flake: update llvmlite and llvm

artiq/compiler/targets.py

@@ -2,7 +2,6 @@
 from artiq.compiler import types, ir
 from llvmlite import ir as ll, binding as llvm
 
-llvm.initialize()
 llvm.initialize_all_targets()
 llvm.initialize_all_asmprinters()
 
@@ -108,35 +107,19 @@ def target_machine(self):
         return llmachine
 
     def optimize(self, llmodule):
-        llpassmgr = llvm.create_module_pass_manager()
-
-        # Register our alias analysis passes.
-        llpassmgr.add_basic_alias_analysis_pass()
-        llpassmgr.add_type_based_alias_analysis_pass()
-
-        # Start by cleaning up after our codegen and exposing as much
-        # information to LLVM as possible.
-        llpassmgr.add_constant_merge_pass()
-        llpassmgr.add_cfg_simplification_pass()
-        llpassmgr.add_instruction_combining_pass()
-        llpassmgr.add_sroa_pass()
-        llpassmgr.add_dead_code_elimination_pass()
-        llpassmgr.add_function_attrs_pass()
-        llpassmgr.add_global_optimizer_pass()
-
-        # Now, actually optimize the code.
-        llpassmgr.add_function_inlining_pass(275)
-        llpassmgr.add_ipsccp_pass()
-        llpassmgr.add_instruction_combining_pass()
-        llpassmgr.add_gvn_pass()
-        llpassmgr.add_cfg_simplification_pass()
-        llpassmgr.add_licm_pass()
-
-        # Clean up after optimizing.
-        llpassmgr.add_dead_arg_elimination_pass()
-        llpassmgr.add_global_dce_pass()
-
-        llpassmgr.run(llmodule)
+        llmachine = self.target_machine()
+        pto = llvm.create_pipeline_tuning_options(size_level=1)
+
+        # This combination of levels gets mapped to the -Os PassManagerBuilder default in
+        # llvmlite 0.45. The mapping is slightly nonsensical in that -Oz (which is more
+        # aggressively optimizing for size at the expense of runtime) is accessed via
+        # speed_level=size_level=2.
+        pto.speed_level = 1
+
+        pb = llvm.create_pass_builder(llmachine, pto)
+        llpassmgr = pb.getModulePassManager()
+
+        llpassmgr.run(llmodule, pb)
 
     def compile(self, module):
         """Compile the module to a relocatable object for this target."""

artiq/firmware/ksupport/ksupport.ld

@@ -71,5 +71,6 @@ SECTIONS
     /DISCARD/ :
     {
         *(.debug*)
+        *(.riscv.attributes)
     }
 }

artiq/test/lit/embedding/annotation.py

@@ -4,7 +4,7 @@
 from artiq.language.core import *
 from artiq.language.types import *
 
-# CHECK: i64 @_Z13testbench.foozz\(i64 %ARG.x, \{ i1, i64 \} %ARG.y\)
+# CHECK-L: i64 @_Z13testbench.foozz(i64
 
 @kernel
 def foo(x: TInt64, y: TInt64 = 1) -> TInt64:

artiq/test/lit/embedding/annotation_py.py

@@ -8,19 +8,19 @@
 from artiq.language.core import *
 from artiq.language.types import *
 
-# CHECK-L: i64 @_Z13testbench.foozz(i64 %ARG.x, { i1, i32 } %ARG.y)
+# CHECK-L: i64 @_Z13testbench.foozz(i64
 
 @kernel
 def foo(x: np.int64, y: np.int32 = 1) -> np.int64:
     print(x + y)
     return x + y
 
-# CHECK-L: void @_Z13testbench.barzz()
+# CHECK-L: void @_Z13testbench.barzz(i32
 @kernel
 def bar(x: np.int32) -> None:
     print(x)
 
-# CHECK-L: @_Z21testbench.unpack_listzz({ i1, i64 }* nocapture writeonly sret({ i1, i64 }) %.1, { i64*, i32 }* %ARG.xs)
+# CHECK-L: @_Z21testbench.unpack_listzz(ptr
 @kernel
 def unpack_list(xs: List[np.int64]) -> Tuple[bool, np.int64]:
     print(xs)

artiq/test/lit/embedding/index_tuple.py

@@ -4,7 +4,7 @@
 from artiq.language.core import *
 from artiq.language.types import *
 
