Remove `spack.target` from code (#46503)

The `spack.target.Target` class is a weird entity, that is just needed to:

1. Sort microarchitectures in lists deterministically
2. Being able to use microarchitectures in hashed containers

This PR removes it, and uses `archspec.cpu.Microarchitecture` directly. To sort lists, we use a proper `key=` when needed. Being able to use `Microarchitecture` objects in sets is achieved by updating the external `archspec`.

Signed-off-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>

28 files changed, 201 insertions, 326 deletions

diff --git a/lib/spack/docs/conf.py b/lib/spack/docs/conf.py
index 95b711db93..c80a661abd 100644
--- a/lib/spack/docs/conf.py
+++ b/lib/spack/docs/conf.py
@@ -219,6 +219,7 @@ nitpick_ignore = [
     ("py:class", "spack.install_test.Pb"),
     ("py:class", "spack.filesystem_view.SimpleFilesystemView"),
     ("py:class", "spack.traverse.EdgeAndDepth"),
+    ("py:class", "archspec.cpu.microarchitecture.Microarchitecture"),
 ]
 
 # The reST default role (used for this markup: `text`) to use for all documents.
diff --git a/lib/spack/external/__init__.py b/lib/spack/external/__init__.py
index b6a35925d5..603cb412c8 100644
--- a/lib/spack/external/__init__.py
+++ b/lib/spack/external/__init__.py
@@ -18,7 +18,7 @@ archspec
 
 * Homepage: https://pypi.python.org/pypi/archspec
 * Usage: Labeling, comparison and detection of microarchitectures
-* Version: 0.2.5-dev (commit cbb1fd5eb397a70d466e5160b393b87b0dbcc78f)
+* Version: 0.2.5-dev (commit bceb39528ac49dd0c876b2e9bf3e7482e9c2be4a)
 
 astunparse
 ----------------
diff --git a/lib/spack/external/archspec/cpu/microarchitecture.py b/lib/spack/external/archspec/cpu/microarchitecture.py
index 1ffe51d918..3f4e30fab5 100644
--- a/lib/spack/external/archspec/cpu/microarchitecture.py
+++ b/lib/spack/external/archspec/cpu/microarchitecture.py
@@ -115,6 +115,9 @@ class Microarchitecture:
             and self.cpu_part == other.cpu_part
         )
 
+    def __hash__(self):
+        return hash(self.name)
+
     @coerce_target_names
     def __ne__(self, other):
         return not self == other
diff --git a/lib/spack/spack/build_environment.py b/lib/spack/spack/build_environment.py
index 26eb0e79f6..fd0adbf39d 100644
--- a/lib/spack/spack/build_environment.py
+++ b/lib/spack/spack/build_environment.py
@@ -45,6 +45,8 @@ from enum import Flag, auto
 from itertools import chain
 from typing import Dict, List, Set, Tuple
 
+import archspec.cpu
+
 import llnl.util.tty as tty
 from llnl.string import plural
 from llnl.util.filesystem import join_path
@@ -358,7 +360,7 @@ def set_compiler_environment_variables(pkg, env):
     _add_werror_handling(keep_werror, env)
 
     # Set the target parameters that the compiler will add
-    isa_arg = spec.architecture.target.optimization_flags(compiler)
+    isa_arg = optimization_flags(compiler, spec.target)
     env.set("SPACK_TARGET_ARGS", isa_arg)
 
     # Trap spack-tracked compiler flags as appropriate.
@@ -403,6 +405,36 @@ def set_compiler_environment_variables(pkg, env):
     return env
 
 
+def optimization_flags(compiler, target):
+    if spack.compilers.is_mixed_toolchain(compiler):
+        msg = (
+            "microarchitecture specific optimizations are not "
+            "supported yet on mixed compiler toolchains [check"
+            f" {compiler.name}@{compiler.version} for further details]"
+        )
+        tty.debug(msg)
+        return ""
+
+    # Try to check if the current compiler comes with a version number or
+    # has an unexpected suffix. If so, treat it as a compiler with a
+    # custom spec.
+    compiler_version = compiler.version
+    version_number, suffix = archspec.cpu.version_components(compiler.version)
+    if not version_number or suffix:
+        try:
+            compiler_version = compiler.real_version
+        except spack.util.executable.ProcessError as e:
+            # log this and just return compiler.version instead
+            tty.debug(str(e))
+
+    try:
+        result = target.optimization_flags(compiler.name, compiler_version.dotted_numeric_string)
+    except (ValueError, archspec.cpu.UnsupportedMicroarchitecture):
+        result = ""
+
+    return result
+
+
 def set_wrapper_variables(pkg, env):
     """Set environment variables used by the Spack compiler wrapper (which have the prefix
     `SPACK_`) and also add the compiler wrappers to PATH.
@@ -783,7 +815,6 @@ def setup_package(pkg, dirty, context: Context = Context.BUILD):
     # Platform specific setup goes before package specific setup. This is for setting
     # defaults like MACOSX_DEPLOYMENT_TARGET on macOS.
     platform = spack.platforms.by_name(pkg.spec.architecture.platform)
-    target = platform.target(pkg.spec.architecture.target)
     platform.setup_platform_environment(pkg, env_mods)
 
     tty.debug("setup_package: grabbing modifications from dependencies")
@@ -808,9 +839,6 @@ def setup_package(pkg, dirty, context: Context = Context.BUILD):
         for mod in pkg.compiler.modules:
             load_module(mod)
 
-    if target and target.module_name:
-        load_module(target.module_name)
-
     load_external_modules(pkg)
 
     implicit_rpaths = pkg.compiler.implicit_rpaths()
diff --git a/lib/spack/spack/compilers/__init__.py b/lib/spack/spack/compilers/__init__.py
index 6b03810826..bfb781685f 100644
--- a/lib/spack/spack/compilers/__init__.py
+++ b/lib/spack/spack/compilers/__init__.py
@@ -558,7 +558,7 @@ def get_compilers(config, cspec=None, arch_spec=None):
         except KeyError:
             # TODO: Check if this exception handling makes sense, or if we
             # TODO: need to change / refactor tests
-            family = arch_spec.target
+            family = str(arch_spec.target)
         except AttributeError:
             assert arch_spec is None
 
