Added Cmm_helpers.Scalar_type #3423
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jvanburen
requested review from
mshinwell and
Gbury
and removed request for
mshinwell
Selection Change CheckThis PR modifies the original selection pass (targeting Mach). |
jvanburen
force-pushed
the
cmm-refactor-unboxed-fields
branch
from
16f1b23 to
443a81c
Compare
#Squashed commit of the following:
#commit c9d7aa6
#Author: Jacob Van Buren <[email protected]>
#Date: Thu Jan 2 14:49:45 2025 -0500
#cleaned up div/mod
#commit 4d9f427
#Author: Jacob Van Buren <[email protected]>
#Date: Thu Jan 2 14:45:42 2025 -0500
#address feedback and simplify division interface
#unified unboxed field getters/setters. This will be useful once we have unboxed integers of different sizes
#formatted
#updated cmm_helpers interface to be more amenable to adding other integer sizes
#Separate test for vectorizer in the CI (#3414)
#* Separate test for vectorizer in the CI
#* Remove vectorizer from "gi" CI job
#CI: simplify the regalloc jobs (#3389)
#ASR 64-bit lane not available in sse instruction (#3413)
#Fix case where parser drops attributes in packed module types. (#3262)
#* Demonstrate dropped attributes in test.
#Signed-off-by: Thomas Del Vecchio <[email protected]>
#* Syntax error on misplaced attribute in packed module types.
#Signed-off-by: Thomas Del Vecchio <[email protected]>
#---------
#Signed-off-by: Thomas Del Vecchio <[email protected]>
#Add attributes to (unsafely) skip jkind check (#3385)
#* Add attributes to (unsafely) skip jkind check
#Add a pair of attributes, [@@unsafe_allow_any_kind_in_intf] and
#[@@unsafe_allow_any_kind_in_impl], which if set on both the impl and the intf
#respectively, skip checking the jkind of the type in a signature against the
#jkind of the type in a struct entirely. This is a more-selective version of the
#`--allow-illegal-crossing` flag, and likely eventually subsumes it.
#Signed-off-by: Aspen Smith <[email protected]>
#* Emit a warning when unsafe_allow_any_kind is added unnecessarily
#Note that this is /only/ done if the attribute is set in both signatures but not
#used - also this is a little over-sensitive (sadly) since this is done during
#sigature inclusion too. A new test covers the over-sensitivity.
#Signed-off-by: Aspen Smith <[email protected]>
#---------
#Signed-off-by: Aspen Smith <[email protected]>
#Add `Variant_with_null` and `Null` variant constructors (#2870)
#* `Variant_with_null`
#* `Null` tagged constructors
#* precise value kind
#* No private re-export
#---------
#Co-authored-by: Diana Kalinichenko <[email protected]>
#Revert "Implement %makearray_dynamic{,_uninit}" (#3408)
#Revert "Implement %makearray_dynamic{,_uninit} (#3317)"
#This reverts commit 6da1dde.
#Upload core files etc upon CI failure (#3405)
#Fix IRC and Greedy allocators (arm64) (#3388)
#Convert float32 constants to int32 in first stage compiler (#3371)
#* convert float32 constants in bytecode output
#* edit
#* edit
#* blocks + test
#* compare against float64 constants
#* tests check proper custom ops
#---------
#Co-authored-by: Diana Kalinichenko <[email protected]>
Squashed commit of the following:
commit 16f1b23
Author: Jacob Van Buren <[email protected]>
Date: Wed Jan 22 15:25:23 2025 -0500
Squashed commit of the following:
commit 7c5c7f0
Author: Jacob Van Buren <[email protected]>
Date: Wed Jan 22 15:24:10 2025 -0500
Squashed commit of the following:
commit a53391d
Author: Max Slater <[email protected]>
Date: Wed Jan 22 13:53:02 2025 -0500
Rename `atomic_cas`/`Compare_and_swap` (#3491)
commit 92b327c
Author: Max Slater <[email protected]>
Date: Tue Jan 21 14:18:56 2025 -0500
Additional operations for int atomics (#3490)
commit 38e792c
Author: Luke Maurer <[email protected]>
Date: Tue Jan 21 17:09:23 2025 +0000
Support `-open Foo` where `Foo` is parameterised (#3489)
The command line
```
ocamlopt -open Foo -parameter P -c bar.ml
```
should be fine, even if `Foo` is itself parameterised by `P`: as usual, we
compile `bar.ml` as if it began with `open! Foo`, and by the subset rule, `Bar`
can refer to `Foo` because it takes at least the same parameters. Unfortunately,
currently we process `-open` before `-parameter`, so when we go to check the
implicit reference to `Foo`, we think there are no parameters, and we report an
error. (Confusingly, the error suggests that the user add `-parameter P` to the
command line.)
