both(x86.cvtt2si.i32.f32, r.rfurm, 0xf3, 0x0f, 0x2c) enc_both(x86.cvtt2si.i32.f32, r.rfurm, 0xf3, 0x0f, 0x2c)
I64.enc(x86.cvtt2si.i64.f32, *r.rfurm.rex(0xf3, 0x0f, 0x2c, w=1)) X86_64.enc(x86.cvtt2si.i64.f32, *r.rfurm.rex(0xf3, 0x0f, 0x2c, w=1))
# cvttsd2si # cvttsd2si
enc_both(x86.cvtt2si.i32.f64, r.rfurm, 0xf2, 0x0f, 0x2c) enc_both(x86.cvtt2si.i32.f64, r.rfurm, 0xf2, 0x0f, 0x2c)
I64.enc(x86.cvtt2si.i64.f64, *r.rfurm.rex(0xf2, 0x0f, 0x2c, w=1)) X86_64.enc(x86.cvtt2si.i64.f64, *r.rfurm.rex(0xf2, 0x0f, 0x2c, w=1))
# Exact square roots. # Exact square roots.
enc_both(base.sqrt.f32, r.furm, 0xf3, 0x0f, 0x51) enc_both(base.sqrt.f32, r.furm, 0xf3, 0x0f, 0x51)