@@ -803,12 +803,11 @@ class CompilerConfigFactory:
         if not spec.architecture:
             host_platform = spack.platforms.host()
             operating_system = host_platform.operating_system("default_os")
-            target = host_platform.target("default_target").microarchitecture
+            target = host_platform.target("default_target")
         else:
             target = spec.architecture.target
             if not target:
                 target = spack.platforms.host().target("default_target")
-            target = target.microarchitecture
 
             operating_system = spec.os
             if not operating_system:
diff --git a/lib/spack/spack/patch.py b/lib/spack/spack/patch.py
index 50393ecfad..a0f4152317 100644
--- a/lib/spack/spack/patch.py
+++ b/lib/spack/spack/patch.py
@@ -15,7 +15,7 @@ from llnl.url import allowed_archive
 
 import spack
 import spack.error
-import spack.fetch_strategy as fs
+import spack.fetch_strategy
 import spack.mirror
 import spack.repo
 import spack.stage
@@ -314,11 +314,15 @@ class UrlPatch(Patch):
 
         # Two checksums, one for compressed file, one for its contents
         if self.archive_sha256 and self.sha256:
-            fetcher: fs.FetchStrategy = fs.FetchAndVerifyExpandedFile(
-                self.url, archive_sha256=self.archive_sha256, expanded_sha256=self.sha256
+            fetcher: spack.fetch_strategy.FetchStrategy = (
+                spack.fetch_strategy.FetchAndVerifyExpandedFile(
+                    self.url, archive_sha256=self.archive_sha256, expanded_sha256=self.sha256
+                )
             )
         else:
-            fetcher = fs.URLFetchStrategy(url=self.url, sha256=self.sha256, expand=False)
+            fetcher = spack.fetch_strategy.URLFetchStrategy(
+                url=self.url, sha256=self.sha256, expand=False
+            )
 
         # The same package can have multiple patches with the same name but
         # with different contents, therefore apply a subset of the hash.
@@ -397,7 +401,7 @@ def from_dict(
         sha256 = dictionary["sha256"]
         checker = Checker(sha256)
         if patch.path and not checker.check(patch.path):
-            raise fs.ChecksumError(
+            raise spack.fetch_strategy.ChecksumError(
                 "sha256 checksum failed for %s" % patch.path,
                 "Expected %s but got %s " % (sha256, checker.sum)
                 + "Patch may have changed since concretization.",
diff --git a/lib/spack/spack/platforms/_platform.py b/lib/spack/spack/platforms/_platform.py
index c165cf9f33..8668958966 100644
--- a/lib/spack/spack/platforms/_platform.py
+++ b/lib/spack/spack/platforms/_platform.py
@@ -4,6 +4,8 @@
 # SPDX-License-Identifier: (Apache-2.0 OR MIT)
 from typing import Optional
 
+import archspec.cpu
+
 import llnl.util.lang
 
 import spack.error
@@ -60,7 +62,7 @@ class Platform:
         self.operating_sys = {}
         self.name = name
 
-    def add_target(self, name, target):
+    def add_target(self, name: str, target: archspec.cpu.Microarchitecture) -> None:
         """Used by the platform specific subclass to list available targets.
         Raises an error if the platform specifies a name
         that is reserved by spack as an alias.
@@ -70,6 +72,10 @@ class Platform:
             raise ValueError(msg.format(name))
         self.targets[name] = target
 
+    def _add_archspec_targets(self):
+        for name, microarchitecture in archspec.cpu.TARGETS.items():
+            self.add_target(name, microarchitecture)
+
     def target(self, name):
         """This is a getter method for the target dictionary
         that handles defaulting based on the values provided by default,
diff --git a/lib/spack/spack/platforms/darwin.py b/lib/spack/spack/platforms/darwin.py
index 1b7a5927f4..7ea6a09c58 100644
--- a/lib/spack/spack/platforms/darwin.py
+++ b/lib/spack/spack/platforms/darwin.py
@@ -7,7 +7,6 @@ import platform as py_platform
 
 import archspec.cpu
 
-import spack.target
 from spack.operating_systems.mac_os import MacOs
 from spack.version import Version
 
@@ -21,9 +20,7 @@ class Darwin(Platform):
 
     def __init__(self):
         super().__init__("darwin")
-
-        for name in archspec.cpu.TARGETS:
-            self.add_target(name, spack.target.Target(name))
+        self._add_archspec_targets()
 
         self.default = archspec.cpu.host().name
         self.front_end = self.default
diff --git a/lib/spack/spack/platforms/freebsd.py b/lib/spack/spack/platforms/freebsd.py
index 4485550789..af9b1a0934 100644
--- a/lib/spack/spack/platforms/freebsd.py
+++ b/lib/spack/spack/platforms/freebsd.py
@@ -6,7 +6,6 @@ import platform
 
 import archspec.cpu
 
-import spack.target
 from spack.operating_systems.freebsd import FreeBSDOs
 
 from ._platform import Platform
@@ -18,8 +17,7 @@ class FreeBSD(Platform):
     def __init__(self):
         super().__init__("freebsd")
 
-        for name in archspec.cpu.TARGETS:
-            self.add_target(name, spack.target.Target(name))
+        self._add_archspec_targets()
 
         # Get specific default
         self.default = archspec.cpu.host().name
diff --git a/lib/spack/spack/platforms/linux.py b/lib/spack/spack/platforms/linux.py
index 2be5b51cb2..aede37d868 100644
--- a/lib/spack/spack/platforms/linux.py
+++ b/lib/spack/spack/platforms/linux.py
@@ -6,7 +6,6 @@ import platform
 
 import archspec.cpu
 
-import spack.target
 from spack.operating_systems.linux_distro import LinuxDistro
 
 from ._platform import Platform
@@ -18,8 +17,7 @@ class Linux(Platform):
     def __init__(self):
         super().__init__("linux")
 