The fix is simple: move the code that processes `-parameter` earlier so that
the initial environment is constructed with the parameters already available.
commit 784dc96
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 21 14:07:41 2025 +0000
Rename [emit.mlp] to [emit.ml] on amd64 (#3488)
commit f3b720a
Author: Zesen Qian <[email protected]>
Date: Tue Jan 21 10:35:35 2025 +0000
Module aliases save locks instead of walking them immediately (#3398)
commit 389a7c3
Author: Nick Barnes <[email protected]>
Date: Mon Jan 20 17:55:51 2025 +0000
Add OCAMLRUNPARAM d= parameter for max # domains (#3487)
Allow maximum number of domains to be specified as a OCAMLRUNPARAM parameter.
(cherry picked from commit f92715f1044aea30ca97e497653c883578f91fe6)
Co-authored-by: KC Sivaramakrishnan <[email protected]>
commit 63767d7
Author: Nick Barnes <[email protected]>
Date: Mon Jan 20 17:45:30 2025 +0000
Add caml_runtime_parameters back (#3468)
Add caml_runtime_parameters back.
commit 5e9975e
Author: Greta Yorsh <[email protected]>
Date: Mon Jan 20 17:15:15 2025 +0000
Emit atomic compare and exchange (#3486)
commit a9821e8
Author: Basile Clément <[email protected]>
Date: Mon Jan 20 15:36:09 2025 +0100
Make patricia trees big-endian (#3438)
This patch switches up the implementation of the `Patricia_tree` module
from little-endian to big-endian, with the main motivation to be able to
implement in-order traversal.
The `caml_int_clz_tagged_to_untagged` and `caml_int_tagged_to_tagged` C
stubs are recognized and replaced with the `clz` instruction when
compiling with flambda2, so they are only used in the boot compiler.
commit b8a9789
Author: Leo White <[email protected]>
Date: Fri Jan 17 13:35:19 2025 +0000
Generate specific instructions for atomics on immediates (#3477)
* Generate specific instructions for atomics on immediates
* Fix formatting
commit 7b93134
Author: Greta Yorsh <[email protected]>
Date: Fri Jan 17 09:30:30 2025 +0000
Vectorizer: add tests (#3456)
* Add tests
* Disable ocamlformat on unboxed tests
* Increase -vectorize-max-block-size for tests
* Fix asssertion failure when vectorizing unboxed int32
* Disable float32 on arm64 (not yet implemented)
* improve gen_dune.ml for the vectorizer tests
Co-authored-by: Xavier Clerc <[email protected]>
commit 6379678
Author: Mark Shinwell <[email protected]>
Date: Thu Jan 16 16:08:29 2025 +0000
Add "-ocamlrunparam" linker flag (#3483)
commit f7b2cbe
Author: Xavier Clerc <[email protected]>
Date: Thu Jan 16 15:25:28 2025 +0000
Bump the version of `actions/upload-artifact` (#3474)
* Bump the version of actions/upload-artifact.
* Ensure artifact names are unique.
* To trigger CI.
