Kajetan Puchalski 0d40574e16

[flang] Inline hlfir.copy_in for trivial types (#138718) ...

hlfir.copy_in implements copying non-contiguous array slices for
functions that take in arrays required to be contiguous through
flang-rt.

For large arrays of trivial types, this can incur overhead compared to a
plain, inlined copy loop.

To address that, add a new InlineHLFIRCopyIn optimisation pass to inline
hlfir.copy_in operations for trivial types.

For the time being, the pattern is only applied in cases where the
copy-in does not require a corresponding copy-out, such as when the
function being called declares the array parameter as intent(in).

Applying this optimisation reduces the runtime of thornado-mini's
DeleptonizationProblem by about 10%.

---------

Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>

2025-06-06 15:10:17 +01:00

..

order_assignments

[flang] Undo the effects of CSE for hlfir.exactly_once. (#140190)

2025-05-20 09:22:05 -07:00

all-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

all.fir

…

any-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

any.fir

…

apply-codegen.fir

…

apply.fir

…

as_expr-codegen-polymorphic.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

as_expr-codegen.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

as_expr.fir

…

assign-bufferize.fir

…

assign-codegen-derived.fir

[flang] update fir.coordinate_of to carry the fields (#127231)

2025-02-28 09:50:05 +01:00

assign-codegen.fir

[flang][nfc] Use llvm memmove intrinsic over regular call (#134294)

2025-04-04 06:13:30 -07:00

assign-side-effects.fir

[flang][hlfir] refine hlfir.assign side effects (#113319)

2024-10-23 12:33:14 +02:00

assign.fir

[flang] replace fir.complex usages with mlir complex (#110850)

2024-10-03 17:10:57 +02:00

associate-codegen.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

associate.fir

[flang] replace fir.complex usages with mlir complex (#110850)

2024-10-03 17:10:57 +02:00

assumed_shape_with_value_keyword.f90

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

assumed-type-actual-args.f90

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

boxchar_emboxing.f90

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

bufferize01.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

bufferize-destroy-for-derived.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

bufferize-end-associate-for-derived.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

bufferize-poly-expr.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

bufferize-workshare.fir

[flang] Introduce hlfir.elemental lowerings to omp.workshare_loop_nest (#104748)

2024-11-21 11:14:21 +09:00

c_ptr_byvalue.f90

[flang] update fir.coordinate_of to carry the fields (#127231)

2025-02-28 09:50:05 +01:00

c-null-ptr-init.f90

…

call_with_poly_dummy.f90

…

char_assign.fir

…

char_extremum-bufferization.fir

[flang][nfc] Use llvm memmove intrinsic over regular call (#134294)

2025-04-04 06:13:30 -07:00

char_extremum.fir

…

concat-bufferization.fir

[flang][nfc] Use llvm memmove intrinsic over regular call (#134294)

2025-04-04 06:13:30 -07:00

concat.fir

…

convert-assign-inside-openacc-recipe.fir

…

copy-in-out-codegen.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

copy-in-out.fir

…

count-lowering-default-int-kinds.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

count-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

count.fir

…

cshift-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

declare-codegen.fir

…

declare.fir

…

designate-codegen-complex-part.fir

[flang] replace fir.complex usages with mlir complex (#110850)

2024-10-03 17:10:57 +02:00

designate-codegen-component-refs.fir

[flang] update fir.coordinate_of to carry the fields (#127231)

2025-02-28 09:50:05 +01:00

designate-codegen.fir

[flang] Fixed designator codegen for contiguous boxes. (#139003)

2025-05-12 18:33:29 -07:00

designate.fir

[flang] replace fir.complex usages with mlir complex (#110850)

2024-10-03 17:10:57 +02:00

destroy-codegen.fir

…

destroy.fir

…

dot_product-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

dot_product.fir

…

dummy_deallocation.f90

…

dummy_scope.fir

…

element-addr.fir

…

element-codegen-issue-118922.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

elemental-codegen-nested.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

elemental-codegen.fir

[flang][hlfir] do not propagate polymorphic temporary as allocatables (#142609)

