Bump LLVM to v21 by giordano · Pull Request #59950 · JuliaLang/julia

giordano · 2025-10-24T22:14:34Z

I won't be able to work on this besides for rebasing the PR and other small edits, I'm opening this PR to keep the ball rolling, but everyone is more than welcome to take it over.

This needs #59946 to start with.

giordano · 2025-10-28T00:57:40Z

Oh, cool, with the changes in #59946 temporarily incorporated here, for the platforms where compilation is successful (macos and freebsd) tests are already passing, that's encouraging.

yuyichao · 2025-10-28T03:45:59Z

The compile error on linux seems to be a Werror for uninitialized variable in LLVM. It could easily be a false positive since it's in the part of the code where LLVM does it's trick with memory layout of objects...

giordano · 2026-03-15T20:54:25Z

I think

[buildroot]/usr/include/llvm/ADT/DenseMap.h: In function 'bool removeAddrspaces(llvm::Module&, AddrspaceRemapFunction)':
[buildroot]/usr/include/llvm/ADT/DenseMap.h:1207:18: error: '*((void*)(& VMap)+72).llvm::SmallDenseMap<std::pair<llvm::Metadata*, long unsigned int>, long unsigned int>::LargeRep::Buckets' may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1207 |     return Small ? getInlineBuckets() : getLargeRep()->Buckets;
      |            ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[buildroot]/usr/include/llvm/ADT/DenseMap.h:1216:18: error: '*((void*)(& VMap)+72).llvm::SmallDenseMap<std::pair<llvm::Metadata*, long unsigned int>, long unsigned int>::LargeRep::NumBuckets' may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1216 |     return Small ? InlineBuckets : getLargeRep()->NumBuckets;
      |            ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

was fixed by llvm/llvm-project#177168, so we'll have to backport that PR to our fork (we want to build a newer patch version with other patches anyway)

giordano · 2026-03-15T20:56:13Z

Windows build fails with

JIT session error: could not register eh-frame: __register_frame function not found
Internal error: Lookup failed: Failed to materialize symbols: { (JuliaOJIT, { japi1_top-level scope_0#0, jlplt_ijl_toplevel_eval_in_2_got, ccall_ijl_toplevel_eval_in_1 }) }
[8908] signal 22: SIGABRT
in expression starting at boot.jl:283
crt_sig_handler at C:/workdir/src\signals-win.c:101
raise at C:\Windows\System32\msvcrt.dll (unknown line)
abort at C:\Windows\System32\msvcrt.dll (unknown line)
operator() at C:/workdir/src\jitlayers.cpp:2048 [inlined]
CallImpl<JuliaOJIT::publishCIs(llvm::ArrayRef<_jl_code_instance_t*>, bool)::<lambda(llvm::Expected<llvm::DenseMap<llvm::orc::SymbolStringPtr, llvm::orc::ExecutorSymbolDef> >)> > at C:/workdir/usr/include/llvm/ADT\FunctionExtras.h:222
_ZN4llvm3orc23AsynchronousSymbolQuery12handleFailedENS_5ErrorE at C:\workdir\usr\bin\libLLVM-21jl.dll (unknown line)
_ZN4llvm3orc16ExecutionSession15OL_notifyFailedERNS0_29MaterializationResponsibilityE at C:\workdir\usr\bin\libLLVM-21jl.dll (unknown line)
_ZN4llvm3orc21LinkGraphLinkingLayer10JITLinkCtx12notifyFailedENS_5ErrorE at C:\workdir\usr\bin\libLLVM-21jl.dll (unknown line)
_ZN4llvm6detail18UniqueFunctionBaseIvJNS_8ExpectedINS_7jitlink20JITLinkMemoryManager14FinalizedAllocEEEEE8CallImplIZNS3_13JITLinkerBase10linkPhase3ESt10unique_ptrIS9_St14default_deleteIS9_EENS2_INS_8DenseMapINS_3orc15SymbolStringPtrENSF_17ExecutorSymbolDefENS_12DenseMapInfoISG_vEENS0_12DenseMapPairISG_SH_EEEEEEE3$_0EEvPvRS6_ at C:\workdir\usr\bin\libLLVM-21jl.dll (unknown line)
operator() at C:/workdir/usr/include/llvm/ADT\FunctionExtras.h:389 [inlined]