220
lib/cretonne/meta/isa/intel/recipes.py
View File

@@ -5,9 +5,11 @@ from __future__ import absolute_import
from cdsl.isa import EncRecipe from cdsl.isa import EncRecipe
from cdsl.predicates import IsSignedInt, IsEqual, Or from cdsl.predicates import IsSignedInt, IsEqual, Or
from cdsl.registers import RegClass from cdsl.registers import RegClass
from base.formats import Unary, UnaryImm, Binary, BinaryImm, MultiAry, NullAry from base.formats import Unary, UnaryImm, UnaryBool, Binary, BinaryImm
from base.formats import MultiAry, NullAry
from base.formats import Trap, Call, IndirectCall, Store, Load from base.formats import Trap, Call, IndirectCall, Store, Load
from base.formats import IntCompare, FloatCompare, IntCond, FloatCond from base.formats import IntCompare, IntCompareImm, FloatCompare
from base.formats import IntCond, FloatCond
from base.formats import IntSelect, IntCondTrap, FloatCondTrap from base.formats import IntSelect, IntCondTrap, FloatCondTrap
from base.formats import Jump, Branch, BranchInt, BranchFloat from base.formats import Jump, Branch, BranchInt, BranchFloat
from base.formats import Ternary, FuncAddr, UnaryGlobalVar from base.formats import Ternary, FuncAddr, UnaryGlobalVar
@@ -277,23 +279,27 @@ null = EncRecipe('null', Unary, size=0, ins=GPR, outs=0, emit='')
# XX opcode, no ModR/M. # XX opcode, no ModR/M.
trap = TailRecipe( trap = TailRecipe(
'trap', Trap, size=0, ins=(), outs=(), 'trap', Trap, size=0, ins=(), outs=(),
emit='PUT_OP(bits, BASE_REX, sink);') emit='''
sink.trap(code, func.srclocs[inst]);
PUT_OP(bits, BASE_REX, sink);
''')
# Macro: conditional jump over a ud2. # Macro: conditional jump over a ud2.
trapif = EncRecipe( trapif = EncRecipe(
'trapif', IntCondTrap, size=4, ins=FLAG.eflags, outs=(), 'trapif', IntCondTrap, size=4, ins=FLAG.rflags, outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
// Jump over a 2-byte ud2. // Jump over a 2-byte ud2.
sink.put1(0x70 | (icc2opc(cond.inverse()) as u8)); sink.put1(0x70 | (icc2opc(cond.inverse()) as u8));
sink.put1(2); sink.put1(2);
// ud2. // ud2.
sink.trap(code, func.srclocs[inst]);
sink.put1(0x0f); sink.put1(0x0f);
sink.put1(0x0b); sink.put1(0x0b);
''') ''')
trapff = EncRecipe( trapff = EncRecipe(
'trapff', FloatCondTrap, size=4, ins=FLAG.eflags, outs=(), 'trapff', FloatCondTrap, size=4, ins=FLAG.rflags, outs=(),
clobbers_flags=False, clobbers_flags=False,
instp=floatccs(FloatCondTrap), instp=floatccs(FloatCondTrap),
emit=''' emit='''
@@ -301,6 +307,7 @@ trapff = EncRecipe(
sink.put1(0x70 | (fcc2opc(cond.inverse()) as u8)); sink.put1(0x70 | (fcc2opc(cond.inverse()) as u8));
sink.put1(2); sink.put1(2);
// ud2. // ud2.
sink.trap(code, func.srclocs[inst]);
sink.put1(0x0f); sink.put1(0x0f);
sink.put1(0x0b); sink.put1(0x0b);
''') ''')
@@ -358,7 +365,7 @@ rfumr = TailRecipe(
''') ''')
# XX /r, but for a unary operator with separate input/output register. # XX /r, but for a unary operator with separate input/output register.
# RM form. # RM form. Clobbers FLAGS.
urm = TailRecipe( urm = TailRecipe(
'urm', Unary, size=1, ins=GPR, outs=GPR, 'urm', Unary, size=1, ins=GPR, outs=GPR,
emit=''' emit='''
@@ -366,10 +373,19 @@ urm = TailRecipe(
modrm_rr(in_reg0, out_reg0, sink); modrm_rr(in_reg0, out_reg0, sink);
''') ''')
# XX /r. Same as urm, but input limited to ABCD. # XX /r. Same as urm, but doesn't clobber FLAGS.
urm_abcd = TailRecipe( urm_noflags = TailRecipe(
'urm_abcd', Unary, size=1, ins=ABCD, outs=GPR, 'urm_noflags', Unary, size=1, ins=GPR, outs=GPR,
when_prefixed=urm, clobbers_flags=False,
emit='''
PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_rr(in_reg0, out_reg0, sink);
''')
# XX /r. Same as urm_noflags, but input limited to ABCD.
urm_noflags_abcd = TailRecipe(
'urm_noflags_abcd', Unary, size=1, ins=ABCD, outs=GPR,
when_prefixed=urm_noflags,
emit=''' emit='''
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_rr(in_reg0, out_reg0, sink); modrm_rr(in_reg0, out_reg0, sink);
@@ -449,6 +465,7 @@ div = TailRecipe(
'div', Ternary, size=1, 'div', Ternary, size=1,
ins=(GPR.rax, GPR.rdx, GPR), outs=(GPR.rax, GPR.rdx), ins=(GPR.rax, GPR.rdx, GPR), outs=(GPR.rax, GPR.rdx),
emit=''' emit='''
sink.trap(TrapCode::IntegerDivisionByZero, func.srclocs[inst]);
PUT_OP(bits, rex1(in_reg2), sink); PUT_OP(bits, rex1(in_reg2), sink);
modrm_r_bits(in_reg2, bits, sink); modrm_r_bits(in_reg2, bits, sink);
''') ''')
@@ -506,6 +523,17 @@ puid = TailRecipe(
sink.put4(imm as u32); sink.put4(imm as u32);
''') ''')
# XX+rd id unary with bool immediate. Note no recipe predicate.
puid_bool = TailRecipe(
'puid_bool', UnaryBool, size=4, ins=(), outs=GPR,
emit='''
// The destination register is encoded in the low bits of the opcode.
// No ModR/M.
PUT_OP(bits | (out_reg0 & 7), rex1(out_reg0), sink);
let imm: u32 = if imm.into() { 1 } else { 0 };
sink.put4(imm);
''')
# XX+rd iq unary with 64-bit immediate. # XX+rd iq unary with 64-bit immediate.
puiq = TailRecipe( puiq = TailRecipe(
'puiq', UnaryImm, size=8, ins=(), outs=GPR, 'puiq', UnaryImm, size=8, ins=(), outs=GPR,
@@ -666,6 +694,9 @@ st = TailRecipe(
instp=IsEqual(Store.offset, 0), instp=IsEqual(Store.offset, 0),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_rm(in_reg1, in_reg0, sink); modrm_rm(in_reg1, in_reg0, sink);
''') ''')
@@ -678,6 +709,9 @@ st_abcd = TailRecipe(
when_prefixed=st, when_prefixed=st,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_rm(in_reg1, in_reg0, sink); modrm_rm(in_reg1, in_reg0, sink);
''') ''')
@@ -688,6 +722,9 @@ fst = TailRecipe(
instp=IsEqual(Store.offset, 0), instp=IsEqual(Store.offset, 0),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_rm(in_reg1, in_reg0, sink); modrm_rm(in_reg1, in_reg0, sink);
''') ''')
@@ -698,6 +735,9 @@ stDisp8 = TailRecipe(
instp=IsSignedInt(Store.offset, 8), instp=IsSignedInt(Store.offset, 8),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp8(in_reg1, in_reg0, sink); modrm_disp8(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -709,6 +749,9 @@ stDisp8_abcd = TailRecipe(
when_prefixed=stDisp8, when_prefixed=stDisp8,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp8(in_reg1, in_reg0, sink); modrm_disp8(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -719,6 +762,9 @@ fstDisp8 = TailRecipe(
instp=IsSignedInt(Store.offset, 8), instp=IsSignedInt(Store.offset, 8),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp8(in_reg1, in_reg0, sink); modrm_disp8(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -730,6 +776,9 @@ stDisp32 = TailRecipe(
'stDisp32', Store, size=5, ins=(GPR, GPR_DEREF_SAFE), outs=(), 'stDisp32', Store, size=5, ins=(GPR, GPR_DEREF_SAFE), outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp32(in_reg1, in_reg0, sink); modrm_disp32(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -740,6 +789,9 @@ stDisp32_abcd = TailRecipe(