* Try with commit hash.
commit afb8a55
Author: Mark Shinwell <[email protected]>
Date: Thu Jan 16 14:03:37 2025 +0000
Move two macOS CI controllers to runtime5 (#3482)
commit aae5c40
Author: Mark Shinwell <[email protected]>
Date: Thu Jan 16 13:50:34 2025 +0000
Fix error in caml_get_init_stack_wsize (#3481)
commit 525868c
Author: dkalinichenko-js <[email protected]>
Date: Wed Jan 15 17:47:30 2025 -0500
Use null pointers for `or_null` (#3267)
* runtime changes
* runtime4 changes
* Change `CODE_UNBOXED_INT64` and `CODE_NULL`
* make `Is_block` an inline function
* redefine `Is_long`
* fix
* Change `CODE_UNBOXED_INT64` back
* optimize `Is_block`/`Is_long`
* `null_tag` for `caml_obj_tag`
* consistent naming
* slightly more reassuring comment
* `inline` is unnecessary and might break `#define inline`
* optimization incorrect in presence of nulls
* Constructors and pattern-matching
* Bytecode compilation
* `or_null` is `Variant_or_null`
* Accept tests
* Runtime tests
* Delete obsolete or_null test
---------
Co-authored-by: Diana Kalinichenko <[email protected]>
commit 9796b21
Author: dkalinichenko-js <[email protected]>
Date: Wed Jan 15 17:47:13 2025 -0500
Runtime changes for `or_null` (#3265)
* runtime changes
* runtime4 changes
* Change `CODE_UNBOXED_INT64` and `CODE_NULL`
* make `Is_block` an inline function
* redefine `Is_long`
* fix
* Change `CODE_UNBOXED_INT64` back
* optimize `Is_block`/`Is_long`
* `null_tag` for `caml_obj_tag`
* consistent naming
* slightly more reassuring comment
* `inline` is unnecessary and might break `#define inline`
* optimization incorrect in presence of nulls
---------
Co-authored-by: Diana Kalinichenko <[email protected]>
commit df4a6e0
Author: Chris Casinghino <[email protected]>
Date: Wed Jan 15 13:08:53 2025 -0500
Bump magic numbers for 5.2.0minus-5 (#3478)
commit d1c8d85
Author: Mark Shinwell <[email protected]>
Date: Wed Jan 15 16:44:39 2025 +0000
Peek and poke (#3309)
commit f8caad4
Author: Greta Yorsh <[email protected]>
Date: Wed Jan 15 16:00:39 2025 +0000
Vectorizer: xmm register can hold ocaml values (#3455)
* Add [Valx2] to [Cmm.machtype_component]
* Vectorizer generates [Valx2]
* Record live offsets of [Valx2] in the frametable
For runtime4, xmm register are below [gc_regs], use negative offsets.
* Move [types_are_compatible] from [Reg] to [Proc]
This information has to be in sync with register classes, stack
slock classes, and emit for move instructions.
commit 34a7873
Author: Zesen Qian <[email protected]>
Date: Wed Jan 15 13:13:18 2025 +0000
Improve coherence of modality zapping (#3462)
commit 1a6a9d3
Author: Mark Shinwell <[email protected]>
Date: Wed Jan 15 12:06:24 2025 +0000
Fix caml_obj_with_tag (#3465)
commit bc5110a
Author: Nick Barnes <[email protected]>
Date: Wed Jan 15 11:52:25 2025 +0000
Reset the pacing of major collection after any synchronous major GC (#3463)
Reset the pacing of major collection at the end of any synchronous major collection.
commit 9faf700
Author: Mark Shinwell <[email protected]>
Date: Wed Jan 15 11:32:59 2025 +0000
Remove remnants of caml_obj_truncate (#3469)
commit ff9430b
Author: Luke Maurer <[email protected]>
Date: Wed Jan 15 11:31:22 2025 +0000
Mangle instance symbol names using `____` rather than `___` (#3472)
Apparently there are libraries around that have names ending in single
underscores, leading to ambiguous symbol names if we use triple
underscores to delimit instances. Other choices are possible but this PR
opts for newly-developed quadruple-underscore technology.
