| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| lfortran-0.69.0.tar.gz | 2026-11-10 | 5.1 MB | |
| README.md | 2026-11-10 | 57.3 kB | |
| v0.69.0 source code.tar.gz | 2026-11-10 | 8.4 MB | |
| v0.69.0 source code.zip | 2026-11-10 | 13.6 MB | |
| Totals: 4 Items | 27.2 MB | 11 | |
What's Changed
- fix: provide unsigned binary128 conversion builtins in the runtime by @certik in https://github.com/lfortran/lfortran/pull/15969
- skills: add make-release skill by @certik in https://github.com/lfortran/lfortran/pull/15966
- integration_tests: add run_tests.py --exclude to skip tests at configure time by @certik in https://github.com/lfortran/lfortran/pull/15972
- ci: skip real128_int_to_real_02 on LLVM 7 and 8 by @certik in https://github.com/lfortran/lfortran/pull/15976
- fix: stop a GPU program when the runtime finds no device by @certik in https://github.com/lfortran/lfortran/pull/15979
- Fix: reject subscripts on function references (#15973) by @usefahmed07 in https://github.com/lfortran/lfortran/pull/15978
- fix: import a private specific of a generic under the generic's name by @assem2002 in https://github.com/lfortran/lfortran/pull/15323
- Fix: count only CLOCK_MONOTONIC calls in test_system_clock (#15977) by @usefahmed07 in https://github.com/lfortran/lfortran/pull/15980
- fix: create temporaries for zero-size array expressions by @chaitanya-bhardwaj in https://github.com/lfortran/lfortran/pull/15122
- fix: create section temporaries for complex comparison operands by @certik in https://github.com/lfortran/lfortran/pull/15987
- test: implied-do loop in a list-directed internal read with iostat by @certik in https://github.com/lfortran/lfortran/pull/15986
- test: merge of derived-type values inside an elemental function by @certik in https://github.com/lfortran/lfortran/pull/15991
- test: list-directed internal read into an allocatable character array dummy by @certik in https://github.com/lfortran/lfortran/pull/15985
- test: write an implied-do list of trimmed assumed-shape elements in one record by @certik in https://github.com/lfortran/lfortran/pull/15984
- test: cover the unallocated-LHS hint for internal function results by @certik in https://github.com/lfortran/lfortran/pull/15983
- fix: allow omitting default-initialized derived-type components in structure constructors by @certik in https://github.com/lfortran/lfortran/pull/15989
- fix: take the shape of elemental overloaded operators from the array operand by @certik in https://github.com/lfortran/lfortran/pull/16022
- fix: copy in non-contiguous sections of fixed-size arrays by @certik in https://github.com/lfortran/lfortran/pull/15990
- fix: resolve type-bound generic subroutines mixing pass and nopass specifics by @schizophrenicmaniac in https://github.com/lfortran/lfortran/pull/15861
- refactor: capture a cell's stdout and stderr by @certik in https://github.com/lfortran/lfortran/pull/16027
- fix: evaluate matmul of constant arrays at compile time by @certik in https://github.com/lfortran/lfortran/pull/15993
- feat: serve a session with a built-in Jupyter kernel by @certik in https://github.com/lfortran/lfortran/pull/16030
- feat: make the interactive prompt a Jupyter client by @certik in https://github.com/lfortran/lfortran/pull/16032
- Ignore trailing blank spaces in the file name of inquire by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16016
- fix: fold kind-16 real constants in implicit casts, PARAMETER statements and unary minus by @certik in https://github.com/lfortran/lfortran/pull/16002
- feat: share cell execution with the Xeus kernel by @certik in https://github.com/lfortran/lfortran/pull/16033
- fix: keep random_number results in [0, 1) by @certik in https://github.com/lfortran/lfortran/pull/16075
- fix: give explicit-shape arrays a zero extent when ub < lb by @certik in https://github.com/lfortran/lfortran/pull/16026
- ci: skip real128 cast execution on LLVM 7 and 8 by @certik in https://github.com/lfortran/lfortran/pull/16088
- Keep an array's physical type in duplicate_type (#15839 item 5) by @certik in https://github.com/lfortran/lfortran/pull/16080
- Query LLVM types by declaration instead of building a Var (#15839 item 4, part 1) by @certik in https://github.com/lfortran/lfortran/pull/16041
- fix: SHAPE of explicit-shape arrays after a bound is redefined; test entry setup by @certik in https://github.com/lfortran/lfortran/pull/16092
- fix: keep statements hoisted in a procedure with ENTRY in front of their statement by @certik in https://github.com/lfortran/lfortran/pull/16089
- fix: fold integer to real(16) conversions in constant expressions by @certik in https://github.com/lfortran/lfortran/pull/16048
- fix: keep symbols when mangled names reuse names of the same scope by @certik in https://github.com/lfortran/lfortran/pull/16134
- fix: shape check of array sections with leading scalar subscripts by @certik in https://github.com/lfortran/lfortran/pull/16114
- Choose the physical type of a string array in make_Array_t_util (#15839 item 5, part 2) by @certik in https://github.com/lfortran/lfortran/pull/16083
- fix: keep the descriptor of a SAVE allocatable or pointer array static by @certik in https://github.com/lfortran/lfortran/pull/16110
- Stop building throwaway ASR types in the LLVM backend (#15839 item 4, part 2) by @certik in https://github.com/lfortran/lfortran/pull/16044
- fix: fmt duplicate DO labels and fixed-form options by @certik in https://github.com/lfortran/lfortran/pull/16094
- fix: store array constants from the lower bounds of the variable by @certik in https://github.com/lfortran/lfortran/pull/16107
- fix: link intrinsic module procedures defined in several units by @certik in https://github.com/lfortran/lfortran/pull/16139
- fix: give c_loc of a rank-1 assumed-size array a descriptor pointer type by @certik in https://github.com/lfortran/lfortran/pull/16140
- fix: do not load through a c_ptr value in c_associated by @certik in https://github.com/lfortran/lfortran/pull/16148
- feat: end the kernel when the prompt that started it exits by @certik in https://github.com/lfortran/lfortran/pull/16037
- fix: set DEALLOCATE STAT= to zero on success by @certik in https://github.com/lfortran/lfortran/pull/16111
- perf: accumulate reductions over elementwise expressions without a temporary by @certik in https://github.com/lfortran/lfortran/pull/16141
- refactor: fix explicit-shape bounds on entry right after declarations (closer to Flang) by @certik in https://github.com/lfortran/lfortran/pull/16115
- fix: preserve allocatable module arrays across interactive cells by @certik in https://github.com/lfortran/lfortran/pull/16224
- fix: support assumed-rank c_sizeof and clarify intrinsic diagnostics by @certik in https://github.com/lfortran/lfortran/pull/16226
- fix: forward Emscripten settings to the driver by @certik in https://github.com/lfortran/lfortran/pull/16268
- fix: implement FORMAT leading-zero controls by @certik in https://github.com/lfortran/lfortran/pull/16278
- fix: choose the physical type of an array declared by a dimension statement by @certik in https://github.com/lfortran/lfortran/pull/16171
