- ghc-commits - Haskell.org

[Git][ghc/ghc] Pushed new branch wip/sg/first-class-strict
by Sebastian Graf (＠sgraf812) 17 Jun '26

17 Jun '26

Sebastian Graf pushed new branch wip/sg/first-class-strict at Glasgow Haskell Compiler / GHC -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/tree/wip/sg/first-class-strict You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/marge_bot_batch_merge_job] 6 commits: Make the order of usages deterministic
by Marge Bot (＠marge-bot) 17 Jun '26

17 Jun '26

Marge Bot pushed to branch wip/marge_bot_batch_merge_job at Glasgow Haskell Compiler / GHC Commits: d216412b by Ian-Woo Kim at 2026-06-16T20:25:57-04:00 Make the order of usages deterministic It has been observed that the ordering of usages can be non-determinstic in parallel builds. Therefore, this contribution introduces sorting of usages based on a platform- and race-independent sorting criterion. Resolves #26877. Co-authored-by: Wolfgang Jeltsch <wolfgang(a)well-typed.com> - - - - - 8e1cc105 by Wolfgang Jeltsch at 2026-06-16T20:25:57-04:00 Change the descriptions of two existing changelog entries The descriptions now describe the changes in a user-friendly manner, as opposed to describing the contributions that led to these changes in a developer-friendly manner. - - - - - 636c1c7a by Ian Duncan at 2026-06-16T20:26:50-04:00 AArch64: use SXTH, not SXTW, for W32 signExtendReg signExtendReg was using SXTH (sign-extend halfword, 16-bit) for W32-to-W64 sign extension. This should be SXTW (sign-extend word, 32-bit). SXTH only sign-extends the lower 16 bits, producing wrong results for 32-bit values whose bit 15 differs from bit 31. Other fixes: - At sub-W64, code gen for MO_S_Mul2 should use W32 registers for SMULL source operands as per the ARM spec (SMULL Xd, Wn, Wm), and not W64. - Ensure signExtendReg uses the source width for the source operand in SXTW/SXTH/SXTB instructions. GNU as requires sxtw Xd,Wn (not sxtw Xd,Xn), while LLVM's integrated assembler on macOS is lenient. - Fix overflow flag computation for `MO_S_Mul2`. The overflow bit was exactly inverted for sub-W64 operands. Fixes #26978 and #27047 - - - - - b734c75d by Igor Ranieri at 2026-06-16T20:27:32-04:00 haddock: Update CONTRIBUTING with missing step, add missing test dependency - - - - - fa8c0c26 by Luite Stegeman at 2026-06-16T23:04:29-04:00 tag inference: don't confuse functions with their return values inferTagRhs was mixing up taggedness for closures and return values for function closures. We really shouldn't assign TagTuple to a properly tagged function returning a tuple. We fix this by keeping track of functions (TagFun) separately from values (TagVal) and keeping track of their return value. TagFun is also used for join points. fixes #27005 - - - - - 5c1fd991 by Sebastian Graf at 2026-06-16T23:04:30-04:00 Seed the simplifier's in-scope set for open expressions simplifyExpr simplifies expressions typed at the GHCi prompt and the results of Template Haskell splices. Such an expression may be open: at a GHCi debugger breakpoint its free variables include RuntimeUnk skolems standing for as-yet-unknown types. The simplifier began with an in-scope set holding only the wildcard binder, so when it instantiated the unsafeCoerce# wrapper that GHCi builds around a result, it formed a substitution whose range mentioned a free skolem that was not in scope. That breaks the substitution invariant and, in a compiler built with assertions, trips substTy's sanity check. Seed the initial in-scope set with the free variables of the expression. For a closed expression this adds nothing. See Note [Seed the in-scope set for open expressions]. Fixes #17833 and its duplicate #21118. - - - - - 46 changed files: - + changelog.d/T17833 - + changelog.d/T26978 - + changelog.d/T27047 - + changelog.d/deterministic-usage-order - changelog.d/module-graph-reuse-in-downsweep - changelog.d/more-efficient-home-unit-imports-finding - + changelog.d/tag-inference-27005 - compiler/GHC/CmmToAsm/AArch64/CodeGen.hs - compiler/GHC/CmmToAsm/AArch64/Instr.hs - compiler/GHC/CmmToAsm/AArch64/Ppr.hs - compiler/GHC/Core/Opt/Simplify.hs - compiler/GHC/Core/Utils.hs - compiler/GHC/HsToCore/Usage.hs - compiler/GHC/Stg/EnforceEpt.hs - compiler/GHC/Stg/EnforceEpt/Rewrite.hs - compiler/GHC/Stg/EnforceEpt/TagSig.hs - compiler/GHC/Stg/EnforceEpt/Types.hs - compiler/GHC/Unit/Module/Deps.hs - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.asm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.cmm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.asm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.cmm - testsuite/tests/codeGen/should_gen_asm/all.T - + testsuite/tests/codeGen/should_run/aarch64-sxtw-cmm.cmm - + testsuite/tests/codeGen/should_run/aarch64-sxtw-run.hs - + testsuite/tests/codeGen/should_run/aarch64-sxtw-run.stdout - testsuite/tests/codeGen/should_run/all.T - testsuite/tests/core-to-stg/T14895.stderr - testsuite/tests/ghci.debugger/scripts/all.T - testsuite/tests/simplCore/should_compile/T4201.stdout - + testsuite/tests/simplCore/should_run/T27005.hs - + testsuite/tests/simplCore/should_run/T27005.stdout - + testsuite/tests/simplCore/should_run/T27005_aux.hs - testsuite/tests/simplCore/should_run/all.T - testsuite/tests/simplStg/should_compile/T24806.hs - testsuite/tests/simplStg/should_compile/T24806.stderr - + testsuite/tests/simplStg/should_compile/T27005b.hs - + testsuite/tests/simplStg/should_compile/T27005b.stderr - testsuite/tests/simplStg/should_compile/all.T - testsuite/tests/simplStg/should_compile/inferTags004.hs - testsuite/tests/simplStg/should_compile/inferTags004.stderr - + testsuite/tests/simplStg/should_run/T27005a.hs - + testsuite/tests/simplStg/should_run/T27005a.stdout - testsuite/tests/simplStg/should_run/all.T - utils/haddock/CONTRIBUTING.md - utils/haddock/haddock-test/haddock-test.cabal The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/798b1ee0061d72004286a5a5f92898… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/798b1ee0061d72004286a5a5f92898… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][master] haddock: Update CONTRIBUTING with missing step, add missing test
by Marge Bot (＠marge-bot) 17 Jun '26

17 Jun '26

Marge Bot pushed to branch master at Glasgow Haskell Compiler / GHC Commits: b734c75d by Igor Ranieri at 2026-06-16T20:27:32-04:00 haddock: Update CONTRIBUTING with missing step, add missing test dependency - - - - - 2 changed files: - utils/haddock/CONTRIBUTING.md - utils/haddock/haddock-test/haddock-test.cabal Changes: ===================================== utils/haddock/CONTRIBUTING.md ===================================== @@ -33,6 +33,14 @@ $ ./hadrian/build -j --flavour=Quick --freeze1 _build/stage1/bin/haddock Currently, this cannot be done with hadrian but has to be done with a `cabal-install` built from `master`. +First, generate the `.haddock` interface files for GHC's bundled libraries: + +``` +$ ./hadrian/build -j docs-haddock +``` + +Then run the tests: + ``` cabal test -w <absolute-path-to-ghc-repo>/_build/stage1/bin/ghc ``` ===================================== utils/haddock/haddock-test/haddock-test.cabal ===================================== @@ -16,7 +16,7 @@ library default-language: Haskell2010 ghc-options: -Wall hs-source-dirs: src - build-depends: base >= 4.3 && < 4.24, bytestring, directory, process, filepath, Cabal + build-depends: base >= 4.3 && < 4.24, bytestring, directory, process, filepath, Cabal, containers exposed-modules: Test.Haddock View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b734c75d68e572192f0ca2efb875ad1… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b734c75d68e572192f0ca2efb875ad1… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][master] AArch64: use SXTH, not SXTW, for W32 signExtendReg
by Marge Bot (＠marge-bot) 17 Jun '26