commit 9984700
Author: Vincent Laviron <[email protected]>
Date: Wed Jan 15 11:03:43 2025 +0100
Port upstream PRs 11542 and 12505 to runtime4 (#3431)
fix #11482: random crash in large closure allocation (#11542)
Co-authored-by: Damien Doligez <[email protected]>
commit 058c4db
Author: Mark Shinwell <[email protected]>
Date: Tue Jan 14 22:19:32 2025 +0000
Enable all makearray_dynamic tests on runtime4 (#3470)
commit ba15ee5
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 20:08:31 2025 +0000
Vectorize [Ifloatarithmem] (#3452)
* Add [Isimd_mem] to [Arch.Specific] and emit [addpd] with memory arg
and similar instructions
* Vectorize [Ifloatarithmem]
When the memory alignment is known to be 128-bit (currently, never) emits
[addpd], otherwise emits a vector load followed by an arithmetic
instruction.
commit 9755b39
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 18:57:39 2025 +0000
Fix CI failure (#3473)
commit 859949c
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 17:56:53 2025 +0000
Vectorize [Specific.Istore_int] (#3450)
Used for array initialization (amd64)
commit 50f73cb
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 17:38:39 2025 +0000
Do not allow naked pointers (remove configure option) (#3448)
commit b7c8ad3
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 17:07:39 2025 +0000
Vectorizer refactor heuristic for select_and_join (#3449)
* Refactor [Block.find_last_instruction], cache [Computation.last_pos]
* Improve heuristics in [Computation.select_and_join] using [last_pos]
commit 22f81d8
Author: Mark Shinwell <[email protected]>
Date: Tue Jan 14 17:00:29 2025 +0000
Fix mistake in conditional for makearray_dynamic array initialization (#3466)
commit aaaddfb
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 14:57:17 2025 +0000
Vectorizer: propagate alignment of memory accesses (#3451)
Currently it's always 8 but having this argument will help us
consider alignment for new vector sequences.
commit b15d44e
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 14:55:29 2025 +0000
vectorizer: improve debug printout (#3445)
commit 6239156
Author: Stephen Dolan <[email protected]>
Date: Tue Jan 14 13:43:21 2025 +0000
Better hugepage alignment of stacks and heap (#3384)
Co-authored-by: Mark Shinwell <[email protected]>
commit 677d79a
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 13:23:56 2025 +0000
Backend dune copy and directive (#3467)
* Remove unused line directive from [dune]
* Use [copy_files#] to copy files from ARCH and add a file directive
* Remove existing file directives
commit 314b131
Author: Stephen Dolan <[email protected]>
Date: Tue Jan 14 11:58:41 2025 +0000
Bound stack size in expect tests (#3439)
commit 02774f8
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 10:52:50 2025 +0000
all_deps is reflexive (#3464)
commit 117a0a0
Author: Stephen Dolan <[email protected]>
Date: Tue Jan 14 10:44:15 2025 +0000
Stub implementation of new custom memory API (#3437)
commit 4f30aac
Author: Greta Yorsh <[email protected]>
Date: Tue Jan 14 10:31:48 2025 +0000
Vectorizer bug fix: address argument of memory operations (#3446)
Fix bug: use address arg of the first instruction in a group
... not the last!
Only matters for arrays at the moment, where the address offset
argument is not always the same register.
commit cc91e2b
Author: Vincent Laviron <[email protected]>
Date: Tue Jan 14 11:26:59 2025 +0100
caml_update_dummy: fail on closure blocks (#3429)
commit 17a01a9
Author: Mark Shinwell <[email protected]>
Date: Tue Jan 14 10:07:39 2025 +0000
Implement %array_element_size_in_bytes (#3367)
Co-authored-by: Chris Casinghino <[email protected]>
commit b487f71
Author: Greta Yorsh <[email protected]>
Date: Mon Jan 13 14:05:25 2025 +0000
Runtime: make types explicit when reading [gc_regs] (#3453)
Runtime4: make types explicit when reading [gc_regs].