- fix: allow ALLOCATE of an already associated pointer by @certik in https://github.com/lfortran/lfortran/pull/16217
- fix: honor ALLOCATE SOURCE= and MOLD= followed by other alloc-opts by @certik in https://github.com/lfortran/lfortran/pull/16237
- fix: name interfaces of external procedures by their global name by @certik in https://github.com/lfortran/lfortran/pull/16165
- fix: give fixed-length character components static storage of their length by @certik in https://github.com/lfortran/lfortran/pull/16158
- fix: let EXTERNAL hide a host-associated generic interface by @certik in https://github.com/lfortran/lfortran/pull/16238
- perf: speed up matrix-vector and vector-matrix matmul by @certik in https://github.com/lfortran/lfortran/pull/16166
- fix: lower unary minus of complex and .not. in accumulated reductions by @certik in https://github.com/lfortran/lfortran/pull/16249
- test: check exit diagnostics on the flushing built-in kernel by @certik in https://github.com/lfortran/lfortran/pull/16293
- fix: persist implicit none across interactive cells by @certik in https://github.com/lfortran/lfortran/pull/16251
- fix: SIZE of assumed-size arrays (valid DIM and diagnostics) by @certik in https://github.com/lfortran/lfortran/pull/16236
- fix: same-file external procedure calls with implicit argument casting and from modules by @certik in https://github.com/lfortran/lfortran/pull/16172
- fix: eliminate safe elemental intrinsic argument temporaries by @certik in https://github.com/lfortran/lfortran/pull/16296
- fix: parse allocate statements in interactive cells by @certik in https://github.com/lfortran/lfortran/pull/16294
- Pass declarations to LLVM helpers instead of a Var (#15839 item 4, part 3) by @certik in https://github.com/lfortran/lfortran/pull/16053
- perf: avoid temporaries for identical array sections in elementwise assignments by @certik in https://github.com/lfortran/lfortran/pull/16253
- fix: reject selected_real_kind calls without arguments by @certik in https://github.com/lfortran/lfortran/pull/16229
- perf: add conservative LLVM alias information under --fast by @certik in https://github.com/lfortran/lfortran/pull/16298
- fix: refresh shared-session output while editing the prompt by @certik in https://github.com/lfortran/lfortran/pull/16304
- fix: remove safe component operand copies and preserve result alias safety by @certik in https://github.com/lfortran/lfortran/pull/16299
- fix: stream native cell output during execution by @certik in https://github.com/lfortran/lfortran/pull/16309
- refactor: merge explicit-shape entry-bound passes by @certik in https://github.com/lfortran/lfortran/pull/16295
- fix: validate constant character lengths in declaration contexts by @certik in https://github.com/lfortran/lfortran/pull/16271
- fix: validate INQUIRE DELIM output destinations by @certik in https://github.com/lfortran/lfortran/pull/16230
- fix: reject associated variable assignments in pure procedures by @certik in https://github.com/lfortran/lfortran/pull/16252
- fix: isolate WASM evaluator symbol namespaces by @certik in https://github.com/lfortran/lfortran/pull/16333
- fix: allocate derived-copy loop counters in the entry block by @certik in https://github.com/lfortran/lfortran/pull/16334
- fix: emit semicolon for module PARAMETER in C++ backend by @jschueller in https://github.com/lfortran/lfortran/pull/16064
- fix: support dummy procedures in C/C++ backends by @jschueller in https://github.com/lfortran/lfortran/pull/16058
- fix: allow a SAVE statement before the type declaration by @assem2002 in https://github.com/lfortran/lfortran/pull/15705
- test: regression test for extending a COMMON block from a module by @atharvaajmera in https://github.com/lfortran/lfortran/pull/14489
- fix: use the comparison's logical kind for an integer comparison temporary by @schizophrenicmaniac in https://github.com/lfortran/lfortran/pull/15258
- fix: build and test on GCC 15 (ubuntu-latest 26.04) by @certik in https://github.com/lfortran/lfortran/pull/16373
- test: initialize the strings compared in string_37 by @certik in https://github.com/lfortran/lfortran/pull/16384
- fix: use a constant DO loop step in C/C++ only when it is an integer by @certik in https://github.com/lfortran/lfortran/pull/16385
- fix: isolate OpenCoarrays reference toolchain with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16374
- fix: placement-new array members of generated LSP unions by @certik in https://github.com/lfortran/lfortran/pull/16394
- test: avoid using undefined units after failed OPEN by @certik in https://github.com/lfortran/lfortran/pull/16399
- fix: locate intent-out array finalizer calls in debug info by @certik in https://github.com/lfortran/lfortran/pull/16320
- build: provide native CMake LFortran support in Pixi LLVM 11 by @certik in https://github.com/lfortran/lfortran/pull/16318
- fix: preserve bounds in descriptor-to-data array copies by @certik in https://github.com/lfortran/lfortran/pull/16327
- perf: reuse private array function result storage by @certik in https://github.com/lfortran/lfortran/pull/16322
- perf: index safe multidimensional in-place sections through their bases by @certik in https://github.com/lfortran/lfortran/pull/16325
- perf: use statement-local stack storage for numeric temporaries by @certik in https://github.com/lfortran/lfortran/pull/16328
- perf: give eligible constructor temporaries statement lifetimes by @certik in https://github.com/lfortran/lfortran/pull/16339
- fix: set current scope in C backend visit_Program by @certik in https://github.com/lfortran/lfortran/pull/16402
- perf: avoid eager allocator access under native fast mode by @certik in https://github.com/lfortran/lfortran/pull/16338
- perf: simplify fast rank-one SIZE expressions by @certik in https://github.com/lfortran/lfortran/pull/16352
- Verify struct signatures and read them in the LLVM backend (#15839 item 4, part 4) by @certik in https://github.com/lfortran/lfortran/pull/16069
- perf: reuse safe repeated pure array calls in fast assignments by @certik in https://github.com/lfortran/lfortran/pull/16356
- perf: expose owned component aliasing and hoist descriptors after loop unswitching by @certik in https://github.com/lfortran/lfortran/pull/16363
- fix: exclude the reduced dimension from COUNT/PARITY/NORM2 output loops by @certik in https://github.com/lfortran/lfortran/pull/16408
- fix: remove loop_vectorise, which miscompiled --fast C output by @certik in https://github.com/lfortran/lfortran/pull/16409
- ci: compile macOS C test files with $CC by @certik in https://github.com/lfortran/lfortran/pull/16419
- build: pin all Pixi dependencies to exact versions by @certik in https://github.com/lfortran/lfortran/pull/16418
- fix: compile the Metal runtime with the linker path's clang by @certik in https://github.com/lfortran/lfortran/pull/16420
- build: add the CI platform build tools to the Pixi environments by @certik in https://github.com/lfortran/lfortran/pull/16422
- build: take build1.sh's source directory from LFORTRAN_SOURCE_DIR by @certik in https://github.com/lfortran/lfortran/pull/16421