when_prefixed=stDisp32, when_prefixed=stDisp32,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp32(in_reg1, in_reg0, sink); modrm_disp32(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -749,6 +801,9 @@ fstDisp32 = TailRecipe(
'fstDisp32', Store, size=5, ins=(FPR, GPR_DEREF_SAFE), outs=(), 'fstDisp32', Store, size=5, ins=(FPR, GPR_DEREF_SAFE), outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_disp32(in_reg1, in_reg0, sink); modrm_disp32(in_reg1, in_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -756,8 +811,8 @@ fstDisp32 = TailRecipe(
''') ''')
# Unary spill with SIB and 32-bit displacement. # Unary spill with SIB and 32-bit displacement.
spSib32 = TailRecipe( spillSib32 = TailRecipe(
'spSib32', Unary, size=6, ins=GPR, outs=StackGPR32, 'spillSib32', Unary, size=6, ins=GPR, outs=StackGPR32,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let base = stk_base(out_stk0.base); let base = stk_base(out_stk0.base);
@@ -766,8 +821,10 @@ spSib32 = TailRecipe(
sib_noindex(base, sink); sib_noindex(base, sink);
sink.put4(out_stk0.offset as u32); sink.put4(out_stk0.offset as u32);
''') ''')
fspSib32 = TailRecipe(
'fspSib32', Unary, size=6, ins=FPR, outs=StackFPR32, # Like spillSib32, but targeting an FPR rather than a GPR.
fspillSib32 = TailRecipe(
'fspillSib32', Unary, size=6, ins=FPR, outs=StackFPR32,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let base = stk_base(out_stk0.base); let base = stk_base(out_stk0.base);
@@ -778,8 +835,8 @@ fspSib32 = TailRecipe(
''') ''')
# Regspill using RSP-relative addressing. # Regspill using RSP-relative addressing.
rsp32 = TailRecipe( regspill32 = TailRecipe(
'rsp32', RegSpill, size=6, ins=GPR, outs=(), 'regspill32', RegSpill, size=6, ins=GPR, outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let dst = StackRef::sp(dst, &func.stack_slots); let dst = StackRef::sp(dst, &func.stack_slots);
@@ -789,8 +846,10 @@ rsp32 = TailRecipe(
sib_noindex(base, sink); sib_noindex(base, sink);
sink.put4(dst.offset as u32); sink.put4(dst.offset as u32);
''') ''')
frsp32 = TailRecipe(
'frsp32', RegSpill, size=6, ins=FPR, outs=(), # Like regspill32, but targeting an FPR rather than a GPR.
fregspill32 = TailRecipe(
'fregspill32', RegSpill, size=6, ins=FPR, outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let dst = StackRef::sp(dst, &func.stack_slots); let dst = StackRef::sp(dst, &func.stack_slots);
@@ -811,6 +870,9 @@ ld = TailRecipe(
instp=IsEqual(Load.offset, 0), instp=IsEqual(Load.offset, 0),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_rm(in_reg0, out_reg0, sink); modrm_rm(in_reg0, out_reg0, sink);
''') ''')
@@ -821,6 +883,9 @@ fld = TailRecipe(
instp=IsEqual(Load.offset, 0), instp=IsEqual(Load.offset, 0),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_rm(in_reg0, out_reg0, sink); modrm_rm(in_reg0, out_reg0, sink);
''') ''')
@@ -831,6 +896,9 @@ ldDisp8 = TailRecipe(
instp=IsSignedInt(Load.offset, 8), instp=IsSignedInt(Load.offset, 8),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_disp8(in_reg0, out_reg0, sink); modrm_disp8(in_reg0, out_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -843,6 +911,9 @@ fldDisp8 = TailRecipe(
instp=IsSignedInt(Load.offset, 8), instp=IsSignedInt(Load.offset, 8),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_disp8(in_reg0, out_reg0, sink); modrm_disp8(in_reg0, out_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -855,6 +926,9 @@ ldDisp32 = TailRecipe(
instp=IsSignedInt(Load.offset, 32), instp=IsSignedInt(Load.offset, 32),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_disp32(in_reg0, out_reg0, sink); modrm_disp32(in_reg0, out_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -867,6 +941,9 @@ fldDisp32 = TailRecipe(
instp=IsSignedInt(Load.offset, 32), instp=IsSignedInt(Load.offset, 32),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
if !flags.notrap() {
sink.trap(TrapCode::HeapOutOfBounds, func.srclocs[inst]);
}
PUT_OP(bits, rex2(in_reg0, out_reg0), sink); PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
modrm_disp32(in_reg0, out_reg0, sink); modrm_disp32(in_reg0, out_reg0, sink);
let offset: i32 = offset.into(); let offset: i32 = offset.into();
@@ -874,8 +951,8 @@ fldDisp32 = TailRecipe(
''') ''')
# Unary fill with SIB and 32-bit displacement. # Unary fill with SIB and 32-bit displacement.
fiSib32 = TailRecipe( fillSib32 = TailRecipe(
'fiSib32', Unary, size=6, ins=StackGPR32, outs=GPR, 'fillSib32', Unary, size=6, ins=StackGPR32, outs=GPR,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let base = stk_base(in_stk0.base); let base = stk_base(in_stk0.base);
@@ -884,8 +961,10 @@ fiSib32 = TailRecipe(
sib_noindex(base, sink); sib_noindex(base, sink);
sink.put4(in_stk0.offset as u32); sink.put4(in_stk0.offset as u32);
''') ''')
ffiSib32 = TailRecipe(
'ffiSib32', Unary, size=6, ins=StackFPR32, outs=FPR, # Like fillSib32, but targeting an FPR rather than a GPR.
ffillSib32 = TailRecipe(
'ffillSib32', Unary, size=6, ins=StackFPR32, outs=FPR,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let base = stk_base(in_stk0.base); let base = stk_base(in_stk0.base);
@@ -896,8 +975,8 @@ ffiSib32 = TailRecipe(
''') ''')
# Regfill with RSP-relative 32-bit displacement. # Regfill with RSP-relative 32-bit displacement.
rfi32 = TailRecipe( regfill32 = TailRecipe(
'rfi32', RegFill, size=6, ins=StackGPR32, outs=(), 'regfill32', RegFill, size=6, ins=StackGPR32, outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let src = StackRef::sp(src, &func.stack_slots); let src = StackRef::sp(src, &func.stack_slots);
@@ -907,8 +986,10 @@ rfi32 = TailRecipe(
sib_noindex(base, sink); sib_noindex(base, sink);
sink.put4(src.offset as u32); sink.put4(src.offset as u32);
''') ''')
frfi32 = TailRecipe(
'frfi32', RegFill, size=6, ins=StackFPR32, outs=(), # Like regfill32, but targeting an FPR rather than a GPR.
fregfill32 = TailRecipe(
'fregfill32', RegFill, size=6, ins=StackFPR32, outs=(),
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
let src = StackRef::sp(src, &func.stack_slots); let src = StackRef::sp(src, &func.stack_slots);
@@ -977,7 +1058,7 @@ jmpd = TailRecipe(
''') ''')
brib = TailRecipe( brib = TailRecipe(
'brib', BranchInt, size=1, ins=FLAG.eflags, outs=(), 'brib', BranchInt, size=1, ins=FLAG.rflags, outs=(),
branch_range=8, branch_range=8,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -986,7 +1067,7 @@ brib = TailRecipe(
''') ''')
brid = TailRecipe( brid = TailRecipe(
'brid', BranchInt, size=4, ins=FLAG.eflags, outs=(), 'brid', BranchInt, size=4, ins=FLAG.rflags, outs=(),
branch_range=32, branch_range=32,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -995,7 +1076,7 @@ brid = TailRecipe(
''') ''')
brfb = TailRecipe( brfb = TailRecipe(
'brfb', BranchFloat, size=1, ins=FLAG.eflags, outs=(), 'brfb', BranchFloat, size=1, ins=FLAG.rflags, outs=(),
branch_range=8, branch_range=8,
clobbers_flags=False, clobbers_flags=False,
