GitLab

Marge Bot pushed to branch wip/marge_bot_batch_merge_job at Glasgow Haskell Compiler / GHC

Commits:

cdb74049

by Cheng Shao at 2026-01-22T14:52:36-05:00

rts: avoid Cmm loop to initialize Array#/SmallArray#

Previously, `newArray#`/`newSmallArray#` called an RTS C function to
allocate the `Array#`/`SmallArray#`, then used a Cmm loop to
initialize the elements. Cmm doesn't have native for-loop so the code
is a bit awkward, and it's less efficient than a C loop, since the C
compiler can effectively vectorize the loop with optimizations.

So this patch moves the loop that initializes the elements to the C
side. `allocateMutArrPtrs`/`allocateSmallMutArrPtrs` now takes a new
`init` argument and initializes the elements if `init` is non-NULL.

4c784f00

by Cheng Shao at 2026-01-22T14:53:19-05:00

Fix testsuite run for +ipe flavour transformer

This patch makes the +ipe flavour transformer pass the entire
testsuite:

- An RTS debug option `-DI` is added, the IPE trace information is now
  only printed with `-DI`. The test cases that do require IPE trace
  are now run with `-DI`.
- The testsuite config option `ghc_with_ipe` is added, enabled when
  running the testsuite with `+ipe`, which skips a few tests that are
  sensitive to eventlog output, allocation patterns etc that can fail
  under `+ipe`.

This is the first step towards #26799.

Co-authored-by: Codex <codex@openai.com>

33c700f6

by Ben Gamari at 2026-01-22T15:26:06-05:00

hadrian: Bump QuickCheck upper bound

This patch bumps QuickCheck upper bound to 2.18. selftest rule
manually tested to work with current latest QuickCheck-2.17.1.0.

b1a60194

by Zubin Duggal at 2026-01-22T15:26:12-05:00

Add genindex to index.rst. This adds a link to the index in the navigation bar.

Fixes #26437

21 changed files:

docs/users_guide/index.rst
docs/users_guide/runtime_control.rst
hadrian/hadrian.cabal
hadrian/src/Flavour.hs
libraries/ghc-compact/tests/all.T
libraries/ghc-internal/tests/backtraces/all.T
rts/AllocArray.c
rts/AllocArray.h
rts/ClosureTable.c
rts/Heap.c
rts/PrimOps.cmm
rts/RtsFlags.c
rts/Threads.c
rts/Trace.c
rts/Weak.c
rts/include/rts/Flags.h
testsuite/driver/testglobals.py
testsuite/driver/testlib.py
testsuite/tests/rts/Makefile
testsuite/tests/rts/all.T
testsuite/tests/rts/ipe/all.T

Changes:

docs/users_guide/index.rst

@@ -29,6 +29,7 @@ Contents:
     eventlog-formats
     glossary
     editing-guide
 +   genindex
  Indices and tables

docs/users_guide/runtime_control.rst

@@ -1588,6 +1588,7 @@ recommended for everyday use!
  .. rts-flag::  -Ds  DEBUG: scheduler
  .. rts-flag::  -Di  DEBUG: interpreter
 +.. rts-flag::  -DI  DEBUG: IPE
  .. rts-flag::  -Dw  DEBUG: weak
  .. rts-flag::  -DG  DEBUG: gccafs
  .. rts-flag::  -Dg  DEBUG: gc

hadrian/hadrian.cabal

@@ -190,4 +190,4 @@ executable hadrian
      if flag(selftest)
        other-modules:   Rules.Selftest
        cpp-options:     -DHADRIAN_ENABLE_SELFTEST
 -      build-depends:   QuickCheck           >= 2.6     && < 2.15
 +      build-depends:   QuickCheck           >= 2.6     && < 2.18