2025-06-06 09:53:41 +02:00

elemental-cse.fir

…

elemental-shallow-copy.fir

…

elemental-with-empty-check-codegen.fir

[flang] Reset all extents to zero for empty hlfir.elemental loops. (#124867)

2025-01-29 12:03:05 -08:00

elemental.fir

…

elsewhere.fir

…

eval_in_mem-codegen.fir

[flang][hlfir] add hlfir.eval_in_mem operation (#118067)

2024-12-02 09:52:29 +01:00

eval_in_mem.fir

[flang][hlfir] add hlfir.eval_in_mem operation (#118067)

2024-12-02 09:52:29 +01:00

expr-type.fir

…

extents-of-shape-of.f90

…

forall_mask.fir

…

forall-index.fir

…

forall.fir

…

get_length_codegen.fir

[flang] Allow non-index length parameter on exprs fed into hlfir.get_length. (#124827)

2025-01-29 12:01:13 -08:00

get_length.fir

…

getextent-codegen.fir

[flang] Support non-index shape in hlfir.get_extent codegen. (#124622)

2025-01-28 09:25:36 -08:00

getextent.fir

…

hlfir-flags.f90

[flang][driver] rename flang-new to flang (#110023)

2024-10-10 09:26:04 +01:00

inline-elemental.f90

…

inline-elemental.fir

…

inline-hlfir-assign.fir

[flang] Extract hlfir.assign inlining from opt-bufferization. (#121544)

2025-01-03 08:33:14 -08:00

inline-hlfir-copy-in.fir

[flang] Inline hlfir.copy_in for trivial types (#138718)

2025-06-06 15:10:17 +01:00

invalid.fir

[flang] Added hlfir.reshape definition/lowering/codegen. (#124226)

2025-01-27 18:14:02 -08:00

matmul_transpose.fir

…

matmul-lowering.fir

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

matmul.fir

…

maxloc-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

maxloc.fir

…

maxval-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

maxval.fir

…

memory-effects.fir

…

minloc-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

minloc.fir

…

minval-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

minval-maxval-issue-134308.fir

[flang] Avoid optimizing min and max if not valid type (#134972)

2025-04-15 10:14:58 +01:00

minval.fir

…

mul_transpose.f90

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

no_reassoc-codegen.fir

…

no_reassoc.fir

…

no-block-merging.fir

…

null-codegen.fir

…

null.fir

…

opt-array-slice-assign.fir

[flang] Improve disjoint/identical slices recognition in opt-bufferization. (#119780)

2024-12-13 13:08:02 -08:00

opt-bufferization-eval_in_mem-with-associate.fir

[flang] Treat hlfir.associate as Allocate for FIR alias analysis. (#139004)

2025-05-12 18:34:12 -07:00

opt-bufferization-eval_in_mem.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

opt-bufferization-leslie3d.fir

…

opt-bufferization-non-realloc-assignment.fir

[flang] Assume matching shapes in elemental assignment with non-realloc lhs. (#118552)

2024-12-05 09:34:32 -08:00

opt-bufferization-same-ptr-elemental.fir

[flang] Improve designate/elemental indices match in opt-bufferization. (#121371)

2025-01-03 08:17:52 -08:00

opt-bufferization-skip-volatile.fir

[flang] Skip opt-bufferization when memory effect does not have an associated value (#140781)

2025-05-22 06:50:25 -07:00

opt-bufferization.fir

[flang] Extract hlfir.assign inlining from opt-bufferization. (#121544)

2025-01-03 08:33:14 -08:00

opt-scalar-assign.fir

[flang] replace fir.complex usages with mlir complex (#110850)

2024-10-03 17:10:57 +02:00

opt-variable-assign-omp.fir

[mlir][OpenMP][flang] make private variable allocation implicit in omp.private (#124019)