-# CHECK-L: void @_Z16testbench.unpackzz({ i32, { i8*, i32 }, i1 } %ARG.x)
+# CHECK-L: void @_Z16testbench.unpackzz({ i32, { ptr, i32 }, i1 } %ARG.x)
 
 @kernel
 def unpack(x):

artiq/test/lit/embedding/invariant_propagation.py

@@ -15,8 +15,8 @@ def run(self):
         if self.foo:
             print("bar")
         else:
-            # Make sure all the code for this branch will be completely elided:
-            # CHECK-NOT: baz
+            # The kernel_invariant annotation should generate !invariant.load metadata
+            # CHECK: invariant.load
             print("baz")
 
 obj = Class()

artiq/test/lit/embedding/subkernel_message_recv.py

@@ -9,13 +9,13 @@ def entrypoint():
     # CHECK: call void @subkernel_load_run\(i32 1, i8 1, i1 true\), !dbg !.
     message_pass()
     # CHECK-NOT: call void @subkernel_send_message\(i32 ., i1 false, i8 1, i8 1, .*\), !dbg !.
-    # CHECK: call i8 @subkernel_await_message\(i32 2, i64 -1, { i8\*, i32 }\* nonnull .*, i8 1, i8 1\), !dbg !.
+    # CHECK: call i8 @subkernel_await_message\(i32 2, i64 -1, ptr nonnull .*, i8 1, i8 1\), !dbg !.
     subkernel_recv("message", TInt32)
 
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
 # CHECK-NOT-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
-# CHECK-L: declare i8 @subkernel_await_message(i32, i64, { i8*, i32 }*, i8, i8) local_unnamed_addr
+# CHECK-L: declare i8 @subkernel_await_message(i32, i64, ptr, i8, i8) local_unnamed_addr
 # CHECK-NOT-L: declare void @subkernel_await_finish(i32, i64) local_unnamed_addr
 @subkernel(destination=1)
 def message_pass() -> TNone:

artiq/test/lit/embedding/subkernel_message_send.py

@@ -14,7 +14,7 @@ def entrypoint():
 
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
-# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
+# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, ptr, ptr) local_unnamed_addr
 @subkernel(destination=1)
 def message_pass() -> TNone:
     subkernel_recv("message", TInt32)

artiq/test/lit/embedding/subkernel_return.py

@@ -9,13 +9,13 @@ def entrypoint():
     # CHECK: call void @subkernel_load_run\(i32 1, i8 1, i1 true\), !dbg !.
     # CHECK-NOT: call void @subkernel_send_message\(.*\), !dbg !.
     returning()
-    # CHECK: call i8 @subkernel_await_message\(i32 1, i64 -1, { i8\*, i32 }\* nonnull .*, i8 1, i8 1\), !dbg !.
+    # CHECK: call i8 @subkernel_await_message\(i32 1, i64 -1, ptr nonnull .*, i8 1, i8 1\), !dbg !.
     # CHECK: call void @subkernel_await_finish\(i32 1, i64 -1\), !dbg !.
     subkernel_await(returning)
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
 # CHECK-NOT-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
-# CHECK-L: declare i8 @subkernel_await_message(i32, i64, { i8*, i32 }*, i8, i8) local_unnamed_addr
+# CHECK-L: declare i8 @subkernel_await_message(i32, i64, ptr, i8, i8) local_unnamed_addr
 # CHECK-L: declare void @subkernel_await_finish(i32, i64) local_unnamed_addr
 @subkernel(destination=1)
 def returning() -> TInt32:

artiq/test/lit/embedding/subkernel_self_args.py

@@ -22,4 +22,4 @@ def entrypoint():
     a.kernel_entrypoint()
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
-# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
+# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, ptr, ptr) local_unnamed_addr

artiq/test/lit/embedding/subkernel_with_arg.py

@@ -12,7 +12,7 @@ def entrypoint():
 