hadrian/src/Flavour.hs

@@ -384,9 +384,15 @@ omitPragmas = addArgs
  -- | Build stage2 dependencies with options to enable IPE debugging
  -- information.
  enableIPE :: Flavour -> Flavour
 -enableIPE = addArgs
 -    $ notStage0 ? builder (Ghc CompileHs)
 -    ? pure ["-finfo-table-map", "-fdistinct-constructor-tables"]
 +enableIPE =
 +  addArgs $
 +    mconcat
 +      [ notStage0
 +          ? builder (Ghc CompileHs)
 +          ? pure
 +            ["-finfo-table-map", "-fdistinct-constructor-tables"],
 +        builder Testsuite ? arg "--config=ghc_with_ipe=True"
 +      ]
  enableLateCCS :: Flavour -> Flavour
  enableLateCCS = addArgs

libraries/ghc-compact/tests/all.T

@@ -20,7 +20,8 @@ test('compact_gc', [fragile_for(17253, ['ghci']), ignore_stdout], compile_and_ru
  # this test computes closure sizes and those are affected
  # by the ghci and prof ways, because of BCOs and profiling headers.
  # Optimization levels slightly change what is/isn't shared so only run in normal mode
 -test('compact_share', only_ways(['normal']), compile_and_run, [''])
 +test('compact_share', [only_ways(['normal']), when(ghc_with_ipe(), skip)], # IPE changes allocation/layout affecting compactSize output.
 +     compile_and_run, [''])
  test('compact_bench', [ ignore_stdout, extra_run_opts('100') ],
                         compile_and_run, [''])
  test('T17044', normal, compile_and_run, [''])

libraries/ghc-internal/tests/backtraces/all.T

@@ -2,5 +2,5 @@ test('T14532a', [], compile_and_run, [''])
  test('T14532b', [], compile_and_run, [''])
  test('T26507', [ when(have_profiling(), extra_ways(['prof']))
                 , when(js_arch(), skip)
 -	       , exit_code(1)], compile_and_run, [''])
+-
 +               , when(ghc_with_ipe(), skip) # IPE builds include an IPE backtrace section on stderr.
 +               , exit_code(1)], compile_and_run, [''])

rts/AllocArray.c

@@ -5,6 +5,7 @@
  StgMutArrPtrs *allocateMutArrPtrs (Capability *cap,
                                     StgWord nelements,
 +                                   StgClosure *init,
                                     CostCentreStack *ccs USED_IF_PROFILING)
+ {
      /* All sizes in words */
@@ -25,6 +26,12 @@ StgMutArrPtrs *allocateMutArrPtrs (Capability *cap,
      arr->ptrs = nelements;
      arr->size = arrsize;
 +    if (init != NULL) {
 +        for (StgWord i = 0; i < nelements; ++i) {
 +            arr->payload[i] = init;
 +        }
 +    }
++
      /* Initialize the card array. Note that memset needs sizes in bytes. */
      memset(&(arr->payload[nelements]), 0, mutArrPtrsCards(nelements));
@@ -33,6 +40,7 @@ StgMutArrPtrs *allocateMutArrPtrs (Capability *cap,
  StgSmallMutArrPtrs *allocateSmallMutArrPtrs (Capability *cap,
                                               StgWord nelements,
 +                                             StgClosure *init,
                                               CostCentreStack *ccs
                                                 USED_IF_PROFILING)
+ {
@@ -47,6 +55,13 @@ StgSmallMutArrPtrs *allocateSmallMutArrPtrs (Capability *cap,
      /* No write barrier needed since this is a new allocation. */
      SET_HDR(arr, &stg_SMALL_MUT_ARR_PTRS_DIRTY_info, ccs);
      arr->ptrs = nelements;
++
 +    if (init != NULL) {
 +        for (StgWord i = 0; i < nelements; ++i) {
 +            arr->payload[i] = init;
 +        }
 +    }
++
      return arr;
+ }

rts/AllocArray.h

@@ -21,16 +21,19 @@
   */
  /* Allocate a StgMutArrPtrs for a given number of elements. It is allocated in
 - * the DIRTY state.
 + * the DIRTY state. If init is non-NULL, initialize payload elements to init.
   */
  StgMutArrPtrs *allocateMutArrPtrs (Capability *cap,
                                     StgWord nelements,
 +                                   StgClosure *init,
                                     CostCentreStack *ccs);
 -/* Allocate a StgSmallMutArrPtrs for a given number of elements.
 +/* Allocate a StgSmallMutArrPtrs for a given number of elements. If init is
 + * non-NULL, initialize payload elements to init.
   */
  StgSmallMutArrPtrs *allocateSmallMutArrPtrs (Capability *cap,
                                               StgWord nelements,
 +                                             StgClosure *init,
                                               CostCentreStack *ccs);
  /* Allocate a StgArrBytes for a given number of bytes.