-        for name in archspec.cpu.TARGETS:
-            self.add_target(name, spack.target.Target(name))
+        self._add_archspec_targets()
 
         # Get specific default
         self.default = archspec.cpu.host().name
diff --git a/lib/spack/spack/platforms/test.py b/lib/spack/spack/platforms/test.py
index 9ead66ab27..8633efc7cb 100644
--- a/lib/spack/spack/platforms/test.py
+++ b/lib/spack/spack/platforms/test.py
@@ -4,8 +4,9 @@
 # SPDX-License-Identifier: (Apache-2.0 OR MIT)
 import platform
 
+import archspec.cpu
+
 import spack.operating_systems
-import spack.target
 
 from ._platform import Platform
 
@@ -32,8 +33,8 @@ class Test(Platform):
     def __init__(self, name=None):
         name = name or "test"
         super().__init__(name)
-        self.add_target(self.default, spack.target.Target(self.default))
-        self.add_target(self.front_end, spack.target.Target(self.front_end))
+        self.add_target(self.default, archspec.cpu.TARGETS[self.default])
+        self.add_target(self.front_end, archspec.cpu.TARGETS[self.front_end])
 
         self.add_operating_system(
             self.default_os, spack.operating_systems.OperatingSystem("debian", 6)
diff --git a/lib/spack/spack/platforms/windows.py b/lib/spack/spack/platforms/windows.py
index c00382e198..8cc89b4b19 100755
--- a/lib/spack/spack/platforms/windows.py
+++ b/lib/spack/spack/platforms/windows.py
@@ -7,7 +7,6 @@ import platform
 
 import archspec.cpu
 
-import spack.target
 from spack.operating_systems.windows_os import WindowsOs
 
 from ._platform import Platform
@@ -18,9 +17,7 @@ class Windows(Platform):
 
     def __init__(self):
         super().__init__("windows")
-
-        for name in archspec.cpu.TARGETS:
-            self.add_target(name, spack.target.Target(name))
+        self._add_archspec_targets()
 
         self.default = archspec.cpu.host().name
         self.front_end = self.default
diff --git a/lib/spack/spack/solver/asp.py b/lib/spack/spack/solver/asp.py
index d8d052f311..3fd8b9cc8d 100644
--- a/lib/spack/spack/solver/asp.py
+++ b/lib/spack/spack/solver/asp.py
@@ -2479,7 +2479,7 @@ class SpackSolverSetup:
             return allowed_targets
 
         cache = {}
-        for target_constraint in sorted(self.target_constraints):
+        for target_constraint in sorted(self.target_constraints, key=lambda x: x.name):
             # Construct the list of allowed targets for this constraint
             allowed_targets = []
             for single_constraint in str(target_constraint).split(","):
@@ -3237,7 +3237,7 @@ class CompilerParser:
         candidate = KnownCompiler(
             spec=spec.compiler,
             os=str(spec.architecture.os),
-            target=str(spec.architecture.target.microarchitecture.family),
+            target=str(spec.architecture.target.family),
             available=False,
             compiler_obj=None,
         )
diff --git a/lib/spack/spack/spec.py b/lib/spack/spack/spec.py
index fb1a05f37c..1c566349a4 100644
--- a/lib/spack/spack/spec.py
+++ b/lib/spack/spack/spec.py
@@ -61,6 +61,8 @@ import socket
 import warnings
 from typing import Any, Callable, Dict, List, Match, Optional, Set, Tuple, Union
 
+import archspec.cpu
+
 import llnl.path
 import llnl.string
 import llnl.util.filesystem as fs
@@ -82,7 +84,6 @@ import spack.provider_index
 import spack.repo
 import spack.solver
 import spack.store
-import spack.target
 import spack.traverse as traverse
 import spack.util.executable
 import spack.util.hash
@@ -213,6 +214,12 @@ def ensure_modern_format_string(fmt: str) -> None:
         )
 
 
+def _make_microarchitecture(name: str) -> archspec.cpu.Microarchitecture:
+    if isinstance(name, archspec.cpu.Microarchitecture):
+        return name
+    return archspec.cpu.TARGETS.get(name, archspec.cpu.generic_microarchitecture(name))
+
+
 @lang.lazy_lexicographic_ordering
 class ArchSpec:
     """Aggregate the target platform, the operating system and the target microarchitecture."""
@@ -301,7 +308,10 @@ class ArchSpec:
     def _cmp_iter(self):
         yield self.platform
         yield self.os
-        yield self.target
+        if self.target is None:
+            yield self.target
+        else:
+            yield self.target.name
 
     @property
     def platform(self):
@@ -360,10 +370,10 @@ class ArchSpec:
         # will assumed to be the host machine's platform.
 
         def target_or_none(t):
-            if isinstance(t, spack.target.Target):
+            if isinstance(t, archspec.cpu.Microarchitecture):
                 return t
             if t and t != "None":
-                return spack.target.Target(t)
+                return _make_microarchitecture(t)
             return None
 
         value = target_or_none(value)
@@ -452,10 +462,11 @@ class ArchSpec:
         results = self._target_intersection(other)
         attribute_str = ",".join(results)
 
-        if self.target == attribute_str:
+        intersection_target = _make_microarchitecture(attribute_str)
+        if self.target == intersection_target:
             return False
 