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
-# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
+# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, ptr, ptr) local_unnamed_addr
 @subkernel(destination=1)
 def accept_arg(arg: TInt32) -> TNone:
     pass

artiq/test/lit/embedding/subkernel_with_opt_arg.py

@@ -15,7 +15,7 @@ def entrypoint():
 
 
 # CHECK-L: declare void @subkernel_load_run(i32, i8, i1) local_unnamed_addr
-# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, { i8*, i32 }*, i8**) local_unnamed_addr
+# CHECK-L: declare void @subkernel_send_message(i32, i1, i8, i8, ptr, ptr) local_unnamed_addr
 @subkernel(destination=1)
 def accept_arg(arg_a, arg_b=5) -> TNone:
     pass

artiq/test/lit/embedding/syscall_arg_attrs.py

@@ -11,20 +11,20 @@
 
 @kernel
 def entrypoint():
-    # CHECK: call void @accept_str\({ i8\*, i32 }
+    # CHECK: call void @accept_str\({ ptr, i32 }
     accept_str("foo")
 
-    # CHECK: call void @return_str\({ i8\*, i32 }\* nonnull sret
+    # CHECK: call void @return_str\(ptr nonnull sret
     return_str()
 
 
-# CHECK: declare void @accept_str\({ i8\*, i32 }\)
+# CHECK: declare void @accept_str\({ ptr, i32 }\)
 @syscall
 def accept_str(name: TStr) -> TNone:
     pass
 
 
-# CHECK: declare void @return_str\({ i8\*, i32 }\* sret\({ i8\*, i32 }\)\)
+# CHECK: declare void @return_str\(ptr sret\({ ptr, i32 }\)\)
 @syscall
 def return_str() -> TStr:
     pass

artiq/test/lit/embedding/syscall_flags.py

@@ -6,7 +6,7 @@
 
 # CHECK: call void @foo\(\)(, !dbg !\d+)?
 
-# CHECK-L: ; Function Attrs: inaccessiblememonly nounwind
+# CHECK-L: ; Function Attrs: nounwind memory(inaccessiblemem: readwrite)
 # CHECK-NEXT-L: declare void @foo()
 
 @syscall(flags={"nounwind", "nowrite"})
@@ -15,7 +15,7 @@ def foo() -> TNone:
 
 # sret nowrite functions shouldn't be marked inaccessiblememonly.
 # CHECK-L: ; Function Attrs: nounwind
-# CHECK-NEXT-L: declare void @bar({ i32, i64 }* sret({ i32, i64 }))
+# CHECK-NEXT-L: declare void @bar(ptr sret({ i32, i64 }))
 @syscall(flags={"nounwind", "nowrite"})
 def bar() -> TTuple([TInt32, TInt64]):
     pass

flake.nix

@@ -74,8 +74,6 @@
       QML2_IMPORT_PATH = "${qtbase}/${qtQmlPrefix}";
     };
 
-    inherit (pkgs.callPackage ./llvm {}) llvm_15 lld_15 llvmPackages_15;
-
     rust = pkgs.rust-bin.nightly."2021-09-01".default.override {
       extensions = ["rust-src"];
       targets = [];
@@ -158,30 +156,6 @@
       '';
     };
 
-    llvmlite-new = pkgs.python3Packages.buildPythonPackage rec {
-      pname = "llvmlite";
-      version = "0.44.0";
-      src = pkgs.fetchFromGitHub {
-        owner = "numba";
-        repo = "llvmlite";
-        rev = "v${version}";
-        sha256 = "sha256-ZIA/JfK9ZP00Zn6SZuPus30Xw10hn3DArHCkzBZAUV0=";
-      };
-      pyproject = true;
-      build-system = [pkgs.python3Packages.setuptools];
-      nativeBuildInputs = [llvm_15];
-      # Disable static linking
-      # https://github.com/numba/llvmlite/issues/93
-      postPatch = ''
-        substituteInPlace ffi/Makefile.linux --replace "-static-libstdc++" ""
-        substituteInPlace llvmlite/tests/test_binding.py --replace "test_linux" "nope"
-      '';
-      # Set directory containing llvm-config binary
-      preConfigure = ''
-        export LLVM_CONFIG=${llvm_15.dev}/bin/llvm-config
-      '';
-    };
-
     artiq-upstream = pkgs.python3Packages.buildPythonPackage rec {
       pname = "artiq";
       version = artiqVersion;
@@ -195,10 +169,9 @@
       '';
 