17 Jun '26

Marge Bot pushed to branch master at Glasgow Haskell Compiler / GHC Commits: 636c1c7a by Ian Duncan at 2026-06-16T20:26:50-04:00 AArch64: use SXTH, not SXTW, for W32 signExtendReg signExtendReg was using SXTH (sign-extend halfword, 16-bit) for W32-to-W64 sign extension. This should be SXTW (sign-extend word, 32-bit). SXTH only sign-extends the lower 16 bits, producing wrong results for 32-bit values whose bit 15 differs from bit 31. Other fixes: - At sub-W64, code gen for MO_S_Mul2 should use W32 registers for SMULL source operands as per the ARM spec (SMULL Xd, Wn, Wm), and not W64. - Ensure signExtendReg uses the source width for the source operand in SXTW/SXTH/SXTB instructions. GNU as requires sxtw Xd,Wn (not sxtw Xd,Xn), while LLVM's integrated assembler on macOS is lenient. - Fix overflow flag computation for `MO_S_Mul2`. The overflow bit was exactly inverted for sub-W64 operands. Fixes #26978 and #27047 - - - - - 14 changed files: - + changelog.d/T26978 - + changelog.d/T27047 - compiler/GHC/CmmToAsm/AArch64/CodeGen.hs - compiler/GHC/CmmToAsm/AArch64/Instr.hs - compiler/GHC/CmmToAsm/AArch64/Ppr.hs - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.asm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.cmm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.asm - + testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.cmm - testsuite/tests/codeGen/should_gen_asm/all.T - + testsuite/tests/codeGen/should_run/aarch64-sxtw-cmm.cmm - + testsuite/tests/codeGen/should_run/aarch64-sxtw-run.hs - + testsuite/tests/codeGen/should_run/aarch64-sxtw-run.stdout - testsuite/tests/codeGen/should_run/all.T Changes: ===================================== changelog.d/T26978 ===================================== @@ -0,0 +1,14 @@ +section: compiler +issues: #26978 +mrs: !15619 +synopsis: + AArch64 code generation: use ``SXTH`` instead of ``SXTW`` for ``W32`` sign extension +description: + Fix the AArch64 code generator incorrectly using using ``SXTH`` + (sign-extend halfword, 16-bit) for ``W32``-to-``W64`` sign extension, and + always use the source width for the source operand in sign extension + instructions. + + Also ensures that sub-W64 code generation for ``MO_S_Mul2`` uses ``W32`` + registers as mandated by the ARM specification. + ===================================== changelog.d/T27047 ===================================== @@ -0,0 +1,9 @@ +section: compiler +issues: #27047 +mrs: !15619 +synopsis: + Fix incorrect overflow bit for ``MUL2`` on AArch64 for sub-``W64`` operands. +description: + This fixes the computation of the overflow bit for ``MO_S_Mul2`` on + sub-``W64`` operands; the old condition was exactly inverted. + ===================================== compiler/GHC/CmmToAsm/AArch64/CodeGen.hs ===================================== @@ -1894,15 +1894,15 @@ signExtendReg w w' r = W64 -> noop W32 | w' == W32 -> noop - | otherwise -> extend SXTH + | otherwise -> extend SXTW W16 -> extend SXTH W8 -> extend SXTB _ -> panic "intOp" where noop = return (r, nilOL) extend instr = do - r' <- getNewRegNat II64 - return (r', unitOL $ instr (OpReg w' r') (OpReg w' r)) + r' <- getNewRegNat (intFormat w') + return (r', unitOL $ instr (OpReg w' r') (OpReg w r)) -- | Instructions to truncate the value in the given register from width @w@ -- down to width @w'@. @@ -2333,41 +2333,35 @@ genCCall target dest_regs arg_regs = do let lo = getRegisterReg platform (CmmLocal dst_lo) hi = getRegisterReg platform (CmmLocal dst_hi) nd = getRegisterReg platform (CmmLocal dst_needed) - -- Do everything in a full 64 bit registers w' = platformWordWidth platform - (reg_a, code_a') <- signExtendReg w w' reg_a' - (reg_b, code_b') <- signExtendReg w w' reg_b' + -- Sign-extend inputs to W32 for SMULL (Xd = Wn * Wm). + -- signExtendReg always allocates a fresh temp for w < W32, + -- and is a noop for W32 (safe: SMULL reads both sources + -- atomically before writing the destination). + (reg_a, code_a') <- signExtendReg w W32 reg_a' + (reg_b, code_b') <- signExtendReg w W32 reg_b' return $ code_a `appOL` code_b `appOL` code_a' `appOL` code_b' `snocOL` - -- the low 2w' of lo contains the full multiplication; - -- eg: int8 * int8 -> int16 result - -- so lo is in the last w of the register, and hi is in the second w. - SMULL (OpReg w' lo) (OpReg w' reg_a) (OpReg w' reg_b) `snocOL` - -- Make sure we hold onto the sign bits for dst_needed - ASR (OpReg w' hi) (OpReg w' lo) (OpImm (ImmInt $ widthInBits w)) `appOL` - -- lo can now be truncated so we can get at it's top bit easily. + -- SMULL Xd, Wn, Wm: multiply two W32 values producing a + -- 64-bit result. The low w bits of lo contain the truncated + -- product, and hi gets the overflow (sign extension bits). + SMULL (OpReg w' lo) (OpReg W32 reg_a) (OpReg W32 reg_b) `snocOL` + ASR (OpReg w' hi) (OpReg w' lo) (OpImm (ImmInt $ widthInBits w)) `appOL` truncateReg w' w lo `snocOL` - -- Note the use of CMN (compare negative), not CMP: we want to - -- test if the top half is negative one and the top - -- bit of the bottom half is positive one. eg: - -- hi = 0b1111_1111 (actually 64 bits) - -- lo = 0b1010_1111 (-81, so the result didn't need the top half) - -- lo' = ASR(lo,7) (second reg of SMN) - -- = 0b0000_0001 (theeshift gives us 1 for negative, - -- and 0 for positive) - -- hi == -lo'? - -- 0b1111_1111 == 0b1111_1111 (yes, top half is just overflow) - -- Another way to think of this is if hi + lo' == 0, which is what - -- CMN really is under the hood. + -- CMN (compare negative) tests hi + lo' == 0, i.e. hi == -lo'. + -- lo' = LSR(lo, w-1) gives 1 if lo is negative, 0 if positive. + -- No overflow iff hi is the sign extension of lo: + -- lo positive (bit w-1 = 0) => lo' = 0, need hi == 0 + -- lo negative (bit w-1 = 1) => lo' = 1, need hi == -1 + -- CMN sets Z when hi + lo' == 0 (no overflow), so we use + -- NE to set nd = 1 when overflow occurred. CMN (OpReg w' hi) (OpRegShift w' lo SLSR (widthInBits w - 1)) `snocOL` - -- Set dst_needed to 1 if hi and lo' were (negatively) equal - CSET (OpReg w' nd) EQ `appOL` - -- Finally truncate hi to drop any extraneous sign bits. + CSET (OpReg w' nd) NE `appOL` truncateReg w' w hi -- Can't handle > 64 bit operands | otherwise -> unsupported (MO_S_Mul2 w) ===================================== compiler/GHC/CmmToAsm/AArch64/Instr.hs ===================================== @@ -105,6 +105,7 @@ regUsageOfInstr platform instr = case instr of SXTB dst src -> usage (regOp src, regOp dst) UXTB dst src -> usage (regOp src, regOp dst) SXTH dst src -> usage (regOp src, regOp dst) + SXTW dst src -> usage (regOp src, regOp dst) UXTH dst src -> usage (regOp src, regOp dst) CLZ dst src -> usage (regOp src, regOp dst) RBIT dst src -> usage (regOp src, regOp dst) @@ -296,6 +297,7 @@ patchRegsOfInstr instr env = case instr of SXTB o1 o2 -> SXTB (patchOp o1) (patchOp o2) UXTB o1 o2 -> UXTB (patchOp o1) (patchOp o2) SXTH o1 o2 -> SXTH (patchOp o1) (patchOp o2) + SXTW o1 o2 -> SXTW (patchOp o1) (patchOp o2) UXTH o1 o2 -> UXTH (patchOp o1) (patchOp o2) CLZ o1 o2 -> CLZ (patchOp o1) (patchOp o2) RBIT o1 o2 -> RBIT (patchOp o1) (patchOp o2) @@ -699,8 +701,7 @@ data Instr | UXTB Operand Operand | SXTH Operand Operand | UXTH Operand Operand - -- | SXTW Operand Operand - -- | SXTX Operand Operand + | SXTW Operand Operand | PUSH_STACK_FRAME | POP_STACK_FRAME -- 1. Arithmetic Instructions ---------------------------------------------- @@ -857,6 +858,7 @@ instrCon i = SXTB{} -> "SXTB" UXTB{} -> "UXTB" SXTH{} -> "SXTH" + SXTW{} -> "SXTW" UXTH{} -> "UXTH" PUSH_STACK_FRAME{} -> "PUSH_STACK_FRAME" POP_STACK_FRAME{} -> "POP_STACK_FRAME" ===================================== compiler/GHC/CmmToAsm/AArch64/Ppr.hs ===================================== @@ -516,6 +516,7 @@ pprInstr platform instr = case instr of SXTB o1 o2 -> op2 (text "\tsxtb") o1 o2 UXTB o1 o2 -> op2 (text "\tuxtb") o1 o2 SXTH o1 o2 -> op2 (text "\tsxth") o1 o2 + SXTW o1 o2 -> op2 (text "\tsxtw") o1 o2 UXTH o1 o2 -> op2 (text "\tuxth") o1 o2 -- 3. Logical and Move Instructions ------------------------------------------ ===================================== testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.asm ===================================== @@ -0,0 +1 @@ +sxth ===================================== testsuite/tests/codeGen/should_gen_asm/aarch64-sxth-mul2.cmm ===================================== @@ -0,0 +1,12 @@ +#include "Cmm.h" + +// Exercises MO_S_Mul2 W16: signExtendReg W16 W32 must emit SXTH to +// sign-extend the 16-bit inputs to 32-bit before SMULL. +testMul2W16 (W_ buffer) { + I16 a, b, needed, hi, lo; + a = I16[buffer]; + b = I16[buffer + 2]; + (needed, hi, lo) = prim %mul2_16(a, b); + I16[buffer + 4] = lo; + return(); +} ===================================== testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.asm ===================================== @@ -0,0 +1 @@ +smull ===================================== testsuite/tests/codeGen/should_gen_asm/aarch64-sxtw.cmm ===================================== @@ -0,0 +1,13 @@ +#include "Cmm.h" + +// Exercises MO_S_Mul2 W32, which uses SMULL (Xd = Wn * Wm). +// After the fix, signExtendReg W32 W32 is a noop (no SXTW needed), +// but the SMULL instruction itself must still appear. +testMul2W32 (W_ buffer) { + I32 a, b, needed, hi, lo; + a = I32[buffer]; + b = I32[buffer + 4]; + (needed, hi, lo) = prim %mul2_32(a, b); + I32[buffer + 8] = lo; + return(); +} ===================================== testsuite/tests/codeGen/should_gen_asm/all.T ===================================== @@ -31,3 +31,5 @@ is_aarch64_codegen = [ # AArch64-specific tests test('aarch64-ushr-subword', is_aarch64_codegen, compile_grep_asm, ['hs', True, '-O']) test('aarch64-shl-subword', is_aarch64_codegen, compile_grep_asm, ['hs', True, '-O']) +test('aarch64-sxtw', is_aarch64_codegen, compile_grep_asm, ['cmm', True, '']) +test('aarch64-sxth-mul2', is_aarch64_codegen, compile_grep_asm, ['cmm', True, '']) ===================================== testsuite/tests/codeGen/should_run/aarch64-sxtw-cmm.cmm ===================================== @@ -0,0 +1,41 @@ +#include "Cmm.h" + +// MO_S_Mul2 W32: signExtendReg W32 W32 is a noop (inputs already W32). +// Tests the SMULL path with values whose bit 15 != bit 31 (the old SXTH +// bug would corrupt these). +runMul2W32zh (W_ x) { + I32 a, b, needed, hi, lo; + a = %lobits32(x); + b = 2::I32; + (needed, hi, lo) = prim %mul2_32(a, b); + return(TO_W_(lo)); +} + +// MO_S_Mul2 W16: signExtendReg W16 W32 emits SXTH before SMULL. +runMul2W16zh (W_ x) { + I16 a, b, needed, hi, lo; + a = %lobits16(x); + b = 2::I16; + (needed, hi, lo) = prim %mul2_16(a, b); + return(TO_W_(lo)); +} + +// MO_S_Mul2 W8: signExtendReg W8 W32 emits SXTB before SMULL. +runMul2W8zh (W_ x) { + I8 a, b, needed, hi, lo; + a = %lobits8(x); + b = 2::I8; + (needed, hi, lo) = prim %mul2_8(a, b); + return(TO_W_(lo)); +} + +// MO_S_Mul2 W16 returning (needed, hi, lo) as a packed triple. +// Tests the CSET NE overflow detection: needed=1 when overflow occurs. +// Returns (needed << 32) | (lo & 0xFFFF) so the Haskell side can inspect both. +runMul2W16Overflowzh (W_ x, W_ y) { + I16 a, b, needed, hi, lo; + a = %lobits16(x); + b = %lobits16(y); + (needed, hi, lo) = prim %mul2_16(a, b); + return(TO_W_(needed) * 0x100000000 + (TO_W_(lo) `and` 0xFFFF)); +} ===================================== testsuite/tests/codeGen/should_run/aarch64-sxtw-run.hs ===================================== @@ -0,0 +1,60 @@ +{-# LANGUAGE ForeignFunctionInterface #-} +{-# LANGUAGE GHCForeignImportPrim #-} +{-# LANGUAGE MagicHash #-} +{-# LANGUAGE UnliftedFFITypes #-} +module Main where + +import GHC.Exts +import Data.Bits (shiftR, (.&.)) + +foreign import prim "runMul2W32zh" runMul2W32# :: Word# -> Word# +foreign import prim "runMul2W16zh" runMul2W16# :: Word# -> Word# +foreign import prim "runMul2W8zh" runMul2W8# :: Word# -> Word# +foreign import prim "runMul2W16Overflowzh" + runMul2W16Overflow# :: Word# -> Word# -> Word# + +mul2W32 :: Word -> Word +mul2W32 (W# x) = W# (runMul2W32# x) + +mul2W16 :: Word -> Word +mul2W16 (W# x) = W# (runMul2W16# x) + +mul2W8 :: Word -> Word +mul2W8 (W# x) = W# (runMul2W8# x) + +mul2W16Overflow :: Word -> Word -> (Word, Word) +mul2W16Overflow (W# x) (W# y) = + let r = W# (runMul2W16Overflow# x y) + needed = r `shiftR` 32 + lo = r .&. 0xFFFF + in (needed, lo) + +main :: IO () +main = do + -- W32: 50000 has bit 15 set, bit 31 clear. The old SXTH bug would + -- sign-extend from bit 15, corrupting the value before SMULL. + putStrLn $ "W32 50000*2 = " ++ show (mul2W32 50000) + putStrLn $ "W32 1*2 = " ++ show (mul2W32 1) + putStrLn $ "W32 70000*2 = " ++ show (mul2W32 70000) + + -- W16: exercises signExtendReg W16 W32 (SXTH before SMULL). + putStrLn $ "W16 200*2 = " ++ show (mul2W16 200) + putStrLn $ "W16 100*2 = " ++ show (mul2W16 100) + + -- W8: exercises signExtendReg W8 W32 (SXTB before SMULL). + putStrLn $ "W8 50*2 = " ++ show (mul2W8 50) + putStrLn $ "W8 3*2 = " ++ show (mul2W8 3) + + -- Overflow detection (CSET NE fix): + -- 200 * 200 = 40000, overflows I16 [-32768, 32767]. needed=1. + let (n1, lo1) = mul2W16Overflow 200 200 + putStrLn $ "W16 200*200: needed=" ++ show n1 ++ " lo=" ++ show lo1 + -- 100 * 2 = 200, fits in I16. needed=0. + let (n2, lo2) = mul2W16Overflow 100 2 + putStrLn $ "W16 100*2: needed=" ++ show n2 ++ " lo=" ++ show lo2 + -- 181 * 181 = 32761, just under 32767. needed=0. + let (n3, lo3) = mul2W16Overflow 181 181 + putStrLn $ "W16 181*181: needed=" ++ show n3 ++ " lo=" ++ show lo3 + -- 182 * 182 = 33124, just over 32767. needed=1. + let (n4, lo4) = mul2W16Overflow 182 182 + putStrLn $ "W16 182*182: needed=" ++ show n4 ++ " lo=" ++ show lo4 ===================================== testsuite/tests/codeGen/should_run/aarch64-sxtw-run.stdout ===================================== @@ -0,0 +1,11 @@ +W32 50000*2 = 100000 +W32 1*2 = 2 +W32 70000*2 = 140000 +W16 200*2 = 400 +W16 100*2 = 200 +W8 50*2 = 100 +W8 3*2 = 6 +W16 200*200: needed=1 lo=40000 +W16 100*2: needed=0 lo=200 +W16 181*181: needed=0 lo=32761 +W16 182*182: needed=1 lo=33124 ===================================== testsuite/tests/codeGen/should_run/all.T ===================================== @@ -289,3 +289,9 @@ test('T27072w', [req_c, js_skip, when(opsys('darwin'), skip)], # AArch64-specific runtime tests test('aarch64-ushr-subword-run', [unless(arch('aarch64'), skip)], compile_and_run, ['-O']) test('aarch64-subword-ops', [unless(arch('aarch64'), skip)], compile_and_run, ['-O']) +test('aarch64-sxtw-run', + [ extra_files(['aarch64-sxtw-cmm.cmm']), + unless(arch('aarch64'), skip), + when(unregisterised(), skip)], + multi_compile_and_run, + ['aarch64-sxtw-run', [('aarch64-sxtw-cmm.cmm', '')], '-O']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/636c1c7ae47495f022affa501ea0a40… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/636c1c7ae47495f022affa501ea0a40… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][master] 2 commits: Make the order of usages deterministic
by Marge Bot (＠marge-bot) 17 Jun '26