rts/ClosureTable.c

@@ -46,7 +46,7 @@ bool enlargeClosureTable(Capability *cap, ClosureTable *t, int newcapacity)
      ASSERT(newcapacity > oldcapacity);
      StgMutArrPtrs *newarr;
 -    newarr = allocateMutArrPtrs(cap, newcapacity, CCS_SYSTEM_OR_NULL);
 +    newarr = allocateMutArrPtrs(cap, newcapacity, NULL, CCS_SYSTEM_OR_NULL);
      if (RTS_UNLIKELY(newarr == NULL)) return false;
      StgArrBytes *newfree;
@@ -276,4 +276,3 @@ static bool isCompactClosureTable(ClosureTable *t)
+     }
      return isCompact;
+ }
+-

rts/Heap.c

@@ -279,7 +279,7 @@ StgMutArrPtrs *heap_view_closurePtrs(Capability *cap, StgClosure *closure) {
      StgClosure **ptrs = (StgClosure **) stgMallocBytes(sizeof(StgClosure *) * size, "heap_view_closurePtrs");
      StgWord nptrs = collect_pointers(closure, ptrs);
 -    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, nptrs, cap->r.rCCCS);
 +    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, nptrs, NULL, cap->r.rCCCS);
      if (RTS_UNLIKELY(arr == NULL)) goto end;
      SET_INFO((StgClosure *) arr, &stg_MUT_ARR_PTRS_FROZEN_CLEAN_info);

rts/PrimOps.cmm

@@ -386,24 +386,11 @@ stg_newArrayzh ( W_ n /* words */, gcptr init )
      again: MAYBE_GC(again);
 -    ("ptr" arr) = ccall allocateMutArrPtrs(MyCapability() "ptr", n, CCCS);
 +    ("ptr" arr) = ccall allocateMutArrPtrs(MyCapability() "ptr", n, init "ptr", CCCS);
      if (arr == NULL) (likely: False) {
          jump stg_raisezh(HsIface_heapOverflow_closure(W_[ghc_hs_iface]));
+     }
 -    // Initialise all elements of the array with the value init
 -    W_ p;
 -    p = arr + SIZEOF_StgMutArrPtrs;
 -    // Avoid the shift for `WDS(n)` in the inner loop
 -    W_ limit;
 -    limit = arr + SIZEOF_StgMutArrPtrs + WDS(n);
 -  for:
 -    if (p < limit) (likely: True) {
 -        W_[p] = init;
 -        p = p + WDS(1);
 -        goto for;
 -    }
+-
      return (arr);
+ }
@@ -496,24 +483,11 @@ stg_newSmallArrayzh ( W_ n /* words */, gcptr init )
      again: MAYBE_GC(again);
 -    ("ptr" arr) = ccall allocateSmallMutArrPtrs(MyCapability() "ptr", n, CCCS);
 +    ("ptr" arr) = ccall allocateSmallMutArrPtrs(MyCapability() "ptr", n, init "ptr", CCCS);
      if (arr == NULL) (likely: False) {
          jump stg_raisezh(HsIface_heapOverflow_closure(W_[ghc_hs_iface]));
+     }
 -    // Initialise all elements of the array with the value init
 -    W_ p;
 -    p = arr + SIZEOF_StgSmallMutArrPtrs;
 -    // Avoid the shift for `WDS(n)` in the inner loop
 -    W_ limit;
 -    limit = arr + SIZEOF_StgSmallMutArrPtrs + WDS(n);
 -  for:
 -    if (p < limit) (likely: True) {
 -        W_[p] = init;
 -        p = p + WDS(1);
 -        goto for;
 -    }
+-
      return (arr);
+ }