- perf: reuse contiguous array specializations in descriptor entries by @certik in https://github.com/lfortran/lfortran/pull/16343
- Keep assumed length on ClassToIntrinsic cast in character type guard by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16341
- Evaluate the kind of an allocate type-spec instead of reading only literals by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16438
- test: require LLVM 18 for the parent-loop descriptor hoisting check by @certik in https://github.com/lfortran/lfortran/pull/16428
- perf: keep loops rolled under --fast instead of runtime unrolling by @certik in https://github.com/lfortran/lfortran/pull/16431
- fix: disable upper-bound loop unrolling on LLVM 7 and 8 by @certik in https://github.com/lfortran/lfortran/pull/16447
- fix: support logical operators in DO CONCURRENT REDUCE by @certik in https://github.com/lfortran/lfortran/pull/16439
- fix: load a procedure pointer actual argument once by @certik in https://github.com/lfortran/lfortran/pull/16413
- refactor: make normalized DO-loop heads an ASR invariant by @certik in https://github.com/lfortran/lfortran/pull/16410
- gpu: emit C for loops for counted DO loops of device code by @certik in https://github.com/lfortran/lfortran/pull/16411
- llvm: lower DO loops natively and stop running do_loops (#16398 Step 3) by @certik in https://github.com/lfortran/lfortran/pull/16416
- c, cpp: write DO loops by the C-family rules (#16398 Step 4) by @certik in https://github.com/lfortran/lfortran/pull/16469
- wasm: lower DO loops natively and stop running do_loops (#16398 Step 5) by @certik in https://github.com/lfortran/lfortran/pull/16470
- transform_optional_argument_functions: pass the host's presence flag for a host-associated optional dummy by @certik in https://github.com/lfortran/lfortran/pull/16458
- openmp: match call arguments with the dummies after replacing program arrays by @certik in https://github.com/lfortran/lfortran/pull/16459
- ASR nodes for host frames: Reference, BindReference, HostAssociated by @certik in https://github.com/lfortran/lfortran/pull/16461
- array_struct_temporary: match arguments replaced by temporaries with the dummies by @certik in https://github.com/lfortran/lfortran/pull/16460
- mlir: lower DO loops natively and stop running do_loops (#16398 Step 6) by @certik in https://github.com/lfortran/lfortran/pull/16472
- Skip the do loop head type check on unresolved externals by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16476
- x86: lower DO loops natively and stop running do_loops (#16398 Step 7) by @certik in https://github.com/lfortran/lfortran/pull/16477
- ci: split ci/build.sh into ci/source_tarball.sh and ci/check_build.sh by @certik in https://github.com/lfortran/lfortran/pull/16423
- nested_vars: give internal procedures the host instance through a frame by @certik in https://github.com/lfortran/lfortran/pull/16462
- fix: preserve array section base pointers in fast device lowering by @certik in https://github.com/lfortran/lfortran/pull/16457
- Emit DoLoop in loop_unroll and intent_out_deallocate (#16398 Step 8) by @certik in https://github.com/lfortran/lfortran/pull/16481
- build: run build0.sh with Bash when the Prefix shell is not installed by @certik in https://github.com/lfortran/lfortran/pull/16480
- ci: build the platform compilers with Pixi by @certik in https://github.com/lfortran/lfortran/pull/15865
- nested_vars: call internal procedures used as values through thunks with a frame slot by @certik in https://github.com/lfortran/lfortran/pull/16463
- Delete the do_loops pass and replace_doloop (#16398 Step 9) by @certik in https://github.com/lfortran/lfortran/pull/16482
- nested_vars: keep the declared bounds of assumed-size arrays of the host by @certik in https://github.com/lfortran/lfortran/pull/16464
- nested_vars: reach assumed-size arrays of derived type of the host by @certik in https://github.com/lfortran/lfortran/pull/16466
- nested_vars: reach assumed-size character arrays of the host by @certik in https://github.com/lfortran/lfortran/pull/16465
- Remove --use-loop-variable-after-loop (#16398 follow-up 1) by @certik in https://github.com/lfortran/lfortran/pull/16487
- nested_vars: procedure values of internal procedures through thunk pools by @certik in https://github.com/lfortran/lfortran/pull/16417
- ci: use setup-pixi v0.11.3 and Pixi v0.83.0 everywhere by @certik in https://github.com/lfortran/lfortran/pull/16483
- ci: use setup-micromamba v3.2.1 and micromamba 2.10.0 everywhere by @certik in https://github.com/lfortran/lfortran/pull/16484
- nested_vars: default to 8 thunks per internal procedure by @certik in https://github.com/lfortran/lfortran/pull/16490
- perf: mark signed integer arithmetic nsw under --fast by @certik in https://github.com/lfortran/lfortran/pull/16493
- build: link lfortran_lib with the threads library by @certik in https://github.com/lfortran/lfortran/pull/16494
- pixi: add fpm and the third-party application environments by @certik in https://github.com/lfortran/lfortran/pull/16496
- pixi: add an llvm19-wasm environment for the LLVM WASM backends by @certik in https://github.com/lfortran/lfortran/pull/16499
- pixi: add a nollvm environment for builds without LLVM by @certik in https://github.com/lfortran/lfortran/pull/16495
- Suggest not to use the DO loop variable after the loop (#16398 follow-up 2) by @certik in https://github.com/lfortran/lfortran/pull/16488
- gpu_offload_preflight: build the extent scope once per symbol table by @certik in https://github.com/lfortran/lfortran/pull/16502
- ci: name the micromamba environment in the out-of-source build by @certik in https://github.com/lfortran/lfortran/pull/16546
- remove the loop_unroll pass by @certik in https://github.com/lfortran/lfortran/pull/16505
- intrinsic_function: give the PARITY result argument empty extents by @certik in https://github.com/lfortran/lfortran/pull/16504
- nested_vars: do not finalize the result a thunk hands to its caller by @certik in https://github.com/lfortran/lfortran/pull/16500
- asr_verify: check host-frame invariants after nested_vars by @certik in https://github.com/lfortran/lfortran/pull/16501
- fix: allocate stack automatic arrays as elements with LLVM >= 15 by @certik in https://github.com/lfortran/lfortran/pull/16510
- test: check the kind of size(x, dim) of an explicit-shape dummy by @certik in https://github.com/lfortran/lfortran/pull/16524
- pixi: link with GCC in the LLVM 7 and 8 environments by @certik in https://github.com/lfortran/lfortran/pull/16518
- pixi: build the WASM targets with Pixi's tools only by @certik in https://github.com/lfortran/lfortran/pull/16534
- pixi: provide the LSP tests' dependencies and an lsp_tests task by @certik in https://github.com/lfortran/lfortran/pull/16535
- fix: keep elemental defined assignment to an array section in place by @certik in https://github.com/lfortran/lfortran/pull/16528
- fix: subtract the lower bound before narrowing a GPU subscript by @certik in https://github.com/lfortran/lfortran/pull/16530
- pixi: add a docs environment by @certik in https://github.com/lfortran/lfortran/pull/16533
- Keep sequence associated actuals with function-call subscripts in place by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16479