-        self.target = attribute_str
+        self.target = intersection_target
         return True
 
     def _target_intersection(self, other):
@@ -473,7 +484,7 @@ class ArchSpec:
                     # s_target_range is a concrete target
                     # get a microarchitecture reference for at least one side
                     # of each comparison so we can use archspec comparators
-                    s_comp = spack.target.Target(s_min).microarchitecture
+                    s_comp = _make_microarchitecture(s_min)
                     if not o_sep:
                         if s_min == o_min:
                             results.append(s_min)
@@ -481,21 +492,21 @@ class ArchSpec:
                         results.append(s_min)
                 elif not o_sep:
                     # "cast" to microarchitecture
-                    o_comp = spack.target.Target(o_min).microarchitecture
+                    o_comp = _make_microarchitecture(o_min)
                     if (not s_min or o_comp >= s_min) and (not s_max or o_comp <= s_max):
                         results.append(o_min)
                 else:
                     # Take intersection of two ranges
                     # Lots of comparisons needed
-                    _s_min = spack.target.Target(s_min).microarchitecture
-                    _s_max = spack.target.Target(s_max).microarchitecture
-                    _o_min = spack.target.Target(o_min).microarchitecture
-                    _o_max = spack.target.Target(o_max).microarchitecture
+                    _s_min = _make_microarchitecture(s_min)
+                    _s_max = _make_microarchitecture(s_max)
+                    _o_min = _make_microarchitecture(o_min)
+                    _o_max = _make_microarchitecture(o_max)
 
                     n_min = s_min if _s_min >= _o_min else o_min
                     n_max = s_max if _s_max <= _o_max else o_max
-                    _n_min = spack.target.Target(n_min).microarchitecture
-                    _n_max = spack.target.Target(n_max).microarchitecture
+                    _n_min = _make_microarchitecture(n_min)
+                    _n_max = _make_microarchitecture(n_max)
                     if _n_min == _n_max:
                         results.append(n_min)
                     elif not n_min or not n_max or _n_min < _n_max:
@@ -548,12 +559,18 @@ class ArchSpec:
         )
 
     def to_dict(self):
+        # Generic targets represent either an architecture family (like x86_64)
+        # or a custom micro-architecture
+        if self.target.vendor == "generic":
+            target_data = str(self.target)
+        else:
+            # Get rid of compiler flag information before turning the uarch into a dict
+            uarch_dict = self.target.to_dict()
+            uarch_dict.pop("compilers", None)
+            target_data = syaml.syaml_dict(uarch_dict.items())
+
         d = syaml.syaml_dict(
-            [
-                ("platform", self.platform),
-                ("platform_os", self.os),
-                ("target", self.target.to_dict_or_value()),
-            ]
+            [("platform", self.platform), ("platform_os", self.os), ("target", target_data)]
         )
         return syaml.syaml_dict([("arch", d)])
 
@@ -561,7 +578,10 @@ class ArchSpec:
     def from_dict(d):
         """Import an ArchSpec from raw YAML/JSON data"""
         arch = d["arch"]
-        target = spack.target.Target.from_dict_or_value(arch["target"])
+        target_name = arch["target"]
+        if not isinstance(target_name, str):
+            target_name = target_name["name"]
+        target = _make_microarchitecture(target_name)
         return ArchSpec((arch["platform"], arch["platform_os"], target))
 
     def __str__(self):
@@ -4120,9 +4140,7 @@ class Spec:
 
     @property
     def target(self):
-        # This property returns the underlying microarchitecture object
-        # to give to the attribute the appropriate comparison semantic
-        return self.architecture.target.microarchitecture
+        return self.architecture.target
 
     @property
     def build_spec(self):
diff --git a/lib/spack/spack/stage.py b/lib/spack/spack/stage.py
index ee3b4e95e5..8b4efcf387 100644
--- a/lib/spack/spack/stage.py
+++ b/lib/spack/spack/stage.py
@@ -33,7 +33,6 @@ from llnl.util.tty.color import colorize
 import spack.caches
 import spack.config
 import spack.error
-import spack.fetch_strategy as fs
 import spack.mirror
 import spack.resource
 import spack.spec
@@ -43,6 +42,7 @@ import spack.util.parallel
 import spack.util.path as sup
 import spack.util.pattern as pattern
 import spack.util.url as url_util
+from spack import fetch_strategy as fs  # breaks a cycle
 from spack.util.crypto import bit_length, prefix_bits
 from spack.util.editor import editor, executable
 from spack.version import StandardVersion, VersionList
@@ -352,8 +352,8 @@ class Stage(LockableStagingDir):
         url_or_fetch_strategy,
         *,
         name=None,
-        mirror_paths: Optional[spack.mirror.MirrorLayout] = None,
-        mirrors: Optional[Iterable[spack.mirror.Mirror]] = None,
+        mirror_paths: Optional["spack.mirror.MirrorLayout"] = None,
+        mirrors: Optional[Iterable["spack.mirror.Mirror"]] = None,
         keep=False,
         path=None,
         lock=True,
@@ -464,7 +464,7 @@ class Stage(LockableStagingDir):
         """Returns the well-known source directory path."""
         return os.path.join(self.path, _source_path_subdir)
 
-    def _generate_fetchers(self, mirror_only=False) -> Generator[fs.FetchStrategy, None, None]:
+    def _generate_fetchers(self, mirror_only=False) -> Generator["fs.FetchStrategy", None, None]:
         fetchers: List[fs.FetchStrategy] = []
         if not mirror_only:
             fetchers.append(self.default_fetcher)
@@ -600,7 +600,7 @@ class Stage(LockableStagingDir):
         spack.caches.FETCH_CACHE.store(self.fetcher, self.mirror_layout.path)
 
     def cache_mirror(
-        self, mirror: spack.caches.MirrorCache, stats: spack.mirror.MirrorStats
+        self, mirror: "spack.caches.MirrorCache", stats: "spack.mirror.MirrorStats"
     ) -> None:
         """Perform a fetch if the resource is not already cached
 