rts/RtsFlags.c

@@ -209,6 +209,8 @@ void initRtsFlagsDefaults(void)
      RtsFlags.DebugFlags.numa            = false;
      RtsFlags.DebugFlags.compact         = false;
      RtsFlags.DebugFlags.continuation    = false;
 +    RtsFlags.DebugFlags.iomanager       = false;
 +    RtsFlags.DebugFlags.ipe             = false;
  #if defined(PROFILING)
      RtsFlags.CcFlags.doCostCentres      = COST_CENTRES_NONE;
@@ -482,6 +484,7 @@ usage_text[] = {
  #if defined(DEBUG)
  "  -Ds  DEBUG: scheduler",
  "  -Di  DEBUG: interpreter",
 +"  -DI  DEBUG: IPE",
  "  -Dw  DEBUG: weak",
  "  -DG  DEBUG: gccafs",
  "  -Dg  DEBUG: gc",
@@ -2311,6 +2314,9 @@ static void read_debug_flags(const char* arg)
          case 'o':
              RtsFlags.DebugFlags.iomanager = true;
              break;
 +        case 'I':
 +            RtsFlags.DebugFlags.ipe = true;
 +            break;
          default:
              bad_option( arg );
+         }

rts/Threads.c

@@ -894,7 +894,7 @@ StgMutArrPtrs *listThreads(Capability *cap)
+     }
      // Allocate a suitably-sized array...
 -    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, n_threads, cap->r.rCCCS);
 +    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, n_threads, NULL, cap->r.rCCCS);
      if (RTS_UNLIKELY(arr == NULL)) goto end;
      // Populate it...

rts/Trace.c

@@ -685,7 +685,8 @@ void traceHeapProfSampleString(const char *label, StgWord residency)
  void traceIPE(const InfoProvEnt *ipe)
+ {
  #if defined(DEBUG)
 -    if (RtsFlags.TraceFlags.tracing == TRACE_STDERR) {
 +    if (RtsFlags.TraceFlags.tracing == TRACE_STDERR
 +        && RtsFlags.DebugFlags.ipe) {
          ACQUIRE_LOCK(&trace_utx);
          char closure_desc_buf[CLOSURE_DESC_BUFFER_SIZE] = {};

rts/Weak.c

@@ -146,7 +146,7 @@ scheduleFinalizers(Capability *cap, StgWeak *list)
      debugTrace(DEBUG_weak, "weak: batching %d finalizers", n);
 -    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, n, CCS_SYSTEM_OR_NULL);
 +    StgMutArrPtrs *arr = allocateMutArrPtrs(cap, n, NULL, CCS_SYSTEM_OR_NULL);
      if (RTS_UNLIKELY(arr == NULL)) exitHeapOverflow();
      // No write barrier needed here; this array is only going to referred to by this core.
      SET_INFO((StgClosure *) arr, &stg_MUT_ARR_PTRS_FROZEN_CLEAN_info);

rts/include/rts/Flags.h

@@ -118,6 +118,7 @@ typedef struct _DEBUG_FLAGS {
      bool compact;        /* 'C' */
      bool continuation;   /* 'k' */
      bool iomanager;      /* 'o' */
 +    bool ipe;            /* 'I' */
  } DEBUG_FLAGS;
  /* See Note [Synchronization of flags and base APIs] */

testsuite/driver/testglobals.py

@@ -72,6 +72,10 @@ class TestConfig:
          # Was the compiler compiled with -debug?
          self.debug_rts = False
 +        # Were the compiler + libraries built with IPE-related options
 +        # (e.g. -finfo-table-map, -fdistinct-constructor-tables)?
 +        self.ghc_with_ipe = False
++
          # Was the compiler compiled with LLVM?
          self.ghc_built_by_llvm = False

testsuite/driver/testlib.py

@@ -1074,6 +1074,9 @@ def have_profiling( ) -> bool:
  def have_threaded( ) -> bool:
      return config.ghc_with_threaded_rts
 +def ghc_with_ipe( ) -> bool:
 +    return config.ghc_with_ipe
++
  def in_tree_compiler( ) -> bool:
      return config.in_tree_compiler