instp=floatccs(BranchFloat), instp=floatccs(BranchFloat),
@@ -1005,7 +1086,7 @@ brfb = TailRecipe(
''') ''')
brfd = TailRecipe( brfd = TailRecipe(
'brfd', BranchFloat, size=4, ins=FLAG.eflags, outs=(), 'brfd', BranchFloat, size=4, ins=FLAG.rflags, outs=(),
branch_range=32, branch_range=32,
clobbers_flags=False, clobbers_flags=False,
instp=floatccs(BranchFloat), instp=floatccs(BranchFloat),
@@ -1025,7 +1106,7 @@ brfd = TailRecipe(
# #
seti = TailRecipe( seti = TailRecipe(
'seti', IntCond, size=1, ins=FLAG.eflags, outs=GPR, 'seti', IntCond, size=1, ins=FLAG.rflags, outs=GPR,
requires_prefix=True, requires_prefix=True,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -1033,7 +1114,7 @@ seti = TailRecipe(
modrm_r_bits(out_reg0, bits, sink); modrm_r_bits(out_reg0, bits, sink);
''') ''')
seti_abcd = TailRecipe( seti_abcd = TailRecipe(
'seti_abcd', IntCond, size=1, ins=FLAG.eflags, outs=ABCD, 'seti_abcd', IntCond, size=1, ins=FLAG.rflags, outs=ABCD,
when_prefixed=seti, when_prefixed=seti,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -1042,7 +1123,7 @@ seti_abcd = TailRecipe(
''') ''')
setf = TailRecipe( setf = TailRecipe(
'setf', FloatCond, size=1, ins=FLAG.eflags, outs=GPR, 'setf', FloatCond, size=1, ins=FLAG.rflags, outs=GPR,
requires_prefix=True, requires_prefix=True,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -1050,7 +1131,7 @@ setf = TailRecipe(
modrm_r_bits(out_reg0, bits, sink); modrm_r_bits(out_reg0, bits, sink);
''') ''')
setf_abcd = TailRecipe( setf_abcd = TailRecipe(
'setf_abcd', FloatCond, size=1, ins=FLAG.eflags, outs=ABCD, 'setf_abcd', FloatCond, size=1, ins=FLAG.rflags, outs=ABCD,
when_prefixed=setf, when_prefixed=setf,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -1064,7 +1145,7 @@ setf_abcd = TailRecipe(
# 1 byte, modrm(r,r), is after the opcode # 1 byte, modrm(r,r), is after the opcode
# #
cmov = TailRecipe( cmov = TailRecipe(
'cmov', IntSelect, size=1, ins=(FLAG.eflags, GPR, GPR), outs=2, 'cmov', IntSelect, size=1, ins=(FLAG.rflags, GPR, GPR), outs=2,
requires_prefix=False, requires_prefix=False,
clobbers_flags=False, clobbers_flags=False,
emit=''' emit='''
@@ -1076,7 +1157,7 @@ cmov = TailRecipe(
# Bit scan forwards and reverse # Bit scan forwards and reverse
# #
bsf_and_bsr = TailRecipe( bsf_and_bsr = TailRecipe(
'bsf_and_bsr', Unary, size=1, ins=GPR, outs=(GPR, FLAG.eflags), 'bsf_and_bsr', Unary, size=1, ins=GPR, outs=(GPR, FLAG.rflags),
requires_prefix=False, requires_prefix=False,
clobbers_flags=True, clobbers_flags=True,
emit=''' emit='''
@@ -1090,7 +1171,7 @@ bsf_and_bsr = TailRecipe(
# XX /r, MR form. Compare two GPR registers and set flags. # XX /r, MR form. Compare two GPR registers and set flags.
rcmp = TailRecipe( rcmp = TailRecipe(
'rcmp', Binary, size=1, ins=(GPR, GPR), outs=FLAG.eflags, 'rcmp', Binary, size=1, ins=(GPR, GPR), outs=FLAG.rflags,
emit=''' emit='''
PUT_OP(bits, rex2(in_reg0, in_reg1), sink); PUT_OP(bits, rex2(in_reg0, in_reg1), sink);
modrm_rr(in_reg0, in_reg1, sink); modrm_rr(in_reg0, in_reg1, sink);
@@ -1098,7 +1179,7 @@ rcmp = TailRecipe(
# XX /r, RM form. Compare two FPR registers and set flags. # XX /r, RM form. Compare two FPR registers and set flags.
fcmp = TailRecipe( fcmp = TailRecipe(
'fcmp', Binary, size=1, ins=(FPR, FPR), outs=FLAG.eflags, 'fcmp', Binary, size=1, ins=(FPR, FPR), outs=FLAG.rflags,
emit=''' emit='''
PUT_OP(bits, rex2(in_reg1, in_reg0), sink); PUT_OP(bits, rex2(in_reg1, in_reg0), sink);
modrm_rr(in_reg1, in_reg0, sink); modrm_rr(in_reg1, in_reg0, sink);
@@ -1106,7 +1187,7 @@ fcmp = TailRecipe(
# XX /n, MI form with imm8. # XX /n, MI form with imm8.
rcmpib = TailRecipe( rcmpib = TailRecipe(
'rcmpib', BinaryImm, size=2, ins=GPR, outs=FLAG.eflags, 'rcmpib', BinaryImm, size=2, ins=GPR, outs=FLAG.rflags,
instp=IsSignedInt(BinaryImm.imm, 8), instp=IsSignedInt(BinaryImm.imm, 8),
emit=''' emit='''
PUT_OP(bits, rex1(in_reg0), sink); PUT_OP(bits, rex1(in_reg0), sink);
@@ -1117,7 +1198,7 @@ rcmpib = TailRecipe(
# XX /n, MI form with imm32. # XX /n, MI form with imm32.
rcmpid = TailRecipe( rcmpid = TailRecipe(
'rcmpid', BinaryImm, size=5, ins=GPR, outs=FLAG.eflags, 'rcmpid', BinaryImm, size=5, ins=GPR, outs=FLAG.rflags,
instp=IsSignedInt(BinaryImm.imm, 32), instp=IsSignedInt(BinaryImm.imm, 32),
emit=''' emit='''
PUT_OP(bits, rex1(in_reg0), sink); PUT_OP(bits, rex1(in_reg0), sink);
@@ -1128,7 +1209,7 @@ rcmpid = TailRecipe(
# Same as rcmp, but second operand is the stack pointer. # Same as rcmp, but second operand is the stack pointer.
rcmp_sp = TailRecipe( rcmp_sp = TailRecipe(
'rcmp_sp', Unary, size=1, ins=GPR, outs=FLAG.eflags, 'rcmp_sp', Unary, size=1, ins=GPR, outs=FLAG.rflags,
emit=''' emit='''
PUT_OP(bits, rex2(in_reg0, RU::rsp.into()), sink); PUT_OP(bits, rex2(in_reg0, RU::rsp.into()), sink);
modrm_rr(in_reg0, RU::rsp.into(), sink); modrm_rr(in_reg0, RU::rsp.into(), sink);
@@ -1289,12 +1370,67 @@ icscc = TailRecipe(
modrm_rr(out_reg0, 0, sink); modrm_rr(out_reg0, 0, sink);
''') ''')
icsccib = TailRecipe(
'icsccib', IntCompareImm, size=2 + 3, ins=GPR, outs=ABCD,
instp=IsSignedInt(IntCompareImm.imm, 8),
emit='''
// Comparison instruction.
PUT_OP(bits, rex1(in_reg0), sink);
modrm_r_bits(in_reg0, bits, sink);
let imm: i64 = imm.into();
sink.put1(imm as u8);
// `setCC` instruction, no REX.
use ir::condcodes::IntCC::*;
let setcc = match cond {
Equal => 0x94,
NotEqual => 0x95,
SignedLessThan => 0x9c,
SignedGreaterThanOrEqual => 0x9d,
SignedGreaterThan => 0x9f,
SignedLessThanOrEqual => 0x9e,
UnsignedLessThan => 0x92,
UnsignedGreaterThanOrEqual => 0x93,
UnsignedGreaterThan => 0x97,
UnsignedLessThanOrEqual => 0x96,
};
sink.put1(0x0f);
sink.put1(setcc);
modrm_rr(out_reg0, 0, sink);
''')
icsccid = TailRecipe(
'icsccid', IntCompareImm, size=5 + 3, ins=GPR, outs=ABCD,
instp=IsSignedInt(IntCompareImm.imm, 32),
emit='''
// Comparison instruction.
PUT_OP(bits, rex1(in_reg0), sink);
modrm_r_bits(in_reg0, bits, sink);
let imm: i64 = imm.into();
sink.put4(imm as u32);
// `setCC` instruction, no REX.
use ir::condcodes::IntCC::*;
let setcc = match cond {
Equal => 0x94,
NotEqual => 0x95,
SignedLessThan => 0x9c,
SignedGreaterThanOrEqual => 0x9d,
SignedGreaterThan => 0x9f,
SignedLessThanOrEqual => 0x9e,
UnsignedLessThan => 0x92,
UnsignedGreaterThanOrEqual => 0x93,
UnsignedGreaterThan => 0x97,
UnsignedLessThanOrEqual => 0x96,
};
sink.put1(0x0f);
sink.put1(setcc);
modrm_rr(out_reg0, 0, sink);
''')
# Make a FloatCompare instruction predicate with the supported condition codes. # Make a FloatCompare instruction predicate with the supported condition codes.
# Same thing for floating point. # Same thing for floating point.
# #
# The ucomiss/ucomisd instructions set the EFLAGS bits CF/PF/CF like this: # The ucomiss/ucomisd instructions set the FLAGS bits CF/PF/CF like this:
# #
# ZPC OSA # ZPC OSA
# UN 111 000 # UN 111 000