17 Jun '26

Marge Bot pushed to branch master at Glasgow Haskell Compiler / GHC Commits: d216412b by Ian-Woo Kim at 2026-06-16T20:25:57-04:00 Make the order of usages deterministic It has been observed that the ordering of usages can be non-determinstic in parallel builds. Therefore, this contribution introduces sorting of usages based on a platform- and race-independent sorting criterion. Resolves #26877. Co-authored-by: Wolfgang Jeltsch <wolfgang(a)well-typed.com> - - - - - 8e1cc105 by Wolfgang Jeltsch at 2026-06-16T20:25:57-04:00 Change the descriptions of two existing changelog entries The descriptions now describe the changes in a user-friendly manner, as opposed to describing the contributions that led to these changes in a developer-friendly manner. - - - - - 5 changed files: - + changelog.d/deterministic-usage-order - changelog.d/module-graph-reuse-in-downsweep - changelog.d/more-efficient-home-unit-imports-finding - compiler/GHC/HsToCore/Usage.hs - compiler/GHC/Unit/Module/Deps.hs Changes: ===================================== changelog.d/deterministic-usage-order ===================================== @@ -0,0 +1,8 @@ +section: compiler +synopsis: Make the order of usages deterministic +issues: #26877 +mrs: !15484 +description: { + The order in which usages appear in interface files is now + deterministic. +} ===================================== changelog.d/module-graph-reuse-in-downsweep ===================================== @@ -3,7 +3,7 @@ synopsis: Allow `downsweep` to use nodes of an existing module graph issues: #27054 mrs: !16028 description: { - This contribution enables `downsweep` to use the nodes of a module - graph obtained from a previous downsweeping round, which allows GHC - API applications to build module graphs somewhat incrementally. + The `downsweep` operation can now use the nodes of a module graph + obtained from a previous downsweeping round, which allows GHC API + applications to build module graphs somewhat incrementally. } ===================================== changelog.d/more-efficient-home-unit-imports-finding ===================================== @@ -3,13 +3,10 @@ synopsis: Introduce a cache of home module name providers issues: #27055 mrs: !15888 description: { - This contribution optimizes the algorithm for finding out which home - unit provides the module that a certain import declaration refers - to. The previous approach has been to simply search all home units - in no particular order. This change introduces a cache that allows - for efficiently determining those complete home units that provide a - certain module name and changes the module-finding algorithm such - that it searches these units before the other home units. This leads - to significant performance improvements in situations where there - are lots of home units. + For finding out which home unit provides the module that a certain + import declaration refers to, we now use a better algorithm. The + previous approach was to simply search all home units in no + particular order. The new algorithm first searches those complete + home units that provide a certain module name, before searching the + other home units. } ===================================== compiler/GHC/HsToCore/Usage.hs ===================================== @@ -70,6 +70,8 @@ data UsageConfig = UsageConfig { uc_safe_implicit_imps_req :: !Bool -- ^ Are all implicit imports required to be safe for this Safe Haskell mode? } +-- | Build the list of 'Usage's that drive recompilation checking. +-- The resulting list is deterministically sorted. mkUsageInfo :: UsageConfig -> Plugins -> FinderCache -> UnitEnv -> Module -> ImportedMods -> [ImportUserSpec] -> NameSet -> [FilePath] -> [FilePath] -> [(Module, Fingerprint)] -> [LinkableUsage] -> PkgsLoaded @@ -100,10 +102,10 @@ mkUsageInfo uc plugins fc unit_env } | (mod, hash) <- merged ] ++ object_usages - usages `seqList` return usages - -- seq the list of Usages returned: occasionally these - -- don't get evaluated for a while and we can end up hanging on to - -- the entire collection of Ifaces. + usages `seqList` return (sortBy stableUsageCmp usages) + -- The use of 'seqList' is important because occasionally the returned list + -- is not evaluated for a while, so that with too much laziness here we + -- could end up hanging on to the entire collection of 'Iface's. {- Note [Plugin dependencies] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ===================================== compiler/GHC/Unit/Module/Deps.hs ===================================== @@ -1,5 +1,8 @@ {-# LANGUAGE PatternSynonyms #-} {-# LANGUAGE DerivingVia #-} + +{-# OPTIONS_GHC -Wwarn=incomplete-record-selectors #-} + -- | Dependencies and Usage of a module module GHC.Unit.Module.Deps ( Dependencies(dep_direct_mods @@ -17,6 +20,7 @@ module GHC.Unit.Module.Deps , noDependencies , pprDeps , Usage (..) + , stableUsageCmp , HomeModImport (..) , HomeModImportedAvails (..) , ImportAvails (..) @@ -52,6 +56,7 @@ import GHC.Utils.Outputable import Control.DeepSeq import Data.Bifunctor import qualified Data.Foldable as Foldable +import Data.Function (on) import Data.List (sortBy, sort, partition) import Data.Set (Set) import qualified Data.Set as Set @@ -497,6 +502,47 @@ instance Binary Usage where i -> error ("Binary.get(Usage): " ++ show i) +-- | Compares 'Usage's by constructor and, if the constructors are the same, by +-- identifying strings, to achieve a predictable ordering. +stableUsageCmp :: Usage -> Usage -> Ordering +stableUsageCmp + usage1@UsagePackageModule {} + usage2@UsagePackageModule {} + = (compare `on` usg_mod) usage1 usage2 +stableUsageCmp + usage1@UsageHomeModule {} + usage2@UsageHomeModule {} + = (compare `on` Module <$> usg_unit_id <*> usg_mod_name) usage1 usage2 +stableUsageCmp + usage1@UsageFile {} + usage2@UsageFile {} + = (lexicalCompareFS `on` usg_file_path) usage1 usage2 +stableUsageCmp + usage1@UsageDirectory {} + usage2@UsageDirectory {} + = (lexicalCompareFS `on` usg_dir_path) usage1 usage2 +stableUsageCmp + usage1@UsageHomeModuleBytecode {} + usage2@UsageHomeModuleBytecode {} + = (compare `on` Module <$> usg_unit_id <*> usg_mod_name) usage1 usage2 +stableUsageCmp + usage1@UsageMergedRequirement {} + usage2@UsageMergedRequirement {} + = (compare `on` usg_mod) usage1 usage2 +stableUsageCmp + usage1 + usage2 + = (compare `on` constructorIndex) usage1 usage2 + where + + constructorIndex :: Usage -> Int + constructorIndex UsagePackageModule {} = 0 + constructorIndex UsageHomeModule {} = 1 + constructorIndex UsageFile {} = 2 + constructorIndex UsageDirectory {} = 3 + constructorIndex UsageHomeModuleBytecode {} = 4 + constructorIndex UsageMergedRequirement {} = 5 + -- | Records the imports that we depend on from a home module, -- for recompilation checking. -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/a3fa10e0c93a58f46815c4fab7522f… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/a3fa10e0c93a58f46815c4fab7522f… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/T27383] Desugar a `case` scrutinee only once (#27383, #20251)
by Sebastian Graf (＠sgraf812) 16 Jun '26