2025-01-31 09:35:26 +00:00

optional_dummy.f90

[flang] Do not produce result for void runtime call (#123155)

2025-01-16 12:34:38 -08:00

parent_comp-codegen.fir

…

parent_comp.fir

…

product-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

product.fir

…

propagate-contiguous-attribute.fir

[flang] Propagate contiguous attribute through HLFIR. (#138797)

2025-05-12 18:33:47 -07:00

region-assign.fir

…

reshape-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

reshape.fir

[flang] Added hlfir.reshape definition/lowering/codegen. (#124226)

2025-01-27 18:14:02 -08:00

set_length-codegen.fir

…

set_length.fir

…

shapeof-lowering.fir

…

shapeof.fir

[flang] Simple folding for hlfir.shape_of. (#119649)

2024-12-12 10:38:34 -08:00

simplify-hlfir-intrinsics-all.fir

Reland [flang] Generalized simplification of HLFIR reduction ops. (#136071) (#136246)

2025-04-18 11:56:07 -07:00

simplify-hlfir-intrinsics-any.fir

Reland [flang] Generalized simplification of HLFIR reduction ops. (#136071) (#136246)

2025-04-18 11:56:07 -07:00

simplify-hlfir-intrinsics-count.fir

Reland [flang] Generalized simplification of HLFIR reduction ops. (#136071) (#136246)

2025-04-18 11:56:07 -07:00

simplify-hlfir-intrinsics-cshift.fir

[flang] Added alternative inlining code for hlfir.cshift. (#129176)

2025-03-03 09:58:20 -08:00

simplify-hlfir-intrinsics-dotproduct.fir

Reland "[flang] Inline hlfir.dot_product. (#123143)" (#123385)

2025-01-17 12:09:44 -08:00

simplify-hlfir-intrinsics-matmul.fir

[flang] Inline hlfir.matmul[_transpose]. (#122821)

2025-01-15 08:42:57 -08:00

simplify-hlfir-intrinsics-maxloc.fir

[flang] Fetch the initial reduction value from the input array. (#136790)

2025-04-30 13:53:26 -07:00

simplify-hlfir-intrinsics-maxval.fir

[flang] Fetch the initial reduction value from the input array. (#136790)

2025-04-30 13:53:26 -07:00

simplify-hlfir-intrinsics-minloc.fir

[flang] Fetch the initial reduction value from the input array. (#136790)

2025-04-30 13:53:26 -07:00

simplify-hlfir-intrinsics-minval.fir

[flang] Fetch the initial reduction value from the input array. (#136790)

2025-04-30 13:53:26 -07:00

simplify-hlfir-intrinsics-reshape.fir

[flang] Inline hlfir.reshape as hlfir.elemental. (#124683)

2025-01-30 07:46:12 -08:00

simplify-hlfir-intrinsics-sum.fir

[flang][NFC] Removed unneeded engineering option. (#122305)

2025-01-10 09:28:03 -08:00

simplify-hlfir-intrinsics.fir

[flang] Turn SimplifyHLFIRIntrinsics into a greedy rewriter. (#119946)

2024-12-16 08:00:29 -08:00

sum-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

sum.fir

…

transpose-lowering.fir

[flang][NFC] use hlfir.declare first result when both results are raw pointers (#132261)

2025-03-21 11:41:04 +01:00

transpose.fir

…

volatile1.fir

[flang] Hide strict volatility checks behind flag (#138183)

2025-05-02 09:03:20 -07:00

volatile2.fir

[flang] Hide strict volatility checks behind flag (#138183)

2025-05-02 09:03:20 -07:00

volatile3.fir

[flang] Hide strict volatility checks behind flag (#138183)

2025-05-02 09:03:20 -07:00

volatile4.fir

[flang] Hide strict volatility checks behind flag (#138183)

2025-05-02 09:03:20 -07:00

volatile.fir

[flang] Hide strict volatility checks behind flag (#138183)

2025-05-02 09:03:20 -07:00

where.fir

…