I'll need help for that one.

giordano · 2026-03-16T13:22:02Z

32-bit Windows somehow builds successfully, but then there's an interesting test failure

Error in testset ccall:
Test Failed at C:\buildkite-agent\builds\win2k22-amdci6-4\julialang\julia-master\julia-722c9d9693\share\julia\test\ccall.jl:669
  Expression: x == a + b * 1
   Evaluated: -78957614000159160166582917303 == 168812324694683145004459105838176403456

julia/test/ccall.jl

Lines 664 to 669 in 3360c8c

    
           a = Int128(0x7f00123456789abc)<<64 + typemax(UInt64) 
        
           b = Int64(1) 
        
           x = ccall((:test_128, libccalltest), Int128, (Int128, Int64), a, b) 
        
           @test x == a + b*1

is that an ABI problem with Int128 on 32-bit systems?

gbaraldi · 2026-03-16T13:54:33Z

@xal-0 JIT session error: could not register eh-frame: __register_frame function not found
Any clue?

giordano · 2026-03-17T01:19:13Z

After Sam's fix to the 64-bit Windows build, its tests are successful 🥳 Ok, I guess now it's time to start a new rebuild of LLVM in Yggdrasil

giordano · 2026-03-23T11:33:28Z

/cache/build/tester-amdci4-10/julialang/julia-master/src/clangsa/GCChecker.cpp:1053:73: error: no matching function for call to 'clang::ento::SValBuilder::conjureSymbolVal(const clang::Expr*, const clang::LocationContext*, clang::QualType&, unsigned int)'
 1053 |         Call.getOriginExpr(), C.getLocationContext(), QT, C.blockCount());
      |

https://buildkite.com/julialang/julia-master/builds/55880#019d1a76-6b59-47af-badc-653ea1406ce2/L1440

giordano · 2026-03-23T11:49:25Z

We're still getting

[buildroot]/usr/include/llvm/ADT/DenseMap.h:1207:18: error: '*((void*)(& VMap)+72).llvm::SmallDenseMap<std::pair<llvm::Metadata*, long unsigned int>, long unsigned int>::LargeRep::Buckets' may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1207 |     return Small ? getInlineBuckets() : getLargeRep()->Buckets;
      |            ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~~

despite backporting llvm/llvm-project#177168, which puzzles me quite a bit. But also, I can't reproduce the warning/error with GCC 15. Edit: I can't reproduce the error with -Werror even with GCC 9 inside BinaryBuilder, so I have no clue where this error is coming from here.

giordano · 2026-03-23T21:42:37Z

I'm temporarily (?) disabling the maybe uninitialized warning for the sake of moving forward because I don't know what to do with it (I can't even reproduce it with gcc from 9 to 15). All 64-bit platforms are looking good, both 32-bit platforms are sad about ccalls involving Int128 (incorrect result on windows, segfault in Linux). Analyzegc needs adapting to new clang API

giordano · 2026-03-24T00:08:29Z

My pattern-matching-powered fix to the analyzegc build failure seems to have worked in the sense that now there are actual errors reported by the analysis:

/cache/build/builder-amdci5-2/julialang/julia-master/src/gc-stock.c:1566:12: error: Dereference of a fixed address [core.FixedAddressDereference]
 1566 |     return *(uintptr_t*)real_addr;
      |            ^
[...]
/cache/build/builder-amdci5-2/julialang/julia-master/src/gc-stock.c:1566:12: error: Dereference of a fixed address [core.FixedAddressDereference]
 1566 |     return *(uintptr_t*)real_addr;
      |            ^
[...]
/cache/build/builder-amdci5-2/julialang/julia-master/usr/include/llvm/ADT/ilist_node_base.h:55:49: error: Access to field 'Parent' results in a dereference of a fixed address [core.FixedAddressDereference]
   55 |   inline ParentTy *getNodeBaseParent() { return Parent; }
      |                                                 ^
[...]
/cache/build/builder-amdci5-2/julialang/julia-master/src/aotcompile.cpp:344:20: error: Passing non-rooted value as argument to function that may GC [julia.GCChecker]
  344 |             list = jl_idset_put_key(list, val, &idx);
      |                    ^                ~~~~

I'll let someone else have fun fixing them 🙃

Edit: my friendly assistant proposed the following patch

diff --git a/src/Makefile b/src/Makefile
index 2f12ff8d80..11c5df6634 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -642,6 +642,8 @@ SA_EXCEPTIONS-jloptions.c                   := -Xanalyzer -analyzer-config -Xana
  # clang doesn't understand that e->vars has the same value in save_env (NULL) and restore_env (assumed non-NULL)
 SA_EXCEPTIONS-subtype.c                     := -Xanalyzer -analyzer-config -Xanalyzer silence-checkers="core.uninitialized.Assign;core.UndefinedBinaryOperatorResult"
 SA_EXCEPTIONS-codegen.c                     := -Xanalyzer -analyzer-config -Xanalyzer silence-checkers="core"
+SA_EXCEPTIONS-gc-stock.c                    := -Xanalyzer -analyzer-config -Xanalyzer silence-checkers="core.FixedAddressDereference"
+SA_EXCEPTIONS-aotcompile.cpp                := -Xanalyzer -analyzer-config -Xanalyzer silence-checkers="core.FixedAddressDereference"
  # these need to be annotated (and possibly fixed)
 SKIP_GC_CHECK := codegen.cpp rtutils.c
 
diff --git a/src/aotcompile.cpp b/src/aotcompile.cpp
index 1fbfd459fc..96220989fa 100644
--- a/src/aotcompile.cpp
+++ b/src/aotcompile.cpp
@@ -341,7 +341,10 @@ public:
         jl_value_t *rval = jl_idset_get(list, keyset, val);
         if (rval == NULL) {
             ssize_t idx;
+            JL_GC_PROMISE_ROOTED(list);
             list = jl_idset_put_key(list, val, &idx);
+            JL_GC_PROMISE_ROOTED(list);
+            JL_GC_PROMISE_ROOTED(keyset);
             keyset = jl_idset_put_idx(list, keyset, idx);
         }
     }

but I can't really judge it, so I'm not even trying to push these changes.

giordano · 2026-03-31T08:17:57Z

All tests green, yay! Now only analyzegc is unhappy.

giordano · 2026-04-04T13:55:33Z

@nanosoldier runtests()

giordano · 2026-04-07T17:06:23Z

@maleadt What's up with nanosoldier?

maleadt · 2026-04-08T05:53:31Z

Down until further notice.

maleadt · 2026-04-09T19:32:48Z

@nanosoldier runtests()

nanosoldier · 2026-04-10T16:56:14Z

The package evaluation job you requested has completed - possible new issues were detected.
The full report is available.

Report summary

❗ Packages that crashed

5 packages crashed only on the current version.

The process was aborted: 1 packages
An internal error was encountered: 1 packages
A segmentation fault happened: 3 packages

458 packages crashed on the previous version too.

✖ Packages that failed

83 packages failed only on the current version.

Package fails to precompile: 60 packages
Package has test failures: 2 packages
Package tests unexpectedly errored: 2 packages
Tests became inactive: 1 packages
Test duration exceeded the time limit: 17 packages
Test log exceeded the size limit: 1 packages

3904 packages failed on the previous version too.

✔ Packages that passed tests

7 packages passed tests only on the current version.

Other: 7 packages

5797 packages passed tests on the previous version too.