2
lib/cretonne/meta/isa/intel/registers.py
View File

@@ -43,7 +43,7 @@ FlagRegs = RegBank(
'Flag registers', 'Flag registers',
units=1, units=1,
pressure_tracking=False, pressure_tracking=False,
names=['eflags']) names=['rflags'])
GPR = RegClass(IntRegs) GPR = RegClass(IntRegs)
# Certain types of deref encodings cannot be used with all registers. # Certain types of deref encodings cannot be used with all registers.

82
lib/cretonne/meta/srcgen.py
View File

@@ -8,9 +8,10 @@ source code.
from __future__ import absolute_import from __future__ import absolute_import
import sys import sys
import os import os
from collections import OrderedDict
try: try:
from typing import Any, List # noqa from typing import Any, List, Set, Tuple # noqa
except ImportError: except ImportError:
pass pass
@@ -146,6 +147,52 @@ class Formatter(object):
for l in parse_multiline(s): for l in parse_multiline(s):
self.line('/// ' + l if l else '///') self.line('/// ' + l if l else '///')
def match(self, m):
# type: (Match) -> None
"""
Add a match expression.
Example:
>>> f = Formatter()
>>> m = Match('x')
>>> m.arm('Orange', ['a', 'b'], 'some body')
>>> m.arm('Yellow', ['a', 'b'], 'some body')
>>> m.arm('Green', ['a', 'b'], 'different body')
>>> m.arm('Blue', ['x', 'y'], 'some body')
>>> f.match(m)
>>> f.writelines()
match x {
Orange { a, b } |
Yellow { a, b } => {
some body
}
Green { a, b } => {
different body
}
Blue { x, y } => {
some body
}
}
"""
with self.indented('match {} {{'.format(m.expr), '}'):
for (fields, body), names in m.arms.items():
with self.indented('', '}'):
names_left = len(names)
for name in names.keys():
fields_str = ', '.join(fields)
if len(fields) != 0:
fields_str = '{{ {} }} '.format(fields_str)
names_left -= 1
if names_left > 0:
suffix = '|'
else:
suffix = '=> {'
self.outdented_line(name + ' ' + fields_str + suffix)
if names_left == 0:
self.multi_line(body)
def _indent(s): def _indent(s):
# type: (str) -> int # type: (str) -> int
@@ -195,3 +242,36 @@ def parse_multiline(s):
while trimmed and not trimmed[0]: while trimmed and not trimmed[0]:
trimmed.pop(0) trimmed.pop(0)
return trimmed return trimmed
class Match(object):
"""
Match formatting class.
Match objects collect all the information needed to emit a Rust `match`
expression, automatically deduplicating overlapping identical arms.
Example:
>>> m = Match('x')
>>> m.arm('Orange', ['a', 'b'], 'some body')
>>> m.arm('Yellow', ['a', 'b'], 'some body')
>>> m.arm('Green', ['a', 'b'], 'different body')
>>> m.arm('Blue', ['x', 'y'], 'some body')
>>> assert(len(m.arms) == 3)
Note that this class is ignorant of Rust types, and considers two fields
with the same name to be equivalent.
"""
def __init__(self, expr):
# type: (str) -> None
self.expr = expr
self.arms = OrderedDict() # type: OrderedDict[Tuple[Tuple[str, ...], str], OrderedDict[str, None]] # noqa
def arm(self, name, fields, body):
# type: (str, List[str], str) -> None
key = (tuple(fields), body)
if key not in self.arms:
self.arms[key] = OrderedDict()
self.arms[key][name] = None

4
lib/cretonne/meta/test_gen_legalizer.py
View File

@@ -148,9 +148,9 @@ class TestRuntimeChecks(TestCase):
self.v5 << vselect(self.v1, self.v3, self.v4), self.v5 << vselect(self.v1, self.v3, self.v4),
) )
x = XForm(r, r) x = XForm(r, r)
tv2_exp = 'Some({}).map(|t: Type| -> t.as_bool())'\ tv2_exp = 'Some({}).map(|t: ir::Type| t.as_bool())'\
.format(self.v2.get_typevar().name) .format(self.v2.get_typevar().name)
tv3_exp = 'Some({}).map(|t: Type| -> t.as_bool())'\ tv3_exp = 'Some({}).map(|t: ir::Type| t.as_bool())'\
.format(self.v3.get_typevar().name) .format(self.v3.get_typevar().name)
self.check_yo_check( self.check_yo_check(

4
lib/cretonne/src/abi.rs
View File

@@ -3,7 +3,7 @@
//! This module provides functions and data structures that are useful for implementing the //! This module provides functions and data structures that are useful for implementing the
//! `TargetIsa::legalize_signature()` method. //! `TargetIsa::legalize_signature()` method.
use ir::{ArgumentLoc, AbiParam, ArgumentExtension, Type}; use ir::{AbiParam, ArgumentExtension, ArgumentLoc, Type};
use std::cmp::Ordering; use std::cmp::Ordering;
use std::vec::Vec; use std::vec::Vec;
@@ -186,8 +186,8 @@ pub fn legalize_abi_value(have: Type, arg: &AbiParam) -> ValueConversion {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;
use ir::types;
use ir::AbiParam; use ir::AbiParam;
use ir::types;
#[test] #[test]
fn legalize() { fn legalize() {