@@ -668,7 +668,7 @@ class Stage(LockableStagingDir):
 class ResourceStage(Stage):
     def __init__(
         self,
-        fetch_strategy: fs.FetchStrategy,
+        fetch_strategy: "fs.FetchStrategy",
         root: Stage,
         resource: spack.resource.Resource,
         **kwargs,
diff --git a/lib/spack/spack/target.py b/lib/spack/spack/target.py
deleted file mode 100644
index 4a78196865..0000000000
--- a/lib/spack/spack/target.py
+++ /dev/null
@@ -1,161 +0,0 @@
-# Copyright 2013-2024 Lawrence Livermore National Security, LLC and other
-# Spack Project Developers. See the top-level COPYRIGHT file for details.
-#
-# SPDX-License-Identifier: (Apache-2.0 OR MIT)
-import functools
-
-import archspec.cpu
-
-import llnl.util.tty as tty
-
-import spack.compiler
-import spack.compilers
-import spack.spec
-import spack.util.executable
-import spack.util.spack_yaml as syaml
-
-
-def _ensure_other_is_target(method):
-    """In a single argument method, ensure that the argument is an
-    instance of ``Target``.
-    """
-
-    @functools.wraps(method)
-    def _impl(self, other):
-        if isinstance(other, str):
-            other = Target(other)
-
-        if not isinstance(other, Target):
-            return NotImplemented
-
-        return method(self, other)
-
-    return _impl
-
-
-class Target:
-    def __init__(self, name, module_name=None):
-        """Target models microarchitectures and their compatibility.
-
-        Args:
-            name (str or Microarchitecture): microarchitecture of the target
-            module_name (str): optional module name to get access to the
-                current target. This is typically used on machines
-                like Cray (e.g. craype-compiler)
-        """
-        if not isinstance(name, archspec.cpu.Microarchitecture):
-            name = archspec.cpu.TARGETS.get(name, archspec.cpu.generic_microarchitecture(name))
-        self.microarchitecture = name
-        self.module_name = module_name
-
-    @property
-    def name(self):
-        return self.microarchitecture.name
-
-    @_ensure_other_is_target
-    def __eq__(self, other):
-        return (
-            self.microarchitecture == other.microarchitecture
-            and self.module_name == other.module_name
-        )
-
-    def __ne__(self, other):
-        # This method is necessary as long as we support Python 2. In Python 3
-        # __ne__ defaults to the implementation below
-        return not self == other
-
-    @_ensure_other_is_target
-    def __lt__(self, other):
-        # TODO: In the future it would be convenient to say
-        # TODO: `spec.architecture.target < other.architecture.target`
-        # TODO: and change the semantic of the comparison operators
-
-        # This is needed to sort deterministically specs in a list.
-        # It doesn't implement a total ordering semantic.
-        return self.microarchitecture.name < other.microarchitecture.name
-
-    def __hash__(self):
-        return hash((self.name, self.module_name))
-
-    @staticmethod
-    def from_dict_or_value(dict_or_value):
-        # A string here represents a generic target (like x86_64 or ppc64) or
-        # a custom micro-architecture
-        if isinstance(dict_or_value, str):
-            return Target(dict_or_value)
-
-        # TODO: From a dict we actually retrieve much more information than
-        # TODO: just the name. We can use that information to reconstruct an
-        # TODO: "old" micro-architecture or check the current definition.
-        target_info = dict_or_value
-        return Target(target_info["name"])
-
-    def to_dict_or_value(self):
-        """Returns a dict or a value representing the current target.
-
-        String values are used to keep backward compatibility with generic
-        targets, like e.g. x86_64 or ppc64. More specific micro-architectures
-        will return a dictionary which contains information on the name,
-        features, vendor, generation and parents of the current target.
-        """
-        # Generic targets represent either an architecture
-        # family (like x86_64) or a custom micro-architecture
-        if self.microarchitecture.vendor == "generic":
-            return str(self)
-
-        # Get rid of compiler flag information before turning the uarch into a dict
-        uarch_dict = self.microarchitecture.to_dict()
-        uarch_dict.pop("compilers", None)
-        return syaml.syaml_dict(uarch_dict.items())
-
-    def __repr__(self):
-        cls_name = self.__class__.__name__
-        fmt = cls_name + "({0}, {1})"
-        return fmt.format(repr(self.microarchitecture), repr(self.module_name))
-
-    def __str__(self):
-        return str(self.microarchitecture)
-
-    def __contains__(self, cpu_flag):
-        return cpu_flag in self.microarchitecture
-
-    def optimization_flags(self, compiler):
-        """Returns the flags needed to optimize for this target using
-        the compiler passed as argument.
-
-        Args:
-            compiler (spack.spec.CompilerSpec or spack.compiler.Compiler): object that
-                contains both the name and the version of the compiler we want to use
-        """
-        # Mixed toolchains are not supported yet
-        if isinstance(compiler, spack.compiler.Compiler):
-            if spack.compilers.is_mixed_toolchain(compiler):
-                msg = (
-                    "microarchitecture specific optimizations are not "
-                    "supported yet on mixed compiler toolchains [check"
-                    " {0.name}@{0.version} for further details]"
-                )
-                tty.debug(msg.format(compiler))
-                return ""
-
-        # Try to check if the current compiler comes with a version number or
-        # has an unexpected suffix. If so, treat it as a compiler with a
-        # custom spec.
-        compiler_version = compiler.version
-        version_number, suffix = archspec.cpu.version_components(compiler.version)
-        if not version_number or suffix:
-            # Try to deduce the underlying version of the compiler, regardless
-            # of its name in compilers.yaml. Depending on where this function
-            # is called we might get either a CompilerSpec or a fully fledged
-            # compiler object.
-            if isinstance(compiler, spack.spec.CompilerSpec):
-                compiler = spack.compilers.compilers_for_spec(compiler).pop()
-            try:
-                compiler_version = compiler.real_version
-            except spack.util.executable.ProcessError as e:
-                # log this and just return compiler.version instead
-                tty.debug(str(e))
-
-        return self.microarchitecture.optimization_flags(
-            compiler.name, compiler_version.dotted_numeric_string
-        )
diff --git a/lib/spack/spack/test/architecture.py b/lib/spack/spack/test/architecture.py
index cbd8a0b9d7..3d489a22cc 100644
--- a/lib/spack/spack/test/architecture.py
+++ b/lib/spack/spack/test/architecture.py
@@ -12,7 +12,6 @@ import spack.compilers
 import spack.concretize
 import spack.operating_systems
 import spack.platforms
-import spack.target
 from spack.spec import ArchSpec, Spec
 