- fix: lower array MOLD= in ALLOCATE as a mold, not as SOURCE= by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16567
- Allow an element of a contiguous ASSOCIATE name as an element sequence by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16019
- fix: number GPU offload iterations in integer(8) by @certik in https://github.com/lfortran/lfortran/pull/16532
- c, cpp: lower host frames (Reference, BindReference, HostAssociated) by @certik in https://github.com/lfortran/lfortran/pull/16506
- Add --fast pass that clones a callee into the host for internal procedure arguments by @certik in https://github.com/lfortran/lfortran/pull/16512
- fix: deep copy arrays of a type with an allocatable array component of its own type by @certik in https://github.com/lfortran/lfortran/pull/16526
- fix: load allocatable/pointer component sources in allocate(source=) by @certik in https://github.com/lfortran/lfortran/pull/16529
- ci: run the no-LLVM and MLIR jobs with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16579
- fix: pass non-pointer target arrays to pointer, intent(in) dummies by @certik in https://github.com/lfortran/lfortran/pull/16583
- test: keep ILP64 file fixtures local to each test worktree by @certik in https://github.com/lfortran/lfortran/pull/16548
- fix: preserve signed integer kinds when folding IBSET by @certik in https://github.com/lfortran/lfortran/pull/16551
- fix: preserve INQUIRE SIZE kinds and execution-local conversions by @certik in https://github.com/lfortran/lfortran/pull/16552
- fix: bound runtime RESHAPE temporary lifetime by @certik in https://github.com/lfortran/lfortran/pull/16553
- fix: discover GPU helper calls in nested BLOCK scopes by @certik in https://github.com/lfortran/lfortran/pull/16558
- Write back CFI descriptor for character arrays after bind(C) calls by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16433
- fix: lower scalar LOG to native GPU operations by @certik in https://github.com/lfortran/lfortran/pull/16559
- build: run build0.sh generators with the Python found by CMake by @certik in https://github.com/lfortran/lfortran/pull/16589
- fix: forward possibly absent optional actuals to pointer, intent(in) dummies of functions by @certik in https://github.com/lfortran/lfortran/pull/16598
- fix: skip intent(out) initialization of absent optional dummies by @certik in https://github.com/lfortran/lfortran/pull/16602
- semantics: keep the implicit interface of a procedure pointer and type each call at its call site by @certik in https://github.com/lfortran/lfortran/pull/16557
- fix: finalize host locals before giving back the frame slot by @certik in https://github.com/lfortran/lfortran/pull/16576
- fix: parse line tables without blank-line separators by @certik in https://github.com/lfortran/lfortran/pull/16590
- test: initialize hs(1) in pointer_intent_in_03 by @certik in https://github.com/lfortran/lfortran/pull/16605
- feat: run GPU launches of more than the chunk limit in chunks (step 1 of [#16569]) by @certik in https://github.com/lfortran/lfortran/pull/16601
- ci: build LFortran to WASM and the JupyterLite kernel with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16582
- ci: run the LLVM WASM jobs with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16581
- ci: run the Release, tarball and docs jobs with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16591
- fix: offload integer(1) and integer(2) to the GPU by @certik in https://github.com/lfortran/lfortran/pull/16614
- fix: address GPU kernel arrays of more than huge(0) elements in 64 bits by @certik in https://github.com/lfortran/lfortran/pull/16547
- ci: drop the platform jobs' brew, pip and Kokkos installs by @certik in https://github.com/lfortran/lfortran/pull/16635
- pixi: move shell into its own feature and add llvm11-noshell by @certik in https://github.com/lfortran/lfortran/pull/16639
- fix: give the end of a BLOCK construct a debug location by @certik in https://github.com/lfortran/lfortran/pull/16641
- fix: emit scalar TRANSFER (BitCast) in GPU kernels by @certik in https://github.com/lfortran/lfortran/pull/16647
- test: C interoperability of procedure values and remaining [#16382] regressions by @certik in https://github.com/lfortran/lfortran/pull/16633
- fix: unify integer kinds of GPU device-local allocation sizes under --descriptor-index-64 by @certik in https://github.com/lfortran/lfortran/pull/16634
- fix: copy gpu arguments with the descriptor index kind by @certik in https://github.com/lfortran/lfortran/pull/16615
- fix: keep do concurrent in a BLOCK inside a parallel loop sequential by @certik in https://github.com/lfortran/lfortran/pull/16640
- ci: build splpak with --fast instead of fpm's release profile by @certik in https://github.com/lfortran/lfortran/pull/16659
- fix: keep array-expression physical casts in the OpenMP pass by @certik in https://github.com/lfortran/lfortran/pull/16668
- fix: reject a nonintrinsic elemental procedure as a procedure pointer target by @certik in https://github.com/lfortran/lfortran/pull/16670
- fix: compute integer(8) values in Metal kernels by @certik in https://github.com/lfortran/lfortran/pull/16658
- fix: copy host array extents used as gpu kernel SUM and MATMUL bounds by @certik in https://github.com/lfortran/lfortran/pull/16650
- fix: scalarize elemental calls and scalar broadcasts in GPU SUM inlining by @certik in https://github.com/lfortran/lfortran/pull/16693
- fix: give gpu size-from-type replacements the kind of the size by @certik in https://github.com/lfortran/lfortran/pull/16643
- --fast: pass host frames explicitly to specialized procedures (part of [#16665]) by @certik in https://github.com/lfortran/lfortran/pull/16700
- fix: lower opaque procedure types with the complex result ABI by @certik in https://github.com/lfortran/lfortran/pull/16692
- fix: redirect calls with pointer actuals to a specialized dummy procedure by @certik in https://github.com/lfortran/lfortran/pull/16695
- fix: wait for GPU writes through pointers with unknown targets by @certik in https://github.com/lfortran/lfortran/pull/16694
- fix: namelist arrays of derived types with descriptor storage by @certik in https://github.com/lfortran/lfortran/pull/16696
- fix: require NOPASS on procedure pointer components with implicit interface by @certik in https://github.com/lfortran/lfortran/pull/16688
- Create the missing descriptor before bind(C) copy-in of an allocatable array by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16474
- Set elem_len to len*kind for character arrays by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16707
- Allocate the polymorphic element wrapper in the entry block by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16706
- fix: stop metal launches with arrays of more than huge(0) elements by @certik in https://github.com/lfortran/lfortran/pull/16714
- feat: run GPU launches with reductions in chunks (step 2 of [#16569]) by @certik in https://github.com/lfortran/lfortran/pull/16708
- fix: count GPU kernel array expressions in the offset kind under --descriptor-index-64 by @certik in https://github.com/lfortran/lfortran/pull/16717