16 Jun '26

Sebastian Graf pushed to branch wip/T27383 at Glasgow Haskell Compiler / GHC Commits: 0e7c8aca by Sebastian Graf at 2026-06-16T23:47:03+02:00 Desugar a `case` scrutinee only once (#27383, #20251) In `dsExpr` for `HsCase` we desugared the scrutinee /twice/: once to build the Core `case` itself, and again inside `matchWrapper`, which re-desugared the source scrutinee (via `addHsScrutTmCs`) purely to record long-distance information for the pattern-match checker. For a single `case` that is merely wasteful. But for nested cases it is catastrophic. Consider case (case (case e of ... ) of ... ) of ... Desugaring the outer scrutinee desugars the middle `case` twice, each of which desugars the inner `case` twice, and so on. The work doubles at every level, so desugaring takes O(2^n) time in the nesting depth. That is the blowup reported in #27383; it is also what makes the machine-generated program in #20251 take an age to compile. The fix is simple. `matchWrapper` is handed the scrutinee anyway, so we give it the Core expression we have /already/ desugared, and record the long-distance term constraint with `addCoreScrutTmCs` instead of re-desugaring from source. This is just what `matchSinglePatVar` already does for single-pattern matches. So: * `matchWrapper` now takes `Maybe [CoreExpr]` rather than `Maybe [LHsExpr GhcTc]`. * The `HsCase` equation of `dsExpr` passes the already-desugared `core_discrim`; the arrow desugarer passes its match variables. * `addHsScrutTmCs` had no other use, so it is gone. Desugaring is now linear in the nesting depth. (The coverage checker still runs `simpleOptExpr` over each scrutinee, which leaves the total at O(n^2); that is ample.) The long-distance information itself is unchanged: the checker sees precisely the Core that backs the generated code. Test: deSugar/should_compile/T27383 - - - - - 8 changed files: - + changelog.d/fix-exponential-case-desugar-27383 - compiler/GHC/HsToCore/Arrows.hs - compiler/GHC/HsToCore/Expr.hs - compiler/GHC/HsToCore/Match.hs - compiler/GHC/HsToCore/Match.hs-boot - compiler/GHC/HsToCore/Pmc.hs - + testsuite/tests/deSugar/should_compile/T27383.hs - testsuite/tests/deSugar/should_compile/all.T Changes: ===================================== changelog.d/fix-exponential-case-desugar-27383 ===================================== @@ -0,0 +1,6 @@ +section: compiler +synopsis: Fix exponential-time desugaring of nested ``case`` expressions. + The scrutinee is no longer desugared a second time when recording + long-distance information for the pattern-match checker. +issues: #27383 #20251 +mrs: !16203 ===================================== compiler/GHC/HsToCore/Arrows.hs ===================================== @@ -553,7 +553,7 @@ dsCmd ids local_vars stack_ty res_ty in_ty = envStackType env_ids stack_ty' discrims = map nlHsVar arg_ids (discrim_vars, matching_code) - <- matchWrapper (ArrowMatchCtxt match_ctxt) (Just discrims) match' + <- matchWrapper (ArrowMatchCtxt match_ctxt) (Just (map Var arg_ids)) match' core_body <- flip (bind_vars discrim_vars) matching_code <$> traverse dsLExpr discrims ===================================== compiler/GHC/HsToCore/Expr.hs ===================================== @@ -421,7 +421,7 @@ dsExpr (HsPragE _ (HsPragSCC _ cc) expr) dsExpr (HsCase ctxt discrim matches) = do { core_discrim <- dsLExpr discrim - ; ([discrim_var], matching_code) <- matchWrapper ctxt (Just [discrim]) matches + ; ([discrim_var], matching_code) <- matchWrapper ctxt (Just [core_discrim]) matches ; return (bindNonRec discrim_var core_discrim matching_code) } -- Pepe: The binds are in scope in the body but NOT in the binding group ===================================== compiler/GHC/HsToCore/Match.hs ===================================== @@ -761,7 +761,7 @@ Call @match@ with all of this information! matchWrapper :: HsMatchContextRn -- ^ For shadowing warning messages - -> Maybe [LHsExpr GhcTc] -- ^ Scrutinee(s) + -> Maybe [CoreExpr] -- ^ Already-desugared scrutinee(s) -- see Note [matchWrapper scrutinees] -> MatchGroup GhcTc (LHsExpr GhcTc) -- ^ Matches being desugared -> DsM ([Id], CoreExpr) -- ^ Results (usually passed to 'match') @@ -818,7 +818,7 @@ matchWrapper ctxt scrs (MG { mg_alts = L _ matches -- pmc for pattern synonyms -- See Note [Long-distance information] in GHC.HsToCore.Pmc - then addHsScrutTmCs (concat scrs) new_vars $ + then addCoreScrutTmCs (concat scrs) new_vars $ pmcMatches origin (DsMatchContext ctxt locn) new_vars matches -- When we're not doing PM checks on the match group, ===================================== compiler/GHC/HsToCore/Match.hs-boot ===================================== @@ -16,7 +16,7 @@ match :: [Id] matchWrapper :: HsMatchContextRn - -> Maybe [LHsExpr GhcTc] + -> Maybe [CoreExpr] -> MatchGroup GhcTc (LHsExpr GhcTc) -> DsM ([Id], CoreExpr) ===================================== compiler/GHC/HsToCore/Pmc.hs ===================================== @@ -34,7 +34,7 @@ module GHC.HsToCore.Pmc ( isMatchContextPmChecked, isMatchContextPmChecked_SinglePat, -- See Note [Long-distance information] - addTyCs, addCoreScrutTmCs, addHsScrutTmCs, getLdiNablas, + addTyCs, addCoreScrutTmCs, getLdiNablas, getNFirstUncovered ) where @@ -59,7 +59,6 @@ import GHC.Utils.Panic import GHC.Types.Var (EvVar, Var (..)) import GHC.Types.Id.Info import GHC.Tc.Utils.TcType (evVarPred) -import {-# SOURCE #-} GHC.HsToCore.Expr (dsLExpr) import GHC.HsToCore.Monad import GHC.Data.Bag import GHC.Data.OrdList @@ -684,12 +683,6 @@ addCoreScrutTmCs (scr:scrs) (x:xs) k addPhiCtsNablas nablas (unitBag (PhiCoreCt x scr)) addCoreScrutTmCs _ _ _ = panic "addCoreScrutTmCs: numbers of scrutinees and match ids differ" --- | 'addCoreScrutTmCs', but desugars the 'LHsExpr's first. -addHsScrutTmCs :: [LHsExpr GhcTc] -> [Id] -> DsM a -> DsM a -addHsScrutTmCs scrs vars k = do - scr_es <- traverse dsLExpr scrs - addCoreScrutTmCs scr_es vars k - {- Note [Long-distance information] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Consider @@ -716,8 +709,8 @@ To achieve similar reasoning in the coverage checker: which the call sites put into place with 'GHC.HsToCore.Monad.updPmNablas'. * Call sites also extend this set with facts from type-constraint dictionaries, - case scrutinees, etc. with the exported functions 'addTyCs', 'addCoreScrutTmCs' - and 'addHsScrutTmCs'. + case scrutinees, etc. with the exported functions 'addTyCs' and + 'addCoreScrutTmCs'. Note [Recovering from unsatisfiable pattern-matching constraints] ===================================== testsuite/tests/deSugar/should_compile/T27383.hs ===================================== @@ -0,0 +1,9 @@ +-- Regression test for #27383 (and #20251): deeply nested cases whose +-- scrutinees are themselves cases must desugar in polynomial (not +-- exponential) time. With the bug the scrutinee is desugared twice at +-- every level, so compiler allocation (tracked by this performance test) +-- grows as O(2^n) in the nesting depth. +module T27383 where + +zero :: Int +zero = (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case 0 of x500 -> x500) of x499 -> x499) of x498 -> x498) of x497 -> x497) of x496 -> x496) of x495 -> x495) of x494 -> x494) of x493 -> x493) of x492 -> x492) of x491 -> x491) of x490 -> x490) of x489 -> x489) of x488 -> x488) of x487 -> x487) of x486 -> x486) of x485 -> x485) of x484 -> x484) of x483 -> x483) of x482 -> x482) of x481 -> x481) of x480 -> x480) of x479 -> x479) of x478 -> x478) of x477 -> x477) of x476 -> x476) of x475 -> x475) of x474 -> x474) of x473 -> x473) of x472 -> x472) of x471 -> x471) of x470 -> x470) of x469 -> x469) of x468 -> x468) of x467 -> x467) of x466 -> x466) of x465 -> x465) of x464 -> x464) of x463 -> x463) of x462 -> x462) of x461 -> x461) of x460 -> x460) of x459 -> x459) of x458 -> x458) of x457 -> x457) of x456 -> x456) of x455 -> x455) of x454 -> x454) of x453 -> x453) of x452 -> x452) of x451 -> x451) of x450 -> x450) of x449 -> x449) of x448 -> x448) of x447 -> x447) of x446 -> x446) of x445 -> x445) of x444 -> x444) of x443 -> x443) of x442 -> x442) of