 
@@ -84,25 +83,6 @@ def test_operating_system_conversion_to_dict():
 
 
 @pytest.mark.parametrize(
-    "cpu_flag,target_name",
-    [
-        # Test that specific flags can be used in queries
-        ("ssse3", "haswell"),
-        ("popcnt", "nehalem"),
-        ("avx512f", "skylake_avx512"),
-        ("avx512ifma", "icelake"),
-        # Test that proxy flags can be used in queries too
-        ("sse3", "nehalem"),
-        ("avx512", "skylake_avx512"),
-        ("avx512", "icelake"),
-    ],
-)
-def test_target_container_semantic(cpu_flag, target_name):
-    target = spack.target.Target(target_name)
-    assert cpu_flag in target
-
-
-@pytest.mark.parametrize(
     "item,architecture_str",
     [
         # We can search the architecture string representation
@@ -118,68 +98,6 @@ def test_arch_spec_container_semantic(item, architecture_str):
     assert item in architecture
 
 
-@pytest.mark.parametrize(
-    "compiler_spec,target_name,expected_flags",
-    [
-        # Homogeneous compilers
-        ("gcc@4.7.2", "ivybridge", "-march=core-avx-i -mtune=core-avx-i"),
-        ("clang@3.5", "x86_64", "-march=x86-64 -mtune=generic"),
-        ("apple-clang@9.1.0", "x86_64", "-march=x86-64"),
-        # Mixed toolchain
-        ("clang@8.0.0", "broadwell", ""),
-    ],
-)
-@pytest.mark.filterwarnings("ignore:microarchitecture specific")
-@pytest.mark.not_on_windows("Windows doesn't support the compiler wrapper")
-def test_optimization_flags(compiler_spec, target_name, expected_flags, compiler_factory):
-    target = spack.target.Target(target_name)
-    compiler_dict = compiler_factory(spec=compiler_spec, operating_system="")["compiler"]
-    if compiler_spec == "clang@8.0.0":
-        compiler_dict["paths"] = {
-            "cc": "/path/to/clang-8",
-            "cxx": "/path/to/clang++-8",
-            "f77": "/path/to/gfortran-9",
-            "fc": "/path/to/gfortran-9",
-        }
-    compiler = spack.compilers.compiler_from_dict(compiler_dict)
-
-    opt_flags = target.optimization_flags(compiler)
-    assert opt_flags == expected_flags
-
-
-@pytest.mark.parametrize(
-    "compiler_str,real_version,target_str,expected_flags",
-    [
-        ("gcc@=9.2.0", None, "haswell", "-march=haswell -mtune=haswell"),
-        # Check that custom string versions are accepted
-        ("gcc@=10foo", "9.2.0", "icelake", "-march=icelake-client -mtune=icelake-client"),
-        # Check that we run version detection (4.4.0 doesn't support icelake)
-        ("gcc@=4.4.0-special", "9.2.0", "icelake", "-march=icelake-client -mtune=icelake-client"),
-        # Check that the special case for Apple's clang is treated correctly
-        # i.e. it won't try to detect the version again
-        ("apple-clang@=9.1.0", None, "x86_64", "-march=x86-64"),
-    ],
-)
-def test_optimization_flags_with_custom_versions(
-    compiler_str,
-    real_version,
-    target_str,
-    expected_flags,
-    monkeypatch,
-    mutable_config,
-    compiler_factory,
-):
-    target = spack.target.Target(target_str)
-    compiler_dict = compiler_factory(spec=compiler_str, operating_system="redhat6")
-    mutable_config.set("compilers", [compiler_dict])
-    if real_version:
-        monkeypatch.setattr(spack.compiler.Compiler, "get_real_version", lambda x: real_version)
-    compiler = spack.compilers.compiler_from_dict(compiler_dict["compiler"])
-
-    opt_flags = target.optimization_flags(compiler)
-    assert opt_flags == expected_flags
-
-
 @pytest.mark.regression("15306")
 @pytest.mark.parametrize(
     "architecture_tuple,constraint_tuple",
diff --git a/lib/spack/spack/test/build_environment.py b/lib/spack/spack/test/build_environment.py
index c3ccaee029..f435a893a1 100644
--- a/lib/spack/spack/test/build_environment.py
+++ b/lib/spack/spack/test/build_environment.py
@@ -9,6 +9,8 @@ import sys
 
 import pytest
 
+import archspec.cpu
+
 from llnl.path import Path, convert_to_platform_path
 from llnl.util.filesystem import HeaderList, LibraryList
 