- fix: recover GPU private array bounds from declared shapes by @certik in https://github.com/lfortran/lfortran/pull/16713
- fix: give internal procedures internal linkage by @certik in https://github.com/lfortran/lfortran/pull/16689
- fix: diagnose duplicate IMPLICIT NONE without --implicit-typing by @certik in https://github.com/lfortran/lfortran/pull/16671
- fix: namelist arrays of derived types with array components and extended types by @certik in https://github.com/lfortran/lfortran/pull/16697
- fix: preserve GPU identifier identity for Metal and CUDA by @certik in https://github.com/lfortran/lfortran/pull/16716
- fix: enforce implicit none (external) by @certik in https://github.com/lfortran/lfortran/pull/16672
- ci: run the out-of-source build with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16657
- docker: build the image with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16638
- perf: avoid materializing scalar reduction call actuals by @certik in https://github.com/lfortran/lfortran/pull/16733
- feat: decide GPU launch chunking at compile time for constant extents (step 3 of [#16569]) by @certik in https://github.com/lfortran/lfortran/pull/16728
- fix: finalize host locals before the frame slot when bounds inquire about them by @certik in https://github.com/lfortran/lfortran/pull/16729
- codegen: initialize the step that extract_value may not set by @certik in https://github.com/lfortran/lfortran/pull/16736
- feat(gpu): offload do concurrent to AMD GPUs with --gpu=hip by @certik in https://github.com/lfortran/lfortran/pull/16738
- fix: add ElementalChecks and report allocatable dummies of elemental procedures by @assem2002 in https://github.com/lfortran/lfortran/pull/16742
- fix: skip ASR verify after a reported error under --continue-compilation by @assem2002 in https://github.com/lfortran/lfortran/pull/16748
- fix: preserve GPU launches with absent optional captures by @certik in https://github.com/lfortran/lfortran/pull/16726
- fix: chunk Metal GPU launches of more than huge(0) iterations by @certik in https://github.com/lfortran/lfortran/pull/16727
- ci: move the in-tree compiler copy into ci/stage_build.sh by @certik in https://github.com/lfortran/lfortran/pull/16730
- ci: make the Caffeine tests work with Pixi's toolchain by @certik in https://github.com/lfortran/lfortran/pull/16731
- ci: install the applications' dependencies before the application tests by @certik in https://github.com/lfortran/lfortran/pull/16732
- fix: skip inactive GPU private workspace allocation by @certik in https://github.com/lfortran/lfortran/pull/16740
- fix: skip ASR verify in load_module after a reported error by @assem2002 in https://github.com/lfortran/lfortran/pull/16764
- fix: ishft/shiftl/shiftr/shifta with shift equal to bit_size by @certik in https://github.com/lfortran/lfortran/pull/16766
- perf: fuse private maps into ordered GPU reductions by @certik in https://github.com/lfortran/lfortran/pull/16768
- perf: number GPU iterations of integer(4) nests in 32 bits per chunk (step 4 of [#16569]) by @certik in https://github.com/lfortran/lfortran/pull/16739
- fix(gpu): shift right logically for ishft and shiftr on the device by @certik in https://github.com/lfortran/lfortran/pull/16759
- fix(gpu): do not advise the HIP driver about memory used in place by @certik in https://github.com/lfortran/lfortran/pull/16775
- fix(gpu): keep freed memory mapped in the unified memory model by @certik in https://github.com/lfortran/lfortran/pull/16776
- fix: reject access to private type-bound procedures outside their module by @certik in https://github.com/lfortran/lfortran/pull/16770
- ci: activate Pixi environments after ccache-action by @certik in https://github.com/lfortran/lfortran/pull/16791
- fix(gpu): report the last-level cache of an AMD GPU, not its L2 by @certik in https://github.com/lfortran/lfortran/pull/16777
- fix(gpu): wait for an APU by blocking rather than spinning by @certik in https://github.com/lfortran/lfortran/pull/16778
- perf(gpu): run a reduction in a grid-stride loop by @certik in https://github.com/lfortran/lfortran/pull/16779
- perf(gpu): unroll short sequential device loops on HIP, also with a run-time start by @certik in https://github.com/lfortran/lfortran/pull/16780
- fix(gpu): give a type inquiry in a kernel its value by @certik in https://github.com/lfortran/lfortran/pull/16789
- fix(gpu): give the size of a gathered section to an inlined procedure by @certik in https://github.com/lfortran/lfortran/pull/16790
- perf(gpu): compute 1/sqrt(x) as rsqrt(x) under --fast by @certik in https://github.com/lfortran/lfortran/pull/16785
- perf(gpu): wait for a launch that reads host memory only where the host changes it by @certik in https://github.com/lfortran/lfortran/pull/16788
- ci: run the compiler compatibility jobs with Pixi by @certik in https://github.com/lfortran/lfortran/pull/16755
- fix: pass a null pointer for INQUIRE without POS= by @certik in https://github.com/lfortran/lfortran/pull/16794
- perf: fuse adjacent loop nests by @certik in https://github.com/lfortran/lfortran/pull/16773
- Update reference results by @certik in https://github.com/lfortran/lfortran/pull/16803
- test: normalize COMDATs in reference test output by @certik in https://github.com/lfortran/lfortran/pull/16809
- ci: drop leftovers of the micromamba compatibility job by @certik in https://github.com/lfortran/lfortran/pull/16802
- fix: convert reduction DIM to the descriptor index kind by @certik in https://github.com/lfortran/lfortran/pull/16797
- fix: allocate I/O statement temporaries in the entry block by @certik in https://github.com/lfortran/lfortran/pull/16793
- perf: remove the fma ASR pass by @certik in https://github.com/lfortran/lfortran/pull/16804
- ci: build Caffeine with Xcode's Clang on macOS by @certik in https://github.com/lfortran/lfortran/pull/16801
- fix: count packed GPU component data in the kernel's offset kind by @certik in https://github.com/lfortran/lfortran/pull/16799
- fix: compute gpu component subscripts with 64-bit lower bounds for every rank by @certik in https://github.com/lfortran/lfortran/pull/16813
- perf(gpu): run fewer threads for a reduction on an NVIDIA GPU by @certik in https://github.com/lfortran/lfortran/pull/16816
- fix: ElementalChecks group 1: dummy argument attributes by @assem2002 in https://github.com/lfortran/lfortran/pull/16743
- fix: use the index kind for all array constructor sizes by @certik in https://github.com/lfortran/lfortran/pull/16798
- ci: remove the last installs outside Pixi by @certik in https://github.com/lfortran/lfortran/pull/16817
- pixi: pin the latest patch release of LLVM 19, 21, 22 and 23 by @certik in https://github.com/lfortran/lfortran/pull/16818
- pixi: provide binutils and pkg-config by @certik in https://github.com/lfortran/lfortran/pull/16822
- fix: return 64-bit INQUIRE POS= values by @certik in https://github.com/lfortran/lfortran/pull/16820
- fix: avoid rescanning the file in INQUIRE(POS=) on sequential units by @certik in https://github.com/lfortran/lfortran/pull/16821