➖ Packages that were skipped altogether

1409 packages were skipped on the previous version too.

giordano · 2026-04-10T18:02:02Z

Of the 83 failing packages, I think

Package fails to precompile: 60 packages

are just packages depending on LLVM.jl which doesn't support LLVM 21 yet?

gbaraldi · 2026-04-20T15:43:02Z

Notably if you increase the size the regression goes away, so I'm not super worried.

oscardssmith · 2026-04-21T08:32:14Z

@giordano can we put a geomean on those graphs? Currenlty it's hard to get a "are the changes overall good or bad"

giordano · 2026-04-21T08:54:31Z

https://tealquaternion.camdvr.org/compare.html?start=dd0802b7462871a7d454af069fa0677c20d2edb0&end=0f2031fd150d453d8e86990e500f26c78c43adfa&stat=min-wall-time shows an overall mean (and also mean and ranges for regressions and improvements, separately), although I suggested @Zentrik showing the median rather than (or in addition to) the mean. The page also gives a "significance factor" to the outliers.

maleadt · 2026-04-22T20:17:54Z

LLVM.jl v9.5.0 supports LLVM 21, so PkgEval should be happier.

giordano · 2026-04-22T20:35:45Z

@nanosoldier runtests()

giordano · 2026-04-23T09:20:59Z

https://buildkite.com/julialang/julia-master/builds/56699#019db6e8-5381-4dc5-a597-1b52a270a056/L2218

/cache/build/builder-amdci5-1/julialang/julia-master/src/debuginfo.cpp:374:49: error: Calling potential safepoint as SimpleFunctionCall from function annotated JL_NOTSAFEPOINT [julia.GCChecker]
  374 |                 codeinst = (jl_code_instance_t*)jl_as_global_root((jl_value_t*)codeinst, 1);
      |                                                 ^

nanosoldier · 2026-04-23T12:26:04Z

The package evaluation job you requested has completed - possible new issues were detected.
The full report is available.

Report summary

❗ Packages that crashed

3 packages crashed only on the current version.

An internal error was encountered: 2 packages
A segmentation fault happened: 1 packages

246 packages crashed on the previous version too.

✖ Packages that failed

21 packages failed only on the current version.

Package fails to precompile: 2 packages
Package has test failures: 1 packages
Package tests unexpectedly errored: 5 packages
Tests became inactive: 1 packages
Test duration exceeded the time limit: 12 packages

1179 packages failed on the previous version too.

✔ Packages that passed tests

8 packages passed tests only on the current version.

Other: 8 packages

5830 packages passed tests on the previous version too.

~ Packages that at least loaded

3539 packages successfully loaded on the previous version too.

➖ Packages that were skipped altogether

4 packages were skipped only on the current version.

Package could not be installed: 4 packages

902 packages were skipped on the previous version too.

giordano · 2026-04-23T15:39:34Z

Summary:

Convex1d: flaky test (self-appreciation for RNG seed being printed to screen, otherwise it'd have been very hard to reproduce)
FMICore: segfault
SIMD: Clang_unified_jll broken
IRStructurizer: LLVM_jll broken
FileCheck: LLVM_jll broken
Clang: libclang not defined in Clang.LibClang
GenericDecMats: timeout
GPUCompiler: LLVM_jll
YFinance: flaky test (429 Too Many Requests errors), should probably be removed from PkgEval
Oxygen: timeout
RepliBuild: libclang not defined in Clang.LibClang
DMRGenie: timeout
SmoQyHankelCorrCleaner: timeout
Tesserae: timeout
Sundials: timeout
FSimBase: timeout
SciBmad: error: during type inference
ModelOrderReduction: timeout
EnsembleKalmanProcesses: timeout
GongBetaAdrenergicSignaling: timeout
PeriodicMatrices: timeout
TensorCategories: timeout

Apart from possibly SciBmad (and not sure it's due to this PR), the other errors are mainly timeouts and some LLVM-related JLLs not working (probably the packages need to allow the LLVM v21 builds). This is looking pretty good, but analyzegc needs to be fixed still.

maleadt · 2026-04-24T07:13:40Z

SIMD: Clang_unified_jll broken

Clang: libclang not defined in Clang.LibClang

Clang.jl also uses Clang_unified_jll, which just needs to be rebuilt for LLVM 21.