testsuite/tests/rts/Makefile

@@ -140,7 +140,7 @@ T20199:
  .PHONY: EventlogOutput_IPE
  EventlogOutput_IPE:
  	"$(TEST_HC)" $(TEST_HC_OPTS) -debug -finfo-table-map -v0 EventlogOutput.hs
 -	./EventlogOutput +RTS -va 2> EventlogOutput_IPE.stderr.log
 +	./EventlogOutput +RTS -va -DI 2> EventlogOutput_IPE.stderr.log
  	grep "IPE:" EventlogOutput_IPE.stderr.log
  .PHONY: T23142

testsuite/tests/rts/all.T

@@ -535,6 +535,7 @@ test('T13676',
  test('InitEventLogging',
       [ only_ways(['normal'])
       , extra_run_opts('+RTS -RTS')
 +     , when(ghc_with_ipe(), skip) # IPE builds can change eventlog writer call counts.
       , req_c
       ],
       compile_and_run, ['InitEventLogging_c.c'])
@@ -588,6 +589,7 @@ test('cloneThreadStack', [req_c, only_ways(['threaded1']), extra_ways(['threaded
  test('decodeMyStack',
    [ omit_ghci, js_broken(22261) # cloneMyStack# not yet implemented
 +  , when(ghc_with_ipe(), skip) # IPE builds can change decoded stack output.
    ], compile_and_run, ['-finfo-table-map'])
  # Options:
@@ -595,6 +597,7 @@ test('decodeMyStack',
  test('decodeMyStack_underflowFrames',
    [ extra_run_opts('+RTS -kc8K -RTS')
    , omit_ghci, js_broken(22261) # cloneMyStack# not yet implemented
 +  , when(ghc_with_ipe(), skip) # IPE builds can change decoded stack layout/length.
    ], compile_and_run, ['-finfo-table-map -rtsopts'])
  # -finfo-table-map intentionally missing
@@ -602,6 +605,7 @@ test('decodeMyStack_emptyListForMissingFlag',
    [ ignore_stdout
    , ignore_stderr
    , js_broken(22261) # cloneMyStack# not yet implemented
 +  , when(ghc_with_ipe(), skip) # IPE builds can populate IPE info even without -finfo-table-map on this module.
    ], compile_and_run, [''])
  # Tests RTS flag parsing. Skipped on JS as it uses a distinct RTS.
@@ -646,7 +650,7 @@ test('T25280', [unless(opsys('linux'),skip),req_process,js_skip], compile_and_ru
  test('T25560', [req_c_rts, ignore_stderr], compile_and_run, [''])
  test('TestProddableBlockSet', [req_c_rts], multimod_compile_and_run, ['TestProddableBlockSet.c', '-no-hs-main'])
 -test('T22859',
 +test('T22859',
       [js_skip,
        # This test is vulnerable to changes in allocation behaviour, so we disable it in some ways
        when(arch('wasm32'), skip),

testsuite/tests/rts/ipe/all.T

@@ -8,7 +8,7 @@ test('ipeMap', [extra_files(['ipe_lib.c', 'ipe_lib.h']), c_src, omit_ghci], comp
  test('ipeEventLog',
       [ c_src,
         extra_files(['ipe_lib.c', 'ipe_lib.h']),
 -       extra_run_opts('+RTS -va -RTS'),
 +       extra_run_opts('+RTS -va -DI -RTS'),
         grep_errmsg('table_name_'),
         only_ways(debug_ways),
         normalise_errmsg_fun(noCapabilityOutputFilter),
@@ -24,7 +24,7 @@ test('ipeEventLog',
  test('ipeEventLog_fromMap',
       [ c_src,
         extra_files(['ipe_lib.c', 'ipe_lib.h']),
 -       extra_run_opts('+RTS -va -RTS'),
 +       extra_run_opts('+RTS -va -DI -RTS'),
         grep_errmsg('table_name_'),
         only_ways(debug_ways),
         normalise_errmsg_fun(noCapabilityOutputFilter),
@@ -34,4 +34,3 @@ test('ipeEventLog_fromMap',
         when(opsys('darwin'), fragile(0))
       ],
       compile_and_run, ['ipe_lib.c'])
+-