- fix: call a module generic subroutine from an interactive cell by @certik in https://github.com/lfortran/lfortran/pull/16832
- fix(gpu): compute per-thread workspace offsets in 64 bits by @certik in https://github.com/lfortran/lfortran/pull/16800
- fix: ElementalChecks group 2: function result by @assem2002 in https://github.com/lfortran/lfortran/pull/16746
- fix: give a streaming REPL cell the main thread's stack by @certik in https://github.com/lfortran/lfortran/pull/16831
- fix: print array sections with one format in interactive mode by @certik in https://github.com/lfortran/lfortran/pull/16830
- fix: pass scalar descriptors to allocatable assumed-rank dummies by @certik in https://github.com/lfortran/lfortran/pull/16844
- test: pass a 64-bit POS= to _lfortran_inquire in the runtime test by @certik in https://github.com/lfortran/lfortran/pull/16856
- doc: build from Git with Pixi only; delete the Conda environment files by @certik in https://github.com/lfortran/lfortran/pull/16823
- fix: keep a slow subscriber's latest output in a shared session by @certik in https://github.com/lfortran/lfortran/pull/16846
- fix: PureChecks class and pure rules for elemental procedures (ElementalChecks group 3) by @assem2002 in https://github.com/lfortran/lfortran/pull/16749
- fix: decline GPU kernels whose parameters hold a deferred-length string by @certik in https://github.com/lfortran/lfortran/pull/16860
- feat(gpu): offload do concurrent logical reductions by @certik in https://github.com/lfortran/lfortran/pull/16853
- fix(gpu): give size, lbound and ubound of kind 8 their type in kernels by @certik in https://github.com/lfortran/lfortran/pull/16839
- fix: compute the GPU index-set split frame in integer(8) by @certik in https://github.com/lfortran/lfortran/pull/16833
- fix: compare section bounds strictly in the extent-1 shape shortcut by @certik in https://github.com/lfortran/lfortran/pull/16852
- fix: compute implied-do sizes of array constructors in one integer kind by @certik in https://github.com/lfortran/lfortran/pull/16834
- fix: retain source type when allocating empty polymorphic arrays by @certik in https://github.com/lfortran/lfortran/pull/16857
- fix: keep descriptor originals of array_by_data procedures in interactive mode by @certik in https://github.com/lfortran/lfortran/pull/16858
- fix: terminate a formatted record left open by a non-advancing write by @certik in https://github.com/lfortran/lfortran/pull/16864
- fix: reject POS= on non-stream units in READ/WRITE by @certik in https://github.com/lfortran/lfortran/pull/16866
- fix: print INQUIRE statements in --show-fortran by @certik in https://github.com/lfortran/lfortran/pull/16865
- fix: take queued kernel requests one per client per pass by @certik in https://github.com/lfortran/lfortran/pull/16829
- fix: validate INQUIRE SIZE= and POS= are scalar integer variables by @certik in https://github.com/lfortran/lfortran/pull/16867
- fix: pass and read pointer intrinsic arguments by value by @certik in https://github.com/lfortran/lfortran/pull/16895
- fix: keep output after a partly sent oversized message by @certik in https://github.com/lfortran/lfortran/pull/16899
- fix: return allocatable intrinsic array results into a temporary by @certik in https://github.com/lfortran/lfortran/pull/16897
- fix: keep program-scoped derived types in the OpenMP outline by @certik in https://github.com/lfortran/lfortran/pull/16905
- fix: return the actual argument's address from c_loc of a type(*) dummy by @certik in https://github.com/lfortran/lfortran/pull/16903
- Evaluate elementwise reductions and same-element section assignments in place by @certik in https://github.com/lfortran/lfortran/pull/16893
- fix: keep only linkable user procedures in array_by_data across units by @certik in https://github.com/lfortran/lfortran/pull/16906
- fix: ElementalChecks group 4: uses of elemental procedures by @assem2002 in https://github.com/lfortran/lfortran/pull/16753
- fix: lend the base's shape to an element selection inside a component chain by @certik in https://github.com/lfortran/lfortran/pull/16915
- refactor: insert the dropped-output notice in make_room by @certik in https://github.com/lfortran/lfortran/pull/16898
- fix: give overloaded intrinsic operator results lower bound 1 by @assem2002 in https://github.com/lfortran/lfortran/pull/15590
- fix: join the LSP server's threads on shutdown by @certik in https://github.com/lfortran/lfortran/pull/16920
- fix: use sequence association for explicit-shape dummy argument checks by @certik in https://github.com/lfortran/lfortran/pull/16894
- fix: keep an interactive global's derived-type initializer by @certik in https://github.com/lfortran/lfortran/pull/16902
- perf(gpu): keep the copy of what launches only read until the host waits by @certik in https://github.com/lfortran/lfortran/pull/16841
- perf: prefer a known CPU model over generic tuning by @certik in https://github.com/lfortran/lfortran/pull/16917
- feat(gpu): managed memory model for allocatable arrays on NVIDIA GPUs by @certik in https://github.com/lfortran/lfortran/pull/16782
- fix(gpu): do not sync around launches because a call was handed their arrays by @certik in https://github.com/lfortran/lfortran/pull/16944
- Convert character array descriptors in bind(C) procedures called from C by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16781
- Handle disassociated character pointer arrays in bind(C) calls by @Jainam-not-a-robot in https://github.com/lfortran/lfortran/pull/16774
- perf: hoist fixed-size BLOCK allocas out of loops by @certik in https://github.com/lfortran/lfortran/pull/16916
- fix: diagnose polymorphic arguments of c_loc and c_f_pointer by @certik in https://github.com/lfortran/lfortran/pull/16900
- fix: initialize interactive cell array globals statically by @certik in https://github.com/lfortran/lfortran/pull/16947
- fix: do not fold a structure constructor that gives an allocatable component a value by @certik in https://github.com/lfortran/lfortran/pull/16948
- fix: relink derived types copied into the interactive cell chain by @certik in https://github.com/lfortran/lfortran/pull/16957
- fix: use declared lower bounds when folding a named-constant array section by @certik in https://github.com/lfortran/lfortran/pull/16960
- fix: formatted read into a pointer or host-associated string by @certik in https://github.com/lfortran/lfortran/pull/16961
- fix: index a pointer array component reached through a component by @certik in https://github.com/lfortran/lfortran/pull/16966
- fix: build the CUDA GPU runtime against the CUDA 13 runtime API by @certik in https://github.com/lfortran/lfortran/pull/16974
- fix: hand reduction accumulators to the device write-discard by @certik in https://github.com/lfortran/lfortran/pull/16988
- fix: import an inaccessible derived type under a name no user can spell by @certik in https://github.com/lfortran/lfortran/pull/17014
- fix: do not read back a scalar a later statement assigns first by @certik in https://github.com/lfortran/lfortran/pull/16989
- fix: declare interface bodies with class dummies and never abstract interfaces in LLVM by @certik in https://github.com/lfortran/lfortran/pull/16968