IRStructurizer: LLVM_jll broken

FileCheck: LLVM_jll broken

GPUCompiler: LLVM_jll

This breaks because we currently rely on version bounds to select an appropriate version of LLVM_jll.jl, which obviously doesn't work when a single Julia release covers multiple LLVM versions. It probably warrants a LLVM_unified_jll just like Clang_unified_jll (even though these are botches).

EDIT: both fixed.

giordano · 2026-04-26T13:38:09Z

can we put a geomean on those graphs? Currenlty it's hard to get a "are the changes overall good or bad"

https://tealquaternion.camdvr.org/compare.html?start=dd0802b7462871a7d454af069fa0677c20d2edb0&end=0f2031fd150d453d8e86990e500f26c78c43adfa&stat=min-wall-time now shows geometric mean and median for all categories (regressions, improvements, all), and they both point to an overall slight improvement:

Thanks @Zentrik for implementing it!

…EPOINT region JITDebugInfoRegistry::registerJITObject is annotated JL_NOTSAFEPOINT but called jl_as_global_root, which acquires global_roots_lock and may safepoint. Move the promotion into the jl_register_jit_object wrapper, iterating Info.ci_funcs directly so the JL_NOTSAFEPOINT body no longer needs to touch the global roots table. This pull request was written with the assistance of generative AI (Claude Opus 4.7). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

giordano · 2026-05-08T10:13:53Z

All green! Should do/test anything else here (benchmark/pkgeval again?)?

gbaraldi · 2026-05-08T17:52:58Z

I think this is good to merge

IanButterworth · 2026-05-08T18:36:19Z

Was the non-squash merge intentional here?

oscardssmith · 2026-05-08T19:29:46Z

no.

#60339 was auto-closed when #59950 merged and can't be reopened. Co-authored-by: Claude <noreply@anthropic.com>

JuliaLang#60339 was auto-closed when JuliaLang#59950 merged and can't be reopened. Co-authored-by: Claude <noreply@anthropic.com>

giordano added building Build system, or building Julia or its dependencies external dependencies Involves LLVM, OpenBLAS, or other linked libraries stdlib Julia's standard library JLLs compiler:llvm For issues that relate to LLVM labels Oct 24, 2025

yuyichao force-pushed the mg/llvm-21 branch from e056726 to 7d303e3 Compare October 27, 2025 20:01

giordano force-pushed the mg/llvm-21 branch from 7d303e3 to 722c9d9 Compare March 15, 2026 20:45

IanButterworth mentioned this pull request Mar 15, 2026

jitlayers: Use GlobalISel on AArch64 at -O0/-O1 #60339

Closed

giordano force-pushed the mg/llvm-21 branch from fdb3d8b to 3542185 Compare March 23, 2026 11:30

giordano mentioned this pull request Mar 28, 2026

Incorrect uitofp / sitofp results for Int128 (runtime intrinsics) #61419

Closed

adienes mentioned this pull request Apr 6, 2026

stricter shapes for setindex #59025

Open

giordano added the needs pkgeval Tests for all registered packages should be run with this change label Apr 8, 2026

adienes removed needs nanosoldier run This PR should have benchmarks run on it needs pkgeval Tests for all registered packages should be run with this change labels Apr 20, 2026