6
lib/cretonne/src/bforest/map.rs
View File

@@ -1,8 +1,8 @@
//! Forest of maps. //! Forest of maps.
use super::{Comparator, Forest, Node, NodeData, NodePool, Path, INNER_SIZE};
use packed_option::PackedOption; use packed_option::PackedOption;
use std::marker::PhantomData; use std::marker::PhantomData;
use super::{INNER_SIZE, Comparator, Forest, NodePool, Node, NodeData, Path};
/// Tag type defining forest types for a map. /// Tag type defining forest types for a map.
struct MapTypes<K, V, C>(PhantomData<(K, V, C)>); struct MapTypes<K, V, C>(PhantomData<(K, V, C)>);
@@ -424,10 +424,10 @@ where
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::super::NodeData;
use super::*;
use std::mem; use std::mem;
use std::vec::Vec; use std::vec::Vec;
use super::*;
use super::super::NodeData;
#[test] #[test]
fn node_size() { fn node_size() {

4
lib/cretonne/src/bforest/mod.rs
View File

@@ -22,8 +22,8 @@ mod path;
mod pool; mod pool;
mod set; mod set;
pub use self::map::{MapForest, Map, MapCursor, MapIter}; pub use self::map::{Map, MapCursor, MapForest, MapIter};
pub use self::set::{SetForest, Set, SetCursor, SetIter}; pub use self::set::{Set, SetCursor, SetForest, SetIter};
use self::node::NodeData; use self::node::NodeData;
use self::path::Path; use self::path::Path;

4
lib/cretonne/src/bforest/node.rs
View File

@@ -1,8 +1,8 @@
//! B+-tree nodes. //! B+-tree nodes.
use super::{slice_insert, slice_shift, Forest, Node, SetValue, INNER_SIZE};
use std::borrow::{Borrow, BorrowMut}; use std::borrow::{Borrow, BorrowMut};
use std::fmt; use std::fmt;
use super::{Forest, Node, INNER_SIZE, SetValue, slice_insert, slice_shift};
/// B+-tree node. /// B+-tree node.
/// ///
@@ -579,9 +579,9 @@ where
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::*;
use std::mem; use std::mem;
use std::string::ToString; use std::string::ToString;
use super::*;
// Forest impl for a set implementation. // Forest impl for a set implementation.
struct TF(); struct TF();