- fix: allow a pointer intent(in) dummy as actual to a modifying dummy by @certik in https://github.com/lfortran/lfortran/pull/17006
- feat(gpu): turn XNACK on for HIP programs unless the user set HSA_XNACK by @certik in https://github.com/lfortran/lfortran/pull/16969
- examples: add the Benchmarks Game n-body sources and measurements by @certik in https://github.com/lfortran/lfortran/pull/16919
- fix: replay array_by_data on earlier interactive cells' procedures by @certik in https://github.com/lfortran/lfortran/pull/16949
- perf: forward a block-private device map into its reads by @certik in https://github.com/lfortran/lfortran/pull/16951
- perf: time each Metal kernel on the GPU with LFORTRAN_GPU_PROFILE=2 by @certik in https://github.com/lfortran/lfortran/pull/16962
- fix: prevent lost shutdown wakeups in the LSP message queue by @certik in https://github.com/lfortran/lfortran/pull/16914
- perf(gpu): keep the copies of arrays the host leaves unchanged past a wait by @certik in https://github.com/lfortran/lfortran/pull/16963
- fix: match the passed object at its pass position in type-bound generics by @certik in https://github.com/lfortran/lfortran/pull/17007
- fix: mangle global procedure definitions with --mangle-underscore-external by @certik in https://github.com/lfortran/lfortran/pull/17025
- fix: fold implied-do initializers of BLOCK declarations by @certik in https://github.com/lfortran/lfortran/pull/17028
- fix: allow a component of a named constant as a derived-type initializer by @certik in https://github.com/lfortran/lfortran/pull/17036
- fix: make c_long_double match the target C long double by @certik in https://github.com/lfortran/lfortran/pull/17033
- perf(gpu): managed memory model for allocatable arrays on Metal by @certik in https://github.com/lfortran/lfortran/pull/16970
- fix: privatize arrays defined by full-extent sections in do concurrent offload by @certik in https://github.com/lfortran/lfortran/pull/17019
- perf(gpu): unroll device loops by the divisor of their packed index by @certik in https://github.com/lfortran/lfortran/pull/16993
- fix: let an entity-decl char-length override the type-spec length by @certik in https://github.com/lfortran/lfortran/pull/17020
- fix: remap component and type symbols in pass_array_by_data copies by @certik in https://github.com/lfortran/lfortran/pull/17027
- fix: splice backslash-newline everywhere in the C preprocessor by @certik in https://github.com/lfortran/lfortran/pull/17026
- perf: use the AVX-512 estimate for double precision 1/sqrt in reductions by @certik in https://github.com/lfortran/lfortran/pull/16921
- test: run array_section_assign_10 without fast math under --fast by @certik in https://github.com/lfortran/lfortran/pull/17068
- fix: reuse the parent's vtable slot for a re-declared deferred binding by @certik in https://github.com/lfortran/lfortran/pull/17029
- fix: give interactive character globals external linkage by @certik in https://github.com/lfortran/lfortran/pull/17037
- fix: support array sections in namelist input by @certik in https://github.com/lfortran/lfortran/pull/17041
- fix: implicitly save local derived-type arrays with constant array initializers by @certik in https://github.com/lfortran/lfortran/pull/17039
- fix: lower assumed-rank to assumed-size array physical cast by @certik in https://github.com/lfortran/lfortran/pull/17079
- perf(gpu): let procedures return with their launches pending by @certik in https://github.com/lfortran/lfortran/pull/17040
- perf(gpu): compute small integer powers of device code as products by @certik in https://github.com/lfortran/lfortran/pull/17066
- test: exclude the reciprocal square root IR tests from the WASM build by @certik in https://github.com/lfortran/lfortran/pull/17126
- perf: enable the SLP vectorizer in the LLVM 17+ pass pipeline by @certik in https://github.com/lfortran/lfortran/pull/17093
- perf: allow the backend to fuse multiplies and adds under --fast by @certik in https://github.com/lfortran/lfortran/pull/17095
- perf(gpu): fold device checks that the ranges of their integers decide by @certik in https://github.com/lfortran/lfortran/pull/17057
- perf(gpu): remove the host waits of a decode loop that keeps its host steps on the device by @certik in https://github.com/lfortran/lfortran/pull/17097
- fix: reference a parent component whose type is declared in a program by @certik in https://github.com/lfortran/lfortran/pull/17038
- fix: do not override an inaccessible private binding of the parent type by @certik in https://github.com/lfortran/lfortran/pull/17069
- fix: tokenize ABSTRACT INTERFACE, IMPORT and PROCEDURE declarations in fixed form by @certik in https://github.com/lfortran/lfortran/pull/17091
- fix: accept deferred-length allocatable iomsg= in READ/WRITE by @certik in https://github.com/lfortran/lfortran/pull/17096
- fix: fold implied-do array constructors of derived types by @certik in https://github.com/lfortran/lfortran/pull/17099
- fix: use the dynamic type of polymorphic arrays in same_type_as and extends_type_of by @certik in https://github.com/lfortran/lfortran/pull/17100
- fix: bind interactive interface bodies and EXTERNAL to earlier cells' procedures by @certik in https://github.com/lfortran/lfortran/pull/17124
- fix: diagnose ambiguous generic interfaces (C1514) by @certik in https://github.com/lfortran/lfortran/pull/17021
- fix(gpu): call no deferring twins of a module compiled separately for the host by @certik in https://github.com/lfortran/lfortran/pull/17142
- perf(gpu): share inner loops between the lanes of a warp or SIMD group on NVIDIA and Metal GPUs by @certik in https://github.com/lfortran/lfortran/pull/16787
- fix(gpu): make offloaded kernel names unique across translation units by @certik in https://github.com/lfortran/lfortran/pull/17187
- perf(gpu): inline small loop functions into kernels and unroll the lanes' loops on Metal by @certik in https://github.com/lfortran/lfortran/pull/17121
- ci: do not run implied_do_loops61 on LLVM 7 and 8 by @certik in https://github.com/lfortran/lfortran/pull/17136
- test: run the interactive EXTERNAL tests only where the JIT supports them (WASM, LLVM 8) by @certik in https://github.com/lfortran/lfortran/pull/17140
- ci: provide ccache/sccache from Pixi and pin ccache-action by @certik in https://github.com/lfortran/lfortran/pull/17189
- perf: rewrite SLP transpose-gather sums of squares as horizontal adds by @certik in https://github.com/lfortran/lfortran/pull/16958
- fix: copy struct member symbols into external procedure interfaces by @certik in https://github.com/lfortran/lfortran/pull/17180
- fix: assign IOMSG= when a namelist READ fails by @certik in https://github.com/lfortran/lfortran/pull/17067
- fix: allow a component of an array element in a DATA implied-do by @certik in https://github.com/lfortran/lfortran/pull/17089
- fix: lower SELECT CASE nested at any depth by @assem2002 in https://github.com/lfortran/lfortran/pull/16941
- perf: prove zero fourth components of horizontal reductions over loops by @certik in https://github.com/lfortran/lfortran/pull/17204