giordano mentioned this pull request Apr 22, 2026

LLVM: Bump LLVMExtra. JuliaPackaging/Yggdrasil#13554

Merged

Merge branch 'master' into mg/llvm-21

c0c9a57

giordano mentioned this pull request Apr 23, 2026

Flaky "minimize no domain" test jw3126/Convex1d.jl#1

Open

giordano marked this pull request as ready for review April 23, 2026 15:39

This was referenced Apr 24, 2026

Switch to a unified LLVM_utils JLL. JuliaLLVM/FileCheck.jl#3

Merged

LLVM: Bump Clang_unified for LLVM 21. JuliaPackaging/Yggdrasil#13570

Merged

giordano mentioned this pull request Apr 26, 2026

Show median instead of mean JuliaCI/julia-perf#40

Closed

gbaraldi and others added 2 commits May 6, 2026 18:51

Merge branch 'master' into mg/llvm-21

14dda4d

oscardssmith merged commit c053ab3 into master May 8, 2026
9 checks passed

oscardssmith deleted the mg/llvm-21 branch May 8, 2026 17:57

IanButterworth mentioned this pull request May 8, 2026

jitlayers: Use GlobalISel on AArch64 at -O0/-O1 #61758

Merged

IanButterworth added a commit that referenced this pull request May 11, 2026

jitlayers: Use GlobalISel on AArch64 at -O0/-O1 (#61758)

5dee674

#60339 was auto-closed when #59950 merged and can't be reopened. Co-authored-by: Claude <noreply@anthropic.com>

hardikxk pushed a commit to hardikxk/julia that referenced this pull request May 18, 2026

jitlayers: Use GlobalISel on AArch64 at -O0/-O1 (JuliaLang#61758)

9a23a07

JuliaLang#60339 was auto-closed when JuliaLang#59950 merged and can't be reopened. Co-authored-by: Claude <noreply@anthropic.com>

Uh oh!

Conversation

giordano commented Oct 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

giordano commented Oct 28, 2025

Uh oh!

yuyichao commented Oct 28, 2025

Uh oh!

giordano commented Mar 15, 2026

Uh oh!

giordano commented Mar 15, 2026

Uh oh!

giordano commented Mar 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

gbaraldi commented Mar 16, 2026

Uh oh!

giordano commented Mar 17, 2026

Uh oh!

giordano commented Mar 23, 2026

Uh oh!

giordano commented Mar 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

giordano commented Mar 23, 2026

Uh oh!

giordano commented Mar 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

giordano commented Mar 31, 2026

Uh oh!

giordano commented Apr 4, 2026

Uh oh!

giordano commented Apr 7, 2026

Uh oh!

maleadt commented Apr 8, 2026

Uh oh!

maleadt commented Apr 9, 2026

Uh oh!

nanosoldier commented Apr 10, 2026

❗ Packages that crashed

✖ Packages that failed

✔ Packages that passed tests

➖ Packages that were skipped altogether

Uh oh!

giordano commented Apr 10, 2026

Uh oh!

gbaraldi commented Apr 20, 2026

Uh oh!

oscardssmith commented Apr 21, 2026

Uh oh!

giordano commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

maleadt commented Apr 22, 2026

Uh oh!

giordano commented Apr 22, 2026

Uh oh!

giordano commented Apr 23, 2026

Uh oh!

nanosoldier commented Apr 23, 2026

❗ Packages that crashed

✖ Packages that failed

✔ Packages that passed tests

~ Packages that at least loaded

➖ Packages that were skipped altogether

Uh oh!

giordano commented Apr 23, 2026

Uh oh!

maleadt commented Apr 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

giordano commented Apr 26, 2026

Uh oh!

giordano commented May 8, 2026

Uh oh!

gbaraldi commented May 8, 2026

Uh oh!

Uh oh!

giordano commented Oct 24, 2025 •

edited

Loading

giordano commented Mar 16, 2026 •

edited

Loading

giordano commented Mar 23, 2026 •

edited

Loading

giordano commented Mar 24, 2026 •

edited

Loading

giordano commented Apr 21, 2026 •

edited

Loading

maleadt commented Apr 24, 2026 •

edited

Loading