59
lib/cretonne/src/bforest/path.rs
View File

@@ -1,9 +1,9 @@
//! A path from the root of a B+-tree to a leaf node. //! A path from the root of a B+-tree to a leaf node.
use super::node::Removed;
use super::{slice_insert, slice_shift, Comparator, Forest, Node, NodeData, NodePool, MAX_PATH};
use std::borrow::Borrow; use std::borrow::Borrow;
use std::marker::PhantomData; use std::marker::PhantomData;
use super::{Forest, Node, NodeData, NodePool, MAX_PATH, Comparator, slice_insert, slice_shift};
use super::node::Removed;
#[cfg(test)] #[cfg(test)]
use std::fmt; use std::fmt;
@@ -55,8 +55,8 @@ impl<F: Forest> Path<F> {
for level in 0.. { for level in 0.. {
self.size = level + 1; self.size = level + 1;
self.node[level] = node; self.node[level] = node;
match &pool[node] { match pool[node] {
&NodeData::Inner { size, keys, tree } => { NodeData::Inner { size, keys, tree } => {
// Invariant: `tree[i]` contains keys smaller than // Invariant: `tree[i]` contains keys smaller than
// `keys[i]`, greater or equal to `keys[i-1]`. // `keys[i]`, greater or equal to `keys[i-1]`.
let i = match comp.search(key, &keys[0..size.into()]) { let i = match comp.search(key, &keys[0..size.into()]) {
@@ -68,7 +68,7 @@ impl<F: Forest> Path<F> {
self.entry[level] = i as u8; self.entry[level] = i as u8;
node = tree[i]; node = tree[i];
} }
&NodeData::Leaf { size, keys, vals } => { NodeData::Leaf { size, keys, vals } => {
// For a leaf we want either the found key or an insert position. // For a leaf we want either the found key or an insert position.
return match comp.search(key, &keys.borrow()[0..size.into()]) { return match comp.search(key, &keys.borrow()[0..size.into()]) {
Ok(i) => { Ok(i) => {
@@ -81,7 +81,7 @@ impl<F: Forest> Path<F> {
} }
}; };
} }
&NodeData::Free { .. } => panic!("Free {} reached from {}", node, root), NodeData::Free { .. } => panic!("Free {} reached from {}", node, root),
} }
} }
unreachable!(); unreachable!();
@@ -94,10 +94,10 @@ impl<F: Forest> Path<F> {
self.size = level + 1; self.size = level + 1;
self.node[level] = node; self.node[level] = node;
self.entry[level] = 0; self.entry[level] = 0;
match &pool[node] { match pool[node] {
&NodeData::Inner { tree, .. } => node = tree[0], NodeData::Inner { tree, .. } => node = tree[0],
&NodeData::Leaf { keys, vals, .. } => return (keys.borrow()[0], vals.borrow()[0]), NodeData::Leaf { keys, vals, .. } => return (keys.borrow()[0], vals.borrow()[0]),
&NodeData::Free { .. } => panic!("Free {} reached from {}", node, root), NodeData::Free { .. } => panic!("Free {} reached from {}", node, root),
} }
} }
unreachable!(); unreachable!();
@@ -205,17 +205,17 @@ impl<F: Forest> Path<F> {
let mut node = root; let mut node = root;
for l in level.. { for l in level.. {
self.node[l] = node; self.node[l] = node;
match &pool[node] { match pool[node] {
&NodeData::Inner { size, ref tree, .. } => { NodeData::Inner { size, ref tree, .. } => {
self.entry[l] = size; self.entry[l] = size;
node = tree[usize::from(size)]; node = tree[usize::from(size)];
} }
&NodeData::Leaf { size, .. } => { NodeData::Leaf { size, .. } => {
self.entry[l] = size - 1; self.entry[l] = size - 1;
self.size = l + 1; self.size = l + 1;
break; break;
} }
&NodeData::Free { .. } => panic!("Free {} reached from {}", node, root), NodeData::Free { .. } => panic!("Free {} reached from {}", node, root),
} }
} }
node node
@@ -405,8 +405,8 @@ impl<F: Forest> Path<F> {
let crit_key = pool[self.leaf_node()].leaf_crit_key(); let crit_key = pool[self.leaf_node()].leaf_crit_key();
let crit_node = self.node[crit_level]; let crit_node = self.node[crit_level];
match &mut pool[crit_node] { match pool[crit_node] {
&mut NodeData::Inner { size, ref mut keys, .. } => { NodeData::Inner { size, ref mut keys, .. } => {
debug_assert!(crit_kidx < size); debug_assert!(crit_kidx < size);
keys[usize::from(crit_kidx)] = crit_key; keys[usize::from(crit_kidx)] = crit_key;
} }
@@ -414,7 +414,6 @@ impl<F: Forest> Path<F> {
} }
} }
/// Given that the current leaf node is in an unhealthy (underflowed or even empty) status, /// Given that the current leaf node is in an unhealthy (underflowed or even empty) status,
/// balance it with sibling nodes. /// balance it with sibling nodes.
/// ///
@@ -437,7 +436,7 @@ impl<F: Forest> Path<F> {
// Discard the root node if it has shrunk to a single sub-tree. // Discard the root node if it has shrunk to a single sub-tree.
let mut ns = 0; let mut ns = 0;
while let &NodeData::Inner { size: 0, ref tree, .. } = &pool[self.node[ns]] { while let NodeData::Inner { size: 0, ref tree, .. } = pool[self.node[ns]] {
ns += 1; ns += 1;
self.node[ns] = tree[0]; self.node[ns] = tree[0];
} }
@@ -529,14 +528,12 @@ impl<F: Forest> Path<F> {
// current entry[level] was one off the end of the node, it will now point at a proper // current entry[level] was one off the end of the node, it will now point at a proper
// entry. // entry.
debug_assert!(usize::from(self.entry[level]) < pool[self.node[level]].entries()); debug_assert!(usize::from(self.entry[level]) < pool[self.node[level]].entries());
} else { } else if usize::from(self.entry[level]) >= pool[self.node[level]].entries() {
// There's no right sibling at this level, so the node can't be rebalanced. // There's no right sibling at this level, so the node can't be rebalanced.
// Check if we are in an off-the-end position. // Check if we are in an off-the-end position.
if usize::from(self.entry[level]) >= pool[self.node[level]].entries() {
self.size = 0; self.size = 0;
} }
} }
}
/// The current node at `level` has become empty. /// The current node at `level` has become empty.
/// ///
@@ -581,8 +578,8 @@ impl<F: Forest> Path<F> {
/// ///
/// Returns `None` if the current node is a right-most node so no right sibling exists. /// Returns `None` if the current node is a right-most node so no right sibling exists.
fn right_sibling_branch_level(&self, level: usize, pool: &NodePool<F>) -> Option<usize> { fn right_sibling_branch_level(&self, level: usize, pool: &NodePool<F>) -> Option<usize> {
(0..level).rposition(|l| match &pool[self.node[l]] { (0..level).rposition(|l| match pool[self.node[l]] {
&NodeData::Inner { size, .. } => self.entry[l] < size, NodeData::Inner { size, .. } => self.entry[l] < size,
_ => panic!("Expected inner node"), _ => panic!("Expected inner node"),
}) })
} }
@@ -622,8 +619,8 @@ impl<F: Forest> Path<F> {
let bl = self.right_sibling_branch_level(level, pool).expect( let bl = self.right_sibling_branch_level(level, pool).expect(
"No right sibling exists", "No right sibling exists",
); );
match &mut pool[self.node[bl]] { match pool[self.node[bl]] {
&mut NodeData::Inner { ref mut keys, .. } => { NodeData::Inner { ref mut keys, .. } => {
keys[usize::from(self.entry[bl])] = crit_key; keys[usize::from(self.entry[bl])] = crit_key;
} }
_ => panic!("Expected inner node"), _ => panic!("Expected inner node"),
@@ -647,8 +644,8 @@ impl<F: Forest> Path<F> {
/// Check the internal consistency of this path. /// Check the internal consistency of this path.
pub fn verify(&self, pool: &NodePool<F>) { pub fn verify(&self, pool: &NodePool<F>) {
for level in 0..self.size { for level in 0..self.size {
match &pool[self.node[level]] { match pool[self.node[level]] {
&NodeData::Inner { size, tree, .. } => { NodeData::Inner { size, tree, .. } => {
assert!( assert!(
level < self.size - 1, level < self.size - 1,
"Expected leaf node at level {}", "Expected leaf node at level {}",
@@ -668,7 +665,7 @@ impl<F: Forest> Path<F> {
level level
); );
} }
&NodeData::Leaf { size, .. } => { NodeData::Leaf { size, .. } => {
assert_eq!(level, self.size - 1, "Expected inner node"); assert_eq!(level, self.size - 1, "Expected inner node");
assert!( assert!(
self.entry[level] <= size, self.entry[level] <= size,
@@ -677,7 +674,7 @@ impl<F: Forest> Path<F> {
size, size,
); );
} }
&NodeData::Free { .. } => { NodeData::Free { .. } => {
panic!("Free {} in path", self.node[level]); panic!("Free {} in path", self.node[level]);
} }
} }
@@ -702,9 +699,9 @@ impl<F: Forest> fmt::Display for Path<F> {
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use std::cmp::Ordering; use super::super::{Forest, NodeData, NodePool};
use super::*; use super::*;
use super::super::{Forest, NodePool, NodeData}; use std::cmp::Ordering;
struct TC(); struct TC();