       nativeBuildInputs = [pkgs.qt6.wrapQtAppsHook];
-      # keep llvm_x and lld_x in sync with llvmlite
       propagatedBuildInputs =
-        [llvm_15 lld_15 sipyco.packages.x86_64-linux.sipyco pythonparser llvmlite-new pkgs.qt6.qtsvg artiq-comtools.packages.x86_64-linux.artiq-comtools]
-        ++ (with pkgs.python3Packages; [pyqtgraph pygit2 numpy dateutil scipy prettytable pyserial levenshtein h5py pyqt6 qasync tqdm lmdb jsonschema platformdirs]);
+        [pkgs.llvm_20 pkgs.lld_20 sipyco.packages.x86_64-linux.sipyco pythonparser pkgs.qt6.qtsvg artiq-comtools.packages.x86_64-linux.artiq-comtools]
+        ++ (with pkgs.python3Packages; [llvmlite pyqtgraph pygit2 numpy dateutil scipy prettytable pyserial levenshtein h5py pyqt6 qasync tqdm lmdb jsonschema platformdirs]);
 
       dontWrapQtApps = true;
       postFixup = ''
@@ -219,10 +192,10 @@
         "--set FONTCONFIG_FILE ${pkgs.fontconfig.out}/etc/fonts/fonts.conf"
       ];
 
-      # FIXME: automatically propagate lld_15 llvm_15 dependencies
+      # FIXME: automatically propagate lld_20 llvm_20 dependencies
       # cacert is required in the check stage only, as certificates are to be
       # obtained from system elsewhere
-      nativeCheckInputs = [lld_15 llvm_15 libartiq-support pkgs.lit pkgs.outputcheck pkgs.cacert];
+      nativeCheckInputs = with pkgs; [lld_20 llvm_20 lit outputcheck cacert] ++ [libartiq-support];
       checkPhase = ''
         python -m unittest discover -v artiq.test
 
@@ -308,9 +281,9 @@
         nativeBuildInputs = [
           (pkgs.python3.withPackages (ps: [migen misoc (artiq.withExperimentalFeatures experimentalFeatures) ps.packaging]))
           rust
-          llvm_15
-          lld_15
-          llvmPackages_15.clang-unwrapped
+          pkgs.llvm_20
+          pkgs.lld_20
+          pkgs.llvmPackages_20.clang-unwrapped
           vivado
         ];
         overrideMain = _: {
@@ -498,6 +471,9 @@
           [
             git
             lit
+            lld_20
+            llvm_20
+            llvmPackages_20.clang-unwrapped
             outputcheck
             pdf2svg
 
@@ -510,9 +486,6 @@
             (python3.withPackages (ps: [migen misoc microscope ps.packaging ps.paramiko] ++ artiq.propagatedBuildInputs))
           ]
           ++ [
-            llvm_15
-            lld_15
-            llvmPackages_15.clang-unwrapped
             rust
             latex-artiq-manual
             artiq-frontend-dev-wrappers
@@ -538,9 +511,9 @@
         packages = [
           rust
 
-          llvm_15
-          lld_15
-          llvmPackages_15.clang-unwrapped
+          pkgs.llvm_20
+          pkgs.lld_20
+          pkgs.llvmPackages_20.clang-unwrapped
 
           packages.x86_64-linux.vivado
           packages.x86_64-linux.openocd-bscanspi
@@ -584,8 +557,8 @@
                 ]
                 ++ ps.paramiko.optional-dependencies.ed25519
           ))
-          llvm_15
-          lld_15
+          pkgs.llvm_20
+          pkgs.lld_20
           pkgs.openssh
           packages.x86_64-linux.openocd-bscanspi # for the bscanspi bitstreams
         ];