@@ -737,3 +739,64 @@ def test_rpath_with_duplicate_link_deps():
     assert child in rpath_deps
     assert runtime_2 in rpath_deps
     assert runtime_1 not in rpath_deps
+
+
+@pytest.mark.parametrize(
+    "compiler_spec,target_name,expected_flags",
+    [
+        # Homogeneous compilers
+        ("gcc@4.7.2", "ivybridge", "-march=core-avx-i -mtune=core-avx-i"),
+        ("clang@3.5", "x86_64", "-march=x86-64 -mtune=generic"),
+        ("apple-clang@9.1.0", "x86_64", "-march=x86-64"),
+        # Mixed toolchain
+        ("clang@8.0.0", "broadwell", ""),
+    ],
+)
+@pytest.mark.filterwarnings("ignore:microarchitecture specific")
+@pytest.mark.not_on_windows("Windows doesn't support the compiler wrapper")
+def test_optimization_flags(compiler_spec, target_name, expected_flags, compiler_factory):
+    target = archspec.cpu.TARGETS[target_name]
+    compiler_dict = compiler_factory(spec=compiler_spec, operating_system="")["compiler"]
+    if compiler_spec == "clang@8.0.0":
+        compiler_dict["paths"] = {
+            "cc": "/path/to/clang-8",
+            "cxx": "/path/to/clang++-8",
+            "f77": "/path/to/gfortran-9",
+            "fc": "/path/to/gfortran-9",
+        }
+    compiler = spack.compilers.compiler_from_dict(compiler_dict)
+    opt_flags = spack.build_environment.optimization_flags(compiler, target)
+    assert opt_flags == expected_flags
+
+
+@pytest.mark.parametrize(
+    "compiler_str,real_version,target_str,expected_flags",
+    [
+        ("gcc@=9.2.0", None, "haswell", "-march=haswell -mtune=haswell"),
+        # Check that custom string versions are accepted
+        ("gcc@=10foo", "9.2.0", "icelake", "-march=icelake-client -mtune=icelake-client"),
+        # Check that we run version detection (4.4.0 doesn't support icelake)
+        ("gcc@=4.4.0-special", "9.2.0", "icelake", "-march=icelake-client -mtune=icelake-client"),
+        # Check that the special case for Apple's clang is treated correctly
+        # i.e. it won't try to detect the version again
+        ("apple-clang@=9.1.0", None, "x86_64", "-march=x86-64"),
+    ],
+)
+def test_optimization_flags_with_custom_versions(
+    compiler_str,
+    real_version,
+    target_str,
+    expected_flags,
+    monkeypatch,
+    mutable_config,
+    compiler_factory,
+):
+    target = archspec.cpu.TARGETS[target_str]
+    compiler_dict = compiler_factory(spec=compiler_str, operating_system="redhat6")
+    mutable_config.set("compilers", [compiler_dict])
+    if real_version:
+        monkeypatch.setattr(spack.compiler.Compiler, "get_real_version", lambda x: real_version)
+    compiler = spack.compilers.compiler_from_dict(compiler_dict["compiler"])
+
+    opt_flags = spack.build_environment.optimization_flags(compiler, target)
+    assert opt_flags == expected_flags
diff --git a/lib/spack/spack/test/config.py b/lib/spack/spack/test/config.py
index ddf21ffb05..be727f1c56 100644
--- a/lib/spack/spack/test/config.py
+++ b/lib/spack/spack/test/config.py
@@ -407,7 +407,7 @@ def test_substitute_config_variables(mock_low_high_config, monkeypatch):
     ) == os.path.abspath(os.path.join("foo", "test", "bar"))
 
     host_target = spack.platforms.host().target("default_target")
-    host_target_family = str(host_target.microarchitecture.family)
+    host_target_family = str(host_target.family)
     assert spack_path.canonicalize_path(
         os.path.join("foo", "$target_family", "bar")
     ) == os.path.abspath(os.path.join("foo", host_target_family, "bar"))
diff --git a/lib/spack/spack/util/path.py b/lib/spack/spack/util/path.py
index e0fa29dbcc..a91972c79b 100644
--- a/lib/spack/spack/util/path.py
+++ b/lib/spack/spack/util/path.py
@@ -71,7 +71,7 @@ def replacements():
         "operating_system": lambda: arch.os,
         "os": lambda: arch.os,
         "target": lambda: arch.target,
-        "target_family": lambda: arch.target.microarchitecture.family,
+        "target_family": lambda: arch.target.family,
         "date": lambda: date.today().strftime("%Y-%m-%d"),
         "env": lambda: ev.active_environment().path if ev.active_environment() else NOMATCH,
     }
diff --git a/var/spack/repos/builtin/packages/amdfftw/package.py b/var/spack/repos/builtin/packages/amdfftw/package.py
index 6ac400a013..f508c4162b 100644
--- a/var/spack/repos/builtin/packages/amdfftw/package.py
+++ b/var/spack/repos/builtin/packages/amdfftw/package.py
@@ -7,6 +7,7 @@ import os
 
 from llnl.util import tty
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 from spack.pkg.builtin.fftw import FftwBase
 
@@ -213,10 +214,7 @@ class Amdfftw(FftwBase):
         # variable to set AMD_ARCH configure option.
         # Spack user can not directly use AMD_ARCH for this purpose but should
         # use target variable to set appropriate -march option in AMD_ARCH.
-        arch = spec.architecture
-        options.append(
-            "AMD_ARCH={0}".format(arch.target.optimization_flags(spec.compiler).split("=")[-1])
-        )
+        options.append(f"AMD_ARCH={optimization_flags(self.compiler, spec.target)}")
 
         # Specific SIMD support.
         # float and double precisions are supported
diff --git a/var/spack/repos/builtin/packages/arbor/package.py b/var/spack/repos/builtin/packages/arbor/package.py
index 832078c7bd..415914b7ba 100644
--- a/var/spack/repos/builtin/packages/arbor/package.py
+++ b/var/spack/repos/builtin/packages/arbor/package.py
@@ -2,7 +2,7 @@
 # Spack Project Developers. See the top-level COPYRIGHT file for details.
 #
 # SPDX-License-Identifier: (Apache-2.0 OR MIT)
-
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 
@@ -110,21 +110,23 @@ class Arbor(CMakePackage, CudaPackage):
         return ["all", "html"] if "+doc" in self.spec else ["all"]
 
     def cmake_args(self):
+        spec = self.spec
         args = [
             self.define_from_variant("ARB_WITH_ASSERTIONS", "assertions"),
             self.define_from_variant("ARB_WITH_MPI", "mpi"),
             self.define_from_variant("ARB_WITH_PYTHON", "python"),
             self.define_from_variant("ARB_VECTORIZE", "vectorize"),
+            self.define("ARB_ARCH", "none"),
+            self.define("ARB_CXX_FLAGS_TARGET", optimization_flags(self.compiler, spec.target)),
         ]
 