15
lib/cretonne/src/bforest/pool.rs
View File

@@ -1,8 +1,8 @@
//! B+-tree node pool. //! B+-tree node pool.
use super::{Forest, Node, NodeData};
use entity::PrimaryMap; use entity::PrimaryMap;
use std::ops::{Index, IndexMut}; use std::ops::{Index, IndexMut};
use super::{Forest, Node, NodeData};
/// A pool of nodes, including a free list. /// A pool of nodes, including a free list.
pub(super) struct NodePool<F: Forest> { pub(super) struct NodePool<F: Forest> {
@@ -57,6 +57,7 @@ impl<F: Forest> NodePool<F> {
pub fn free_tree(&mut self, node: Node) { pub fn free_tree(&mut self, node: Node) {
if let NodeData::Inner { size, tree, .. } = self[node] { if let NodeData::Inner { size, tree, .. } = self[node] {
// Note that we have to capture `tree` by value to avoid borrow checker trouble. // Note that we have to capture `tree` by value to avoid borrow checker trouble.
#[cfg_attr(feature = "cargo-clippy", allow(needless_range_loop))]
for i in 0..usize::from(size + 1) { for i in 0..usize::from(size + 1) {
// Recursively free sub-trees. This recursion can never be deeper than `MAX_PATH`, // Recursively free sub-trees. This recursion can never be deeper than `MAX_PATH`,
// and since most trees have less than a handful of nodes, it is worthwhile to // and since most trees have less than a handful of nodes, it is worthwhile to
@@ -76,11 +77,11 @@ impl<F: Forest> NodePool<F> {
NodeData<F>: ::std::fmt::Display, NodeData<F>: ::std::fmt::Display,
F::Key: ::std::fmt::Display, F::Key: ::std::fmt::Display,
{ {
use super::Comparator;
use entity::SparseSet;
use std::borrow::Borrow; use std::borrow::Borrow;
use std::cmp::Ordering; use std::cmp::Ordering;
use std::vec::Vec; use std::vec::Vec;
use super::Comparator;
use entity::SparseSet;
// The root node can't be an inner node with just a single sub-tree. It should have been // The root node can't be an inner node with just a single sub-tree. It should have been
// pruned. // pruned.
@@ -105,8 +106,8 @@ impl<F: Forest> NodePool<F> {
); );
let mut lower = lkey; let mut lower = lkey;
match &self[node] { match self[node] {
&NodeData::Inner { size, keys, tree } => { NodeData::Inner { size, keys, tree } => {
let size = size as usize; let size = size as usize;
let capacity = tree.len(); let capacity = tree.len();
let keys = &keys[0..size]; let keys = &keys[0..size];
@@ -148,7 +149,7 @@ impl<F: Forest> NodePool<F> {
lower = upper; lower = upper;
} }
} }
&NodeData::Leaf { size, keys, .. } => { NodeData::Leaf { size, keys, .. } => {
let size = size as usize; let size = size as usize;
let capacity = keys.borrow().len(); let capacity = keys.borrow().len();
let keys = &keys.borrow()[0..size]; let keys = &keys.borrow()[0..size];
@@ -191,7 +192,7 @@ impl<F: Forest> NodePool<F> {
lower = upper; lower = upper;
} }
} }
&NodeData::Free { .. } => panic!("Free {} reached", node), NodeData::Free { .. } => panic!("Free {} reached", node),
} }
} }
} }

6
lib/cretonne/src/bforest/set.rs
View File

@@ -1,8 +1,8 @@
//! Forest of sets. //! Forest of sets.
use super::{Comparator, Forest, Node, NodeData, NodePool, Path, SetValue, INNER_SIZE};
use packed_option::PackedOption; use packed_option::PackedOption;
use std::marker::PhantomData; use std::marker::PhantomData;
use super::{INNER_SIZE, Comparator, Forest, NodePool, Node, NodeData, Path, SetValue};
/// Tag type defining forest types for a set. /// Tag type defining forest types for a set.
struct SetTypes<K, C>(PhantomData<(K, C)>); struct SetTypes<K, C>(PhantomData<(K, C)>);
@@ -351,10 +351,10 @@ where
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::super::NodeData;
use super::*;
use std::mem; use std::mem;
use std::vec::Vec; use std::vec::Vec;
use super::*;
use super::super::NodeData;
#[test] #[test]
fn node_size() { fn node_size() {

25
lib/cretonne/src/binemit/memorysink.rs
View File

@@ -14,13 +14,13 @@
//! relocations to a `RelocSink` trait object. Relocations are less frequent than the //! relocations to a `RelocSink` trait object. Relocations are less frequent than the
//! `CodeSink::put*` methods, so the performance impact of the virtual callbacks is less severe. //! `CodeSink::put*` methods, so the performance impact of the virtual callbacks is less severe.
use ir::{ExternalName, JumpTable}; use super::{Addend, CodeOffset, CodeSink, Reloc};
use super::{CodeSink, CodeOffset, Reloc, Addend}; use ir::{ExternalName, JumpTable, SourceLoc, TrapCode};
use std::ptr::write_unaligned; use std::ptr::write_unaligned;
/// A `CodeSink` that writes binary machine code directly into memory. /// A `CodeSink` that writes binary machine code directly into memory.
/// ///
/// A `MemoryCodeSink` object should be used when emitting a Cretonne IL function into executable /// A `MemoryCodeSink` object should be used when emitting a Cretonne IR function into executable
/// memory. It writes machine code directly to a raw pointer without any bounds checking, so make /// memory. It writes machine code directly to a raw pointer without any bounds checking, so make
/// sure to allocate enough memory for the whole function. The number of bytes required is returned /// sure to allocate enough memory for the whole function. The number of bytes required is returned
/// by the `Context::compile()` function. /// by the `Context::compile()` function.
@@ -33,15 +33,21 @@ pub struct MemoryCodeSink<'a> {
data: *mut u8, data: *mut u8,
offset: isize, offset: isize,
relocs: &'a mut RelocSink, relocs: &'a mut RelocSink,
traps: &'a mut TrapSink,
} }
impl<'a> MemoryCodeSink<'a> { impl<'a> MemoryCodeSink<'a> {
/// Create a new memory code sink that writes a function to the memory pointed to by `data`. /// Create a new memory code sink that writes a function to the memory pointed to by `data`.
pub fn new(data: *mut u8, relocs: &mut RelocSink) -> MemoryCodeSink { pub fn new<'sink>(
data: *mut u8,
relocs: &'sink mut RelocSink,
traps: &'sink mut TrapSink,
) -> MemoryCodeSink<'sink> {
MemoryCodeSink { MemoryCodeSink {
data, data,
offset: 0, offset: 0,
relocs, relocs,
traps,
} }
} }
} }
@@ -58,6 +64,12 @@ pub trait RelocSink {
fn reloc_jt(&mut self, CodeOffset, Reloc, JumpTable); fn reloc_jt(&mut self, CodeOffset, Reloc, JumpTable);
} }
/// A trait for receiving trap codes and offsets.
pub trait TrapSink {
/// Add trap information for a specific offset.
fn trap(&mut self, CodeOffset, SourceLoc, TrapCode);
}
impl<'a> CodeSink for MemoryCodeSink<'a> { impl<'a> CodeSink for MemoryCodeSink<'a> {
fn offset(&self) -> CodeOffset { fn offset(&self) -> CodeOffset {
self.offset as CodeOffset self.offset as CodeOffset
@@ -105,4 +117,9 @@ impl<'a> CodeSink for MemoryCodeSink<'a> {
let ofs = self.offset(); let ofs = self.offset();
self.relocs.reloc_jt(ofs, rel, jt); self.relocs.reloc_jt(ofs, rel, jt);
} }
fn trap(&mut self, code: TrapCode, srcloc: SourceLoc) {
let ofs = self.offset();
self.traps.trap(ofs, srcloc, code);
}
} }

Some files were not shown because too many files have changed in this diff Show More