commit 67e6eb3
Author: Max Slater <[email protected]>
Date: Fri Jan 10 16:17:32 2025 -0500
More capsule API updates (#3440)
commit c7f573f
Author: Mark Shinwell <[email protected]>
Date: Fri Jan 10 18:26:15 2025 +0000
Reinstate %makearray_dynamic (#3460)
commit e1e4fb8
Author: Zesen Qian <[email protected]>
Date: Fri Jan 10 16:15:48 2025 +0000
`portable` lazy allows `nonportable` thunk (#3436)
* portable lazy allows nonportable thunk
* add documentation
* improve documentation
* add examples
* improve comments in test
* say "not stronger"
commit c30ec74
Author: Ryan Tjoa <[email protected]>
Date: Fri Jan 10 10:41:08 2025 -0500
Check for type recursion without boxing (#3407)
commit cb290c5
Author: Greta Yorsh <[email protected]>
Date: Fri Jan 10 11:00:32 2025 +0000
Vectorizer: rename New (#3454)
Rename New to New_vec128 to make the type clear
and distinguish it from the upcoming Valx2
commit bd39e02
Author: Greta Yorsh <[email protected]>
Date: Fri Jan 10 10:24:48 2025 +0000
Add function [DLL.for_all_i] (#3442)
* Add function [DLL.for_all_i]
* Rename to [for_alli] to match existing [mapi] and [iteri]
* Remove unused argument of [aux] in [DLL.for_all*]
commit c048920
Author: Greta Yorsh <[email protected]>
Date: Thu Jan 9 13:16:36 2025 +0000
Cleanup machtype_component size (#3441)
Cleanup size_component
commit 830d5e7
Author: Greta Yorsh <[email protected]>
Date: Thu Jan 9 13:15:59 2025 +0000
Add "dump-vectorize" to OCAMLPARAM (#3443)
Add [dump-vectorize] to OCAMLPARAM for debugging
commit 157c95e
Author: Greta Yorsh <[email protected]>
Date: Thu Jan 9 13:15:33 2025 +0000
Vectorizer bug fix: 128-bit vectorized constant (#3447)
Fix bug: 128-bit vectorized constant high/low correctly ordered
commit 648155d
Author: Greta Yorsh <[email protected]>
Date: Thu Jan 9 13:09:48 2025 +0000
Add [Printreg.reglist] for debugging (#3444)
commit d40254f
Author: Stephen Dolan <[email protected]>
Date: Tue Jan 7 21:25:45 2025 +0000
Move two misplaced files (#3435)
commit 4a0bb69
Author: dkalinichenko-js <[email protected]>
Date: Tue Jan 7 15:34:27 2025 -0500
`Yielding` mode axis (#3283)
* `Yielding` mode axis
* Tests
* fix printing
---------
Co-authored-by: Diana Kalinichenko <[email protected]>
commit 00275e0
Author: Max Slater <[email protected]>
Date: Mon Jan 6 13:05:58 2025 -0500
Unbox_float32 should check custom ops name (#3433)
check sym name
commit 2e49469
Author: Max Slater <[email protected]>
Date: Mon Jan 6 13:05:03 2025 -0500
Make Capsule preserve wrapped exception backtraces (#3421)
* with_password
* portable
* don't use polymorphic parameters
* review
* protect encapsulated from other capsule
* raise wrapped exceptions with existing backtrace
* cr
commit 2de23a5
Author: Ryan Tjoa <[email protected]>
Date: Mon Jan 6 04:04:29 2025 -0500
Fix CI by using `setup-ocaml` v3 for ocamlformat workflow (#3426)
[CI] Use setup-ocaml v3 for ocamlformat workflow
commit eada0f1
Author: Ryan Tjoa <[email protected]>
Date: Fri Jan 3 21:23:23 2025 -0500
Move unboxed records to stable (#3419)
commit a273a33
Author: Jacob Van Buren <[email protected]>
Date: Fri Jan 3 11:17:18 2025 -0500
Changed make fmt to run in parallel (#3422)
changed make fmt to run in parallel
commit 4de5a72
Author: Max Slater <[email protected]>
Date: Thu Jan 2 20:10:08 2025 -0500
Add `Capsule.with_password` (#3420)
commit b084ff3
Author: Greta Yorsh <[email protected]>
Date: Wed Jan 1 15:34:11 2025 +0000
vectorizer: new test (#3418)
Add test for register compatiblity
commit 5549015
Author: Greta Yorsh <[email protected]>
Date: Tue Dec 31 17:20:56 2024 +0000
Vectorizer: check register compatibility (#3412)
Check that registers are compatible when joining computations
…ng between integers types of different widths and signedness. This is in preparation for adding unboxed small integer types.
# Squashed commit of the following:
# commit c9d7aa6
# Author: Jacob Van Buren <[email protected]>
# Date: Thu Jan 2 14:49:45 2025 -0500
# cleaned up div/mod
# commit 4d9f427
# Author: Jacob Van Buren <[email protected]>
# Date: Thu Jan 2 14:45:42 2025 -0500
# address feedback and simplify division interface
# unified unboxed field getters/setters. This will be useful once we have unboxed integers of different sizes
# formatted
# updated cmm_helpers interface to be more amenable to adding other integer sizes
# Separate test for vectorizer in the CI (#3414)
# * Separate test for vectorizer in the CI
# * Remove vectorizer from "gi" CI job
# CI: simplify the regalloc jobs (#3389)
# ASR 64-bit lane not available in sse instruction (#3413)
# Fix case where parser drops attributes in packed module types. (#3262)
# * Demonstrate dropped attributes in test.
# Signed-off-by: Thomas Del Vecchio <[email protected]>
# * Syntax error on misplaced attribute in packed module types.
# Signed-off-by: Thomas Del Vecchio <[email protected]>
# ---------
# Signed-off-by: Thomas Del Vecchio <[email protected]>
# Add attributes to (unsafely) skip jkind check (#3385)
# * Add attributes to (unsafely) skip jkind check
# Add a pair of attributes, [@@unsafe_allow_any_kind_in_intf] and
# [@@unsafe_allow_any_kind_in_impl], which if set on both the impl and the intf
# respectively, skip checking the jkind of the type in a signature against the
# jkind of the type in a struct entirely. This is a more-selective version of the
# `--allow-illegal-crossing` flag, and likely eventually subsumes it.
# Signed-off-by: Aspen Smith <[email protected]>
# * Emit a warning when unsafe_allow_any_kind is added unnecessarily
# Note that this is /only/ done if the attribute is set in both signatures but not
# used - also this is a little over-sensitive (sadly) since this is done during
# sigature inclusion too. A new test covers the over-sensitivity.
# Signed-off-by: Aspen Smith <[email protected]>
# ---------
# Signed-off-by: Aspen Smith <[email protected]>
# Add `Variant_with_null` and `Null` variant constructors (#2870)
# * `Variant_with_null`
# * `Null` tagged constructors
# * precise value kind
# * No private re-export
# ---------
# Co-authored-by: Diana Kalinichenko <[email protected]>
# Revert "Implement %makearray_dynamic{,_uninit}" (#3408)
# Revert "Implement %makearray_dynamic{,_uninit} (#3317)"
# This reverts commit 6da1dde.
# Upload core files etc upon CI failure (#3405)
# Fix IRC and Greedy allocators (arm64) (#3388)
# Convert float32 constants to int32 in first stage compiler (#3371)
# * convert float32 constants in bytecode output
# * edit
# * edit
# * blocks + test
# * compare against float64 constants
# * tests check proper custom ops
# ---------
# Co-authored-by: Diana Kalinichenko <[email protected]>
jvanburen
force-pushed
the
cmm-scalar-type
branch
from
39828c2 to
ae1f27d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Cmm_helpers.Scalar_typeprovides utilities for converting between integers types of different widths and signedness. This is in preparation for adding unboxed small integer types.