- fix: use the argument type for the 1.0 constant in expm1 by @assem2002 in https://github.com/lfortran/lfortran/pull/16940
- fix: resolve defined operators instead of a same-named specific by @certik in https://github.com/lfortran/lfortran/pull/17192
- fix: do not treat a character substring as an external function call by @certik in https://github.com/lfortran/lfortran/pull/17194
- test: cover select type on an assumed-rank class(*) dummy by @certik in https://github.com/lfortran/lfortran/pull/17200
- perf(gpu): share the sums of a quantized matrix-vector product between the lanes by @certik in https://github.com/lfortran/lfortran/pull/17098
- fix: keep an external procedure called with descriptors through its interface by @assem2002 in https://github.com/lfortran/lfortran/pull/16922
- fix: process type(intrinsic-type-spec) declarations as intrinsic types by @certik in https://github.com/lfortran/lfortran/pull/17065
- fix: real(10) constants and output on Apple arm64 by @certik in https://github.com/lfortran/lfortran/pull/17198
- fix: cshift and eoshift into an allocated array with lower bound other than 1 by @assem2002 in https://github.com/lfortran/lfortran/pull/16928
- fix: keep the original access of use-associated names made private or public by @certik in https://github.com/lfortran/lfortran/pull/17137
- fix: copy non-contiguous actuals of contiguous assumed-shape/rank dummies by @certik in https://github.com/lfortran/lfortran/pull/17139
- fix: pass the character data address to type(*) dummies by @certik in https://github.com/lfortran/lfortran/pull/17160
- test: give integration tests that reuse a source their own module directory by @certik in https://github.com/lfortran/lfortran/pull/17202
- fix: include <cstdint> before the LLVM headers in llvm_reductions.cpp (LLVM 15/17/18 build) by @certik in https://github.com/lfortran/lfortran/pull/17210
- fix: move_alloc into a component of a derived-type array element by @assem2002 in https://github.com/lfortran/lfortran/pull/16937
- fix: give an explicit-shape array with a postponed bound a pointer type by @assem2002 in https://github.com/lfortran/lfortran/pull/15316
- fix: keep external procedures that pass_array_by_data specializes by @assem2002 in https://github.com/lfortran/lfortran/pull/17216
- fix: call a user-defined operator(.not.) on a non-logical operand by @assem2002 in https://github.com/lfortran/lfortran/pull/15664
- fix: report a function reference to a generic that resolves to a subroutine by @assem2002 in https://github.com/lfortran/lfortran/pull/15683
- perf: skip the device passes when no GPU device is selected by @assem2002 in https://github.com/lfortran/lfortran/pull/17220
- fix: emit the local descriptor copy in a fixed instruction order by @certik in https://github.com/lfortran/lfortran/pull/17287
- fix: evaluate GPU section subscripts before the element loop by @certik in https://github.com/lfortran/lfortran/pull/17206
- fix: set up derived type module variables without a Fortran program by @krystophny in https://github.com/lfortran/lfortran/pull/17291
- fix: do not give an implicit array conversion a pointer type by @certik in https://github.com/lfortran/lfortran/pull/17207
- fix: compare associated() against allocatable and character targets correctly by @schizophrenicmaniac in https://github.com/lfortran/lfortran/pull/15108
- fix: give c_size_t, c_intptr_t, c_ptrdiff_t and c_long the target's pointer size by @krystophny in https://github.com/lfortran/lfortran/pull/17290
- fix: target the x86-64 baseline CPU on LLVM 7-9 so real(16) uses SSE by @certik in https://github.com/lfortran/lfortran/pull/17151
- fix: copy a polymorphic section into a temporary of its declared type for a component by @schizophrenicmaniac in https://github.com/lfortran/lfortran/pull/14084
- fix: do not inline functions having TARGET dummy arguments by @atharvaajmera in https://github.com/lfortran/lfortran/pull/14705
- fix: keep internal finalizers of later interactive cells by @certik in https://github.com/lfortran/lfortran/pull/17215
- fix: match a procedure actual to an implicit dummy not yet visited by @certik in https://github.com/lfortran/lfortran/pull/17212
- fix: vector subscript then component, c(idx)%n, has the shape of the subscript by @assem2002 in https://github.com/lfortran/lfortran/pull/16923
- fix: accept an array X in ieee_support inquiries by @certik in https://github.com/lfortran/lfortran/pull/17193
- fix: accept a constant expression as the NAME= of bind(c) by @certik in https://github.com/lfortran/lfortran/pull/17146
- fix: keep a vector-subscripted class array a descriptor array by @certik in https://github.com/lfortran/lfortran/pull/17320
- docs: document Windows Conda missing zstd recovery by @certik in https://github.com/lfortran/lfortran/pull/16202
- fix: handle real(10) storage size and mod-file constants by @certik in https://github.com/lfortran/lfortran/pull/17197
- fix: match allocatable and pointer dummies in generic resolution by @certik in https://github.com/lfortran/lfortran/pull/17191
- fix: declare the pointer to an allocatable optional actual contiguous by @certik in https://github.com/lfortran/lfortran/pull/17324
- fix: return kind 16 from selected_real_kind above double precision by @certik in https://github.com/lfortran/lfortran/pull/17338
- fix: diagnose array initializer of a different shape by @certik in https://github.com/lfortran/lfortran/pull/17056
- fix: print FunctionPointerCast in the Fortran dumper by @certik in https://github.com/lfortran/lfortran/pull/17199
- perf(gpu): unroll small inner loops and keep their loads in flight on CUDA by @certik in https://github.com/lfortran/lfortran/pull/16786
- fix: accept a procedure for an INTENT(IN) procedure pointer dummy by @certik in https://github.com/lfortran/lfortran/pull/17339
- ci: build stdlib without quad precision by @certik in https://github.com/lfortran/lfortran/pull/17343
- fix: postpone a bound or length that calls a generic inside an expression by @assem2002 in https://github.com/lfortran/lfortran/pull/15605
- fix: copy back bind(c) allocatable dummies that have no intent by @assem2002 in https://github.com/lfortran/lfortran/pull/17276
- fix: list-directed READ of a last line with no newline returns iostat=-1 by @assem2002 in https://github.com/lfortran/lfortran/pull/17274
- ci: speed up the Linux and macOS Quick platform jobs by @certik in https://github.com/lfortran/lfortran/pull/17205
- fix: evaluate the value of z%im = f() before reading z%re by @assem2002 in https://github.com/lfortran/lfortran/pull/15716
- fix: lbound and ubound of x(idx) are 1 and the size when idx does not start at 1 by @assem2002 in https://github.com/lfortran/lfortran/pull/17271
- fix: keep z%re and z%im as the assignment target by @assem2002 in https://github.com/lfortran/lfortran/pull/17391
- test: add regression tests for 63 issues that work on main by @assem2002 in https://github.com/lfortran/lfortran/pull/17425
New Contributors
- @atharvaajmera made their first contribution in https://github.com/lfortran/lfortran/pull/14489
Full Changelog: https://github.com/lfortran/lfortran/compare/v0.68.0...v0.69.0