x441 -> x441) of x440 -> x440) of x439 -> x439) of x438 -> x438) of x437 -> x437) of x436 -> x436) of x435 -> x435) of x434 -> x434) of x433 -> x433) of x432 -> x432) of x431 -> x431) of x430 -> x430) of x429 -> x429) of x428 -> x428) of x427 -> x427) of x426 -> x426) of x425 -> x425) of x424 -> x424) of x423 -> x423) of x422 -> x422) of x421 -> x421) of x420 -> x420) of x419 -> x419) of x418 -> x418) of x417 -> x417) of x416 -> x416) of x415 -> x415) of x414 -> x414) of x413 -> x413) of x412 -> x412) of x411 -> x411) of x410 -> x410) of x409 -> x409) of x408 -> x408) of x407 -> x407) of x406 -> x406) of x405 -> x405) of x404 -> x404) of x403 -> x403) of x402 -> x402) of x401 -> x401) of x400 -> x400) of x399 -> x399) of x398 -> x398) of x397 -> x397) of x396 -> x396) of x395 -> x395) of x394 -> x394) of x393 -> x393) of x392 -> x392) of x391 -> x391) of x390 -> x390) of x389 -> x389) of x388 -> x388) of x387 -> x387) of x386 -> x386) of x385 -> x385) of x384 -> x384) of x383 -> x383) of x382 -> x382) of x381 -> x381) of x380 -> x380) of x379 -> x379) of x378 -> x378) of x377 -> x377) of x376 -> x376) of x375 -> x375) of x374 -> x374) of x373 -> x373) of x372 -> x372) of x371 -> x371) of x370 -> x370) of x369 -> x369) of x368 -> x368) of x367 -> x367) of x366 -> x366) of x365 -> x365) of x364 -> x364) of x363 -> x363) of x362 -> x362) of x361 -> x361) of x360 -> x360) of x359 -> x359) of x358 -> x358) of x357 -> x357) of x356 -> x356) of x355 -> x355) of x354 -> x354) of x353 -> x353) of x352 -> x352) of x351 -> x351) of x350 -> x350) of x349 -> x349) of x348 -> x348) of x347 -> x347) of x346 -> x346) of x345 -> x345) of x344 -> x344) of x343 -> x343) of x342 -> x342) of x341 -> x341) of x340 -> x340) of x339 -> x339) of x338 -> x338) of x337 -> x337) of x336 -> x336) of x335 -> x335) of x334 -> x334) of x333 -> x333) of x332 -> x332) of x331 -> x331) of x330 -> x330) of x329 -> x329) of x328 -> x328) of x327 -> x327) of x326 -> x326) of x325 -> x325) of x324 -> x324) of x323 -> x323) of x322 -> x322) of x321 -> x321) of x320 -> x320) of x319 -> x319) of x318 -> x318) of x317 -> x317) of x316 -> x316) of x315 -> x315) of x314 -> x314) of x313 -> x313) of x312 -> x312) of x311 -> x311) of x310 -> x310) of x309 -> x309) of x308 -> x308) of x307 -> x307) of x306 -> x306) of x305 -> x305) of x304 -> x304) of x303 -> x303) of x302 -> x302) of x301 -> x301) of x300 -> x300) of x299 -> x299) of x298 -> x298) of x297 -> x297) of x296 -> x296) of x295 -> x295) of x294 -> x294) of x293 -> x293) of x292 -> x292) of x291 -> x291) of x290 -> x290) of x289 -> x289) of x288 -> x288) of x287 -> x287) of x286 -> x286) of x285 -> x285) of x284 -> x284) of x283 -> x283) of x282 -> x282) of x281 -> x281) of x280 -> x280) of x279 -> x279) of x278 -> x278) of x277 -> x277) of x276 -> x276) of x275 -> x275) of x274 -> x274) of x273 -> x273) of x272 -> x272) of x271 -> x271) of x270 -> x270) of x269 -> x269) of x268 -> x268) of x267 -> x267) of x266 -> x266) of x265 -> x265) of x264 -> x264) of x263 -> x263) of x262 -> x262) of x261 -> x261) of x260 -> x260) of x259 -> x259) of x258 -> x258) of x257 -> x257) of x256 -> x256) of x255 -> x255) of x254 -> x254) of x253 -> x253) of x252 -> x252) of x251 -> x251) of x250 -> x250) of x249 -> x249) of x248 -> x248) of x247 -> x247) of x246 -> x246) of x245 -> x245) of x244 -> x244) of x243 -> x243) of x242 -> x242) of x241 -> x241) of x240 -> x240) of x239 -> x239) of x238 -> x238) of x237 -> x237) of x236 -> x236) of x235 -> x235) of x234 -> x234) of x233 -> x233) of x232 -> x232) of x231 -> x231) of x230 -> x230) of x229 -> x229) of x228 -> x228) of x227 -> x227) of x226 -> x226) of x225 -> x225) of x224 -> x224) of x223 -> x223) of x222 -> x222) of x221 -> x221) of x220 -> x220) of x219 -> x219) of x218 -> x218) of x217 -> x217) of x216 -> x216) of x215 -> x215) of x214 -> x214) of x213 -> x213) of x212 -> x212) of x211 -> x211) of x210 -> x210) of x209 -> x209) of x208 -> x208) of x207 -> x207) of x206 -> x206) of x205 -> x205) of x204 -> x204) of x203 -> x203) of x202 -> x202) of x201 -> x201) of x200 -> x200) of x199 -> x199) of x198 -> x198) of x197 -> x197) of x196 -> x196) of x195 -> x195) of x194 -> x194) of x193 -> x193) of x192 -> x192) of x191 -> x191) of x190 -> x190) of x189 -> x189) of x188 -> x188) of x187 -> x187) of x186 -> x186) of x185 -> x185) of x184 -> x184) of x183 -> x183) of x182 -> x182) of x181 -> x181) of x180 -> x180) of x179 -> x179) of x178 -> x178) of x177 -> x177) of x176 -> x176) of x175 -> x175) of x174 -> x174) of x173 -> x173) of x172 -> x172) of x171 -> x171) of x170 -> x170) of x169 -> x169) of x168 -> x168) of x167 -> x167) of x166 -> x166) of x165 -> x165) of x164 -> x164) of x163 -> x163) of x162 -> x162) of x161 -> x161) of x160 -> x160) of x159 -> x159) of x158 -> x158) of x157 -> x157) of x156 -> x156) of x155 -> x155) of x154 -> x154) of x153 -> x153) of x152 -> x152) of x151 -> x151) of x150 -> x150) of x149 -> x149) of x148 -> x148) of x147 -> x147) of x146 -> x146) of x145 -> x145) of x144 -> x144) of x143 -> x143) of x142 -> x142) of x141 -> x141) of x140 -> x140) of x139 -> x139) of x138 -> x138) of x137 -> x137) of x136 -> x136) of x135 -> x135) of x134 -> x134) of x133 -> x133) of x132 -> x132) of x131 -> x131) of x130 -> x130) of x129 -> x129) of x128 -> x128) of x127 -> x127) of x126 -> x126) of x125 -> x125) of x124 -> x124) of x123 -> x123) of x122 -> x122) of x121 -> x121) of x120 -> x120) of x119 -> x119) of x118 -> x118) of x117 -> x117) of x116 -> x116) of x115 -> x115) of x114 -> x114) of x113 -> x113) of x112 -> x112) of x111 -> x111) of x110 -> x110) of x109 -> x109) of x108 -> x108) of x107 -> x107) of x106 -> x106) of x105 -> x105) of x104 -> x104) of x103 -> x103) of x102 -> x102) of x101 -> x101) of x100 -> x100) of x99 -> x99) of x98 -> x98) of x97 -> x97) of x96 -> x96) of x95 -> x95) of x94 -> x94) of x93 -> x93) of x92 -> x92) of x91 -> x91) of x90 -> x90) of x89 -> x89) of x88 -> x88) of x87 -> x87) of x86 -> x86) of x85 -> x85) of x84 -> x84) of x83 -> x83) of x82 -> x82) of x81 -> x81) of x80 -> x80) of x79 -> x79) of x78 -> x78) of x77 -> x77) of x76 -> x76) of x75 -> x75) of x74 -> x74) of x73 -> x73) of x72 -> x72) of x71 -> x71) of x70 -> x70) of x69 -> x69) of x68 -> x68) of x67 -> x67) of x66 -> x66) of x65 -> x65) of x64 -> x64) of x63 -> x63) of x62 -> x62) of x61 -> x61) of x60 -> x60) of x59 -> x59) of x58 -> x58) of x57 -> x57) of x56 -> x56) of x55 -> x55) of x54 -> x54) of x53 -> x53) of x52 -> x52) of x51 -> x51) of x50 -> x50) of x49 -> x49) of x48 -> x48) of x47 -> x47) of x46 -> x46) of x45 -> x45) of x44 -> x44) of x43 -> x43) of x42 -> x42) of x41 -> x41) of x40 -> x40) of x39 -> x39) of x38 -> x38) of x37 -> x37) of x36 -> x36) of x35 -> x35) of x34 -> x34) of x33 -> x33) of x32 -> x32) of x31 -> x31) of x30 -> x30) of x29 -> x29) of x28 -> x28) of x27 -> x27) of x26 -> x26) of x25 -> x25) of x24 -> x24) of x23 -> x23) of x22 -> x22) of x21 -> x21) of x20 -> x20) of x19 -> x19) of x18 -> x18) of x17 -> x17) of x16 -> x16) of x15 -> x15) of x14 -> x14) of x13 -> x13) of x12 -> x12) of x11 -> x11) of x10 -> x10) of x9 -> x9) of x8 -> x8) of x7 -> x7) of x6 -> x6) of x5 -> x5) of x4 -> x4) of x3 -> x3) of x2 -> x2) of x1 -> x1) ===================================== testsuite/tests/deSugar/should_compile/all.T ===================================== @@ -116,3 +116,7 @@ test('T22719', normal, compile, ['-ddump-simpl -dsuppress-uniques -dno-typeable- test('T23550', normal, compile, ['']) test('T24489', normal, compile, ['-O']) test('T25996', normal, compile, ['']) +test('T27383', + [collect_compiler_stats('bytes allocated', 10)], + compile, + ['-Wincomplete-patterns']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/0e7c8acaba5f930f5dffc9660636744… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/0e7c8acaba5f930f5dffc9660636744… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/T17833] Seed the simplifier's in-scope set for open expressions
by Sebastian Graf (＠sgraf812) 16 Jun '26

16 Jun '26

Sebastian Graf pushed to branch wip/T17833 at Glasgow Haskell Compiler / GHC Commits: aa05778e by Sebastian Graf at 2026-06-16T23:26:35+02:00 Seed the simplifier's in-scope set for open expressions simplifyExpr simplifies expressions typed at the GHCi prompt and the results of Template Haskell splices. Such an expression may be open: at a GHCi debugger breakpoint its free variables include RuntimeUnk skolems standing for as-yet-unknown types. The simplifier began with an in-scope set holding only the wildcard binder, so when it instantiated the unsafeCoerce# wrapper that GHCi builds around a result, it formed a substitution whose range mentioned a free skolem that was not in scope. That breaks the substitution invariant and, in a compiler built with assertions, trips substTy's sanity check. Seed the initial in-scope set with the free variables of the expression. For a closed expression this adds nothing. See Note [Seed the in-scope set for open expressions]. Fixes #17833 and its duplicate #21118. - - - - - 3 changed files: - + changelog.d/T17833 - compiler/GHC/Core/Opt/Simplify.hs - testsuite/tests/ghci.debugger/scripts/all.T Changes: ===================================== changelog.d/T17833 ===================================== @@ -0,0 +1,11 @@ +section: compiler +synopsis: Fix a Simplifier panic on open expressions (e.g. in the GHCi debugger) +issues: #17833 #21118 +mrs: !16187 +description: { + ``simplifyExpr`` now seeds the Simplifier's in-scope set with the free + variables of the expression being simplified, restoring the invariant that the + in-scope set contains all of an expression's free variables. This fixes a panic + (a failed ``substTy`` assertion) when an open expression reaches the Simplifier, + as can happen for expressions entered in the GHCi debugger. +} ===================================== compiler/GHC/Core/Opt/Simplify.hs ===================================== @@ -14,6 +14,7 @@ import GHC.Core.Rules import GHC.Core.Ppr ( pprCoreBindings, pprCoreExpr ) import GHC.Core.Opt.OccurAnal ( occurAnalysePgm, occurAnalyseExpr ) import GHC.Core.Stats ( coreBindsSize, coreBindsStats, exprSize ) +import GHC.Core.FVs ( exprFreeVars ) import GHC.Core.Utils ( mkTicks, stripTicksTop ) import GHC.Core.Lint ( LintPassResultConfig, dumpPassResult, lintPassResult ) import GHC.Core.Opt.Simplify.Iteration ( simplTopBinds, simplExpr, simplImpRules ) @@ -74,7 +75,10 @@ simplifyExpr logger euc opts expr ; let fam_envs = ( eps_fam_inst_env eps , extendFamInstEnvList emptyFamInstEnv $ se_fam_inst opts ) - simpl_env = mkSimplEnv (se_mode opts) fam_envs + -- See Note [Seed the in-scope set for open expressions] + simpl_env = setInScopeSet base_env + (getInScope base_env `extendInScopeSetSet` exprFreeVars expr) + base_env = mkSimplEnv (se_mode opts) fam_envs top_env_cfg = se_top_env_cfg opts read_eps_rules = eps_rule_base <$> eucEPS euc read_ruleenv = updExternalPackageRules emptyRuleEnv <$> read_eps_rules @@ -94,6 +98,49 @@ simplifyExpr logger euc opts expr ; return expr' } +{- Note [Seed the in-scope set for open expressions] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +An invariant of the Simplifier is that the in-scope set contains all the free +variables of the expression being simplified. Why? So that the Simplifier +never invents a "fresh" binder that accidentally captures one of those free +variables. + +'simplifyExpr' establishes that invariant by initialising the in-scope set with +the free variables of the expression it is given. For a closed expression this +adds nothing. + +But how can 'simplifyExpr' be handed an /open/ expression in the first place? +The whole-module pipeline never does so: 'simplifyPgm' only ever simplifies +closed bindings. The one caller that can is 'hscCompileCoreExpr' (in +GHC.Driver.Main.Passes), which compiles a single Core expression, and is itself +reached from just two places: + + * GHCi, for an expression typed at the prompt, and in particular in the + debugger (see below); + * Template Haskell, when running a splice (see GHC.Tc.Gen.Splice). + +So this is never on the critical path for mainstream compilation. + +Here is how the debugger hands us an open expression (test break006). We are +stopped at a breakpoint in + mymap f (x:xs) = f x : ... +where the debugger knows that 'x :: Int' but not yet the result type of 'f'. So +it invents a RuntimeUnk skolem 'a' and gives 'f :: Int -> a'. The user then +types + let y = f x +GHCi type-checks that and, to hand the result back to the interpreter (see +GHC.Tc.Module.tcGhciStmts), wraps it as + returnIO @[Any] [unsafeCoerce# @LiftedRep @LiftedRep @a @Any y] +in which the skolem 'a' is free. + +When the Simplifier instantiates that 'unsafeCoerce#', it builds a substitution +mapping unsafeCoerce#'s quantified type variables to (LiftedRep, LiftedRep, a, +Any), whose range therefore mentions 'a'. The in-scope set of a substitution +must contain the free variables of its range (see Note [The substitution +invariant] in GHC.Core.TyCo.Subst). Without the seeding above that invariant is +broken, and in a compiler built with assertions substTy's sanity check fails +(#17833, and its duplicate #21118). -} + simplExprGently :: SimplEnv -> CoreExpr -> SimplM CoreExpr -- ^ Simplifies an expression by doing occurrence analysis, then simplification, -- and repeating (twice currently), because one pass alone leaves tons of crud. ===================================== testsuite/tests/ghci.debugger/scripts/all.T ===================================== @@ -52,8 +52,7 @@ test('break002', extra_files(['../Test2.hs']), ghci_script, ['break002.script']) test('break003', extra_files(['../Test3.hs']), ghci_script, ['break003.script']) test('break005', extra_files(['../QSort.hs']), ghci_script, ['break005.script']) test('break006', - [ when(compiler_debugged(), expect_broken(17833)), - extra_files(['../Test3.hs'])], + [extra_files(['../Test3.hs'])], ghci_script, ['break006.script']) test('break007', extra_files(['Break007.hs']), ghci_script, ['break007.script']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/aa05778e10156f3b571ce429e3b3c20… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/aa05778e10156f3b571ce429e3b3c20… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/T27383] Desugar a `case` scrutinee only once (#27383, #20251)
by Sebastian Graf (＠sgraf812) 16 Jun '26

16 Jun '26

Sebastian Graf pushed to branch wip/T27383 at Glasgow Haskell Compiler / GHC Commits: 331843b8 by Sebastian Graf at 2026-06-16T23:25:26+02:00 Desugar a `case` scrutinee only once (#27383, #20251) In `dsExpr` for `HsCase` we desugared the scrutinee /twice/: once to build the Core `case` itself, and again inside `matchWrapper`, which re-desugared the source scrutinee (via `addHsScrutTmCs`) purely to record long-distance information for the pattern-match checker. For a single `case` that is merely wasteful. But for nested cases it is catastrophic. Consider case (case (case e of ... ) of ... ) of ... Desugaring the outer scrutinee desugars the middle `case` twice, each of which desugars the inner `case` twice, and so on. The work doubles at every level, so desugaring takes O(2^n) time in the nesting depth. That is the blowup reported in #27383; it is also what makes the machine-generated program in #20251 take an age to compile. The fix is simple. `matchWrapper` is handed the scrutinee anyway, so we give it the Core expression we have /already/ desugared, and record the long-distance term constraint with `addCoreScrutTmCs` instead of re-desugaring from source. This is just what `matchSinglePatVar` already does for single-pattern matches. So: * `matchWrapper` now takes `Maybe [CoreExpr]` rather than `Maybe [LHsExpr GhcTc]`. * The `HsCase` equation of `dsExpr` passes the already-desugared `core_discrim`; the arrow desugarer passes its match variables. * `addHsScrutTmCs` had no other use, so it is gone. Desugaring is now linear in the nesting depth. (The coverage checker still runs `simpleOptExpr` over each scrutinee, which leaves the total at O(n^2); that is ample.) The long-distance information itself is unchanged: the checker sees precisely the Core that backs the generated code. Test: deSugar/should_compile/T27383 - - - - - 8 changed files: - + changelog.d/fix-exponential-case-desugar-27383 - compiler/GHC/HsToCore/Arrows.hs - compiler/GHC/HsToCore/Expr.hs - compiler/GHC/HsToCore/Match.hs - compiler/GHC/HsToCore/Match.hs-boot - compiler/GHC/HsToCore/Pmc.hs - + testsuite/tests/deSugar/should_compile/T27383.hs - testsuite/tests/deSugar/should_compile/all.T Changes: ===================================== changelog.d/fix-exponential-case-desugar-27383 ===================================== @@ -0,0 +1,6 @@ +section: compiler +synopsis: Fix exponential-time desugaring of nested ``case`` expressions. + The scrutinee is no longer desugared a second time when recording + long-distance information for the pattern-match checker. +issues: #27383, #20251 +mrs: !16203 ===================================== compiler/GHC/HsToCore/Arrows.hs ===================================== @@ -553,7 +553,7 @@ dsCmd ids local_vars stack_ty res_ty in_ty = envStackType env_ids stack_ty' discrims = map nlHsVar arg_ids (discrim_vars, matching_code) - <- matchWrapper (ArrowMatchCtxt match_ctxt) (Just discrims) match' + <- matchWrapper (ArrowMatchCtxt match_ctxt) (Just (map Var arg_ids)) match' core_body <- flip (bind_vars discrim_vars) matching_code <$> traverse dsLExpr discrims ===================================== compiler/GHC/HsToCore/Expr.hs ===================================== @@ -421,7 +421,7 @@ dsExpr (HsPragE _ (HsPragSCC _ cc) expr) dsExpr (HsCase ctxt discrim matches) = do { core_discrim <- dsLExpr discrim - ; ([discrim_var], matching_code) <- matchWrapper ctxt (Just [discrim]) matches + ; ([discrim_var], matching_code) <- matchWrapper ctxt (Just [core_discrim]) matches ; return (bindNonRec discrim_var core_discrim matching_code) } -- Pepe: The binds are in scope in the body but NOT in the binding group ===================================== compiler/GHC/HsToCore/Match.hs ===================================== @@ -761,7 +761,7 @@ Call @match@ with all of this information! matchWrapper :: HsMatchContextRn -- ^ For shadowing warning messages - -> Maybe [LHsExpr GhcTc] -- ^ Scrutinee(s) + -> Maybe [CoreExpr] -- ^ Already-desugared scrutinee(s) -- see Note [matchWrapper scrutinees] -> MatchGroup GhcTc (LHsExpr GhcTc) -- ^ Matches being desugared -> DsM ([Id], CoreExpr) -- ^ Results (usually passed to 'match') @@ -818,7 +818,7 @@ matchWrapper ctxt scrs (MG { mg_alts = L _ matches -- pmc for pattern synonyms -- See Note [Long-distance information] in GHC.HsToCore.Pmc - then addHsScrutTmCs (concat scrs) new_vars $ + then addCoreScrutTmCs (concat scrs) new_vars $ pmcMatches origin (DsMatchContext ctxt locn) new_vars matches -- When we're not doing PM checks on the match group, ===================================== compiler/GHC/HsToCore/Match.hs-boot ===================================== @@ -16,7 +16,7 @@ match :: [Id] matchWrapper :: HsMatchContextRn - -> Maybe [LHsExpr GhcTc] + -> Maybe [CoreExpr] -> MatchGroup GhcTc (LHsExpr GhcTc) -> DsM ([Id], CoreExpr) ===================================== compiler/GHC/HsToCore/Pmc.hs ===================================== @@ -34,7 +34,7 @@ module GHC.HsToCore.Pmc ( isMatchContextPmChecked, isMatchContextPmChecked_SinglePat, -- See Note [Long-distance information] - addTyCs, addCoreScrutTmCs, addHsScrutTmCs, getLdiNablas, + addTyCs, addCoreScrutTmCs, getLdiNablas, getNFirstUncovered ) where @@ -59,7 +59,6 @@ import GHC.Utils.Panic import GHC.Types.Var (EvVar, Var (..)) import GHC.Types.Id.Info import GHC.Tc.Utils.TcType (evVarPred) -import {-# SOURCE #-} GHC.HsToCore.Expr (dsLExpr) import GHC.HsToCore.Monad import GHC.Data.Bag import GHC.Data.OrdList @@ -684,12 +683,6 @@ addCoreScrutTmCs (scr:scrs) (x:xs) k addPhiCtsNablas nablas (unitBag (PhiCoreCt x scr)) addCoreScrutTmCs _ _ _ = panic "addCoreScrutTmCs: numbers of scrutinees and match ids differ" --- | 'addCoreScrutTmCs', but desugars the 'LHsExpr's first. -addHsScrutTmCs :: [LHsExpr GhcTc] -> [Id] -> DsM a -> DsM a -addHsScrutTmCs scrs vars k = do - scr_es <- traverse dsLExpr scrs - addCoreScrutTmCs scr_es vars k - {- Note [Long-distance information] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Consider @@ -716,8 +709,8 @@ To achieve similar reasoning in the coverage checker: which the call sites put into place with 'GHC.HsToCore.Monad.updPmNablas'. * Call sites also extend this set with facts from type-constraint dictionaries, - case scrutinees, etc. with the exported functions 'addTyCs', 'addCoreScrutTmCs' - and 'addHsScrutTmCs'. + case scrutinees, etc. with the exported functions 'addTyCs' and + 'addCoreScrutTmCs'. Note [Recovering from unsatisfiable pattern-matching constraints] ===================================== testsuite/tests/deSugar/should_compile/T27383.hs ===================================== @@ -0,0 +1,8 @@ +-- Regression test for #27383 (and #20251): deeply nested cases whose +-- scrutinees are themselves cases must desugar in polynomial (not +-- exponential) time. With the bug, even a few dozen nestings make +-- compilation time out. +module T27383 where + +zero :: Int +zero = (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case (case 0 of x100 -> x100) of x99 -> x99) of x98 -> x98) of x97 -> x97) of x96 -> x96) of x95 -> x95) of x94 -> x94) of x93 -> x93) of x92 -> x92) of x91 -> x91) of x90 -> x90) of x89 -> x89) of x88 -> x88) of x87 -> x87) of x86 -> x86) of x85 -> x85) of x84 -> x84) of x83 -> x83) of x82 -> x82) of x81 -> x81) of x80 -> x80) of x79 -> x79) of x78 -> x78) of x77 -> x77) of x76 -> x76) of x75 -> x75) of x74 -> x74) of x73 -> x73) of x72 -> x72) of x71 -> x71) of x70 -> x70) of x69 -> x69) of x68 -> x68) of x67 -> x67) of x66 -> x66) of x65 -> x65) of x64 -> x64) of x63 -> x63) of x62 -> x62) of x61 -> x61) of x60 -> x60) of x59 -> x59) of x58 -> x58) of x57 -> x57) of x56 -> x56) of x55 -> x55) of x54 -> x54) of x53 -> x53) of x52 -> x52) of x51 -> x51) of x50 -> x50) of x49 -> x49) of x48 -> x48) of x47 -> x47) of x46 -> x46) of x45 -> x45) of x44 -> x44) of x43 -> x43) of x42 -> x42) of x41 -> x41) of x40 -> x40) of x39 -> x39) of x38 -> x38) of x37 -> x37) of x36 -> x36) of x35 -> x35) of x34 -> x34) of x33 -> x33) of x32 -> x32) of x31 -> x31) of x30 -> x30) of x29 -> x29) of x28 -> x28) of x27 -> x27) of x26 -> x26) of x25 -> x25) of x24 -> x24) of x23 -> x23) of x22 -> x22) of x21 -> x21) of x20 -> x20) of x19 -> x19) of x18 -> x18) of x17 -> x17) of x16 -> x16) of x15 -> x15) of x14 -> x14) of x13 -> x13) of x12 -> x12) of x11 -> x11) of x10 -> x10) of x9 -> x9) of x8 -> x8) of x7 -> x7) of x6 -> x6) of x5 -> x5) of x4 -> x4) of x3 -> x3) of x2 -> x2) of x1 -> x1) ===================================== testsuite/tests/deSugar/should_compile/all.T ===================================== @@ -116,3 +116,4 @@ test('T22719', normal, compile, ['-ddump-simpl -dsuppress-uniques -dno-typeable- test('T23550', normal, compile, ['']) test('T24489', normal, compile, ['-O']) test('T25996', normal, compile, ['']) +test('T27383', normal, compile, ['']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/331843b80f79c4ff5df5056a0a2f81a… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/331843b80f79c4ff5df5056a0a2f81a… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/T17833] Seed the simplifier's in-scope set for open expressions
by Sebastian Graf (＠sgraf812) 16 Jun '26

16 Jun '26

Sebastian Graf pushed to branch wip/T17833 at Glasgow Haskell Compiler / GHC Commits: 0cf50721 by Sebastian Graf at 2026-06-16T23:12:08+02:00 Seed the simplifier's in-scope set for open expressions simplifyExpr simplifies expressions typed at the GHCi prompt and the results of Template Haskell splices. Such an expression may be open: at a GHCi debugger breakpoint its free variables include RuntimeUnk skolems standing for as-yet-unknown types. The simplifier began with an in-scope set holding only the wildcard binder, so when it instantiated the unsafeCoerce# wrapper that GHCi builds around a result, it formed a substitution whose range mentioned a free skolem that was not in scope. That breaks the substitution invariant and, in a compiler built with assertions, trips substTy's sanity check. Seed the initial in-scope set with the free variables of the expression. For a closed expression this adds nothing. See Note [Seed the in-scope set for open expressions]. Fixes #17833 and its duplicate #21118. - - - - - 2 changed files: - compiler/GHC/Core/Opt/Simplify.hs - testsuite/tests/ghci.debugger/scripts/all.T Changes: ===================================== compiler/GHC/Core/Opt/Simplify.hs ===================================== @@ -14,6 +14,7 @@ import GHC.Core.Rules import GHC.Core.Ppr ( pprCoreBindings, pprCoreExpr ) import GHC.Core.Opt.OccurAnal ( occurAnalysePgm, occurAnalyseExpr ) import GHC.Core.Stats ( coreBindsSize, coreBindsStats, exprSize ) +import GHC.Core.FVs ( exprFreeVars ) import GHC.Core.Utils ( mkTicks, stripTicksTop ) import GHC.Core.Lint ( LintPassResultConfig, dumpPassResult, lintPassResult ) import GHC.Core.Opt.Simplify.Iteration ( simplTopBinds, simplExpr, simplImpRules ) @@ -74,7 +75,10 @@ simplifyExpr logger euc opts expr ; let fam_envs = ( eps_fam_inst_env eps , extendFamInstEnvList emptyFamInstEnv $ se_fam_inst opts ) - simpl_env = mkSimplEnv (se_mode opts) fam_envs + -- See Note [Seed the in-scope set for open expressions] + simpl_env = setInScopeSet base_env + (getInScope base_env `extendInScopeSetSet` exprFreeVars expr) + base_env = mkSimplEnv (se_mode opts) fam_envs top_env_cfg = se_top_env_cfg opts read_eps_rules = eps_rule_base <$> eucEPS euc read_ruleenv = updExternalPackageRules emptyRuleEnv <$> read_eps_rules @@ -94,6 +98,49 @@ simplifyExpr logger euc opts expr ; return expr' } +{- Note [Seed the in-scope set for open expressions] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +An invariant of the Simplifier is that the in-scope set contains all the free +variables of the expression being simplified. Why? So that the Simplifier +never invents a "fresh" binder that accidentally captures one of those free +variables. + +'simplifyExpr' establishes that invariant by initialising the in-scope set with +the free variables of the expression it is given. For a closed expression this +adds nothing. + +But how can 'simplifyExpr' be handed an /open/ expression in the first place? +The whole-module pipeline never does so: 'simplifyPgm' only ever simplifies +closed bindings. The one caller that can is 'hscCompileCoreExpr' (in +GHC.Driver.Main.Passes), which compiles a single Core expression, and is itself +reached from just two places: + + * GHCi, for an expression typed at the prompt, and in particular in the + debugger (see below); + * Template Haskell, when running a splice (see GHC.Tc.Gen.Splice). + +So this is never on the critical path for mainstream compilation. + +Here is how the debugger hands us an open expression (test break006). We are +stopped at a breakpoint in + mymap f (x:xs) = f x : ... +where the debugger knows that 'x :: Int' but not yet the result type of 'f'. So +it invents a RuntimeUnk skolem 'a' and gives 'f :: Int -> a'. The user then +types + let y = f x +GHCi type-checks that and, to hand the result back to the interpreter (see +GHC.Tc.Module.tcGhciStmts), wraps it as + returnIO @[Any] [unsafeCoerce# @LiftedRep @LiftedRep @a @Any y] +in which the skolem 'a' is free. + +When the Simplifier instantiates that 'unsafeCoerce#', it builds a substitution +mapping unsafeCoerce#'s quantified type variables to (LiftedRep, LiftedRep, a, +Any), whose range therefore mentions 'a'. The in-scope set of a substitution +must contain the free variables of its range (see Note [The substitution +invariant] in GHC.Core.TyCo.Subst). Without the seeding above that invariant is +broken, and in a compiler built with assertions substTy's sanity check fails +(#17833, and its duplicate #21118). -} + simplExprGently :: SimplEnv -> CoreExpr -> SimplM CoreExpr -- ^ Simplifies an expression by doing occurrence analysis, then simplification, -- and repeating (twice currently), because one pass alone leaves tons of crud. ===================================== testsuite/tests/ghci.debugger/scripts/all.T ===================================== @@ -52,8 +52,7 @@ test('break002', extra_files(['../Test2.hs']), ghci_script, ['break002.script']) test('break003', extra_files(['../Test3.hs']), ghci_script, ['break003.script']) test('break005', extra_files(['../QSort.hs']), ghci_script, ['break005.script']) test('break006', - [ when(compiler_debugged(), expect_broken(17833)), - extra_files(['../Test3.hs'])], + [extra_files(['../Test3.hs'])], ghci_script, ['break006.script']) test('break007', extra_files(['Break007.hs']), ghci_script, ['break007.script']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/0cf50721a9c5fc4e67234b66bb6de39… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/0cf50721a9c5fc4e67234b66bb6de39… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/az/exactprint-annotation-rationalisation] 19 commits: Implement CLC proposal #378
by Alan Zimmerman (＠alanz) 16 Jun '26

16 Jun '26

Alan Zimmerman pushed to branch wip/az/exactprint-annotation-rationalisation at Glasgow Haskell Compiler / GHC Commits: bab37cc6 by konsumlamm at 2026-06-13T19:10:21+02:00 Implement CLC proposal #378 Add `Data.Double` and `Data.Float` modules Document that GHC uses IEEE 754 - - - - - fb5246ad by fendor at 2026-06-15T18:07:23-04:00 Drop `preloadClosure` from `UnitState` It is always hard-coded to the same value. Backpack Unit instantiation isn't using it any more. Allows us to simplify the API and get rid of `improveUnit`. - - - - - 291ce3aa by ARATA Mizuki at 2026-06-15T18:08:26-04:00 RISC-V NCG: Zero-extend the result of castFloatToWord32 According to the ISA manual, FMV.X.W sign-extends the result. We need to truncate the result to avoid creating an exotic Word32 value. Fixes #27300 - - - - - 011be91f by ARATA Mizuki at 2026-06-15T18:08:26-04:00 RISC-V NCG: Treat d28-d31 (ft8-ft11) as caller-saved According to the calling convention, the registers d28-d31 (ft8-ft11) are caller-saved. Fixes #27306 - - - - - e8a54713 by ARATA Mizuki at 2026-06-15T18:08:26-04:00 RISC-V NCG: Set rounding mode when emitting `truncate` If we omit the rounding mode for `fcvt`, `dyn` will be used. We do not want that for `truncate`, so we set `rtz`. In other places, we set `rne` because we do not use the dynamic rounding mode. Fixes #27303 - - - - - 9438bec7 by Zubin Duggal at 2026-06-15T18:09:11-04:00 rts: fix validate build with gcc 16. `__attribute__((regparm(1)))` is ignored on x86_64 and now gcc warns that it is ignored: rts/sm/Evac.h:35:1: error: error: ‘regparm’ attribute ignored [-Werror=attributes] See https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=ccead81bbc39668376eb5cf47066a… Fixes #27366 - - - - - 893e6133 by Andrew Lelechenko at 2026-06-15T23:55:36+01:00 base: more NonEmpty zips CLC proposal: https://github.com/haskell/core-libraries-committee/issues/409 - - - - - 1314f2fd by David Eichmann at 2026-06-16T05:46:52-04:00 Hadrian: fix ghc-internal .def file name - - - - - 7f72bcb3 by mangoiv at 2026-06-16T05:47:39-04:00 compiler: ignore camelCase and Eta reduce hlint hints These do not cohere with the style used in GHC. After disabling them, hlint lints are much less noisy again. - - - - - 842bef9f by Alan Zimmerman at 2026-06-16T05:48:25-04:00 EPA: Use standard type family declaration for Anno - - - - - b6d7b3dd by Alan Zimmerman at 2026-06-16T20:13:36+01:00 TTG: Add extension points to HsConDetails Extend HsConDetails as data HsConDetails p arg rec = PrefixCon !(XPrefixCon p) [arg] -- C @t1 @t2 p1 p2 p3 | RecCon !(XRecCon p) rec -- C { x = p1, y = p2 } | InfixCon !(XInfixCon p) arg arg -- p1 `C` p2 | XHsConDetails !(XXHsConDetails p) type family XPrefixCon p type family XRecCon p type family XInfixCon p type family XXHsConDetails p - - - - - 3cb48ae7 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: remove LocatedLI / SrcSpanAnnLI These were used for module export lists, and import decl lists. Replace them with direct capture of the relevant EP Annotations in HsModule and ImportDecl annotation extension points. Also use the HsConDetails RecCon extension point to capture the braces in a record constructor - - - - - c91d8993 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: Remove LocatedC / SrcSpanAnnC Used for contexts - - - - - 3e87ac98 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: Harmonise HsQual/HsQualTy TTG extension annotations - - - - - d0b778c7 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA Remove LocatedLC / LocatedLS LocatedLC/LocatedLS were unused - - - - - efa5ed6a by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: Remove LocatedLW from LStmtLR - - - - - 29c665a0 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: Remove LocatedLW from MatchGroup This is the last usage of LocatedLW / SrcSpanAnnLW - - - - - 8d7a6cc6 by Alan Zimmerman at 2026-06-16T20:15:17+01:00 EPA: Move the 'where' annotation for PatSynBind This allows us to move it out of the MatchGroup exact print annotation too - - - - - 140d9655 by Alan Zimmerman at 2026-06-16T22:09:52+01:00 EPA: Replace AnnListItem with simply [TrailingAnn] Remove the unnecessary wrapper around a single field. - - - - - 133 changed files: - + changelog.d/T27308 - compiler/.hlint.yaml - compiler/GHC/CmmToAsm/RV64/CodeGen.hs - compiler/GHC/CmmToAsm/RV64/Instr.hs - compiler/GHC/CmmToAsm/RV64/Ppr.hs - compiler/GHC/CmmToAsm/RV64/Regs.hs - compiler/GHC/Data/BooleanFormula.hs - compiler/GHC/Data/List/NonEmpty.hs - compiler/GHC/Driver/Backpack.hs - compiler/GHC/Hs.hs - compiler/GHC/Hs/Binds.hs - compiler/GHC/Hs/Decls.hs - compiler/GHC/Hs/Dump.hs - compiler/GHC/Hs/Expr.hs - compiler/GHC/Hs/ImpExp.hs - compiler/GHC/Hs/Instances.hs - compiler/GHC/Hs/Pat.hs - compiler/GHC/Hs/Stats.hs - compiler/GHC/Hs/Type.hs - compiler/GHC/Hs/Utils.hs - compiler/GHC/HsToCore/Docs.hs - compiler/GHC/HsToCore/Match.hs - compiler/GHC/HsToCore/Match/Constructor.hs - compiler/GHC/HsToCore/Pmc/Desugar.hs - compiler/GHC/HsToCore/Quote.hs - compiler/GHC/HsToCore/Ticks.hs - compiler/GHC/Iface/Ext/Ast.hs - compiler/GHC/Iface/Load.hs - compiler/GHC/Iface/Recomp.hs - compiler/GHC/Parser.y - compiler/GHC/Parser/Annotation.hs - compiler/GHC/Parser/PostProcess.hs - compiler/GHC/Parser/PostProcess/Haddock.hs - compiler/GHC/Parser/Types.hs - compiler/GHC/Rename/Bind.hs - compiler/GHC/Rename/Expr.hs - compiler/GHC/Rename/HsType.hs - compiler/GHC/Rename/Module.hs - compiler/GHC/Rename/Names.hs - compiler/GHC/Rename/Pat.hs - compiler/GHC/Rename/Utils.hs - compiler/GHC/Tc/Deriv.hs - compiler/GHC/Tc/Deriv/Generate.hs - compiler/GHC/Tc/Gen/App.hs - compiler/GHC/Tc/Gen/Arrow.hs - compiler/GHC/Tc/Gen/Do.hs - compiler/GHC/Tc/Gen/Export.hs - compiler/GHC/Tc/Gen/Expr.hs - compiler/GHC/Tc/Gen/Head.hs - compiler/GHC/Tc/Gen/HsType.hs - compiler/GHC/Tc/Gen/Match.hs - compiler/GHC/Tc/Gen/Pat.hs - compiler/GHC/Tc/Gen/Sig.hs - compiler/GHC/Tc/Module.hs - compiler/GHC/Tc/TyCl.hs - compiler/GHC/Tc/TyCl/PatSyn.hs - compiler/GHC/Tc/TyCl/Utils.hs - compiler/GHC/Tc/Utils/Backpack.hs - compiler/GHC/Tc/Zonk/Type.hs - compiler/GHC/ThToHs.hs - compiler/GHC/Unit.hs - compiler/GHC/Unit/State.hs - compiler/GHC/Unit/Types.hs - compiler/Language/Haskell/Syntax.hs - compiler/Language/Haskell/Syntax/Binds.hs - compiler/Language/Haskell/Syntax/Decls.hs - compiler/Language/Haskell/Syntax/Expr.hs - compiler/Language/Haskell/Syntax/Extension.hs - compiler/Language/Haskell/Syntax/ImpExp.hs - compiler/Language/Haskell/Syntax/Pat.hs - compiler/Language/Haskell/Syntax/Type.hs - docs/users_guide/bugs.rst - ghc/GHCi/UI.hs - hadrian/src/Rules/Rts.hs - libraries/base/base.cabal.in - libraries/base/changelog.md - + libraries/base/src/Data/Double.hs - + libraries/base/src/Data/Float.hs - libraries/base/src/Data/List/NonEmpty.hs - libraries/ghc-internal/src/GHC/Internal/Float.hs - rts/sm/Evac.h - testsuite/tests/codeGen/should_run/T16617.hs - testsuite/tests/codeGen/should_run/T16617.stdout - testsuite/tests/ghc-api/T25121_status.stdout - testsuite/tests/ghc-api/exactprint/T22919.stderr - testsuite/tests/ghc-api/exactprint/Test20239.stderr - testsuite/tests/ghc-api/exactprint/ZeroWidthSemi.stderr - testsuite/tests/haddock/should_compile_flag_haddock/T17544.stderr - testsuite/tests/haddock/should_compile_flag_haddock/T17544_kw.stderr - testsuite/tests/haddock/should_compile_flag_haddock/T24221.stderr - testsuite/tests/interface-stability/base-exports.stdout - testsuite/tests/interface-stability/base-exports.stdout-javascript-unknown-ghcjs - testsuite/tests/interface-stability/base-exports.stdout-mingw32 - testsuite/tests/module/mod185.stderr - testsuite/tests/parser/should_compile/DumpParsedAst.stderr - testsuite/tests/parser/should_compile/DumpParsedAstComments.stderr - testsuite/tests/parser/should_compile/DumpRenamedAst.stderr - testsuite/tests/parser/should_compile/DumpSemis.stderr - testsuite/tests/parser/should_compile/DumpTypecheckedAst.stderr - testsuite/tests/parser/should_compile/KindSigs.stderr - testsuite/tests/parser/should_compile/T14189.stderr - testsuite/tests/parser/should_compile/T15279.stderr - testsuite/tests/parser/should_compile/T15323.stderr - testsuite/tests/parser/should_compile/T20452.stderr - testsuite/tests/parser/should_compile/T20718.stderr - testsuite/tests/parser/should_compile/T20718b.stderr - testsuite/tests/parser/should_compile/T20846.stderr - testsuite/tests/parser/should_compile/T23315/T23315.stderr - testsuite/tests/parser/should_compile/all.T - testsuite/tests/printer/AnnotationNoListTuplePuns.stdout - testsuite/tests/printer/Makefile - testsuite/tests/printer/T18791.stderr - testsuite/tests/printer/Test20297.stdout - testsuite/tests/printer/Test24533.stdout - testsuite/tests/typecheck/should_compile/T15242.stderr - testsuite/tests/typecheck/should_compile/all.T - utils/check-exact/ExactPrint.hs - utils/check-exact/Main.hs - utils/check-exact/Transform.hs - utils/check-exact/Utils.hs - utils/haddock/haddock-api/src/Haddock/Backends/Hoogle.hs - utils/haddock/haddock-api/src/Haddock/Backends/LaTeX.hs - utils/haddock/haddock-api/src/Haddock/Backends/Xhtml/Decl.hs - utils/haddock/haddock-api/src/Haddock/Convert.hs - utils/haddock/haddock-api/src/Haddock/GhcUtils.hs - utils/haddock/haddock-api/src/Haddock/Interface/Create.hs - utils/haddock/haddock-api/src/Haddock/Interface/LexParseRn.hs - utils/haddock/haddock-api/src/Haddock/Interface/Rename.hs - utils/haddock/haddock-api/src/Haddock/Interface/RenameType.hs - utils/haddock/haddock-api/src/Haddock/Types.hs - utils/haddock/html-test/ref/Hash.html - utils/haddock/html-test/ref/Test.html - utils/haddock/html-test/ref/TypeFamilies3.html The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/79c8e7d059e35a1aad102e997b91ad… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/79c8e7d059e35a1aad102e997b91ad… You're receiving this email because of your account on gitlab.haskell.org.

1 0