         if self.spec.satisfies("+cuda"):
-            args.append("-DARB_GPU=cuda")
-            args.append(self.define_from_variant("ARB_USE_GPU_RNG", "gpu_rng"))
-
-        # query spack for the architecture-specific compiler flags set by its wrapper
-        args.append("-DARB_ARCH=none")
-        opt_flags = self.spec.architecture.target.optimization_flags(self.spec.compiler)
-        args.append("-DARB_CXX_FLAGS_TARGET=" + opt_flags)
+            args.extend(
+                [
+                    self.define("ARB_GPU", "cuda"),
+                    self.define_from_variant("ARB_USE_GPU_RNG", "gpu_rng"),
+                ]
+            )
 
         return args
 
diff --git a/var/spack/repos/builtin/packages/hpcc/package.py b/var/spack/repos/builtin/packages/hpcc/package.py
index 4483426024..e8a9acb6de 100644
--- a/var/spack/repos/builtin/packages/hpcc/package.py
+++ b/var/spack/repos/builtin/packages/hpcc/package.py
@@ -7,6 +7,7 @@ import os
 import platform
 import re
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 
@@ -161,7 +162,7 @@ class Hpcc(MakefilePackage):
         if spec.satisfies("%intel"):
             # with intel-parallel-studio+mpi the '-march' arguments
             # are not passed to icc
-            arch_opt = spec.architecture.target.optimization_flags(spec.compiler)
+            arch_opt = optimization_flags(self.compiler, spec.target)
             self.config["@CCFLAGS@"] = f"-O3 -restrict -ansi-alias -ip {arch_opt}"
             self.config["@CCNOOPT@"] = "-restrict"
         self._write_make_arch(spec, prefix)
diff --git a/var/spack/repos/builtin/packages/lammps/package.py b/var/spack/repos/builtin/packages/lammps/package.py
index 5c1f9c0588..be1697543e 100644
--- a/var/spack/repos/builtin/packages/lammps/package.py
+++ b/var/spack/repos/builtin/packages/lammps/package.py
@@ -5,6 +5,7 @@
 import datetime as dt
 import os
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 
@@ -898,7 +899,7 @@ class Lammps(CMakePackage, CudaPackage, ROCmPackage, PythonExtension):
             args.append(self.define("CMAKE_CXX_FLAGS_RELWITHDEBINFO", cxx_flags))
 
         # Overwrite generic cpu tune option
-        cmake_tune_flags = spec.architecture.target.optimization_flags(spec.compiler)
+        cmake_tune_flags = optimization_flags(self.compiler, spec.target)
         args.append(self.define("CMAKE_TUNE_FLAGS", cmake_tune_flags))
 
         args.append(self.define_from_variant("LAMMPS_SIZES", "lammps_sizes"))
diff --git a/var/spack/repos/builtin/packages/namd/package.py b/var/spack/repos/builtin/packages/namd/package.py
index 2e9fc20f79..47fcb3b03c 100644
--- a/var/spack/repos/builtin/packages/namd/package.py
+++ b/var/spack/repos/builtin/packages/namd/package.py
@@ -9,6 +9,7 @@ import sys
 
 import llnl.util.tty as tty
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 
@@ -175,7 +176,7 @@ class Namd(MakefilePackage, CudaPackage, ROCmPackage):
                 # this options are take from the default provided
                 # configuration files
                 # https://github.com/UIUC-PPL/charm/pull/2778
-                archopt = spec.architecture.target.optimization_flags(spec.compiler)
+                archopt = optimization_flags(self.compiler, spec.target)
 
                 if self.spec.satisfies("^charmpp@:6.10.1"):
                     optims_opts = {
diff --git a/var/spack/repos/builtin/packages/neuron/package.py b/var/spack/repos/builtin/packages/neuron/package.py
index 2887ea9342..2cb16fdd03 100644
--- a/var/spack/repos/builtin/packages/neuron/package.py
+++ b/var/spack/repos/builtin/packages/neuron/package.py
@@ -3,6 +3,7 @@
 #
 # SPDX-License-Identifier: (Apache-2.0 OR MIT)
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 
@@ -149,7 +150,7 @@ class Neuron(CMakePackage):
 
         # add cpu arch specific optimisation flags to CMake so that they are passed
         # to embedded Makefile that neuron has for compiling MOD files
-        compilation_flags = self.spec.architecture.target.optimization_flags(self.spec.compiler)
+        compilation_flags = optimization_flags(self.compiler, self.spec.target)
         args.append(self.define("CMAKE_CXX_FLAGS", compilation_flags))
 
         return args
diff --git a/var/spack/repos/builtin/packages/py-tensorflow/package.py b/var/spack/repos/builtin/packages/py-tensorflow/package.py
index 9935be8b29..dfe4f72e6b 100644
--- a/var/spack/repos/builtin/packages/py-tensorflow/package.py
+++ b/var/spack/repos/builtin/packages/py-tensorflow/package.py
@@ -8,6 +8,7 @@ import os
 import sys
 import tempfile
 
+from spack.build_environment import optimization_flags
 from spack.package import *
 
 rocm_dependencies = [
@@ -656,7 +657,7 @@ class PyTensorflow(Package, CudaPackage, ROCmPackage, PythonExtension):
 
         # Please specify optimization flags to use during compilation when
         # bazel option '--config=opt' is specified
-        env.set("CC_OPT_FLAGS", spec.architecture.target.optimization_flags(spec.compiler))
+        env.set("CC_OPT_FLAGS", optimization_flags(self.compiler, spec.target))
 
         # Would you like to interactively configure ./WORKSPACE for
         # Android builds?
diff --git a/var/spack/repos/builtin/packages/wgl/package.py b/var/spack/repos/builtin/packages/wgl/package.py
index b420cdcf64..9d02f3c139 100644
--- a/var/spack/repos/builtin/packages/wgl/package.py
+++ b/var/spack/repos/builtin/packages/wgl/package.py
@@ -83,7 +83,7 @@ class Wgl(Package):
         return variants
 
     def _spec_arch_to_sdk_arch(self):
-        spec_arch = str(self.spec.architecture.target.microarchitecture.family).lower()
+        spec_arch = str(self.spec.architecture.target.family).lower()
         _64bit = "64" in spec_arch
         arm = "arm" in spec_arch
         if arm: