From 1f1b4cf275e9b47a286b83c590f4fb0d96303267 Mon Sep 17 00:00:00 2001
From: Xavier Routh <xrouth2@illinois.edu>
Date: Thu, 30 Jan 2025 15:20:13 -0600
Subject: [PATCH] Forkify + Loop Canonicalization + Initial Fork Fission
---
Cargo.lock | 164 +++-
Cargo.toml | 8 +-
hercules_cg/src/cpu.rs | 4 +-
hercules_ir/src/ir.rs | 30 +
hercules_ir/src/verify.rs | 9 +-
hercules_opt/Cargo.toml | 3 +
hercules_opt/src/ccp.rs | 4 +-
hercules_opt/src/device_placement.rs | 3 -
hercules_opt/src/editor.rs | 18 +
hercules_opt/src/fork_concat_split.rs | 4 +-
hercules_opt/src/fork_guard_elim.rs | 323 +++++---
hercules_opt/src/fork_transforms.rs | 540 ++++++++++++++
hercules_opt/src/forkify.rs | 704 +++++++++++++-----
hercules_opt/src/gcm.rs | 4 +-
hercules_opt/src/ivar.rs | 602 +++++++++++++++
hercules_opt/src/lib.rs | 6 +-
hercules_opt/src/outline.rs | 1 -
hercules_opt/src/schedule.rs | 28 +-
hercules_opt/src/unforkify.rs | 208 +++++-
hercules_opt/src/utils.rs | 105 +++
hercules_samples/matmul/build.rs | 2 +-
hercules_test/hercules_interpreter/Cargo.toml | 5 +-
.../hercules_interpreter/src/interpreter.rs | 444 +++++++----
hercules_test/hercules_interpreter/src/lib.rs | 69 +-
.../hercules_interpreter/src/main.rs | 28 -
.../hercules_interpreter/src/value.rs | 82 +-
hercules_test/hercules_tests/Cargo.toml | 1 +
.../tests/fork_transform_tests.rs | 99 +++
.../hercules_tests/tests/forkify_tests.rs | 294 ++++++++
.../hercules_tests/tests/interpreter_tests.rs | 63 ++
.../hercules_tests/tests/loop_tests.rs | 436 +++++++++++
.../hercules_tests/tests/opt_tests.rs | 397 +++++-----
hercules_test/test_inputs/2d_fork.hir | 8 +
hercules_test/test_inputs/3d_fork.hir | 8 +
hercules_test/test_inputs/5d_fork.hir | 8 +
.../fork_fission/inner_control.hir | 15 +
.../fork_fission/inner_loop.hir | 23 +
.../intermediate_buffer_simple.hir | 10 +
.../fork_transforms/fork_fission/simple1.hir | 13 +
.../fork_transforms/fork_fission/simple2.hir | 19 +
.../fork_transforms/fork_fission/tricky.hir | 13 +
.../fork_transforms/fork_fusion.hir | 0
.../fork_transforms/fork_interchange.hir | 0
.../fork_transforms/matmul_int.hir | 18 +
.../fork_transforms/tiled_matmul_int.hir | 18 +
.../test_inputs/forkify/alternate_bounds.hir | 16 +
.../test_inputs/forkify/broken_sum.hir | 16 +
.../forkify/control_after_condition.hir | 25 +
.../forkify/control_before_condition.hir | 25 +
.../expected_fails.hir/bad_3nest_return.hir | 35 +
.../expected_fails.hir/bad_loop_tid_sum.hir | 16 +
.../test_inputs/forkify/inner_fork.hir | 22 +
.../forkify/inner_fork_complex.hir | 32 +
.../test_inputs/forkify/loop_array_sum.hir | 16 +
.../test_inputs/forkify/loop_simple_iv.hir | 12 +
.../test_inputs/forkify/loop_sum.hir | 16 +
.../test_inputs/forkify/loop_tid_sum.hir | 16 +
.../test_inputs/forkify/merged_phi_cycle.hir | 18 +
.../test_inputs/forkify/nested_loop2.hir | 25 +
.../test_inputs/forkify/nested_tid_sum.hir | 25 +
.../test_inputs/forkify/nested_tid_sum_2.hir | 26 +
.../test_inputs/forkify/phi_loop4.hir | 16 +
.../test_inputs/forkify/split_phi_cycle.hir | 16 +
.../test_inputs/forkify/super_nested_loop.hir | 35 +
.../loop_analysis/alternate_bounds.hir | 14 +
.../alternate_bounds_internal_control.hir | 23 +
.../alternate_bounds_internal_control2.hir | 21 +
.../alternate_bounds_nested_do_loop.hir | 28 +
.../alternate_bounds_nested_do_loop2.hir | 25 +
.../alternate_bounds_nested_do_loop_array.hir | 28 +
...lternate_bounds_nested_do_loop_guarded.hir | 40 +
.../alternate_bounds_use_after_loop.hir | 21 +
.../alternate_bounds_use_after_loop2.hir | 21 +
...alternate_bounds_use_after_loop_no_tid.hir | 17 +
...lternate_bounds_use_after_loop_no_tid2.hir | 19 +
.../test_inputs/loop_analysis/broken_sum.hir | 16 +
.../loop_analysis/do_loop_far_guard.hir | 14 +
.../loop_analysis/do_loop_immediate_guard.hir | 21 +
.../loop_analysis/do_loop_no_guard.hir | 15 +
.../loop_analysis/do_while_separate_body.hir | 16 +
.../loop_analysis/do_while_separate_body2.hir | 18 +
.../loop_analysis/loop_array_sum.hir | 16 +
.../loop_analysis/loop_body_count.hir | 16 +
.../test_inputs/loop_analysis/loop_sum.hir | 16 +
.../loop_analysis/loop_trip_count.hir | 19 +
.../loop_analysis/loop_trip_count_tuple.hir | 19 +
hercules_test/test_inputs/matmul_int.hir | 21 +-
juno_samples/cava/build.rs | 1 -
juno_samples/cava/src/main.rs | 5 +-
juno_samples/matmul/build.rs | 2 +
juno_samples/matmul/src/main.rs | 9 +-
juno_samples/matmul/src/sched.sch | 76 ++
juno_scheduler/Cargo.toml | 2 +
juno_scheduler/src/compile.rs | 6 +-
juno_scheduler/src/default.rs | 7 +-
juno_scheduler/src/ir.rs | 6 +-
juno_scheduler/src/lib.rs | 43 +-
juno_scheduler/src/pm.rs | 153 +++-
juno_utils/src/stringtab.rs | 4 +-
99 files changed, 5199 insertions(+), 828 deletions(-)
delete mode 100644 hercules_opt/src/device_placement.rs
create mode 100644 hercules_opt/src/fork_transforms.rs
create mode 100644 hercules_opt/src/ivar.rs
delete mode 100644 hercules_test/hercules_interpreter/src/main.rs
create mode 100644 hercules_test/hercules_tests/tests/fork_transform_tests.rs
create mode 100644 hercules_test/hercules_tests/tests/forkify_tests.rs
create mode 100644 hercules_test/hercules_tests/tests/interpreter_tests.rs
create mode 100644 hercules_test/hercules_tests/tests/loop_tests.rs
create mode 100644 hercules_test/test_inputs/2d_fork.hir
create mode 100644 hercules_test/test_inputs/3d_fork.hir
create mode 100644 hercules_test/test_inputs/5d_fork.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/inner_control.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/inner_loop.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/intermediate_buffer_simple.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/simple1.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/simple2.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fission/tricky.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_fusion.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/fork_interchange.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/matmul_int.hir
create mode 100644 hercules_test/test_inputs/fork_transforms/tiled_matmul_int.hir
create mode 100644 hercules_test/test_inputs/forkify/alternate_bounds.hir
create mode 100644 hercules_test/test_inputs/forkify/broken_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/control_after_condition.hir
create mode 100644 hercules_test/test_inputs/forkify/control_before_condition.hir
create mode 100644 hercules_test/test_inputs/forkify/expected_fails.hir/bad_3nest_return.hir
create mode 100644 hercules_test/test_inputs/forkify/expected_fails.hir/bad_loop_tid_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/inner_fork.hir
create mode 100644 hercules_test/test_inputs/forkify/inner_fork_complex.hir
create mode 100644 hercules_test/test_inputs/forkify/loop_array_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/loop_simple_iv.hir
create mode 100644 hercules_test/test_inputs/forkify/loop_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/loop_tid_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/merged_phi_cycle.hir
create mode 100644 hercules_test/test_inputs/forkify/nested_loop2.hir
create mode 100644 hercules_test/test_inputs/forkify/nested_tid_sum.hir
create mode 100644 hercules_test/test_inputs/forkify/nested_tid_sum_2.hir
create mode 100644 hercules_test/test_inputs/forkify/phi_loop4.hir
create mode 100644 hercules_test/test_inputs/forkify/split_phi_cycle.hir
create mode 100644 hercules_test/test_inputs/forkify/super_nested_loop.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control2.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop2.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_array.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_guarded.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop2.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid2.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/broken_sum.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/do_loop_far_guard.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/do_loop_immediate_guard.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/do_loop_no_guard.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/do_while_separate_body.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/do_while_separate_body2.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/loop_array_sum.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/loop_body_count.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/loop_sum.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/loop_trip_count.hir
create mode 100644 hercules_test/test_inputs/loop_analysis/loop_trip_count_tuple.hir
create mode 100644 juno_samples/matmul/src/sched.sch
diff --git a/Cargo.lock b/Cargo.lock
index 8bb64bd2..623fc35c 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -259,6 +259,12 @@ dependencies = [
"arrayvec",
]
+[[package]]
+name = "bimap"
+version = "0.6.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "230c5f1ca6a325a32553f8640d31ac9b49f2411e901e427570154868b46da4f7"
+
[[package]]
name = "bincode"
version = "1.3.3"
@@ -282,9 +288,9 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
[[package]]
name = "bitflags"
-version = "2.7.0"
+version = "2.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1be3f42a67d6d345ecd59f675f3f012d6974981560836e938c22b424b85ce1be"
+checksum = "8f68f53c83ab957f72c32642f3868eec03eb974d1fb82e453128456482613d36"
[[package]]
name = "bitstream-io"
@@ -366,9 +372,9 @@ dependencies = [
[[package]]
name = "cc"
-version = "1.2.9"
+version = "1.2.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8293772165d9345bdaaa39b45b2109591e63fe5e6fbc23c6ff930a048aa310b"
+checksum = "13208fcbb66eaeffe09b99fffbe1af420f00a7b35aa99ad683dfc1aa76145229"
dependencies = [
"jobserver",
"libc",
@@ -418,9 +424,9 @@ dependencies = [
[[package]]
name = "clap"
-version = "4.5.26"
+version = "4.5.27"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a8eb5e908ef3a6efbe1ed62520fb7287959888c88485abe072543190ecc66783"
+checksum = "769b0145982b4b48713e01ec42d61614425f27b7058bda7180a3a41f30104796"
dependencies = [
"clap_builder",
"clap_derive",
@@ -428,9 +434,9 @@ dependencies = [
[[package]]
name = "clap_builder"
-version = "4.5.26"
+version = "4.5.27"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "96b01801b5fc6a0a232407abc821660c9c6d25a1cafc0d4f85f29fb8d9afc121"
+checksum = "1b26884eb4b57140e4d2d93652abfa49498b938b3c9179f9fc487b0acc3edad7"
dependencies = [
"anstream",
"anstyle",
@@ -525,9 +531,9 @@ checksum = "d0a5c400df2834b80a4c3327b3aad3a4c4cd4de0629063962b03235697506a28"
[[package]]
name = "crunchy"
-version = "0.2.2"
+version = "0.2.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7a81dae078cea95a014a339291cec439d2f232ebe854a9d672b796c6afafa9b7"
+checksum = "43da5946c66ffcc7745f48db692ffbb10a83bfe0afd96235c5c2a4fb23994929"
[[package]]
name = "deranged"
@@ -538,6 +544,26 @@ dependencies = [
"powerfmt",
]
+[[package]]
+name = "derive_more"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4a9b99b9cbbe49445b21764dc0625032a89b145a2642e67603e1c936f5458d05"
+dependencies = [
+ "derive_more-impl",
+]
+
+[[package]]
+name = "derive_more-impl"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cb7330aeadfbe296029522e6c40f315320aba36fc43a5b3632f3795348f3bd22"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
[[package]]
name = "dot"
version = "0.1.0"
@@ -819,6 +845,23 @@ dependencies = [
"serde",
]
+[[package]]
+name = "hercules_interpreter"
+version = "0.1.0"
+dependencies = [
+ "bitvec",
+ "clap",
+ "derive_more",
+ "hercules_ir",
+ "hercules_opt",
+ "itertools 0.14.0",
+ "juno_scheduler",
+ "ordered-float",
+ "postcard",
+ "rand",
+ "serde",
+]
+
[[package]]
name = "hercules_ir"
version = "0.1.0"
@@ -834,14 +877,17 @@ dependencies = [
name = "hercules_opt"
version = "0.1.0"
dependencies = [
+ "bimap",
"bitvec",
"either",
"hercules_cg",
"hercules_ir",
"itertools 0.14.0",
+ "nestify",
"ordered-float",
"postcard",
"serde",
+ "slotmap",
"take_mut",
"tempfile",
"union-find",
@@ -851,6 +897,21 @@ dependencies = [
name = "hercules_rt"
version = "0.1.0"
+[[package]]
+name = "hercules_tests"
+version = "0.1.0"
+dependencies = [
+ "bitvec",
+ "clap",
+ "hercules_interpreter",
+ "hercules_ir",
+ "hercules_opt",
+ "itertools 0.14.0",
+ "juno_scheduler",
+ "ordered-float",
+ "rand",
+]
+
[[package]]
name = "hermit-abi"
version = "0.4.0"
@@ -898,9 +959,9 @@ checksum = "d0263a3d970d5c054ed9312c0057b4f3bde9c0b33836d3637361d4a9e6e7a408"
[[package]]
name = "indexmap"
-version = "2.7.0"
+version = "2.7.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "62f822373a4fe84d4bb149bf54e584a7f4abec90e072ed49cda0edea5b95471f"
+checksum = "8c9c992b02b5b4c94ea26e32fe5bccb7aa7d9f390ab5c1221ff895bc7ea8b652"
dependencies = [
"equivalent",
"hashbrown",
@@ -1094,6 +1155,8 @@ dependencies = [
"juno_utils",
"lrlex",
"lrpar",
+ "postcard",
+ "serde",
"tempfile",
]
@@ -1157,7 +1220,7 @@ version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c0ff37bd590ca25063e35af745c343cb7a0271906fb7b37e4813e8f79f00268d"
dependencies = [
- "bitflags 2.7.0",
+ "bitflags 2.8.0",
"libc",
"redox_syscall",
]
@@ -1180,9 +1243,9 @@ dependencies = [
[[package]]
name = "log"
-version = "0.4.22"
+version = "0.4.25"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a7a70ba024b9dc04c27ea2f0c0548feb474ec5c54bba33a7f72f873a39d07b24"
+checksum = "04cbf5b083de1c7e0222a7a51dbfdba1cbe1c6ab0b15e29fff3f6c077fd9cd9f"
dependencies = [
"value-bag",
]
@@ -1286,14 +1349,26 @@ checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a"
[[package]]
name = "miniz_oxide"
-version = "0.8.2"
+version = "0.8.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4ffbe83022cedc1d264172192511ae958937694cd57ce297164951b8b3568394"
+checksum = "b8402cab7aefae129c6977bb0ff1b8fd9a04eb5b51efc50a70bea51cda0c7924"
dependencies = [
"adler2",
"simd-adler32",
]
+[[package]]
+name = "nestify"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41d7249f7122d4e8a40f3b1b1850b763d2f864bf8e4b712427f024f8a167ea17"
+dependencies = [
+ "proc-macro-error",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.96",
+]
+
[[package]]
name = "new_debug_unreachable"
version = "1.0.6"
@@ -1547,6 +1622,30 @@ dependencies = [
"zerocopy",
]
+[[package]]
+name = "proc-macro-error"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "da25490ff9892aab3fcf7c36f08cfb902dd3e71ca0f9f9517bea02a73a5ce38c"
+dependencies = [
+ "proc-macro-error-attr",
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+ "version_check",
+]
+
+[[package]]
+name = "proc-macro-error-attr"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a1be40180e52ecc98ad80b184934baf3d0d29f979574e439af5a55274b35f869"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "version_check",
+]
+
[[package]]
name = "proc-macro2"
version = "1.0.93"
@@ -1713,7 +1812,7 @@ version = "0.5.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "03a862b389f93e68874fbf580b9de08dd02facb9a788ebadaf4a3fd33cf58834"
dependencies = [
- "bitflags 2.7.0",
+ "bitflags 2.8.0",
]
[[package]]
@@ -1762,11 +1861,11 @@ dependencies = [
[[package]]
name = "rustix"
-version = "0.38.43"
+version = "0.38.44"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a78891ee6bf2340288408954ac787aa063d8e8817e9f53abb37c695c6d834ef6"
+checksum = "fdb5bc1ae2baa591800df16c9ca78619bf65c0488b41b96ccec5d11220d8c154"
dependencies = [
- "bitflags 2.7.0",
+ "bitflags 2.8.0",
"errno",
"libc",
"linux-raw-sys",
@@ -1787,9 +1886,9 @@ checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
[[package]]
name = "semver"
-version = "1.0.24"
+version = "1.0.25"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3cb6eb87a131f756572d7fb904f6e7b68633f09cca868c5df1c4b8d1a694bbba"
+checksum = "f79dfe2d285b0488816f30e700a7438c5a73d816b5b7d3ac72fbc48b0d185e03"
[[package]]
name = "serde"
@@ -1856,6 +1955,15 @@ dependencies = [
"autocfg",
]
+[[package]]
+name = "slotmap"
+version = "1.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dbff4acf519f630b3a3ddcfaea6c06b42174d9a44bc70c620e9ed1649d58b82a"
+dependencies = [
+ "version_check",
+]
+
[[package]]
name = "smallvec"
version = "1.13.2"
@@ -2084,9 +2192,9 @@ checksum = "e672c95779cf947c5311f83787af4fa8fffd12fb27e4993211a84bdfd9610f9c"
[[package]]
name = "unicode-ident"
-version = "1.0.14"
+version = "1.0.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "adb9e6ca4f869e1180728b7950e35922a7fc6397f7b641499e8f3ef06e50dc83"
+checksum = "11cd88e12b17c6494200a9c1b683a04fcac9573ed74cd1b62aeb2727c5592243"
[[package]]
name = "unicode-width"
@@ -2140,6 +2248,12 @@ version = "0.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "852e951cb7832cb45cb1169900d19760cfa39b82bc0ea9c0e5a14ae88411c98b"
+[[package]]
+name = "version_check"
+version = "0.9.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
+
[[package]]
name = "vob"
version = "3.0.3"
diff --git a/Cargo.toml b/Cargo.toml
index 6adefdba..d31c59f7 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -11,8 +11,8 @@ members = [
"juno_scheduler",
"juno_build",
- #"hercules_test/hercules_interpreter",
- #"hercules_test/hercules_tests",
+ "hercules_test/hercules_interpreter",
+ "hercules_test/hercules_tests",
"hercules_samples/dot",
"hercules_samples/matmul",
@@ -26,7 +26,7 @@ members = [
"juno_samples/nested_ccp",
"juno_samples/antideps",
"juno_samples/implicit_clone",
- "juno_samples/cava",
+ "juno_samples/cava",
"juno_samples/concat",
- "juno_samples/schedule_test",
+ "juno_samples/schedule_test",
]
diff --git a/hercules_cg/src/cpu.rs b/hercules_cg/src/cpu.rs
index 3750c4f6..53743852 100644
--- a/hercules_cg/src/cpu.rs
+++ b/hercules_cg/src/cpu.rs
@@ -615,14 +615,14 @@ impl<'a> CPUContext<'a> {
)?,
DynamicConstant::Min(left, right) => write!(
body,
- " %dc{} = call @llvm.umin.i64(i64%dc{},i64%dc{})\n",
+ " %dc{} = call i64 @llvm.umin.i64(i64%dc{},i64%dc{})\n",
dc.idx(),
left.idx(),
right.idx()
)?,
DynamicConstant::Max(left, right) => write!(
body,
- " %dc{} = call @llvm.umax.i64(i64%dc{},i64%dc{})\n",
+ " %dc{} = call i64 @llvm.umax.i64(i64%dc{},i64%dc{})\n",
dc.idx(),
left.idx(),
right.idx()
diff --git a/hercules_ir/src/ir.rs b/hercules_ir/src/ir.rs
index 0b271e85..bf7806dc 100644
--- a/hercules_ir/src/ir.rs
+++ b/hercules_ir/src/ir.rs
@@ -1356,6 +1356,36 @@ impl Node {
}
}
+ pub fn is_zero_dc(&self, dynamic_constants: &Vec<DynamicConstant>) -> bool {
+ if let Node::DynamicConstant { id } = self
+ && dynamic_constants[id.idx()].try_constant() == Some(0)
+ {
+ true
+ } else {
+ false
+ }
+ }
+
+ pub fn is_one_dc(&self, dynamic_constants: &Vec<DynamicConstant>) -> bool {
+ if let Node::DynamicConstant { id } = self
+ && dynamic_constants[id.idx()].try_constant() == Some(1)
+ {
+ true
+ } else {
+ false
+ }
+ }
+
+ pub fn is_one_constant(&self, constants: &Vec<Constant>) -> bool {
+ if let Node::Constant { id } = self
+ && constants[id.idx()].is_one()
+ {
+ true
+ } else {
+ false
+ }
+ }
+
pub fn try_projection(&self, branch: usize) -> Option<NodeID> {
if let Node::Projection { control, selection } = self
&& branch == *selection
diff --git a/hercules_ir/src/verify.rs b/hercules_ir/src/verify.rs
index 5ee5f1d2..f188932e 100644
--- a/hercules_ir/src/verify.rs
+++ b/hercules_ir/src/verify.rs
@@ -129,6 +129,7 @@ fn verify_structure(
match function.nodes[user.idx()] {
Node::Parameter { index: _ }
| Node::Constant { id: _ }
+ | Node::Undef { ty: _ }
| Node::DynamicConstant { id: _ } => {}
_ => {
if function.nodes[user.idx()].is_control() {
@@ -300,7 +301,7 @@ fn verify_structure(
Err("Call node's control input must be a region node.")?;
}
}
- // Collect nodes must depend on a join node.
+ // Reduce nodes must depend on a join node.
Node::Reduce {
control,
init: _,
@@ -308,7 +309,7 @@ fn verify_structure(
} => {
if let Node::Join { control: _ } = function.nodes[control.idx()] {
} else {
- Err("Collect node's control input must be a join node.")?;
+ Err("Reduce node's control input must be a join node.")?;
}
}
// Return nodes must have no users.
@@ -498,8 +499,8 @@ fn verify_dominance_relationships(
// Every use of a thread ID must be postdominated by
// the thread ID's fork's corresponding join node. We
// don't need to check for the case where the thread ID
- // flows through the collect node out of the fork-join,
- // because after the collect, the thread ID is no longer
+ // flows through the reduce node out of the fork-join,
+ // because after the reduce, the thread ID is no longer
// considered an immediate control output use.
if postdom.contains(this_id)
&& !postdom.does_dom(*fork_join_map.get(&control).unwrap(), this_id)
diff --git a/hercules_opt/Cargo.toml b/hercules_opt/Cargo.toml
index 9f22884d..1db5dc2f 100644
--- a/hercules_opt/Cargo.toml
+++ b/hercules_opt/Cargo.toml
@@ -11,8 +11,11 @@ tempfile = "*"
either = "*"
itertools = "*"
take_mut = "*"
+slotmap = "*"
union-find = "*"
postcard = { version = "*", features = ["alloc"] }
serde = { version = "*", features = ["derive"] }
hercules_cg = { path = "../hercules_cg" }
hercules_ir = { path = "../hercules_ir" }
+nestify = "*"
+bimap = "*"
\ No newline at end of file
diff --git a/hercules_opt/src/ccp.rs b/hercules_opt/src/ccp.rs
index 68693e8b..92d52a71 100644
--- a/hercules_opt/src/ccp.rs
+++ b/hercules_opt/src/ccp.rs
@@ -677,7 +677,9 @@ fn ccp_flow_function(
(BinaryOperator::RSh, Constant::UnsignedInteger64(left_val), Constant::UnsignedInteger64(right_val)) => Some(Constant::UnsignedInteger64(left_val >> right_val)),
_ => panic!("Unsupported combination of binary operation and constant values. Did typechecking succeed?")
};
- new_cons.map(|c| ConstantLattice::Constant(c)).unwrap_or(ConstantLattice::bottom())
+ new_cons
+ .map(|c| ConstantLattice::Constant(c))
+ .unwrap_or(ConstantLattice::bottom())
} else if (left_constant.is_top() && !right_constant.is_bottom())
|| (!left_constant.is_bottom() && right_constant.is_top())
{
diff --git a/hercules_opt/src/device_placement.rs b/hercules_opt/src/device_placement.rs
deleted file mode 100644
index 2badd69d..00000000
--- a/hercules_opt/src/device_placement.rs
+++ /dev/null
@@ -1,3 +0,0 @@
-use hercules_ir::ir::*;
-
-use crate::*;
diff --git a/hercules_opt/src/editor.rs b/hercules_opt/src/editor.rs
index 8b90710e..39f1184c 100644
--- a/hercules_opt/src/editor.rs
+++ b/hercules_opt/src/editor.rs
@@ -1,3 +1,4 @@
+use std::borrow::Borrow;
use std::cell::{Ref, RefCell};
use std::collections::{BTreeMap, HashSet};
use std::mem::take;
@@ -335,6 +336,10 @@ impl<'a: 'b, 'b> FunctionEditor<'a> {
self.function_id
}
+ pub fn node(&self, node: impl Borrow<NodeID>) -> &Node {
+ &self.function.nodes[node.borrow().idx()]
+ }
+
pub fn get_types(&self) -> Ref<'_, Vec<Type>> {
self.types.borrow()
}
@@ -351,6 +356,15 @@ impl<'a: 'b, 'b> FunctionEditor<'a> {
self.mut_def_use[id.idx()].iter().map(|x| *x)
}
+ pub fn get_uses(&self, id: NodeID) -> impl ExactSizeIterator<Item = NodeID> + '_ {
+ get_uses(&self.function.nodes[id.idx()])
+ .as_ref()
+ .into_iter()
+ .map(|x| *x)
+ .collect::<Vec<_>>()
+ .into_iter()
+ }
+
pub fn get_type(&self, id: TypeID) -> Ref<'_, Type> {
Ref::map(self.types.borrow(), |types| &types[id.idx()])
}
@@ -405,6 +419,10 @@ impl<'a, 'b> FunctionEdit<'a, 'b> {
self.editor.function.nodes.len() + self.added_nodeids.len()
}
+ pub fn copy_node(&mut self, node: NodeID) -> NodeID {
+ self.add_node(self.editor.func().nodes[node.idx()].clone())
+ }
+
pub fn add_node(&mut self, node: Node) -> NodeID {
let id = NodeID::new(self.num_node_ids());
// Added nodes need to have an entry in the def-use map.
diff --git a/hercules_opt/src/fork_concat_split.rs b/hercules_opt/src/fork_concat_split.rs
index 186cd6a6..bb3a2cff 100644
--- a/hercules_opt/src/fork_concat_split.rs
+++ b/hercules_opt/src/fork_concat_split.rs
@@ -7,7 +7,8 @@ use crate::*;
/*
* Split multi-dimensional fork-joins into separate one-dimensional fork-joins.
- * Useful for code generation.
+ * Useful for code generation. A single iteration of `fork_split` only splits
+ * at most one fork-join, it must be called repeatedly to split all fork-joins.
*/
pub fn fork_split(
editor: &mut FunctionEditor,
@@ -135,5 +136,6 @@ pub fn fork_split(
Ok(edit)
});
+ break;
}
}
diff --git a/hercules_opt/src/fork_guard_elim.rs b/hercules_opt/src/fork_guard_elim.rs
index 842c8308..1abb8967 100644
--- a/hercules_opt/src/fork_guard_elim.rs
+++ b/hercules_opt/src/fork_guard_elim.rs
@@ -1,8 +1,8 @@
use std::collections::{HashMap, HashSet};
-use hercules_ir::get_uses_mut;
-use hercules_ir::ir::*;
-use hercules_ir::ImmutableDefUseMap;
+use hercules_ir::*;
+
+use crate::*;
/*
* This is a Hercules IR transformation that:
@@ -17,6 +17,33 @@ use hercules_ir::ImmutableDefUseMap;
* guard remains and in these cases the guard is no longer needed.
*/
+// Simplify factors through max
+enum Factor {
+ Max(usize, DynamicConstantID),
+ Normal(DynamicConstantID),
+}
+
+impl Factor {
+ fn get_id(&self) -> DynamicConstantID {
+ match self {
+ Factor::Max(_, dynamic_constant_id) => *dynamic_constant_id,
+ Factor::Normal(dynamic_constant_id) => *dynamic_constant_id,
+ }
+ }
+}
+
+struct GuardedFork {
+ fork: NodeID,
+ join: NodeID,
+ guard_if: NodeID,
+ fork_taken_proj: NodeID,
+ fork_skipped_proj: NodeID,
+ guard_pred: NodeID,
+ guard_join_region: NodeID,
+ phi_reduce_map: HashMap<NodeID, NodeID>,
+ factor: Factor, // The factor that matches the guard
+}
+
/* Given a node index and the node itself, return None if the node is not
* a guarded fork where we can eliminate the guard.
* If the node is a fork with a guard we can eliminate returns a tuple of
@@ -28,27 +55,40 @@ use hercules_ir::ImmutableDefUseMap;
* - A map of NodeIDs for the phi nodes to the reduce they should be replaced
* with, and also the region that joins the guard's branches mapping to the
* fork's join NodeID
+ * - If the replication factor is a max that can be eliminated.
*/
fn guarded_fork(
- function: &Function,
- constants: &Vec<Constant>,
+ editor: &mut FunctionEditor,
fork_join_map: &HashMap<NodeID, NodeID>,
- def_use: &ImmutableDefUseMap,
- index: usize,
- node: &Node,
-) -> Option<(
- NodeID,
- Box<[DynamicConstantID]>,
- NodeID,
- NodeID,
- NodeID,
- NodeID,
- HashMap<NodeID, NodeID>,
-)> {
+ node: NodeID,
+) -> Option<GuardedFork> {
+ let function = editor.func();
+
// Identify fork nodes
- let Node::Fork { control, factors } = node else {
+ let Node::Fork { control, factors } = &function.nodes[node.idx()] else {
return None;
};
+
+ let mut factors = factors.iter().enumerate().map(|(idx, dc)| {
+ let DynamicConstant::Max(l, r) = *editor.get_dynamic_constant(*dc) else {
+ return Factor::Normal(*dc);
+ };
+
+ // There really needs to be a better way to work w/ associativity.
+ let binding = [(l, r), (r, l)];
+ let id = binding.iter().find_map(|(a, b)| {
+ let DynamicConstant::Constant(1) = *editor.get_dynamic_constant(*a) else {
+ return None;
+ };
+ Some(b)
+ });
+
+ match id {
+ Some(v) => Factor::Max(idx, *v),
+ None => Factor::Normal(*dc),
+ }
+ });
+
// Whose predecessor is a read from an if
let Node::Projection {
control: if_node,
@@ -70,47 +110,93 @@ fn guarded_fork(
return None;
};
let branch_idx = *selection;
- // branchIdx == 1 means the true branch so we want the condition to be
- // 0 < n or n > 0
- if branch_idx == 1
- && !((op == BinaryOperator::LT
- && function.nodes[left.idx()].is_zero_constant(constants)
- && factors
- .iter()
- .any(|factor| function.nodes[right.idx()].try_dynamic_constant() == Some(*factor)))
- || (op == BinaryOperator::GT
- && function.nodes[right.idx()].is_zero_constant(constants)
- && factors.iter().any(|factor| {
- function.nodes[left.idx()].try_dynamic_constant() == Some(*factor)
- })))
- {
- return None;
- }
- // branchIdx == 0 means the false branch so we want the condition to be
- // n < 0 or 0 > n
- if branch_idx == 0
- && !((op == BinaryOperator::LT
- && factors
- .iter()
- .any(|factor| function.nodes[left.idx()].try_dynamic_constant() == Some(*factor))
- && function.nodes[right.idx()].is_zero_constant(constants))
- || (op == BinaryOperator::GT
- && factors.iter().any(|factor| {
- function.nodes[right.idx()].try_dynamic_constant() == Some(*factor)
- })
- && function.nodes[left.idx()].is_zero_constant(constants)))
- {
- return None;
- }
+
+ let factor = {
+ // branchIdx == 1 means the true branch so we want the condition to be
+ // 0 < n or n > 0
+ if branch_idx == 1 {
+ [
+ (left, BinaryOperator::LT, right),
+ (right, BinaryOperator::GT, left),
+ ]
+ .iter()
+ .find_map(|(pattern_zero, pattern_op, pattern_factor)| {
+ // Match Op
+ if op != *pattern_op {
+ return None;
+ }
+ // Match Zero
+ if !(function.nodes[pattern_zero.idx()].is_zero_constant(&editor.get_constants())
+ || editor
+ .node(pattern_zero)
+ .is_zero_dc(&editor.get_dynamic_constants()))
+ {
+ return None;
+ }
+
+ // Match Factor
+ let factor = factors.find(|factor| {
+ match (
+ &function.nodes[pattern_factor.idx()],
+ &*editor.get_dynamic_constant(factor.get_id()),
+ ) {
+ (Node::Constant { id }, DynamicConstant::Constant(v)) => {
+ let Constant::UnsignedInteger64(pattern_v) = *editor.get_constant(*id)
+ else {
+ return false;
+ };
+ pattern_v == (*v as u64)
+ }
+ (Node::DynamicConstant { id }, _) => *id == factor.get_id(),
+ _ => false,
+ }
+ });
+ factor
+ })
+ }
+ // branchIdx == 0 means the false branch so we want the condition to be
+ // n < 0 or 0 > n
+ else if branch_idx == 0 {
+ [
+ (right, BinaryOperator::LT, left),
+ (left, BinaryOperator::GT, right),
+ ]
+ .iter()
+ .find_map(|(pattern_zero, pattern_op, pattern_factor)| {
+ // Match Op
+ if op != *pattern_op {
+ return None;
+ }
+ // Match Zero
+ if !(function.nodes[pattern_zero.idx()].is_zero_constant(&editor.get_constants())
+ || editor
+ .node(pattern_zero)
+ .is_zero_dc(&editor.get_dynamic_constants()))
+ {
+ return None;
+ }
+
+ // Match Factor
+ let factor = factors.find(|factor| {
+ function.nodes[pattern_factor.idx()].try_dynamic_constant()
+ == Some(factor.get_id())
+ });
+ factor
+ })
+ } else {
+ None
+ }
+ };
+
+ let Some(factor) = factor else { return None };
// Identify the join node and its users
- let join_id = fork_join_map.get(&NodeID::new(index))?;
- let join_users = def_use.get_users(*join_id);
+ let join_id = fork_join_map.get(&node)?;
// Find the unique control use of the join; if it's not a region we can't
// eliminate this guard
- let join_control = join_users
- .iter()
+ let join_control = editor
+ .get_users(*join_id)
.filter(|n| function.nodes[n.idx()].is_region())
.collect::<Vec<_>>();
if join_control.len() != 1 {
@@ -134,7 +220,7 @@ fn guarded_fork(
} else {
return None;
};
- // Other predecessor needs to be the other read from the guard's if
+ // Other predecessor needs to be the other projection from the guard's if
let Node::Projection {
control: if_node2,
ref selection,
@@ -152,14 +238,13 @@ fn guarded_fork(
// Finally, identify the phi nodes associated with the region and match
// them with the reduce nodes of the fork-join
- let reduce_nodes = join_users
- .iter()
+ let reduce_nodes = editor
+ .get_users(*join_id)
.filter(|n| function.nodes[n.idx()].is_reduce())
.collect::<HashSet<_>>();
// Construct a map from phi nodes indices to the reduce node index
- let phi_nodes = def_use
- .get_users(*join_control)
- .iter()
+ let phi_nodes = editor
+ .get_users(join_control)
.filter_map(|n| {
let Node::Phi {
control: _,
@@ -169,25 +254,25 @@ fn guarded_fork(
return None;
};
if data.len() != 2 {
- return Some((*n, None));
+ return Some((n, None));
}
let (init_idx, reduce_node) = if reduce_nodes.contains(&data[0]) {
(1, data[0])
} else if reduce_nodes.contains(&data[1]) {
(0, data[1])
} else {
- return Some((*n, None));
+ return Some((n, None));
};
let Node::Reduce {
control: _, init, ..
} = function.nodes[reduce_node.idx()]
else {
- return Some((*n, None));
+ return Some((n, None));
};
if data[init_idx] != init {
- return Some((*n, None));
+ return Some((n, None));
}
- Some((*n, Some(reduce_node)))
+ Some((n, Some(reduce_node)))
})
.collect::<HashMap<_, _>>();
@@ -197,69 +282,91 @@ fn guarded_fork(
return None;
}
- let mut phi_nodes = phi_nodes
+ let phi_nodes = phi_nodes
.into_iter()
.map(|(phi, red)| (phi, red.unwrap()))
.collect::<HashMap<_, _>>();
- // We also add a map from the region to the join to this map so we only
- // need one map to handle all node replacements in the elimination process
- phi_nodes.insert(*join_control, *join_id);
-
// Finally, we return this node's index along with
// - The replication factor of the fork
// - The if node
// - The true and false reads of the if
// - The guard's predecessor
// - The map from phi nodes to reduce nodes and the region to the join
- Some((
- NodeID::new(index),
- factors.clone(),
- if_node,
- *control,
- other_pred,
- if_pred,
- phi_nodes,
- ))
+ Some(GuardedFork {
+ fork: node,
+ join: *join_id,
+ guard_if: if_node,
+ fork_taken_proj: *control,
+ fork_skipped_proj: other_pred,
+ guard_pred: if_pred,
+ guard_join_region: join_control,
+ phi_reduce_map: phi_nodes,
+ factor,
+ })
}
/*
* Top level function to run fork guard elimination, as described above.
- * Deletes nodes by setting nodes to gravestones. Works with a function already
- * containing gravestones.
*/
-pub fn fork_guard_elim(
- function: &mut Function,
- constants: &Vec<Constant>,
- fork_join_map: &HashMap<NodeID, NodeID>,
- def_use: &ImmutableDefUseMap,
-) {
- let guard_info = function
- .nodes
- .iter()
- .enumerate()
- .filter_map(|(i, n)| guarded_fork(function, constants, fork_join_map, def_use, i, n))
+pub fn fork_guard_elim(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, NodeID>) {
+ let guard_info = editor
+ .node_ids()
+ .filter_map(|node| guarded_fork(editor, fork_join_map, node))
.collect::<Vec<_>>();
- for (fork_node, factors, guard_node, guard_proj1, guard_proj2, guard_pred, map) in guard_info {
- function.nodes[guard_node.idx()] = Node::Start;
- function.nodes[guard_proj1.idx()] = Node::Start;
- function.nodes[guard_proj2.idx()] = Node::Start;
- function.nodes[fork_node.idx()] = Node::Fork {
- control: guard_pred,
- factors,
+ for GuardedFork {
+ fork,
+ join,
+ fork_taken_proj,
+ fork_skipped_proj,
+ guard_pred,
+ phi_reduce_map,
+ factor,
+ guard_if,
+ guard_join_region,
+ } in guard_info
+ {
+ let new_fork_info = if let Factor::Max(idx, dc) = factor {
+ let Node::Fork {
+ control: _,
+ mut factors,
+ } = editor.func().nodes[fork.idx()].clone()
+ else {
+ unreachable!()
+ };
+ factors[idx] = dc;
+ let new_fork = Node::Fork {
+ control: guard_pred,
+ factors,
+ };
+ Some(new_fork)
+ } else {
+ None
};
- for (idx, node) in function.nodes.iter_mut().enumerate() {
- let node_idx = NodeID::new(idx);
- if map.contains_key(&node_idx) {
- *node = Node::Start;
+ editor.edit(|mut edit| {
+ edit =
+ edit.replace_all_uses_where(fork_taken_proj, guard_pred, |usee| *usee == fork)?;
+ edit = edit.delete_node(guard_if)?;
+ edit = edit.delete_node(fork_taken_proj)?;
+ edit = edit.delete_node(fork_skipped_proj)?;
+ edit = edit.replace_all_uses(guard_join_region, join)?;
+ edit = edit.delete_node(guard_join_region)?;
+ // Delete region node
+
+ for (phi, reduce) in phi_reduce_map.iter() {
+ edit = edit.replace_all_uses(*phi, *reduce)?;
+ edit = edit.delete_node(*phi)?;
}
- for u in get_uses_mut(node).as_mut() {
- if let Some(replacement) = map.get(u) {
- **u = *replacement;
- }
+
+ if let Some(new_fork_info) = new_fork_info {
+ let new_fork = edit.add_node(new_fork_info);
+ edit = edit.replace_all_uses(fork, new_fork)?;
+ edit = edit.delete_node(fork)?;
}
- }
+
+ Ok(edit)
+ });
}
}
diff --git a/hercules_opt/src/fork_transforms.rs b/hercules_opt/src/fork_transforms.rs
new file mode 100644
index 00000000..a4605bec
--- /dev/null
+++ b/hercules_opt/src/fork_transforms.rs
@@ -0,0 +1,540 @@
+use std::collections::{HashMap, HashSet};
+
+use bimap::BiMap;
+use itertools::Itertools;
+
+use hercules_ir::*;
+
+use crate::*;
+
+type ForkID = usize;
+
+/** Places each reduce node into its own fork */
+pub fn default_reduce_partition(
+ editor: &FunctionEditor,
+ _fork: NodeID,
+ join: NodeID,
+) -> SparseNodeMap<ForkID> {
+ let mut map = SparseNodeMap::new();
+
+ editor
+ .get_users(join)
+ .filter(|id| editor.func().nodes[id.idx()].is_reduce())
+ .enumerate()
+ .for_each(|(fork, reduce)| {
+ map.insert(reduce, fork);
+ });
+
+ map
+}
+
+// TODO: Refine these conditions.
+/** */
+pub fn find_reduce_dependencies<'a>(
+ function: &'a Function,
+ reduce: NodeID,
+ fork: NodeID,
+) -> impl IntoIterator<Item = NodeID> + 'a {
+ let len = function.nodes.len();
+
+ let mut visited: DenseNodeMap<bool> = vec![false; len];
+ let mut depdendent: DenseNodeMap<bool> = vec![false; len];
+
+ // Does `fork` need to be a parameter here? It never changes. If this was a closure could it just capture it?
+ fn recurse(
+ function: &Function,
+ node: NodeID,
+ fork: NodeID,
+ dependent_map: &mut DenseNodeMap<bool>,
+ visited: &mut DenseNodeMap<bool>,
+ ) -> () {
+ // return through dependent_map {
+
+ if visited[node.idx()] {
+ return;
+ }
+
+ visited[node.idx()] = true;
+
+ if node == fork {
+ dependent_map[node.idx()] = true;
+ return;
+ }
+
+ let binding = get_uses(&function.nodes[node.idx()]);
+ let uses = binding.as_ref();
+
+ for used in uses {
+ recurse(function, *used, fork, dependent_map, visited);
+ }
+
+ dependent_map[node.idx()] = uses.iter().map(|id| dependent_map[id.idx()]).any(|a| a);
+ return;
+ }
+
+ // Note: HACKY, the condition wwe want is 'all nodes on any path from the fork to the reduce (in the forward graph), or the reduce to the fork (in the directed graph)
+ // cycles break this, but we assume for now that the only cycles are ones that involve the reduce node
+ // NOTE: (control may break this (i.e loop inside fork) is a cycle that isn't the reduce)
+ // the current solution is just to mark the reduce as dependent at the start of traversing the graph.
+ depdendent[reduce.idx()] = true;
+
+ recurse(function, reduce, fork, &mut depdendent, &mut visited);
+
+ // Return node IDs that are dependent
+ let ret_val: Vec<_> = depdendent
+ .iter()
+ .enumerate()
+ .filter_map(|(idx, dependent)| {
+ if *dependent {
+ Some(NodeID::new(idx))
+ } else {
+ None
+ }
+ })
+ .collect();
+
+ ret_val
+}
+
+pub fn copy_subgraph(
+ editor: &mut FunctionEditor,
+ subgraph: HashSet<NodeID>,
+) -> (
+ HashSet<NodeID>,
+ HashMap<NodeID, NodeID>,
+ Vec<(NodeID, NodeID)>,
+) // returns all new nodes, a map from old nodes to new nodes, and
+ // a vec of pairs of nodes (old node, outside node) s.t old node -> outside node,
+ // outside means not part of the original subgraph.
+{
+ let mut map: HashMap<NodeID, NodeID> = HashMap::new();
+ let mut new_nodes: HashSet<NodeID> = HashSet::new();
+
+ // Copy nodes
+ for old_id in subgraph.iter() {
+ editor.edit(|mut edit| {
+ let new_id = edit.copy_node(*old_id);
+ map.insert(*old_id, new_id);
+ new_nodes.insert(new_id);
+ Ok(edit)
+ });
+ }
+
+ // Update edges to new nodes
+ for old_id in subgraph.iter() {
+ // Replace all uses of old_id w/ new_id, where the use is in new_node
+ editor.edit(|edit| {
+ edit.replace_all_uses_where(*old_id, map[old_id], |node_id| new_nodes.contains(node_id))
+ });
+ }
+
+ // Get all users that aren't in new_nodes.
+ let mut outside_users = Vec::new();
+
+ for node in new_nodes.iter() {
+ for user in editor.get_users(*node) {
+ if !new_nodes.contains(&user) {
+ outside_users.push((*node, user));
+ }
+ }
+ }
+
+ (new_nodes, map, outside_users)
+}
+
+pub fn fork_fission<'a>(
+ editor: &'a mut FunctionEditor,
+ _control_subgraph: &Subgraph,
+ _types: &Vec<TypeID>,
+ _loop_tree: &LoopTree,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+) -> () {
+ let forks: Vec<_> = editor
+ .func()
+ .nodes
+ .iter()
+ .enumerate()
+ .filter_map(|(idx, node)| {
+ if node.is_fork() {
+ Some(NodeID::new(idx))
+ } else {
+ None
+ }
+ })
+ .collect();
+
+ let control_pred = NodeID::new(0);
+
+ // This does the reduction fission:
+ for fork in forks.clone() {
+ // FIXME: If there is control in between fork and join, don't just give up.
+ let join = fork_join_map[&fork];
+ let join_pred = editor.func().nodes[join.idx()].try_join().unwrap();
+ if join_pred != fork {
+ todo!("Can't do fork fission on nodes with internal control")
+ // Inner control LOOPs are hard
+ // inner control in general *should* work right now without modifications.
+ }
+ let reduce_partition = default_reduce_partition(editor, fork, join);
+ fork_reduce_fission_helper(editor, fork_join_map, reduce_partition, control_pred, fork);
+ }
+}
+
+/** Split a 1D fork into two forks, placing select intermediate data into buffers. */
+pub fn fork_bufferize_fission_helper<'a>(
+ editor: &'a mut FunctionEditor,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+ bufferized_edges: HashSet<(NodeID, NodeID)>, // Describes what intermediate data should be bufferized.
+ _original_control_pred: NodeID, // What the new fork connects to.
+ types: &Vec<TypeID>,
+ fork: NodeID,
+) -> (NodeID, NodeID) {
+ // Returns the two forks that it generates.
+
+ // TODO: Check that bufferized edges src doesn't depend on anything that comes after the fork.
+
+ // Copy fork + control intermediates + join to new fork + join,
+ // How does control get partitioned?
+ // (depending on how it affects the data nodes on each side of the bufferized_edges)
+ // may end up in each loop, fix me later.
+ // place new fork + join after join of first.
+
+ // Only handle fork+joins with no inner control for now.
+
+ // Create fork + join + Thread control
+ let join = fork_join_map[&fork];
+ let mut new_fork_id = NodeID::new(0);
+ let mut new_join_id = NodeID::new(0);
+
+ editor.edit(|mut edit| {
+ new_join_id = edit.add_node(Node::Join { control: fork });
+ let factors = edit.get_node(fork).try_fork().unwrap().1;
+ new_fork_id = edit.add_node(Node::Fork {
+ control: new_join_id,
+ factors: factors.into(),
+ });
+ edit.replace_all_uses_where(fork, new_fork_id, |usee| *usee == join)
+ });
+
+ for (src, dst) in bufferized_edges {
+ // FIXME: Disgusting cloning and allocationing and iterators.
+ let factors: Vec<_> = editor.func().nodes[fork.idx()]
+ .try_fork()
+ .unwrap()
+ .1
+ .iter()
+ .cloned()
+ .collect();
+ editor.edit(|mut edit| {
+ // Create write to buffer
+
+ let thread_stuff_it = factors.into_iter().enumerate();
+
+ // FIxme: try to use unzip here? Idk why it wasn't working.
+ let tids = thread_stuff_it.clone().map(|(dim, _)| {
+ edit.add_node(Node::ThreadID {
+ control: fork,
+ dimension: dim,
+ })
+ });
+
+ let array_dims = thread_stuff_it.clone().map(|(_, factor)| (factor));
+
+ // Assume 1-d fork only for now.
+ // let tid = edit.add_node(Node::ThreadID { control: fork, dimension: 0 });
+ let position_idx = Index::Position(tids.collect::<Vec<_>>().into_boxed_slice());
+ let write = edit.add_node(Node::Write {
+ collect: NodeID::new(0),
+ data: src,
+ indices: vec![position_idx].into(),
+ });
+ let ele_type = types[src.idx()];
+ let empty_buffer = edit.add_type(hercules_ir::Type::Array(
+ ele_type,
+ array_dims.collect::<Vec<_>>().into_boxed_slice(),
+ ));
+ let empty_buffer = edit.add_zero_constant(empty_buffer);
+ let empty_buffer = edit.add_node(Node::Constant { id: empty_buffer });
+ let reduce = Node::Reduce {
+ control: new_join_id,
+ init: empty_buffer,
+ reduct: write,
+ };
+ let reduce = edit.add_node(reduce);
+ // Fix write node
+ edit = edit.replace_all_uses_where(NodeID::new(0), reduce, |usee| *usee == write)?;
+
+ // Create read from buffer
+ let tids = thread_stuff_it.clone().map(|(dim, _)| {
+ edit.add_node(Node::ThreadID {
+ control: new_fork_id,
+ dimension: dim,
+ })
+ });
+
+ let position_idx = Index::Position(tids.collect::<Vec<_>>().into_boxed_slice());
+
+ let read = edit.add_node(Node::Read {
+ collect: reduce,
+ indices: vec![position_idx].into(),
+ });
+
+ edit = edit.replace_all_uses_where(src, read, |usee| *usee == dst)?;
+
+ Ok(edit)
+ });
+ }
+
+ (fork, new_fork_id)
+}
+
+/** Split a 1D fork into a separate fork for each reduction. */
+pub fn fork_reduce_fission_helper<'a>(
+ editor: &'a mut FunctionEditor,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+ reduce_partition: SparseNodeMap<ForkID>, // Describes how the reduces of the fork should be split,
+ original_control_pred: NodeID, // What the new fork connects to.
+
+ fork: NodeID,
+) -> (NodeID, NodeID) {
+ let join = fork_join_map[&fork];
+
+ let mut new_control_pred: NodeID = original_control_pred;
+ // Important edges are: Reduces,
+
+ // NOTE:
+ // Say two reduce are in a fork, s.t reduce A depends on reduce B
+ // If user wants A and B in separate forks:
+ // - we can simply refuse
+ // - or we can duplicate B
+
+ let mut new_fork = NodeID::new(0);
+ let mut new_join = NodeID::new(0);
+
+ // Gets everything between fork & join that this reduce needs. (ALL CONTROL)
+ for reduce in reduce_partition {
+ let reduce = reduce.0;
+
+ let function = editor.func();
+ let subgraph = find_reduce_dependencies(function, reduce, fork);
+
+ let mut subgraph: HashSet<NodeID> = subgraph.into_iter().collect();
+
+ subgraph.insert(join);
+ subgraph.insert(fork);
+ subgraph.insert(reduce);
+
+ let (_, mapping, _) = copy_subgraph(editor, subgraph);
+
+ new_fork = mapping[&fork];
+ new_join = mapping[&join];
+
+ editor.edit(|mut edit| {
+ // Atttach new_fork after control_pred
+ let (old_control_pred, _) = edit.get_node(new_fork).try_fork().unwrap().clone();
+ edit = edit.replace_all_uses_where(old_control_pred, new_control_pred, |usee| {
+ *usee == new_fork
+ })?;
+
+ // Replace uses of reduce
+ edit = edit.replace_all_uses(reduce, mapping[&reduce])?;
+ Ok(edit)
+ });
+
+ new_control_pred = new_join;
+ }
+
+ editor.edit(|mut edit| {
+ // Replace original join w/ new final join
+ edit = edit.replace_all_uses_where(join, new_join, |_| true)?;
+
+ // Delete original join (all reduce users have been moved)
+ edit = edit.delete_node(join)?;
+
+ // Replace all users of original fork, and then delete it, leftover users will be DCE'd.
+ edit = edit.replace_all_uses(fork, new_fork)?;
+ edit.delete_node(fork)
+ });
+
+ (new_fork, new_join)
+}
+
+pub fn fork_coalesce(
+ editor: &mut FunctionEditor,
+ loops: &LoopTree,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+) -> bool {
+ let fork_joins = loops.bottom_up_loops().into_iter().filter_map(|(k, _)| {
+ if editor.func().nodes[k.idx()].is_fork() {
+ Some(k)
+ } else {
+ None
+ }
+ });
+
+ let fork_joins: Vec<_> = fork_joins.collect();
+ // FIXME: Add a postorder traversal to optimize this.
+
+ // FIXME: This could give us two forks that aren't actually ancestors / related, but then the helper will just return false early.
+ // something like: `fork_joins.postorder_iter().windows(2)` is ideal here.
+ for (inner, outer) in fork_joins.iter().cartesian_product(fork_joins.iter()) {
+ if fork_coalesce_helper(editor, *outer, *inner, fork_join_map) {
+ return true;
+ }
+ }
+ return false;
+}
+
+/** Opposite of fork split, takes two fork-joins
+ with no control between them, and merges them into a single fork-join.
+*/
+pub fn fork_coalesce_helper(
+ editor: &mut FunctionEditor,
+ outer_fork: NodeID,
+ inner_fork: NodeID,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+) -> bool {
+ // Check that all reduces in the outer fork are in *simple* cycles with a unique reduce of the inner fork.
+
+ let outer_join = fork_join_map[&outer_fork];
+ let inner_join = fork_join_map[&inner_fork];
+
+ let mut pairs: BiMap<NodeID, NodeID> = BiMap::new(); // Outer <-> Inner
+
+ // FIXME: Iterate all control uses of joins to really collect all reduces
+ // (reduces can be attached to inner control)
+ for outer_reduce in editor
+ .get_users(outer_join)
+ .filter(|node| editor.func().nodes[node.idx()].is_reduce())
+ {
+ // check that inner reduce is of the inner join
+ let (_, _, outer_reduct) = editor.func().nodes[outer_reduce.idx()]
+ .try_reduce()
+ .unwrap();
+
+ let inner_reduce = outer_reduct;
+ let inner_reduce_node = &editor.func().nodes[outer_reduct.idx()];
+
+ let Node::Reduce {
+ control: inner_control,
+ init: inner_init,
+ reduct: _,
+ } = inner_reduce_node
+ else {
+ return false;
+ };
+
+ // FIXME: check this condition better (i.e reduce might not be attached to join)
+ if *inner_control != inner_join {
+ return false;
+ };
+ if *inner_init != outer_reduce {
+ return false;
+ };
+
+ if pairs.contains_left(&outer_reduce) || pairs.contains_right(&inner_reduce) {
+ return false;
+ } else {
+ pairs.insert(outer_reduce, inner_reduce);
+ }
+ }
+
+ // Check for control between join-join and fork-fork
+ let Some(user) = editor
+ .get_users(outer_fork)
+ .filter(|node| editor.func().nodes[node.idx()].is_control())
+ .next()
+ else {
+ return false;
+ };
+
+ if user != inner_fork {
+ return false;
+ }
+
+ let Some(user) = editor
+ .get_users(inner_join)
+ .filter(|node| editor.func().nodes[node.idx()].is_control())
+ .next()
+ else {
+ return false;
+ };
+
+ if user != outer_join {
+ return false;
+ }
+
+ // Checklist:
+ // Increment inner TIDs
+ // Add outer fork's dimension to front of inner fork.
+ // Fuse reductions
+ // - Initializer becomes outer initializer
+ // Replace uses of outer fork w/ inner fork.
+ // Replace uses of outer join w/ inner join.
+ // Delete outer fork-join
+
+ let inner_tids: Vec<NodeID> = editor
+ .get_users(inner_fork)
+ .filter(|node| editor.func().nodes[node.idx()].is_thread_id())
+ .collect();
+
+ let (outer_pred, outer_dims) = editor.func().nodes[outer_fork.idx()].try_fork().unwrap();
+ let (_, inner_dims) = editor.func().nodes[inner_fork.idx()].try_fork().unwrap();
+ let num_outer_dims = outer_dims.len();
+ let mut new_factors = outer_dims.to_vec();
+
+ // CHECKME / FIXME: Might need to be added the other way.
+ new_factors.append(&mut inner_dims.to_vec());
+
+ for tid in inner_tids {
+ let (fork, dim) = editor.func().nodes[tid.idx()].try_thread_id().unwrap();
+ let new_tid = Node::ThreadID {
+ control: fork,
+ dimension: dim + num_outer_dims,
+ };
+
+ editor.edit(|mut edit| {
+ let new_tid = edit.add_node(new_tid);
+ let edit = edit.replace_all_uses(tid, new_tid)?;
+ Ok(edit)
+ });
+ }
+
+ // Fuse Reductions
+ for (outer_reduce, inner_reduce) in pairs {
+ let (_, outer_init, _) = editor.func().nodes[outer_reduce.idx()]
+ .try_reduce()
+ .unwrap();
+ let (_, inner_init, _) = editor.func().nodes[inner_reduce.idx()]
+ .try_reduce()
+ .unwrap();
+ editor.edit(|mut edit| {
+ // Set inner init to outer init.
+ edit =
+ edit.replace_all_uses_where(inner_init, outer_init, |usee| *usee == inner_reduce)?;
+ edit = edit.replace_all_uses(outer_reduce, inner_reduce)?;
+ edit = edit.delete_node(outer_reduce)?;
+
+ Ok(edit)
+ });
+ }
+
+ editor.edit(|mut edit| {
+ let new_fork = Node::Fork {
+ control: outer_pred,
+ factors: new_factors.into(),
+ };
+ let new_fork = edit.add_node(new_fork);
+
+ edit = edit.replace_all_uses(inner_fork, new_fork)?;
+ edit = edit.replace_all_uses(outer_fork, new_fork)?;
+ edit = edit.replace_all_uses(outer_join, inner_join)?;
+ edit = edit.delete_node(outer_join)?;
+ edit = edit.delete_node(inner_fork)?;
+ edit = edit.delete_node(outer_fork)?;
+
+ Ok(edit)
+ });
+
+ true
+}
diff --git a/hercules_opt/src/forkify.rs b/hercules_opt/src/forkify.rs
index fb53a5e4..ce9ac141 100644
--- a/hercules_opt/src/forkify.rs
+++ b/hercules_opt/src/forkify.rs
@@ -1,239 +1,547 @@
+use std::collections::HashMap;
+use std::collections::HashSet;
use std::iter::zip;
+use std::iter::FromIterator;
-use hercules_ir::def_use::*;
-use hercules_ir::ir::*;
-use hercules_ir::loops::*;
+use itertools::Itertools;
+use nestify::nest;
-/*
- * Top level function to convert natural loops with simple induction variables
- * into fork-joins.
+use hercules_ir::*;
+
+use crate::*;
+
+/*
+ * TODO: Forkify currently makes a bunch of small edits - this needs to be
+ * changed so that every loop that gets forkified corresponds to a single edit
+ * + sub-edits. This would allow us to run forkify on a subset of a function.
*/
pub fn forkify(
- function: &mut Function,
- constants: &Vec<Constant>,
- dynamic_constants: &mut Vec<DynamicConstant>,
- def_use: &ImmutableDefUseMap,
+ editor: &mut FunctionEditor,
+ control_subgraph: &Subgraph,
+ fork_join_map: &HashMap<NodeID, NodeID>,
loops: &LoopTree,
-) {
- // Ignore loops that are already fork-joins. TODO: re-calculate def_use per
- // loop, since it's technically invalidated after each individual loop
- // modification.
+) -> bool {
let natural_loops = loops
.bottom_up_loops()
.into_iter()
- .rev()
- .filter(|(k, _)| function.nodes[k.idx()].is_region());
+ .filter(|(k, _)| editor.func().nodes[k.idx()].is_region());
+
+ let natural_loops: Vec<_> = natural_loops.collect();
+
+ for l in natural_loops {
+ // FIXME: Run on all-bottom level loops, as they can be independently optimized without recomputing analyses.
+ if forkify_loop(
+ editor,
+ control_subgraph,
+ fork_join_map,
+ &Loop {
+ header: l.0,
+ control: l.1.clone(),
+ },
+ ) {
+ return true;
+ }
+ }
+ return false;
+}
+
+/** Given a node used as a loop bound, return a dynamic constant ID. */
+pub fn get_node_as_dc(
+ editor: &mut FunctionEditor,
+ node: NodeID,
+) -> Result<DynamicConstantID, String> {
+ // Check for a constant used as loop bound.
+ match editor.node(node) {
+ Node::DynamicConstant {
+ id: dynamic_constant_id,
+ } => Ok(*dynamic_constant_id),
+ Node::Constant { id: constant_id } => {
+ let dc = match *editor.get_constant(*constant_id) {
+ Constant::Integer8(x) => DynamicConstant::Constant(x as _),
+ Constant::Integer16(x) => DynamicConstant::Constant(x as _),
+ Constant::Integer32(x) => DynamicConstant::Constant(x as _),
+ Constant::Integer64(x) => DynamicConstant::Constant(x as _),
+ Constant::UnsignedInteger8(x) => DynamicConstant::Constant(x as _),
+ Constant::UnsignedInteger16(x) => DynamicConstant::Constant(x as _),
+ Constant::UnsignedInteger32(x) => DynamicConstant::Constant(x as _),
+ Constant::UnsignedInteger64(x) => DynamicConstant::Constant(x as _),
+ _ => return Err("Invalid constant as loop bound".to_string()),
+ };
+
+ let mut b = DynamicConstantID::new(0);
+ editor.edit(|mut edit| {
+ b = edit.add_dynamic_constant(dc);
+ Ok(edit)
+ });
+ // Return the ID of the dynamic constant that is generated from the constant
+ // or dynamic constant that is the existing loop bound
+ Ok(b)
+ }
+ _ => Err("Blah".to_owned()),
+ }
+}
+
+/**
+ Top level function to convert natural loops with simple induction variables
+ into fork-joins.
+*/
+pub fn forkify_loop(
+ editor: &mut FunctionEditor,
+ control_subgraph: &Subgraph,
+ _fork_join_map: &HashMap<NodeID, NodeID>,
+ l: &Loop,
+) -> bool {
+ let function = editor.func();
+
+ let Some(loop_condition) = get_loop_exit_conditions(function, l, control_subgraph) else {
+ return false;
+ };
+
+ let LoopExit::Conditional {
+ if_node: loop_if,
+ condition_node,
+ } = loop_condition.clone()
+ else {
+ return false;
+ };
+
+ // Compute loop variance
+ let loop_variance = compute_loop_variance(editor, l);
+ let ivs = compute_induction_vars(editor.func(), l, &loop_variance);
+ let ivs = compute_iv_ranges(editor, l, ivs, &loop_condition);
+ let Some(canonical_iv) = has_canonical_iv(editor, l, &ivs) else {
+ return false;
+ };
- // Detect loops that have a simple loop induction variable. TODO: proper
- // affine analysis to recognize other cases of linear induction variables.
- let affine_loops: Vec<_> = natural_loops
+ // FIXME: Make sure IV is not used outside the loop.
+
+ // Get bound
+ let bound = match canonical_iv {
+ InductionVariable::Basic {
+ node: _,
+ initializer: _,
+ update: _,
+ final_value,
+ } => final_value
+ .map(|final_value| get_node_as_dc(editor, final_value))
+ .and_then(|r| r.ok()),
+ InductionVariable::SCEV(_) => return false,
+ };
+
+ let Some(bound_dc_id) = bound else {
+ return false;
+ };
+
+ let function = editor.func();
+
+ // Check if it is do-while loop.
+ let loop_exit_projection = editor
+ .get_users(loop_if)
+ .filter(|id| !l.control[id.idx()])
+ .next()
+ .unwrap();
+
+ let loop_continue_projection = editor
+ .get_users(loop_if)
+ .filter(|id| l.control[id.idx()])
+ .next()
+ .unwrap();
+
+ let loop_preds: Vec<_> = editor
+ .get_uses(l.header)
+ .filter(|id| !l.control[id.idx()])
+ .collect();
+
+ if loop_preds.len() != 1 {
+ return false;
+ }
+
+ let loop_pred = loop_preds[0];
+
+ if !editor
+ .get_uses(l.header)
+ .contains(&loop_continue_projection)
+ {
+ return false;
+ }
+
+ // Get all phis used outside of the loop, they need to be reductionable.
+ // For now just assume all phis will be phis used outside of the loop, except for the canonical iv.
+ // FIXME: We need a different definiton of `loop_nodes` to check for phis used outside hte loop than the one
+ // we currently have.
+ let loop_nodes = calculate_loop_nodes(editor, l);
+
+ // Check phis to see if they are reductionable, only PHIs depending on the loop are considered,
+ let candidate_phis: Vec<_> = editor
+ .get_users(l.header)
+ .filter(|id| function.nodes[id.idx()].is_phi())
+ .filter(|id| *id != canonical_iv.phi())
+ .collect();
+
+ let reductionable_phis: Vec<_> = analyze_phis(&editor, &l, &candidate_phis, &loop_nodes)
.into_iter()
- .filter_map(|(header, contents)| {
- // Get the single loop contained predecessor of the loop header.
- let header_uses = get_uses(&function.nodes[header.idx()]);
- let mut pred_loop = header_uses.as_ref().iter().filter(|id| contents[id.idx()]);
- let single_pred_loop = pred_loop.next()?;
- if pred_loop.next().is_some() || header_uses.as_ref().len() != 2 {
- return None;
- }
+ .collect();
- // Check for a very particular loop indexing structure.
- let if_ctrl = function.nodes[single_pred_loop.idx()].try_projection(1)?;
- let (_, if_cond) = function.nodes[if_ctrl.idx()].try_if()?;
- let (idx, bound) = function.nodes[if_cond.idx()].try_binary(BinaryOperator::LT)?;
- let (phi, one) = function.nodes[idx.idx()].try_binary(BinaryOperator::Add)?;
- let (should_be_header, pred_datas) = function.nodes[phi.idx()].try_phi()?;
- let one_c_id = function.nodes[one.idx()].try_constant()?;
+ // TODO: Handle multiple loop body lasts.
+ // If there are multiple candidates for loop body last, return false.
+ if editor
+ .get_uses(loop_if)
+ .filter(|id| l.control[id.idx()])
+ .count()
+ > 1
+ {
+ return false;
+ }
- if should_be_header != header || !constants[one_c_id.idx()].is_one() {
- return None;
- }
+ let loop_body_last = editor.get_uses(loop_if).next().unwrap();
- // Check that phi's if predecessor is the add node, and check that the
- // phi's other predecessors are zeros.
- zip(header_uses.as_ref().iter(), pred_datas.iter())
- .position(|(c, d)| *c == *single_pred_loop && *d == idx)?;
- if zip(header_uses.as_ref().iter(), pred_datas.iter())
- .filter(|(c, d)| {
- (**c != *single_pred_loop)
- && !function.nodes[d.idx()].is_zero_constant(constants)
- })
- .count()
- != 0
- {
- return None;
- }
+ if reductionable_phis
+ .iter()
+ .any(|phi| !matches!(phi, LoopPHI::Reductionable { .. }))
+ {
+ return false;
+ }
- // Check for constant used as loop bound. Do this last, since we may
- // create a new dynamic constant here.
- let bound_dc_id =
- if let Some(bound_dc_id) = function.nodes[bound.idx()].try_dynamic_constant() {
- bound_dc_id
- } else if let Some(bound_c_id) = function.nodes[bound.idx()].try_constant() {
- // Create new dynamic constant that reflects this constant.
- let dc = match constants[bound_c_id.idx()] {
- Constant::Integer8(x) => DynamicConstant::Constant(x as _),
- Constant::Integer16(x) => DynamicConstant::Constant(x as _),
- Constant::Integer32(x) => DynamicConstant::Constant(x as _),
- Constant::Integer64(x) => DynamicConstant::Constant(x as _),
- Constant::UnsignedInteger8(x) => DynamicConstant::Constant(x as _),
- Constant::UnsignedInteger16(x) => DynamicConstant::Constant(x as _),
- Constant::UnsignedInteger32(x) => DynamicConstant::Constant(x as _),
- Constant::UnsignedInteger64(x) => DynamicConstant::Constant(x as _),
- _ => return None,
- };
-
- // The new dynamic constant may already be interned.
- let maybe_already_in = dynamic_constants
- .iter()
- .enumerate()
- .find(|(_, x)| **x == dc)
- .map(|(idx, _)| idx);
- if let Some(bound_dc_idx) = maybe_already_in {
- DynamicConstantID::new(bound_dc_idx)
- } else {
- let id = DynamicConstantID::new(dynamic_constants.len());
- dynamic_constants.push(dc);
- id
- }
- } else {
- return None;
- };
+ let phi_latches: Vec<_> = reductionable_phis
+ .iter()
+ .map(|phi| {
+ let LoopPHI::Reductionable {
+ phi: _,
+ data_cycle: _,
+ continue_latch,
+ is_associative: _,
+ } = phi
+ else {
+ unreachable!()
+ };
+ continue_latch
+ })
+ .collect();
- Some((header, phi, contents, bound_dc_id))
+ let stop_on: HashSet<_> = editor
+ .node_ids()
+ .filter(|node| {
+ if editor.node(node).is_phi() {
+ return true;
+ }
+ if editor.node(node).is_reduce() {
+ return true;
+ }
+ if editor.node(node).is_control() {
+ return true;
+ }
+ if phi_latches.contains(&node) {
+ return true;
+ }
+
+ false
})
.collect();
- // Convert affine loops into fork-joins.
- for (header, idx_phi, contents, dc_id) in affine_loops {
- let header_uses = get_uses(&function.nodes[header.idx()]);
- let header_uses: Vec<_> = header_uses.as_ref().into_iter().map(|x| *x).collect();
+ // Outside loop users of IV, then exit;
+ // Unless the outside user is through the loop latch of a reducing phi,
+ // then we know how to replace this edge, so its fine!
+ let iv_users: Vec<_> =
+ walk_all_users_stop_on(canonical_iv.phi(), editor, stop_on.clone()).collect();
- // Get the control portions of the loop that need to be grafted.
- let loop_pred = header_uses
- .iter()
- .filter(|id| !contents[id.idx()])
- .next()
- .unwrap();
- let loop_true_read = header_uses
- .iter()
- .filter(|id| contents[id.idx()])
- .next()
- .unwrap();
- let loop_end = function.nodes[loop_true_read.idx()]
- .try_projection(1)
- .unwrap();
- let loop_false_read = *def_use
- .get_users(loop_end)
- .iter()
- .filter_map(|id| {
- if function.nodes[id.idx()].try_projection(0).is_some() {
- Some(id)
- } else {
- None
- }
- })
- .next()
- .unwrap();
+ if iv_users
+ .iter()
+ .any(|node| !loop_nodes.contains(&node) && *node != loop_if)
+ {
+ return false;
+ }
+
+ // Start Transformation:
+
+ // Graft everything between header and loop condition
+ // Attach join to right before header (after loop_body_last, unless loop body last *is* the header).
+ // Attach fork to right after loop_continue_projection.
+
+ // // Create fork and join nodes:
+ let mut join_id = NodeID::new(0);
+ let mut fork_id = NodeID::new(0);
- // Create fork and join nodes.
+ // Turn dc bound into max (1, bound),
+ let bound_dc_id = {
+ let mut max_id = DynamicConstantID::new(0);
+ editor.edit(|mut edit| {
+ // FIXME: Maybe add_dynamic_constant should intern?
+ let one_id = edit.add_dynamic_constant(DynamicConstant::Constant(1));
+ max_id = edit.add_dynamic_constant(DynamicConstant::Max(one_id, bound_dc_id));
+ Ok(edit)
+ });
+ max_id
+ };
+
+ // FIXME: (@xrouth) double check handling of control in loop body.
+ editor.edit(|mut edit| {
let fork = Node::Fork {
- control: *loop_pred,
- factors: Box::new([dc_id]),
+ control: loop_pred,
+ factors: Box::new([bound_dc_id]),
};
- let fork_id = NodeID::new(function.nodes.len());
- function.nodes.push(fork);
+ fork_id = edit.add_node(fork);
let join = Node::Join {
- control: if header == get_uses(&function.nodes[loop_end.idx()]).as_ref()[0] {
+ control: if l.header == loop_body_last {
fork_id
} else {
- function.nodes[loop_end.idx()].try_if().unwrap().0
+ loop_body_last
},
};
- let join_id = NodeID::new(function.nodes.len());
- function.nodes.push(join);
-
- // Convert reducing phi nodes to reduce nodes.
- let reduction_phis: Vec<_> = def_use
- .get_users(header)
- .iter()
- .filter(|id| **id != idx_phi && function.nodes[id.idx()].is_phi())
- .collect();
- for reduction_phi in reduction_phis {
- // Loop predecessor input to phi is the reduction initializer.
- let init = *zip(
- header_uses.iter(),
- function.nodes[reduction_phi.idx()]
- .try_phi()
- .unwrap()
- .1
- .iter(),
- )
- .filter(|(c, _)| **c == *loop_pred)
- .next()
- .unwrap()
- .1;
-
- // Loop back edge input to phi is the reduction induction variable.
- let reduct = *zip(
- header_uses.iter(),
- function.nodes[reduction_phi.idx()]
- .try_phi()
- .unwrap()
- .1
- .iter(),
- )
- .filter(|(c, _)| **c == *loop_true_read)
- .next()
- .unwrap()
- .1;
-
- // Create reduction node.
+
+ join_id = edit.add_node(join);
+
+ Ok(edit)
+ });
+
+ let function = editor.func();
+ let (_, factors) = function.nodes[fork_id.idx()].try_fork().unwrap();
+ let dimension = factors.len() - 1;
+
+ // Create ThreadID
+ editor.edit(|mut edit| {
+ let thread_id = Node::ThreadID {
+ control: fork_id,
+ dimension: dimension,
+ };
+ let thread_id_id = edit.add_node(thread_id);
+
+ // Replace uses that are inside with the thread id
+ edit = edit.replace_all_uses_where(canonical_iv.phi(), thread_id_id, |node| {
+ loop_nodes.contains(node)
+ })?;
+
+ // Replace uses that are outside with DC - 1. Or just give up.
+ let bound_dc_node = edit.add_node(Node::DynamicConstant { id: bound_dc_id });
+ edit = edit.replace_all_uses_where(canonical_iv.phi(), bound_dc_node, |node| {
+ !loop_nodes.contains(node)
+ })?;
+
+ edit.delete_node(canonical_iv.phi())
+ });
+
+ for reduction_phi in reductionable_phis {
+ let LoopPHI::Reductionable {
+ phi,
+ data_cycle: _,
+ continue_latch,
+ is_associative: _,
+ } = reduction_phi
+ else {
+ panic!();
+ };
+
+ let function = editor.func();
+
+ let init = *zip(
+ editor.get_uses(l.header),
+ function.nodes[phi.idx()].try_phi().unwrap().1.iter(),
+ )
+ .filter(|(c, _)| *c == loop_pred)
+ .next()
+ .unwrap()
+ .1;
+
+ editor.edit(|mut edit| {
let reduce = Node::Reduce {
control: join_id,
init,
- reduct,
+ reduct: continue_latch,
};
- let reduce_id = NodeID::new(function.nodes.len());
- function.nodes.push(reduce);
+ let reduce_id = edit.add_node(reduce);
- // Edit users of phis.
- for user in def_use.get_users(*reduction_phi) {
- get_uses_mut(&mut function.nodes[user.idx()]).map(*reduction_phi, reduce_id);
- }
+ edit = edit.replace_all_uses_where(phi, reduce_id, |usee| *usee != reduce_id)?;
+ edit = edit.replace_all_uses_where(continue_latch, reduce_id, |usee| {
+ !loop_nodes.contains(usee) && *usee != reduce_id
+ })?;
+ edit.delete_node(phi)
+ });
+ }
- // Edit users of uses of phis.
- for user in def_use.get_users(reduct) {
- get_uses_mut(&mut function.nodes[user.idx()]).map(reduct, reduce_id);
- }
+ // Replace all uses of the loop header with the fork
+ editor.edit(|edit| edit.replace_all_uses(l.header, fork_id));
- // Delete reducing phi.
- function.nodes[reduction_phi.idx()] = Node::Start;
- }
+ editor.edit(|edit| edit.replace_all_uses(loop_continue_projection, fork_id));
- // Convert index phi node to thread ID node.
- let thread_id = Node::ThreadID {
- control: fork_id,
- dimension: 0,
- };
- let thread_id_id = NodeID::new(function.nodes.len());
- function.nodes.push(thread_id);
+ editor.edit(|edit| edit.replace_all_uses(loop_exit_projection, join_id));
- for user in def_use.get_users(idx_phi) {
- get_uses_mut(&mut function.nodes[user.idx()]).map(idx_phi, thread_id_id);
- }
- for user in def_use.get_users(header) {
- get_uses_mut(&mut function.nodes[user.idx()]).map(header, fork_id);
- }
- for user in def_use.get_users(loop_false_read) {
- get_uses_mut(&mut function.nodes[user.idx()]).map(loop_false_read, join_id);
- }
+ // Get rid of loop condition
+ // DCE should get these, but delete them ourselves because we are nice :)
+ editor.edit(|mut edit| {
+ edit = edit.delete_node(loop_continue_projection)?;
+ edit = edit.delete_node(condition_node)?; // Might have to get rid of other users of this.
+ edit = edit.delete_node(loop_exit_projection)?;
+ edit = edit.delete_node(loop_if)?;
+ edit = edit.delete_node(l.header)?;
+ Ok(edit)
+ });
+
+ return true;
+}
- function.nodes[idx_phi.idx()] = Node::Start;
- function.nodes[header.idx()] = Node::Start;
- function.nodes[loop_end.idx()] = Node::Start;
- function.nodes[loop_true_read.idx()] = Node::Start;
- function.nodes[loop_false_read.idx()] = Node::Start;
+nest! {
+ #[derive(Debug)]
+ pub enum LoopPHI {
+ Reductionable {
+ phi: NodeID,
+ data_cycle: HashSet<NodeID>, // All nodes in a data cycle with this phi
+ continue_latch: NodeID,
+ is_associative: bool,
+ },
+ LoopDependant(NodeID),
+ UsedByDependant(NodeID),
}
}
+
+impl LoopPHI {
+ pub fn get_phi(&self) -> NodeID {
+ match self {
+ LoopPHI::Reductionable { phi, .. } => *phi,
+ LoopPHI::LoopDependant(node_id) => *node_id,
+ LoopPHI::UsedByDependant(node_id) => *node_id,
+ }
+ }
+}
+
+/**
+Checks some conditions on loop variables that will need to be converted into reductions to be forkified.
+ - The phi is in a cycle *in the loop* with itself.
+ - Every cycle *in the loop* containing the phi does not contain any other phi of the loop header.
+ - The phi does not immediatley (not blocked by another phi or another reduce) use any other phis of the loop header.
+ */
+pub fn analyze_phis<'a>(
+ editor: &'a FunctionEditor,
+ natural_loop: &'a Loop,
+ phis: &'a [NodeID],
+ loop_nodes: &'a HashSet<NodeID>,
+) -> impl Iterator<Item = LoopPHI> + 'a {
+
+ // Find data cycles within the loop of this phi,
+ // Start from the phis loop_continue_latch, and walk its uses until we find the original phi.
+
+ phis.into_iter().map(move |phi| {
+ let stop_on: HashSet<NodeID> = editor
+ .node_ids()
+ .filter(|node| {
+ let data = &editor.func().nodes[node.idx()];
+
+ // External Phi
+ if let Node::Phi { control, data: _ } = data {
+ if *control != natural_loop.header {
+ return true;
+ }
+ }
+
+ // This phi
+ if node == phi {
+ return true;
+ }
+
+ // External Reduce
+ if let Node::Reduce {
+ control,
+ init: _,
+ reduct: _,
+ } = data
+ {
+ if !natural_loop.control[control.idx()] {
+ return true;
+ } else {
+ return false;
+ }
+ }
+
+ // Data Cycles Only
+ if data.is_control() {
+ return true;
+ }
+
+ return false;
+ })
+ .collect();
+
+ let continue_idx = editor
+ .get_uses(natural_loop.header)
+ .position(|node| natural_loop.control[node.idx()])
+ .unwrap();
+
+ let loop_continue_latch = editor.node(phi).try_phi().unwrap().1[continue_idx];
+
+ let uses = walk_all_uses_stop_on(loop_continue_latch, editor, stop_on.clone());
+ let users = walk_all_users_stop_on(*phi, editor, stop_on.clone());
+
+ let other_stop_on: HashSet<NodeID> = editor
+ .node_ids()
+ .filter(|node| {
+ let data = &editor.func().nodes[node.idx()];
+
+ // Phi, Reduce
+ if data.is_phi() {
+ return true;
+ }
+
+ if data.is_reduce() {
+ return true;
+ }
+
+ // External Control
+ if data.is_control() {
+ return true;
+ }
+
+ return false;
+ })
+ .collect();
+
+ let mut uses_for_dependance =
+ walk_all_users_stop_on(loop_continue_latch, editor, other_stop_on);
+
+ let set1: HashSet<_> = HashSet::from_iter(uses);
+ let set2: HashSet<_> = HashSet::from_iter(users);
+
+ let intersection: HashSet<_> = set1.intersection(&set2).cloned().collect();
+
+ // If this phi uses any other phis the node is loop dependant,
+ // we use `phis` because this phi can actually contain the loop iv and its fine.
+ if uses_for_dependance.any(|node| phis.contains(&node) && node != *phi) {
+ LoopPHI::LoopDependant(*phi)
+ } else if intersection.clone().iter().next().is_some() {
+ // PHIs on the frontier of the uses by the candidate phi, i.e in uses_for_dependance need
+ // to have headers that postdominate the loop continue latch. The value of the PHI used needs to be defined
+ // by the time the reduce is triggered (at the end of the loop's internal control).
+
+ // No nodes in data cycles with this phi (in the loop) are used outside the loop, besides the loop_continue_latch.
+ // If some other node in the cycle is used, there is not a valid node to assign it after making the cycle a reduce.
+ if intersection
+ .iter()
+ .filter(|node| **node != loop_continue_latch )
+ .any(|data_node| {
+ editor
+ .get_users(*data_node)
+ .any(|user| !loop_nodes.contains(&user))
+ })
+ {
+ // This phi can be made into a reduce in different ways, if the cycle is associative (contains all the same kind of associative op)
+ // 3) Split the cycle into two phis, add them or multiply them together at the end.
+ // 4) Split the cycle into two reduces, add them or multiply them together at the end.
+ // Somewhere else should handle this.
+ return LoopPHI::LoopDependant(*phi);
+ }
+
+ // FIXME: Do we want to calculate associativity here, there might be a case where this information is used in forkify
+ // i.e as described above.
+ let is_associative = false;
+
+ // No nodes in the data cycle are used outside of the loop, besides the latched value of the phi
+ LoopPHI::Reductionable {
+ phi: *phi,
+ data_cycle: intersection,
+ continue_latch: loop_continue_latch,
+ is_associative,
+ }
+ } else {
+ // No cycles exist, this isn't a reduction.
+ LoopPHI::LoopDependant(*phi)
+ }
+ })
+}
diff --git a/hercules_opt/src/gcm.rs b/hercules_opt/src/gcm.rs
index b5822a51..462d1087 100644
--- a/hercules_opt/src/gcm.rs
+++ b/hercules_opt/src/gcm.rs
@@ -1022,7 +1022,7 @@ fn liveness_dataflow(
* device clones when a single node may potentially be on different devices.
*/
fn color_nodes(
- editor: &mut FunctionEditor,
+ _editor: &mut FunctionEditor,
reverse_postorder: &Vec<NodeID>,
objects: &FunctionCollectionObjects,
object_device_demands: &FunctionObjectDeviceDemands,
@@ -1138,7 +1138,7 @@ fn object_allocation(
typing: &Vec<TypeID>,
node_colors: &FunctionNodeColors,
alignments: &Vec<usize>,
- liveness: &Liveness,
+ _liveness: &Liveness,
backing_allocations: &BackingAllocations,
) -> FunctionBackingAllocation {
let mut fba = BTreeMap::new();
diff --git a/hercules_opt/src/ivar.rs b/hercules_opt/src/ivar.rs
new file mode 100644
index 00000000..f7252d29
--- /dev/null
+++ b/hercules_opt/src/ivar.rs
@@ -0,0 +1,602 @@
+use std::collections::HashSet;
+
+use bitvec::prelude::*;
+use nestify::nest;
+
+use hercules_ir::*;
+
+use crate::*;
+
+#[derive(Debug)]
+pub struct LoopVarianceInfo {
+ pub loop_header: NodeID,
+ pub map: DenseNodeMap<LoopVariance>,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq)]
+pub enum LoopVariance {
+ Unknown,
+ Invariant,
+ Variant,
+}
+
+type NodeVec = BitVec<u8, Lsb0>;
+
+#[derive(Clone, Debug)]
+pub struct Loop {
+ pub header: NodeID,
+ pub control: NodeVec, //
+}
+
+impl Loop {
+ pub fn get_all_nodes(&self) -> NodeVec {
+ let mut all_loop_nodes = self.control.clone();
+ all_loop_nodes.set(self.header.idx(), true);
+ all_loop_nodes
+ }
+}
+
+nest! {
+ #[derive(Clone, Copy, Debug, PartialEq)]*
+ pub enum InductionVariable {
+ pub Basic {
+ node: NodeID,
+ initializer: NodeID,
+ update: NodeID,
+ final_value: Option<NodeID>,
+ },
+ SCEV(NodeID), // TODO @(xrouth)
+ }
+}
+
+impl InductionVariable {
+ pub fn phi(&self) -> NodeID {
+ match self {
+ InductionVariable::Basic {
+ node,
+ initializer: _,
+ update: _,
+ final_value: _,
+ } => *node,
+ InductionVariable::SCEV(_) => todo!(),
+ }
+ }
+}
+
+// TODO: Optimize.
+pub fn calculate_loop_nodes(editor: &FunctionEditor, natural_loop: &Loop) -> HashSet<NodeID> {
+ // Stop on PHIs / reduces outside of loop.
+ let stop_on: HashSet<NodeID> = editor
+ .node_ids()
+ .filter(|node| {
+ let data = &editor.func().nodes[node.idx()];
+
+ // External Phi
+ if let Node::Phi { control, data: _ } = data {
+ if !natural_loop.control[control.idx()] {
+ return true;
+ }
+ }
+ // External Reduce
+ if let Node::Reduce {
+ control,
+ init: _,
+ reduct: _,
+ } = data
+ {
+ if !natural_loop.control[control.idx()] {
+ return true;
+ }
+ }
+
+ // External Control
+ if data.is_control() && !natural_loop.control[node.idx()] {
+ return true;
+ }
+
+ return false;
+ })
+ .collect();
+
+ let phis: Vec<_> = editor
+ .node_ids()
+ .filter(|node| {
+ let Node::Phi { control, data: _ } = editor.func().nodes[node.idx()] else {
+ return false;
+ };
+ natural_loop.control[control.idx()]
+ })
+ .collect();
+
+ let all_users: HashSet<NodeID> = phis
+ .clone()
+ .iter()
+ .flat_map(|phi| walk_all_users_stop_on(*phi, editor, stop_on.clone()))
+ .chain(phis.clone())
+ .collect();
+
+ let all_uses: HashSet<_> = phis
+ .clone()
+ .iter()
+ .flat_map(|phi| walk_all_uses_stop_on(*phi, editor, stop_on.clone()))
+ .chain(phis.clone())
+ .filter(|node| {
+ // Get rid of nodes in stop_on
+ !stop_on.contains(node)
+ })
+ .collect();
+
+ all_users
+ .intersection(&all_uses)
+ .chain(phis.iter())
+ .cloned()
+ .collect()
+}
+
+/** returns PHIs that are on any regions inside the loop. */
+pub fn get_all_loop_phis<'a>(
+ function: &'a Function,
+ l: &'a Loop,
+) -> impl Iterator<Item = NodeID> + 'a {
+ function
+ .nodes
+ .iter()
+ .enumerate()
+ .filter_map(move |(node_id, node)| {
+ if let Some((control, _)) = node.try_phi() {
+ if l.control[control.idx()] {
+ Some(NodeID::new(node_id))
+ } else {
+ None
+ }
+ } else {
+ None
+ }
+ })
+}
+
+// FIXME: Need a trait that Editor and Function both implement, that gives us UseDefInfo
+
+/** Given a loop determine for each data node if the value might change upon each iteration of the loop */
+pub fn compute_loop_variance(editor: &FunctionEditor, l: &Loop) -> LoopVarianceInfo {
+ // Gather all Phi nodes that are controlled by this loop.
+ let mut loop_vars: Vec<NodeID> = vec![];
+
+ for node_id in editor.get_users(l.header) {
+ let node = &editor.func().nodes[node_id.idx()];
+ if let Some((control, _)) = node.try_phi() {
+ if l.control[control.idx()] {
+ loop_vars.push(node_id);
+ }
+ }
+ }
+
+ let len = editor.func().nodes.len();
+
+ let mut all_loop_nodes = l.control.clone();
+
+ all_loop_nodes.set(l.header.idx(), true);
+
+ let mut variance_map: DenseNodeMap<LoopVariance> = vec![LoopVariance::Unknown; len];
+
+ fn recurse(
+ function: &Function,
+ node: NodeID,
+ all_loop_nodes: &BitVec<u8, Lsb0>,
+ variance_map: &mut DenseNodeMap<LoopVariance>,
+ visited: &mut DenseNodeMap<bool>,
+ ) -> LoopVariance {
+ if visited[node.idx()] {
+ return variance_map[node.idx()];
+ }
+
+ visited[node.idx()] = true;
+
+ let node_variance = match variance_map[node.idx()] {
+ LoopVariance::Invariant => LoopVariance::Invariant,
+ LoopVariance::Variant => LoopVariance::Variant,
+ LoopVariance::Unknown => {
+ let mut node_variance = LoopVariance::Invariant;
+
+ // Two conditions cause something to be loop variant:
+ for node_use in get_uses(&function.nodes[node.idx()]).as_ref() {
+ // 1) The use is a PHI *controlled* by the loop
+ if let Some((control, _)) = function.nodes[node_use.idx()].try_phi() {
+ if *all_loop_nodes.get(control.idx()).unwrap() {
+ node_variance = LoopVariance::Variant;
+ break;
+ }
+ }
+
+ // 2) Any of the nodes uses are loop variant
+ if recurse(function, *node_use, all_loop_nodes, variance_map, visited)
+ == LoopVariance::Variant
+ {
+ node_variance = LoopVariance::Variant;
+ break;
+ }
+ }
+
+ variance_map[node.idx()] = node_variance;
+
+ node_variance
+ }
+ };
+
+ return node_variance;
+ }
+
+ let mut visited: DenseNodeMap<bool> = vec![false; len];
+
+ for node in (0..len).map(NodeID::new) {
+ recurse(
+ editor.func(),
+ node,
+ &all_loop_nodes,
+ &mut variance_map,
+ &mut visited,
+ );
+ }
+
+ return LoopVarianceInfo {
+ loop_header: l.header,
+ map: variance_map,
+ };
+}
+
+nest! {
+#[derive(Clone, Copy, Debug, PartialEq)]
+pub enum LoopExit {
+ Conditional {
+ if_node: NodeID,
+ condition_node: NodeID,
+ },
+ Unconditional(NodeID)
+}
+}
+
+pub fn get_loop_exit_conditions(
+ function: &Function,
+ l: &Loop,
+ control_subgraph: &Subgraph,
+) -> Option<LoopExit> {
+ // impl IntoIterator<Item = LoopExit>
+ // DFS Traversal on loop control subgraph until we find a node that is outside the loop, find the last IF on this path.
+ let mut last_if_on_path: DenseNodeMap<Option<NodeID>> = vec![None; function.nodes.len()];
+
+ // FIXME: (@xrouth) THIS IS MOST CERTAINLY BUGGED
+ // this might be bugged... i.e might need to udpate `last if` even if already defined.
+ // needs to be `saturating` kinda, more iterative. May need to visit nodes more than once?
+
+ // FIXME: (@xrouth) Right now we assume only one exit from the loop, later: check for multiple exits on the loop,
+ // either as an assertion here or some other part of forkify or analysis.
+ let mut bag_of_control_nodes = vec![l.header];
+ let mut visited: DenseNodeMap<bool> = vec![false; function.nodes.len()];
+
+ let mut final_if: Option<NodeID> = None;
+
+ // do WFS
+ while !bag_of_control_nodes.is_empty() {
+ let node = bag_of_control_nodes.pop().unwrap();
+ if visited[node.idx()] {
+ continue;
+ }
+ visited[node.idx()] = true;
+
+ final_if = if function.nodes[node.idx()].is_if() {
+ Some(node)
+ } else {
+ last_if_on_path[node.idx()]
+ };
+
+ if !l.control[node.idx()] {
+ break;
+ }
+
+ for succ in control_subgraph.succs(node) {
+ last_if_on_path[succ.idx()] = final_if;
+ bag_of_control_nodes.push(succ.clone());
+ }
+ }
+
+ final_if.map(|v| LoopExit::Conditional {
+ if_node: v,
+ condition_node: if let Node::If { control: _, cond } = function.nodes[v.idx()] {
+ cond
+ } else {
+ unreachable!()
+ },
+ })
+}
+
+pub fn has_const_fields(editor: &FunctionEditor, ivar: InductionVariable) -> bool {
+ match ivar {
+ InductionVariable::Basic {
+ node: _,
+ initializer,
+ update,
+ final_value,
+ } => {
+ if final_value.is_none() {
+ return false;
+ }
+ [initializer, update]
+ .iter()
+ .any(|node| !editor.node(node).is_constant())
+ }
+ InductionVariable::SCEV(_) => false,
+ }
+}
+
+/* Loop has any IV from range 0....N, N can be dynconst iterates +1 per iteration */
+// IVs need to be bounded...
+pub fn has_canonical_iv<'a>(
+ editor: &FunctionEditor,
+ _l: &Loop,
+ ivs: &'a [InductionVariable],
+) -> Option<&'a InductionVariable> {
+ ivs.iter().find(|iv| match iv {
+ InductionVariable::Basic {
+ node: _,
+ initializer,
+ update,
+ final_value,
+ } => {
+ (editor
+ .node(initializer)
+ .is_zero_constant(&editor.get_constants())
+ || editor
+ .node(initializer)
+ .is_zero_dc(&editor.get_dynamic_constants()))
+ && (editor.node(update).is_one_constant(&editor.get_constants())
+ || editor
+ .node(update)
+ .is_one_dc(&editor.get_dynamic_constants()))
+ && (final_value
+ .map(|val| {
+ editor.node(val).is_constant() || editor.node(val).is_dynamic_constant()
+ })
+ .is_some())
+ }
+ InductionVariable::SCEV(_) => false,
+ })
+}
+
+// Need a transformation that forces all IVs to be SCEVs of an IV from range 0...N, +1, else places them in a separate loop?
+pub fn compute_induction_vars(
+ function: &Function,
+ l: &Loop,
+ _loop_variance: &LoopVarianceInfo,
+) -> Vec<InductionVariable> {
+ // 1) Gather PHIs contained in the loop.
+ // FIXME: (@xrouth) Should this just be PHIs controlled by the header?
+ let mut loop_vars: Vec<NodeID> = vec![];
+
+ for (node_id, node) in function.nodes.iter().enumerate() {
+ if let Some((control, _)) = node.try_phi() {
+ if l.control[control.idx()] {
+ loop_vars.push(NodeID::new(node_id));
+ }
+ }
+ }
+
+ // FIXME: (@xrouth) For now, only compute variables that have one assignment,
+ // (look into this:) possibly treat multiple assignment as separate induction variables.
+ let mut induction_variables: Vec<InductionVariable> = vec![];
+
+ /* For each PHI controlled by the loop, check how it is modified */
+
+ // It's initializer needs to be loop invariant, it's update needs to be loop variant.
+ for phi_id in loop_vars {
+ let phi_node = &function.nodes[phi_id.idx()];
+ let (region, data) = phi_node.try_phi().unwrap();
+ let region_node = &function.nodes[region.idx()];
+ let Node::Region {
+ preds: region_inputs,
+ } = region_node
+ else {
+ continue;
+ };
+
+ // The initializer index is the first index of the inputs to the region node of that isn't in the loop. (what is loop_header, wtf...)
+ // FIXME (@xrouth): If there is control flow in the loop, we won't find ... WHAT
+ let Some(initializer_idx) = region_inputs
+ .iter()
+ .position(|&node_id| !l.control[node_id.idx()])
+ else {
+ continue;
+ };
+
+ let initializer_id = data[initializer_idx];
+
+ // Check dynamic constancy:
+ let initializer = &function.nodes[initializer_id.idx()];
+
+ // In the case of a non 0 starting value:
+ // - a new dynamic constant or constant may need to be created that is the difference between the initiailizer and the loop bounds.
+ // Initializer does not necessarily have to be constant, but this is fine for now.
+ if !(initializer.is_dynamic_constant() || initializer.is_constant()) {
+ continue;
+ }
+
+ // Check all data inputs to this phi, that aren't the initializer (i.e the value the comes from control outside of the loop)
+ // For now we expect only one initializer.
+ let data_inputs = data
+ .iter()
+ .filter(|data_id| NodeID::new(initializer_idx) != **data_id);
+
+ for data_id in data_inputs {
+ let node = &function.nodes[data_id.idx()];
+ for bop in [BinaryOperator::Add] {
+ //, BinaryOperator::Mul, BinaryOperator::Sub] {
+ if let Some((a, b)) = node.try_binary(bop) {
+ let iv = [(a, b), (b, a)]
+ .iter()
+ .find_map(|(pattern_phi, pattern_const)| {
+ if *pattern_phi == phi_id
+ && function.nodes[pattern_const.idx()].is_constant()
+ || function.nodes[pattern_const.idx()].is_dynamic_constant()
+ {
+ return Some(InductionVariable::Basic {
+ node: phi_id,
+ initializer: initializer_id,
+ update: b,
+ final_value: None,
+ });
+ } else {
+ None
+ }
+ });
+ if let Some(iv) = iv {
+ induction_variables.push(iv);
+ }
+ }
+ }
+ }
+ }
+
+ induction_variables
+}
+
+// Find loop iterations
+pub fn compute_iv_ranges(
+ editor: &FunctionEditor,
+ l: &Loop,
+ induction_vars: Vec<InductionVariable>,
+ loop_condition: &LoopExit,
+) -> Vec<InductionVariable> {
+ let condition_node = match loop_condition {
+ LoopExit::Conditional {
+ if_node: _,
+ condition_node,
+ } => condition_node,
+ LoopExit::Unconditional(_) => todo!(),
+ };
+
+ // Find IVs used by the loop condition, not across loop iterations.
+ // without leaving the loop.
+ let stop_on: HashSet<_> = editor
+ .node_ids()
+ .filter(|node_id| {
+ if let Node::Phi { control, data: _ } = editor.node(node_id) {
+ *control == l.header
+ } else {
+ false
+ }
+ })
+ .collect();
+
+ // Bound IVs used in loop bound.
+ let loop_bound_uses: HashSet<_> =
+ walk_all_uses_stop_on(*condition_node, editor, stop_on).collect();
+ let (loop_bound_ivs, other_ivs): (Vec<InductionVariable>, Vec<InductionVariable>) =
+ induction_vars
+ .into_iter()
+ .partition(|f| loop_bound_uses.contains(&f.phi()));
+
+ let Some(iv) = loop_bound_ivs.first() else {
+ return other_ivs;
+ };
+
+ if loop_bound_ivs.len() > 1 {
+ return loop_bound_ivs.into_iter().chain(other_ivs).collect();
+ }
+
+ // FIXME: DO linear algerbra to solve for loop bounds with multiple variables involved.
+ let final_value = match &editor.func().nodes[condition_node.idx()] {
+ Node::Phi {
+ control: _,
+ data: _,
+ } => None,
+ Node::Reduce {
+ control: _,
+ init: _,
+ reduct: _,
+ } => None,
+ Node::Parameter { index: _ } => None,
+ Node::Constant { id: _ } => None,
+ Node::Unary { input: _, op: _ } => None,
+ Node::Ternary {
+ first: _,
+ second: _,
+ third: _,
+ op: _,
+ } => None,
+ Node::Binary { left, right, op } => {
+ match op {
+ BinaryOperator::LT => {
+ // Check for a loop guard condition.
+ // left < right
+ if *left == iv.phi()
+ && (editor.func().nodes[right.idx()].is_constant()
+ || editor.func().nodes[right.idx()].is_dynamic_constant())
+ {
+ Some(*right)
+ }
+ // left + const < right,
+ else if let Node::Binary {
+ left: inner_left,
+ right: inner_right,
+ op: _,
+ } = editor.node(left)
+ {
+ let pattern = [(inner_left, inner_right), (inner_right, inner_left)]
+ .iter()
+ .find_map(|(pattern_iv, pattern_constant)| {
+ if iv.phi() == **pattern_iv
+ && (editor.node(*pattern_constant).is_constant()
+ || editor.node(*pattern_constant).is_dynamic_constant())
+ {
+ // FIXME: pattern_constant can be anything >= loop_update expression,
+ let update = match iv {
+ InductionVariable::Basic {
+ node: _,
+ initializer: _,
+ update,
+ final_value: _,
+ } => update,
+ InductionVariable::SCEV(_) => todo!(),
+ };
+ if *pattern_constant == update {
+ Some(*right)
+ } else {
+ None
+ }
+ } else {
+ None
+ }
+ });
+ pattern.iter().cloned().next()
+ } else {
+ None
+ }
+ }
+ BinaryOperator::LTE => None,
+ BinaryOperator::GT => None,
+ BinaryOperator::GTE => None,
+ BinaryOperator::EQ => None,
+ BinaryOperator::NE => None,
+ _ => None,
+ }
+ }
+ _ => None,
+ };
+
+ let basic = match iv {
+ InductionVariable::Basic {
+ node,
+ initializer,
+ update,
+ final_value: _,
+ } => InductionVariable::Basic {
+ node: *node,
+ initializer: *initializer,
+ update: *update,
+ final_value,
+ },
+ InductionVariable::SCEV(_) => todo!(),
+ };
+
+ // Propagate bounds to other IVs.
+ vec![basic].into_iter().chain(other_ivs).collect()
+}
diff --git a/hercules_opt/src/lib.rs b/hercules_opt/src/lib.rs
index 4a90f698..e3cca161 100644
--- a/hercules_opt/src/lib.rs
+++ b/hercules_opt/src/lib.rs
@@ -4,16 +4,17 @@ pub mod ccp;
pub mod crc;
pub mod dce;
pub mod delete_uncalled;
-pub mod device_placement;
pub mod editor;
pub mod float_collections;
pub mod fork_concat_split;
pub mod fork_guard_elim;
+pub mod fork_transforms;
pub mod forkify;
pub mod gcm;
pub mod gvn;
pub mod inline;
pub mod interprocedural_sroa;
+pub mod ivar;
pub mod lift_dc_math;
pub mod outline;
pub mod phi_elim;
@@ -28,16 +29,17 @@ pub use crate::ccp::*;
pub use crate::crc::*;
pub use crate::dce::*;
pub use crate::delete_uncalled::*;
-pub use crate::device_placement::*;
pub use crate::editor::*;
pub use crate::float_collections::*;
pub use crate::fork_concat_split::*;
pub use crate::fork_guard_elim::*;
+pub use crate::fork_transforms::*;
pub use crate::forkify::*;
pub use crate::gcm::*;
pub use crate::gvn::*;
pub use crate::inline::*;
pub use crate::interprocedural_sroa::*;
+pub use crate::ivar::*;
pub use crate::lift_dc_math::*;
pub use crate::outline::*;
pub use crate::phi_elim::*;
diff --git a/hercules_opt/src/outline.rs b/hercules_opt/src/outline.rs
index e59c815d..8fe978c5 100644
--- a/hercules_opt/src/outline.rs
+++ b/hercules_opt/src/outline.rs
@@ -4,7 +4,6 @@ use std::sync::atomic::{AtomicUsize, Ordering};
use hercules_ir::def_use::*;
use hercules_ir::dom::*;
-use hercules_ir::fork_join_analysis::*;
use hercules_ir::ir::*;
use hercules_ir::subgraph::*;
diff --git a/hercules_opt/src/schedule.rs b/hercules_opt/src/schedule.rs
index 2c8209aa..f9f720be 100644
--- a/hercules_opt/src/schedule.rs
+++ b/hercules_opt/src/schedule.rs
@@ -29,7 +29,7 @@ pub fn infer_parallel_fork(editor: &mut FunctionEditor, fork_join_map: &HashMap<
/*
* Infer parallel reductions consisting of a simple cycle between a Reduce node
* and a Write node, where indices of the Write are position indices using the
- * ThreadID nodes attached to the corresponding Fork, and data of the Write is
+ * ThreadID nodes attached to the corresponding Fork, and data of the Write is
* not in the Reduce node's cycle. This procedure also adds the ParallelReduce
* schedule to Reduce nodes reducing over a parallelized Reduce, as long as the
* base Write node also has position indices of the ThreadID of the outer fork.
@@ -37,7 +37,11 @@ pub fn infer_parallel_fork(editor: &mut FunctionEditor, fork_join_map: &HashMap<
* as long as each ThreadID dimension appears in the positional indexing of the
* original Write.
*/
-pub fn infer_parallel_reduce(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, NodeID>, reduce_cycles: &HashMap<NodeID, HashSet<NodeID>>) {
+pub fn infer_parallel_reduce(
+ editor: &mut FunctionEditor,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+ reduce_cycles: &HashMap<NodeID, HashSet<NodeID>>,
+) {
for id in editor.node_ids() {
let func = editor.func();
if !func.nodes[id.idx()].is_reduce() {
@@ -146,11 +150,17 @@ pub fn infer_vectorizable(editor: &mut FunctionEditor, fork_join_map: &HashMap<N
* operation's operands must be the Reduce node, and all other operands must
* not be in the Reduce node's cycle.
*/
-pub fn infer_tight_associative(editor: &mut FunctionEditor, reduce_cycles: &HashMap<NodeID, HashSet<NodeID>>) {
- let is_binop_associative = |op| matches!(op,
- BinaryOperator::Add | BinaryOperator::Or | BinaryOperator::And | BinaryOperator::Xor);
- let is_intrinsic_associative = |intrinsic| matches!(intrinsic,
- Intrinsic::Max | Intrinsic::Min);
+pub fn infer_tight_associative(
+ editor: &mut FunctionEditor,
+ reduce_cycles: &HashMap<NodeID, HashSet<NodeID>>,
+) {
+ let is_binop_associative = |op| {
+ matches!(
+ op,
+ BinaryOperator::Add | BinaryOperator::Or | BinaryOperator::And | BinaryOperator::Xor
+ )
+ };
+ let is_intrinsic_associative = |intrinsic| matches!(intrinsic, Intrinsic::Max | Intrinsic::Min);
for id in editor.node_ids() {
let func = editor.func();
@@ -162,8 +172,8 @@ pub fn infer_tight_associative(editor: &mut FunctionEditor, reduce_cycles: &Hash
&& (matches!(func.nodes[reduct.idx()], Node::Binary { left, right, op }
if ((left == id && !reduce_cycles[&id].contains(&right)) ||
(right == id && !reduce_cycles[&id].contains(&left))) &&
- is_binop_associative(op)) ||
- matches!(&func.nodes[reduct.idx()], Node::IntrinsicCall { intrinsic, args }
+ is_binop_associative(op))
+ || matches!(&func.nodes[reduct.idx()], Node::IntrinsicCall { intrinsic, args }
if (args.contains(&id) && is_intrinsic_associative(*intrinsic) &&
args.iter().filter(|arg| **arg != id).all(|arg| !reduce_cycles[&id].contains(arg)))))
{
diff --git a/hercules_opt/src/unforkify.rs b/hercules_opt/src/unforkify.rs
index a5df7a7c..85ffd233 100644
--- a/hercules_opt/src/unforkify.rs
+++ b/hercules_opt/src/unforkify.rs
@@ -1,17 +1,110 @@
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet};
use std::iter::zip;
-use hercules_ir::ir::*;
+use bitvec::{order::Lsb0, vec::BitVec};
+use hercules_ir::{ir::*, LoopTree};
use crate::*;
+type NodeVec = BitVec<u8, Lsb0>;
+pub fn calculate_fork_nodes(
+ editor: &FunctionEditor,
+ inner_control: &NodeVec,
+ fork: NodeID,
+) -> HashSet<NodeID> {
+ // Stop on PHIs / reduces outside of loop.
+ let stop_on: HashSet<NodeID> = editor
+ .node_ids()
+ .filter(|node| {
+ let data = &editor.func().nodes[node.idx()];
+
+ // External Phi
+ if let Node::Phi { control, data: _ } = data {
+ if match inner_control.get(control.idx()) {
+ Some(v) => !*v, //
+ None => true, // Doesn't exist, must be external
+ } {
+ return true;
+ }
+ }
+ // External Reduce
+ if let Node::Reduce {
+ control,
+ init: _,
+ reduct: _,
+ } = data
+ {
+ if match inner_control.get(control.idx()) {
+ Some(v) => !*v, //
+ None => true, // Doesn't exist, must be external
+ } {
+ return true;
+ }
+ }
+
+ // External Control
+ if data.is_control() {
+ return match inner_control.get(node.idx()) {
+ Some(v) => !*v, //
+ None => true, // Doesn't exist, must be external
+ };
+ }
+ // else
+ return false;
+ })
+ .collect();
+
+ let reduces: Vec<_> = editor
+ .node_ids()
+ .filter(|node| {
+ let Node::Reduce { control, .. } = editor.func().nodes[node.idx()] else {
+ return false;
+ };
+ match inner_control.get(control.idx()) {
+ Some(v) => *v,
+ None => false,
+ }
+ })
+ .chain(
+ editor
+ .get_users(fork)
+ .filter(|node| editor.node(node).is_thread_id()),
+ )
+ .collect();
+
+ let all_users: HashSet<NodeID> = reduces
+ .clone()
+ .iter()
+ .flat_map(|phi| walk_all_users_stop_on(*phi, editor, stop_on.clone()))
+ .chain(reduces.clone())
+ .collect();
+
+ let all_uses: HashSet<_> = reduces
+ .clone()
+ .iter()
+ .flat_map(|phi| walk_all_uses_stop_on(*phi, editor, stop_on.clone()))
+ .chain(reduces)
+ .filter(|node| {
+ // Get rid of nodes in stop_on
+ !stop_on.contains(node)
+ })
+ .collect();
+
+ all_users.intersection(&all_uses).cloned().collect()
+}
/*
* Convert forks back into loops right before codegen when a backend is not
* lowering a fork-join to vector / parallel code. Lowering fork-joins into
* sequential loops in LLVM is actually not entirely trivial, so it's easier to
* just do this transformation within Hercules IR.
*/
-pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, NodeID>) {
+
+// FIXME: Only works on fully split fork nests.
+pub fn unforkify(
+ editor: &mut FunctionEditor,
+ fork_join_map: &HashMap<NodeID, NodeID>,
+ loop_tree: &LoopTree,
+) {
let mut zero_cons_id = ConstantID::new(0);
let mut one_cons_id = ConstantID::new(0);
assert!(editor.edit(|mut edit| {
@@ -25,7 +118,16 @@ pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, No
// control insides of the fork-join should become the successor of the true
// projection node, and what was the use of the join should become a use of
// the new region.
- for (fork, join) in fork_join_map {
+ for l in loop_tree.bottom_up_loops().into_iter().rev() {
+ if !editor.node(l.0).is_fork() {
+ continue;
+ }
+
+ let fork = &l.0;
+ let join = &fork_join_map[&fork];
+
+ let fork_nodes = calculate_fork_nodes(editor, l.1, *fork);
+
let nodes = &editor.func().nodes;
let (fork_control, factors) = nodes[fork.idx()].try_fork().unwrap();
if factors.len() > 1 {
@@ -54,20 +156,43 @@ pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, No
let add_id = NodeID::new(num_nodes + 7);
let dc_id = NodeID::new(num_nodes + 8);
let neq_id = NodeID::new(num_nodes + 9);
- let phi_ids = (num_nodes + 10..num_nodes + 10 + reduces.len()).map(NodeID::new);
+
+ let guard_if_id = NodeID::new(num_nodes + 10);
+ let guard_join_id = NodeID::new(num_nodes + 11);
+ let guard_taken_proj_id = NodeID::new(num_nodes + 12);
+ let guard_skipped_proj_id = NodeID::new(num_nodes + 13);
+ let guard_cond_id = NodeID::new(num_nodes + 14);
+
+ let phi_ids = (num_nodes + 15..num_nodes + 15 + reduces.len()).map(NodeID::new);
+ let s = num_nodes + 15 + reduces.len();
+ let join_phi_ids = (s..s + reduces.len()).map(NodeID::new);
+
+ let guard_cond = Node::Binary {
+ left: zero_id,
+ right: dc_id,
+ op: BinaryOperator::LT,
+ };
+ let guard_if = Node::If {
+ control: fork_control,
+ cond: guard_cond_id,
+ };
+ let guard_taken_proj = Node::Projection {
+ control: guard_if_id,
+ selection: 1,
+ };
+ let guard_skipped_proj = Node::Projection {
+ control: guard_if_id,
+ selection: 0,
+ };
+ let guard_join = Node::Region {
+ preds: Box::new([guard_skipped_proj_id, proj_exit_id]),
+ };
let region = Node::Region {
- preds: Box::new([
- fork_control,
- if join_control == *fork {
- proj_back_id
- } else {
- join_control
- },
- ]),
+ preds: Box::new([guard_taken_proj_id, proj_back_id]),
};
let if_node = Node::If {
- control: region_id,
+ control: join_control,
cond: neq_id,
};
let proj_back = Node::Projection {
@@ -92,19 +217,25 @@ pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, No
let dc = Node::DynamicConstant { id: factors[0] };
let neq = Node::Binary {
op: BinaryOperator::NE,
- left: indvar_id,
+ left: add_id,
right: dc_id,
};
- let phis: Vec<_> = reduces
+ let (phis, join_phis): (Vec<_>, Vec<_>) = reduces
.iter()
.map(|reduce_id| {
let (_, init, reduct) = nodes[reduce_id.idx()].try_reduce().unwrap();
- Node::Phi {
- control: region_id,
- data: Box::new([init, reduct]),
- }
+ (
+ Node::Phi {
+ control: region_id,
+ data: Box::new([init, reduct]),
+ },
+ Node::Phi {
+ control: guard_join_id,
+ data: Box::new([init, reduct]),
+ },
+ )
})
- .collect();
+ .unzip();
editor.edit(|mut edit| {
assert_eq!(edit.add_node(region), region_id);
@@ -117,21 +248,41 @@ pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, No
assert_eq!(edit.add_node(add), add_id);
assert_eq!(edit.add_node(dc), dc_id);
assert_eq!(edit.add_node(neq), neq_id);
- for (phi_id, phi) in zip(phi_ids.clone(), phis) {
- assert_eq!(edit.add_node(phi), phi_id);
+ assert_eq!(edit.add_node(guard_if), guard_if_id);
+ assert_eq!(edit.add_node(guard_join), guard_join_id);
+ assert_eq!(edit.add_node(guard_taken_proj), guard_taken_proj_id);
+ assert_eq!(edit.add_node(guard_skipped_proj), guard_skipped_proj_id);
+ assert_eq!(edit.add_node(guard_cond), guard_cond_id);
+
+ for (phi_id, phi) in zip(phi_ids.clone(), &phis) {
+ assert_eq!(edit.add_node(phi.clone()), phi_id);
+ }
+ for (phi_id, phi) in zip(join_phi_ids.clone(), &join_phis) {
+ assert_eq!(edit.add_node(phi.clone()), phi_id);
}
- edit = edit.replace_all_uses(*fork, proj_back_id)?;
- edit = edit.replace_all_uses(*join, proj_exit_id)?;
+ edit = edit.replace_all_uses(*fork, region_id)?;
+ edit = edit.replace_all_uses_where(*join, guard_join_id, |usee| *usee != if_id)?;
edit.sub_edit(*fork, region_id);
edit.sub_edit(*join, if_id);
for tid in tids.iter() {
edit.sub_edit(*tid, indvar_id);
edit = edit.replace_all_uses(*tid, indvar_id)?;
}
- for (reduce, phi_id) in zip(reduces.iter(), phi_ids) {
+ for (((reduce, phi_id), phi), join_phi_id) in
+ zip(reduces.iter(), phi_ids).zip(phis).zip(join_phi_ids)
+ {
edit.sub_edit(*reduce, phi_id);
- edit = edit.replace_all_uses(*reduce, phi_id)?;
+ let Node::Phi { control: _, data } = phi else {
+ panic!()
+ };
+ edit = edit.replace_all_uses_where(*reduce, join_phi_id, |usee| {
+ !fork_nodes.contains(usee)
+ })?; //, |usee| *usee != *reduct)?;
+ edit = edit.replace_all_uses_where(*reduce, phi_id, |usee| {
+ fork_nodes.contains(usee) || *usee == data[1]
+ })?;
+ edit = edit.delete_node(*reduce)?;
}
edit = edit.delete_node(*fork)?;
@@ -139,9 +290,6 @@ pub fn unforkify(editor: &mut FunctionEditor, fork_join_map: &HashMap<NodeID, No
for tid in tids {
edit = edit.delete_node(tid)?;
}
- for reduce in reduces {
- edit = edit.delete_node(reduce)?;
- }
Ok(edit)
});
diff --git a/hercules_opt/src/utils.rs b/hercules_opt/src/utils.rs
index aa0d53fe..7ad48c1c 100644
--- a/hercules_opt/src/utils.rs
+++ b/hercules_opt/src/utils.rs
@@ -1,7 +1,10 @@
+use std::collections::HashMap;
+use std::collections::HashSet;
use std::iter::zip;
use hercules_ir::def_use::*;
use hercules_ir::ir::*;
+use nestify::nest;
use crate::*;
@@ -376,3 +379,105 @@ pub(crate) fn indices_may_overlap(indices1: &[Index], indices2: &[Index]) -> boo
// may overlap when one indexes a larger sub-value than the other.
true
}
+
+pub type DenseNodeMap<T> = Vec<T>;
+pub type SparseNodeMap<T> = HashMap<NodeID, T>;
+
+nest! {
+//
+#[derive(Clone, Debug)]
+pub struct NodeIterator<'a> {
+ pub direction:
+ #[derive(Clone, Debug, PartialEq)]
+ pub enum Direction {
+ Uses,
+ Users,
+ },
+ visited: DenseNodeMap<bool>,
+ stack: Vec<NodeID>,
+ func: &'a FunctionEditor<'a>, // Maybe this is an enum, def use can be gotten from the function or from the editor.
+ // `stop condition`, then return all nodes that caused stoppage i.e the frontier of the search.
+ stop_on: HashSet<NodeID>, // Don't add neighbors of these.
+}
+}
+
+pub fn walk_all_uses<'a>(node: NodeID, editor: &'a FunctionEditor<'a>) -> NodeIterator<'a> {
+ let len = editor.func().nodes.len();
+ NodeIterator {
+ direction: Direction::Uses,
+ visited: vec![false; len],
+ stack: vec![node],
+ func: editor,
+ stop_on: HashSet::new(),
+ }
+}
+
+pub fn walk_all_users<'a>(node: NodeID, editor: &'a FunctionEditor<'a>) -> NodeIterator<'a> {
+ let len = editor.func().nodes.len();
+ NodeIterator {
+ direction: Direction::Users,
+ visited: vec![false; len],
+ stack: vec![node],
+ func: editor,
+ stop_on: HashSet::new(),
+ }
+}
+
+pub fn walk_all_uses_stop_on<'a>(
+ node: NodeID,
+ editor: &'a FunctionEditor<'a>,
+ stop_on: HashSet<NodeID>,
+) -> NodeIterator<'a> {
+ let len = editor.func().nodes.len();
+ let uses = editor.get_uses(node).collect();
+ NodeIterator {
+ direction: Direction::Uses,
+ visited: vec![false; len],
+ stack: uses,
+ func: editor,
+ stop_on,
+ }
+}
+
+pub fn walk_all_users_stop_on<'a>(
+ node: NodeID,
+ editor: &'a FunctionEditor<'a>,
+ stop_on: HashSet<NodeID>,
+) -> NodeIterator<'a> {
+ let len = editor.func().nodes.len();
+ let users = editor.get_users(node).collect();
+ NodeIterator {
+ direction: Direction::Users,
+ visited: vec![false; len],
+ stack: users,
+ func: editor,
+ stop_on,
+ }
+}
+
+impl<'a> Iterator for NodeIterator<'a> {
+ type Item = NodeID;
+
+ fn next(&mut self) -> Option<Self::Item> {
+ while let Some(current) = self.stack.pop() {
+ if !self.visited[current.idx()] {
+ self.visited[current.idx()] = true;
+
+ if !self.stop_on.contains(¤t) {
+ if self.direction == Direction::Uses {
+ for neighbor in self.func.get_uses(current) {
+ self.stack.push(neighbor)
+ }
+ } else {
+ for neighbor in self.func.get_users(current) {
+ self.stack.push(neighbor)
+ }
+ }
+ }
+
+ return Some(current);
+ }
+ }
+ None
+ }
+}
diff --git a/hercules_samples/matmul/build.rs b/hercules_samples/matmul/build.rs
index f895af86..c15ca97f 100644
--- a/hercules_samples/matmul/build.rs
+++ b/hercules_samples/matmul/build.rs
@@ -4,7 +4,7 @@ fn main() {
JunoCompiler::new()
.ir_in_src("matmul.hir")
.unwrap()
- //.schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
+ // .schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
.schedule_in_src("cpu.sch")
.unwrap()
.build()
diff --git a/hercules_test/hercules_interpreter/Cargo.toml b/hercules_test/hercules_interpreter/Cargo.toml
index d41caff8..6e02b9b8 100644
--- a/hercules_test/hercules_interpreter/Cargo.toml
+++ b/hercules_test/hercules_interpreter/Cargo.toml
@@ -9,6 +9,9 @@ clap = { version = "*", features = ["derive"] }
rand = "*"
hercules_ir = { path = "../../hercules_ir" }
hercules_opt = { path = "../../hercules_opt" }
+juno_scheduler = { path = "../../juno_scheduler" }
itertools = "*"
ordered-float = "*"
-derive_more = {version = "*", features = ["from"]}
\ No newline at end of file
+derive_more = {version = "*", features = ["from"]}
+postcard = { version = "*", features = ["alloc"] }
+serde = { version = "*", features = ["derive"] }
\ No newline at end of file
diff --git a/hercules_test/hercules_interpreter/src/interpreter.rs b/hercules_test/hercules_interpreter/src/interpreter.rs
index 621260e5..a78330e4 100644
--- a/hercules_test/hercules_interpreter/src/interpreter.rs
+++ b/hercules_test/hercules_interpreter/src/interpreter.rs
@@ -1,15 +1,17 @@
-
+use std::collections::hash_map::Entry::Occupied;
use std::collections::HashMap;
use std::panic;
-use std::collections::hash_map::Entry::Occupied;
use itertools::Itertools;
+use std::cmp::{max, min};
use hercules_ir::*;
use crate::value;
use value::*;
+const VERBOSE: bool = false;
+
/* High level design details / discussion for this:
*
* This crate includes tools for interpreting a hercules IR module. Execution model / flow is based on
@@ -41,8 +43,8 @@ pub struct FunctionContext<'a> {
fork_join_nest: &'a HashMap<NodeID, Vec<NodeID>>,
}
-impl <'a> FunctionContext<'a> {
- pub fn new (
+impl<'a> FunctionContext<'a> {
+ pub fn new(
control_subgraph: &'a Subgraph,
def_use: &'a ImmutableDefUseMap,
fork_join_map: &'a HashMap<NodeID, NodeID>, // Map forks -> joins
@@ -57,12 +59,47 @@ impl <'a> FunctionContext<'a> {
}
}
-pub fn dyn_const_value(dc: &DynamicConstant, dyn_const_params: &[usize]) -> usize {
+// TODO: (@xrouth) I feel like this funcitonality should be provided by the manager that holds and allocates dynamic constants & IDs.
+pub fn dyn_const_value(
+ dc: &DynamicConstantID,
+ dyn_const_values: &[DynamicConstant],
+ dyn_const_params: &[usize],
+) -> usize {
+ let dc = &dyn_const_values[dc.idx()];
match dc {
DynamicConstant::Constant(v) => *v,
DynamicConstant::Parameter(v) => dyn_const_params[*v],
+ DynamicConstant::Add(a, b) => {
+ dyn_const_value(a, dyn_const_values, dyn_const_params)
+ + dyn_const_value(b, dyn_const_values, dyn_const_params)
+ }
+ DynamicConstant::Sub(a, b) => {
+ dyn_const_value(a, dyn_const_values, dyn_const_params)
+ - dyn_const_value(b, dyn_const_values, dyn_const_params)
+ }
+ DynamicConstant::Mul(a, b) => {
+ dyn_const_value(a, dyn_const_values, dyn_const_params)
+ * dyn_const_value(b, dyn_const_values, dyn_const_params)
+ }
+ DynamicConstant::Div(a, b) => {
+ dyn_const_value(a, dyn_const_values, dyn_const_params)
+ / dyn_const_value(b, dyn_const_values, dyn_const_params)
+ }
+ DynamicConstant::Rem(a, b) => {
+ dyn_const_value(a, dyn_const_values, dyn_const_params)
+ % dyn_const_value(b, dyn_const_values, dyn_const_params)
+ }
+ DynamicConstant::Max(a, b) => max(
+ dyn_const_value(a, dyn_const_values, dyn_const_params),
+ dyn_const_value(b, dyn_const_values, dyn_const_params),
+ ),
+ DynamicConstant::Min(a, b) => min(
+ dyn_const_value(a, dyn_const_values, dyn_const_params),
+ dyn_const_value(b, dyn_const_values, dyn_const_params),
+ ),
}
}
+
// Each control token stores a current position, and also a mapping of fork nodes -> thread idx.
#[derive(Debug, Clone, Eq, PartialEq)]
pub struct ControlToken {
@@ -78,7 +115,12 @@ pub struct ControlToken {
impl ControlToken {
pub fn moved_to(&self, next: NodeID) -> ControlToken {
- ControlToken { curr: next, prev: self.curr, thread_indicies: self.thread_indicies.clone(), phi_values: self.phi_values.clone() }
+ ControlToken {
+ curr: next,
+ prev: self.curr,
+ thread_indicies: self.thread_indicies.clone(),
+ phi_values: self.phi_values.clone(),
+ }
}
}
impl<'a> FunctionExecutionState<'a> {
@@ -89,7 +131,15 @@ impl<'a> FunctionExecutionState<'a> {
function_contexts: &'a Vec<FunctionContext>,
dynamic_constant_params: Vec<usize>,
) -> Self {
- assert_eq!(args.len(), module.functions[function_id.idx()].param_types.len());
+ println!(
+ "param types: {:?}",
+ module.functions[function_id.idx()].param_types
+ );
+
+ assert_eq!(
+ args.len(),
+ module.functions[function_id.idx()].param_types.len()
+ );
FunctionExecutionState {
args,
@@ -123,15 +173,10 @@ impl<'a> FunctionExecutionState<'a> {
}
/* Drives PHI values of this region for a control token, returns the next control node. */
- pub fn handle_region(
- &mut self,
- token: &mut ControlToken,
- preds: &Box<[NodeID]>,
- ) -> NodeID {
-
+ pub fn handle_region(&mut self, token: &mut ControlToken, preds: &Box<[NodeID]>) -> NodeID {
let prev = token.prev;
let node = token.curr;
-
+
// Gather PHI nodes for this region node.
let phis: Vec<NodeID> = self
.get_def_use()
@@ -178,8 +223,12 @@ impl<'a> FunctionExecutionState<'a> {
.try_phi()
.expect("PANIC: handle_phi on non-phi node.");
let value_node = data[edge];
- // println!("Latching PHI value of node {:?}", value_node.idx());
+
let value = self.handle_data(token, value_node);
+ if VERBOSE {
+ println!("Latching PHI {:?} to {:?}", phi.idx(), value);
+ }
+
(phi, value)
}
@@ -199,26 +248,32 @@ impl<'a> FunctionExecutionState<'a> {
})
.collect();
- for reduction in reduces {
- self.handle_reduction(&token, reduction);
+ for reduction in &reduces {
+ self.handle_reduction(&token, *reduction);
}
let thread_values = self.get_thread_factors(&token, join);
- // dbg!(thread_values.clone());
// This and_modify doesn't do aynthing??
self.join_counters
.entry((thread_values.clone(), join))
.and_modify(|v| *v -= 1);
+ if VERBOSE {
+ println!(
+ "join, thread_values : {:?}, {:?}",
+ join,
+ thread_values.clone()
+ );
+ }
if *self
.join_counters
- .get(&(thread_values, join))
+ .get(&(thread_values.clone(), join))
.expect("PANIC: join counter not initialized")
== 0
{
let curr = token.curr;
token.prev = curr;
- token.thread_indicies.pop(); // Get rid of this thread index.
+ token.thread_indicies.truncate(thread_values.len()); // Get rid of this thread index.
token.curr = self.get_control_subgraph().succs(join).next().unwrap();
Some(token)
} else {
@@ -236,16 +291,36 @@ impl<'a> FunctionExecutionState<'a> {
// Take the top N entries such that it matches the length of the TRF in the control token.
// Get the depth of the control token that is requesting this reduction node.
- let fork_levels = nested_forks.len();
- let len = fork_levels - 1;
+ // Sum over all thread dimensions in nested forks
+ let fork_levels: usize = nested_forks
+ .iter()
+ .map(|ele| {
+ self.get_function().nodes[ele.idx()]
+ .try_fork()
+ .unwrap()
+ .1
+ .len()
+ })
+ .sum();
+
+ let len = if nested_forks.is_empty() {
+ fork_levels - 1
+ } else {
+ fork_levels
+ - (self.get_function().nodes[nested_forks.first().unwrap().idx()]
+ .try_fork()
+ .unwrap()
+ .1
+ .len())
+ };
+
let mut thread_values = token.thread_indicies.clone();
thread_values.truncate(len);
thread_values
}
- pub fn intialize_reduction(&mut self, token_at_fork: &ControlToken, reduce: NodeID) {
-
+ pub fn initialize_reduction(&mut self, token_at_fork: &ControlToken, reduce: NodeID) {
let token = token_at_fork;
let (control, init, _) = &self.get_function().nodes[reduce.idx()]
@@ -255,12 +330,16 @@ impl<'a> FunctionExecutionState<'a> {
let thread_values = self.get_thread_factors(token, *control);
let init = self.handle_data(&token, *init);
- // Q (@xrouth): It is UB to have the initializer depend on things within the fork-join section? do we check for that?
- // A: Should be done in verify (TODO).
+
+ if VERBOSE {
+ println!(
+ "reduction {:?} initialized to: {:?} on thread {:?}",
+ reduce, init, thread_values
+ );
+ }
self.reduce_values
- .entry((thread_values.clone(), reduce))
- .or_insert(init);
+ .insert((thread_values.clone(), reduce), init);
}
// Drive the reduction, this will be invoked for each control token.
@@ -271,58 +350,86 @@ impl<'a> FunctionExecutionState<'a> {
let thread_values = self.get_thread_factors(token, *control);
- // If empty set to default (figure out how to not repeat this check)
- // TODO: (Can we do it upon entry to the fork node?) (YES!)
-
let data = self.handle_data(&token, *reduct);
- /*
- println!(
- "reduction write: {:?}, {:?}, {:?}",
- thread_values, reduce, data
- ); */
+
+ if VERBOSE {
+ println!(
+ "reduction {:?} write of {:?} on thread {:?}",
+ reduce, data, thread_values
+ );
+ }
self.reduce_values.insert((thread_values, reduce), data);
}
pub fn handle_data(&mut self, token: &ControlToken, node: NodeID) -> InterpreterVal {
- // println!("Data Node: {} {:?}", node.idx(), &self.get_function().nodes[node.idx()]);
-
// Partial borrow complaint. :/
- match &self.module.functions[self.function_id.idx()].nodes[node.idx()]{
- Node::Phi { control: _, data: _ } => (*token
+ match &self.module.functions[self.function_id.idx()].nodes[node.idx()] {
+ Node::Phi {
+ control: _,
+ data: _,
+ } => (*token
.phi_values
.get(&node)
- .expect("PANIC: Phi value not latched."))
+ .expect(&format!("PANIC: Phi {:?} value not latched.", node)))
.clone(),
- Node::ThreadID { control } => {
+ Node::ThreadID { control, dimension } => {
// `control` is the fork that drives this node.
- let nesting_level = self
+
+ let nested_forks = self
.get_fork_join_nest()
.get(control)
.expect("PANIC: No nesting information for thread index!")
- .len();
- let v = token.thread_indicies[nesting_level - 1]; // Might have to -1?
+ .clone();
+
+ // Skip forks until we get to this level.
+ // How many forks are outer? idfk.
+ let outer_forks: Vec<NodeID> = nested_forks
+ .iter()
+ .cloned()
+ .take_while(|fork| *fork != node)
+ .collect();
+
+ let fork_levels: usize = outer_forks
+ .iter()
+ .skip(1)
+ .map(|ele| {
+ self.get_function().nodes[ele.idx()]
+ .try_fork()
+ .unwrap()
+ .1
+ .len()
+ })
+ .sum();
+
+ // Dimension might need to instead be dimensions - dimension
+ let v = token.thread_indicies[fork_levels + dimension]; // Might have to -1?
+ if VERBOSE {
+ println!(
+ "node: {:?} gives tid: {:?} for thread: {:?}, dim: {:?}",
+ node, v, token.thread_indicies, dimension
+ );
+ }
InterpreterVal::DynamicConstant((v).into())
}
// If we read from a reduction that is the same depth as this thread, we need to write back to it before anyone else reads from it.
// This probably isn't the exact condition, but somethign similar. Anyways, we achieve correctness by iterating control nodes recursively.
Node::Reduce {
control,
- init,
+ init: _,
reduct: _,
} => {
let thread_values = self.get_thread_factors(token, *control);
- // println!("reduction read: {:?}, {:?}", thread_values, node);
- let entry = self
- .reduce_values
- .entry((thread_values, node));
-
+ let entry = self.reduce_values.entry((thread_values.clone(), node));
+
let val = match entry {
Occupied(v) => v.get().clone(),
- std::collections::hash_map::Entry::Vacant(_) => panic!("Reduce has not been initialized!"),
+ std::collections::hash_map::Entry::Vacant(_) => panic!(
+ "Ctrl token: {:?}, Reduce {:?} has not been initialized!, TV: {:?}",
+ token, node, thread_values
+ ),
};
- // println!("value: {:?}", val.clone());
val
}
Node::Parameter { index } => self.args[*index].clone(),
@@ -333,21 +440,22 @@ impl<'a> FunctionExecutionState<'a> {
&self.module.constants,
&self.module.types,
&self.module.dynamic_constants,
- &self.dynamic_constant_params
+ &self.dynamic_constant_params,
)
}
Node::DynamicConstant { id } => {
- let dyn_con = &self.module.dynamic_constants[id.idx()];
- let v = match dyn_con {
- DynamicConstant::Constant(v) => v,
- DynamicConstant::Parameter(v) => &self.dynamic_constant_params[*v],
- };
+ let v = dyn_const_value(
+ id,
+ &self.module.dynamic_constants,
+ &self.dynamic_constant_params,
+ );
+
// TODO: Figure out what type / semantics are of thread ID and dynamic const.
- InterpreterVal::DynamicConstant((*v).into())
+ InterpreterVal::UnsignedInteger64(v.try_into().expect("too big dyn const!"))
}
Node::Unary { input, op } => {
let val = self.handle_data(token, *input);
- InterpreterVal::unary_op(*op, val)
+ InterpreterVal::unary_op(&self.module.types, *op, val)
}
Node::Binary { left, right, op } => {
let left = self.handle_data(token, *left);
@@ -376,25 +484,26 @@ impl<'a> FunctionExecutionState<'a> {
}
}
Node::Call {
+ control: _,
function,
dynamic_constants,
args,
} => {
+ let args = args
+ .into_iter()
+ .map(|arg_node| self.handle_data(token, *arg_node))
+ .collect();
- let args = args.into_iter()
- .map(|arg_node| self.handle_data(token, *arg_node))
- .collect();
-
-
- let dynamic_constant_params = dynamic_constants.into_iter()
- .map(|id| {
- let dyn_con = &self.module.dynamic_constants[id.idx()];
- let v = match dyn_con {
- DynamicConstant::Constant(v) => *v,
- DynamicConstant::Parameter(v) => self.dynamic_constant_params[*v],
- };
- v
- }).collect_vec();
+ let dynamic_constant_params = dynamic_constants
+ .into_iter()
+ .map(|id| {
+ dyn_const_value(
+ id,
+ &self.module.dynamic_constants,
+ &self.dynamic_constant_params,
+ )
+ })
+ .collect_vec();
let mut state = FunctionExecutionState::new(
args,
@@ -408,8 +517,19 @@ impl<'a> FunctionExecutionState<'a> {
}
Node::Read { collect, indices } => {
let collection = self.handle_data(token, *collect);
+ if let InterpreterVal::Undef(_) = collection {
+ collection
+ } else {
+ let result = self.handle_read(token, collection.clone(), indices);
- self.handle_read(token, collection, indices)
+ if VERBOSE {
+ println!(
+ "{:?} read value : {:?} from {:?}, {:?} at index {:?}",
+ node, result, collect, collection, indices
+ );
+ }
+ result
+ }
}
Node::Write {
collect,
@@ -417,9 +537,14 @@ impl<'a> FunctionExecutionState<'a> {
indices,
} => {
let collection = self.handle_data(token, *collect);
- let data = self.handle_data(token, *data);
- self.handle_write(token, collection, data, indices)
+ if let InterpreterVal::Undef(_) = collection {
+ collection
+ } else {
+ let data = self.handle_data(token, *data);
+ self.handle_write(token, collection, data, indices)
+ }
}
+ Node::Undef { ty } => InterpreterVal::Undef(*ty),
_ => todo!(),
}
}
@@ -431,13 +556,19 @@ impl<'a> FunctionExecutionState<'a> {
data: InterpreterVal,
indices: &[Index],
) -> InterpreterVal {
- let index = &indices[0];
-
// TODO (@xrouth): Recurse on writes correctly
- let val = match index {
- Index::Field(_) => todo!(),
- Index::Variant(_) => todo!(),
- Index::Position(array_indices) => {
+ let val = match indices.first() {
+ Some(Index::Field(idx)) => {
+ if let InterpreterVal::Product(type_id, mut vals) = collection {
+ vals[*idx] = data;
+ InterpreterVal::Product(type_id, vals)
+ } else {
+ panic!("PANIC: Field index on not a product type")
+ }
+ }
+ None => collection,
+ Some(Index::Variant(_)) => todo!(),
+ Some(Index::Position(array_indices)) => {
// Arrays also have inner indices...
// Recover dimensional data from types.
let array_indices: Vec<_> = array_indices
@@ -451,12 +582,21 @@ impl<'a> FunctionExecutionState<'a> {
.try_extents()
.expect("PANIC: wrong type for array")
.into_iter()
- .map(|extent| dyn_const_value(&self.module.dynamic_constants[extent.idx()], &self.dynamic_constant_params))
+ .map(|extent| {
+ dyn_const_value(
+ extent,
+ &self.module.dynamic_constants,
+ &self.dynamic_constant_params,
+ )
+ })
.collect();
let idx = InterpreterVal::array_idx(&extents, &array_indices);
- //println!("idx: {:?}", idx);
- vals[idx] = data;
- InterpreterVal::Array(type_id, vals)
+ if idx >= vals.len() {
+ InterpreterVal::Undef(type_id)
+ } else {
+ vals[idx] = data;
+ InterpreterVal::Array(type_id, vals)
+ }
} else {
panic!("PANIC: Position index on not an array")
}
@@ -484,6 +624,10 @@ impl<'a> FunctionExecutionState<'a> {
.map(|idx| self.handle_data(token, *idx).as_usize())
.collect();
+ if VERBOSE {
+ println!("read at rt indicies: {:?}", array_indices);
+ }
+
// TODO: Implemenet . try_array() and other try_conversions on the InterpreterVal type
if let InterpreterVal::Array(type_id, vals) = collection {
// TODO: Make this its own funciton to reuse w/ array_size
@@ -491,9 +635,20 @@ impl<'a> FunctionExecutionState<'a> {
.try_extents()
.expect("PANIC: wrong type for array")
.into_iter()
- .map(|extent| dyn_const_value(&self.module.dynamic_constants[extent.idx()], &self.dynamic_constant_params))
+ .map(|extent| {
+ dyn_const_value(
+ extent,
+ &self.module.dynamic_constants,
+ &self.dynamic_constant_params,
+ )
+ })
.collect();
- vals[InterpreterVal::array_idx(&extents, &array_indices)].clone()
+ // FIXME: This type may be wrong.
+ let ret = vals
+ .get(InterpreterVal::array_idx(&extents, &array_indices))
+ .unwrap_or(&InterpreterVal::Undef(type_id))
+ .clone();
+ ret
} else {
panic!("PANIC: Position index on not an array")
}
@@ -521,32 +676,40 @@ impl<'a> FunctionExecutionState<'a> {
let mut live_tokens: Vec<ControlToken> = Vec::new();
live_tokens.push(start_token);
-
// To do reduction nodes correctly we have to traverse control tokens in a depth-first fashion (i.e immediately handle spawned threads).
'outer: loop {
- let mut ctrl_token = live_tokens.pop().expect("PANIC: Interpreter ran out of control tokens without returning.");
-
- /* println!(
- "\n\nNew Token at: Control State: {} threads: {:?}, {:?}",
- ctrl_token.curr.idx(),
- ctrl_token.thread_indicies.clone(),
- &self.get_function().nodes[ctrl_token.curr.idx()]
- ); */
+ let mut ctrl_token = live_tokens
+ .pop()
+ .expect("PANIC: Interpreter ran out of control tokens without returning.");
+
+ // TODO: (@xrouth): Enable this + PHI latch logging wi/ a simple debug flag.
+ // Tracking PHI vals and control state is very useful for debugging.
+
+ if VERBOSE {
+ println!(
+ "control token {} {}",
+ ctrl_token.curr.idx(),
+ &self.get_function().nodes[ctrl_token.curr.idx()].lower_case_name()
+ );
+ }
// TODO: Rust is annoying and can't recognize that this is a partial borrow.
- // Can't partial borrow, so need a clone.
+ // Can't partial borrow, so need a clone.
let node = &self.get_function().nodes[ctrl_token.curr.idx()].clone();
let new_tokens = match node {
Node::Start => {
- let next: NodeID = self.get_control_subgraph().succs(ctrl_token.curr).next().unwrap();
+ let next: NodeID = self
+ .get_control_subgraph()
+ .succs(ctrl_token.curr)
+ .next()
+ .unwrap();
let ctrl_token = ctrl_token.moved_to(next);
-
+
vec![ctrl_token]
}
Node::Region { preds } => {
-
- // Updates
+ // Updates
let next = self.handle_region(&mut ctrl_token, &preds);
let ctrl_token = ctrl_token.moved_to(next);
@@ -558,6 +721,7 @@ impl<'a> FunctionExecutionState<'a> {
// Convert condition to usize
let cond: usize = match cond {
InterpreterVal::Boolean(v) => v.into(),
+ InterpreterVal::Undef(_) => panic!("PANIC: Undef reached IF"),
_ => panic!("PANIC: Invalid condition for IF, please typecheck."),
};
@@ -576,7 +740,11 @@ impl<'a> FunctionExecutionState<'a> {
vec![ctrl_token]
}
Node::Projection { .. } => {
- let next: NodeID = self.get_control_subgraph().succs(ctrl_token.curr).next().unwrap();
+ let next: NodeID = self
+ .get_control_subgraph()
+ .succs(ctrl_token.curr)
+ .next()
+ .unwrap();
let ctrl_token = ctrl_token.moved_to(next);
@@ -584,28 +752,52 @@ impl<'a> FunctionExecutionState<'a> {
}
Node::Match { control: _, sum: _ } => todo!(),
- Node::Fork { control: _, factor } => {
+ Node::Fork {
+ control: _,
+ factors,
+ } => {
let fork = ctrl_token.curr;
- let dyn_con = &self.module.dynamic_constants[factor.idx()];
+ // if factors.len() > 1 {
+ // panic!("multi-dimensional forks unimplemented")
+ // }
+
+ let factors = factors
+ .iter()
+ .map(|f| {
+ dyn_const_value(
+ &f,
+ &self.module.dynamic_constants,
+ &self.dynamic_constant_params,
+ )
+ })
+ .rev();
- let thread_factor = match dyn_con {
- DynamicConstant::Constant(v) => v,
- DynamicConstant::Parameter(v) => &self.dynamic_constant_params[*v],
- }.clone();
+ let n_tokens: usize = factors.clone().product();
- // Update control token
- let next = self.get_control_subgraph().succs(ctrl_token.curr).nth(0).unwrap();
+ // Update control token
+ let next = self
+ .get_control_subgraph()
+ .succs(ctrl_token.curr)
+ .nth(0)
+ .unwrap();
let ctrl_token = ctrl_token.moved_to(next);
- let mut tokens_to_add = Vec::with_capacity(thread_factor);
+ let mut tokens_to_add = Vec::with_capacity(n_tokens);
- assert_ne!(thread_factor, 0);
+ assert_ne!(n_tokens, 0);
// Token is at its correct sontrol succesor already.
+
// Add the new thread index.
- for i in 0..(thread_factor) {
+ let num_outer_dims = ctrl_token.thread_indicies.len();
+ for i in 0..n_tokens {
+ let mut temp = i;
let mut new_token = ctrl_token.clone(); // Copy map, curr, prev, etc.
- new_token.thread_indicies.push(i); // Stack of thread indicies
+
+ for (_, dim) in factors.clone().enumerate().rev() {
+ new_token.thread_indicies.insert(num_outer_dims, temp % dim); // Stack of thread indicies
+ temp /= dim;
+ }
tokens_to_add.push(new_token);
}
@@ -630,15 +822,21 @@ impl<'a> FunctionExecutionState<'a> {
}
})
.collect();
-
+
for reduction in reduces {
// TODO: Is this the correct reduction?
- self.intialize_reduction(&ctrl_token, reduction);
+ self.initialize_reduction(&ctrl_token, reduction);
}
+ if VERBOSE {
+ println!(
+ "tf, fork, join, n_tokens: {:?}, {:?}, {:?}, {:?}",
+ thread_factors, fork, join, n_tokens
+ );
+ }
+ self.join_counters.insert((thread_factors, join), n_tokens);
- self.join_counters.insert((thread_factors, join), thread_factor);
-
+ tokens_to_add.reverse();
tokens_to_add
}
Node::Join { control: _ } => {
@@ -653,7 +851,6 @@ impl<'a> FunctionExecutionState<'a> {
}
Node::Return { control: _, data } => {
let result = self.handle_data(&ctrl_token, *data);
- // println!("result = {:?}", result);
break 'outer result;
}
_ => {
@@ -664,9 +861,6 @@ impl<'a> FunctionExecutionState<'a> {
for i in new_tokens {
live_tokens.push(i);
}
-
}
}
}
-
-
diff --git a/hercules_test/hercules_interpreter/src/lib.rs b/hercules_test/hercules_interpreter/src/lib.rs
index 89fae51a..66f8c4ea 100644
--- a/hercules_test/hercules_interpreter/src/lib.rs
+++ b/hercules_test/hercules_interpreter/src/lib.rs
@@ -2,16 +2,23 @@ pub mod interpreter;
pub mod value;
use std::fs::File;
+use std::io::Read;
use hercules_ir::Module;
use hercules_ir::TypeID;
+use hercules_ir::ID;
+
+pub use juno_scheduler::PassManager;
pub use crate::interpreter::*;
pub use crate::value::*;
-// Get a vec of
-pub fn into_interp_val(module: &Module, wrapper: InterpreterWrapper, target_ty_id: TypeID) -> InterpreterVal
-{
+// Get a vec of
+pub fn into_interp_val(
+ module: &Module,
+ wrapper: InterpreterWrapper,
+ target_ty_id: TypeID,
+) -> InterpreterVal {
match wrapper {
InterpreterWrapper::Boolean(v) => InterpreterVal::Boolean(v),
InterpreterWrapper::Integer8(v) => InterpreterVal::Integer8(v),
@@ -29,31 +36,25 @@ pub fn into_interp_val(module: &Module, wrapper: InterpreterWrapper, target_ty_i
InterpreterWrapper::Array(array) => {
let ty = &module.types[target_ty_id.idx()];
- let ele_type = ty.try_element_type().expect("PANIC: Type ID");
- // unwrap -> map to rust type, check
-
+ ty.try_element_type()
+ .expect("PANIC: Invalid parameter type");
+
let mut values = vec![];
-
+
for i in 0..array.len() {
values.push(into_interp_val(module, array[i].clone(), TypeID::new(0)));
}
-
+
InterpreterVal::Array(target_ty_id, values.into_boxed_slice())
}
}
-}
-
-pub fn array_from_interp_val<T: Clone>(module: &Module, interp_val: InterpreterVal) -> Vec<T>
- where value::InterpreterVal: Into<T>
-{
- vec![]
}
// Recursively turns rt args into interpreter wrappers.
#[macro_export]
macro_rules! parse_rt_args {
($arg:expr) => {
- {
+ {
let mut values: Vec<InterpreterWrapper> = vec![];
@@ -63,7 +64,7 @@ macro_rules! parse_rt_args {
}
};
( $arg:expr, $($tail_args:expr), +) => {
- {
+ {
let mut values: Vec<InterpreterWrapper> = vec![];
values.push($arg.into());
@@ -85,9 +86,16 @@ pub fn parse_file(path: &str) -> Module {
module
}
+pub fn parse_module_from_hbin(path: &str) -> hercules_ir::ir::Module {
+ let mut file = File::open(path).expect("PANIC: Unable to open input file.");
+ let mut buffer = vec![];
+ file.read_to_end(&mut buffer).unwrap();
+ postcard::from_bytes(&buffer).unwrap()
+}
+
#[macro_export]
macro_rules! interp_module {
- ($module:ident, $dynamic_constants:expr, $($args:expr), *) => {
+ ($module:ident, $entry_func:expr, $dynamic_constants:expr, $($args:expr), *) => {
{
//let hir_file = String::from($path);
@@ -96,27 +104,22 @@ macro_rules! interp_module {
let dynamic_constants: Vec<usize> = $dynamic_constants.into();
let module = $module.clone(); //parse_file(hir_file);
- let mut pm = hercules_opt::pass::PassManager::new(module);
- pm.add_pass(hercules_opt::pass::Pass::Verify);
-
- pm.run_passes();
+ let mut pm = PassManager::new(module);
+ pm.make_typing();
pm.make_reverse_postorders();
pm.make_doms();
pm.make_fork_join_maps();
pm.make_fork_join_nests();
pm.make_control_subgraphs();
- pm.make_plans();
let reverse_postorders = pm.reverse_postorders.as_ref().unwrap().clone();
let doms = pm.doms.as_ref().unwrap().clone();
let fork_join_maps = pm.fork_join_maps.as_ref().unwrap().clone();
let fork_join_nests = pm.fork_join_nests.as_ref().unwrap().clone();
- let plans = pm.plans.as_ref().unwrap().clone();
let control_subgraphs = pm.control_subgraphs.as_ref().unwrap().clone();
let def_uses = pm.def_uses.as_ref().unwrap().clone();
let module = pm.get_module();
-
let mut function_contexts = vec![];
for idx in 0..module.functions.len() {
@@ -129,7 +132,7 @@ macro_rules! interp_module {
function_contexts.push(context);
}
- let function_number = 0;
+ let function_number = $entry_func;
let parameter_types = &module.functions[function_number].param_types;
@@ -148,27 +151,19 @@ macro_rules! interp_module {
};
}
-
#[macro_export]
macro_rules! interp_file_with_passes {
($path:literal, $dynamic_constants:expr, $passes:expr, $($args:expr), *) => {
{
let module = parse_file($path);
-
- let result_before = interp_module!(module, $dynamic_constants, $($args), *);
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
- for pass in $passes {
- pm.add_pass(pass);
- }
+ let result_before = interp_module!(module, $dynamic_constants, $($args), *);
- pm.run_passes();
+ let module = run_schedule_on_hercules(module, None).unwrap();
- let module = pm.get_module();
- let result_after = interp_module!(module, $dynamic_constants, $($args), *);
+ let result_after = interp_module!(module, $dynamic_constants, $($args), *);
assert_eq!(result_after, result_before);
}
};
-}
\ No newline at end of file
+}
diff --git a/hercules_test/hercules_interpreter/src/main.rs b/hercules_test/hercules_interpreter/src/main.rs
deleted file mode 100644
index 5db31cd7..00000000
--- a/hercules_test/hercules_interpreter/src/main.rs
+++ /dev/null
@@ -1,28 +0,0 @@
-use std::fs::File;
-use std::io::prelude::*;
-
-use clap::Parser;
-
-use hercules_ir::*;
-
-use hercules_interpreter::interpreter::*;
-use hercules_interpreter::*;
-use hercules_interpreter::value;
-
-#[derive(Parser, Debug)]
-#[command(author, version, about, long_about = None)]
-struct Args {
- hir_file: String,
-
- #[arg(short, long, default_value_t = String::new())]
- output: String,
-}
-
-fn main() {
- let args = Args::parse();
- let module = parse_file(&args.hir_file);
- let ret_val = interp_module!(module, [2, 3, 4], 1, 3);
-
- println!("ret val: {:?}", ret_val);
-}
-
diff --git a/hercules_test/hercules_interpreter/src/value.rs b/hercules_test/hercules_interpreter/src/value.rs
index e032bd5b..53911e05 100644
--- a/hercules_test/hercules_interpreter/src/value.rs
+++ b/hercules_test/hercules_interpreter/src/value.rs
@@ -29,6 +29,7 @@ pub enum InterpreterVal {
Float32(ordered_float::OrderedFloat<f32>),
Float64(ordered_float::OrderedFloat<f64>),
+ Undef(TypeID),
Product(TypeID, Box<[InterpreterVal]>),
Summation(TypeID, u32, Box<[InterpreterVal]>),
Array(TypeID, Box<[InterpreterVal]>), // TypeID of the array Type (not the element type)
@@ -76,6 +77,7 @@ where
{
fn from(value: Vec<T>) -> Self {
let mut values = vec![];
+ values.reserve(value.len());
for i in 0..value.len() {
values.push(value[i].clone().into());
}
@@ -89,8 +91,23 @@ where
{
fn from(value: &[T]) -> Self {
let mut values = vec![];
+ values.reserve(value.len());
for i in 0..value.len() {
- values[i] = value[i].clone().into()
+ values.push(value[i].clone().into());
+ }
+ InterpreterWrapper::Array(values.into_boxed_slice())
+ }
+}
+
+impl<T> From<Box<[T]>> for InterpreterWrapper
+where
+ T: Into<InterpreterWrapper> + Clone,
+{
+ fn from(value: Box<[T]>) -> Self {
+ let mut values = vec![];
+ values.reserve(value.len());
+ for i in 0..value.len() {
+ values.push(value[i].clone().into());
}
InterpreterWrapper::Array(values.into_boxed_slice())
}
@@ -138,7 +155,10 @@ impl<'a> InterpreterVal {
Constant::Float32(v) => Self::Float32(v),
Constant::Float64(v) => Self::Float64(v),
- Constant::Product(_, _) => todo!(),
+ Constant::Product(ref type_id, ref constant_ids) => {
+ // Self::Product((), ())
+ todo!()
+ }
Constant::Summation(_, _, _) => todo!(),
Constant::Array(type_id) => {
// TODO: This is currently only implemented for arrays of primitive types, implement zero initializers for other types.
@@ -149,7 +169,7 @@ impl<'a> InterpreterVal {
.expect("PANIC: wrong type for array")
.into_iter()
.map(|extent| {
- dyn_const_value(&dynamic_constants[extent.idx()], &dynamic_constant_params)
+ dyn_const_value(extent, &dynamic_constants, &dynamic_constant_params)
})
.collect();
@@ -193,6 +213,14 @@ impl<'a> InterpreterVal {
left: InterpreterVal,
right: InterpreterVal,
) -> InterpreterVal {
+ // If either are undef, propogate undef
+ if let InterpreterVal::Undef(v) = left {
+ return InterpreterVal::Undef(v);
+ }
+ if let InterpreterVal::Undef(v) = right {
+ return InterpreterVal::Undef(v);
+ }
+
// Do some type conversion first.
let left = match left {
InterpreterVal::DynamicConstant(v) => match right {
@@ -213,7 +241,8 @@ impl<'a> InterpreterVal {
InterpreterVal::UnsignedInteger64(v.try_into().unwrap())
}
InterpreterVal::DynamicConstant(_) => {
- panic!("PANIC: Some math on dynamic constants is unimplemented")
+ InterpreterVal::UnsignedInteger64(v.try_into().unwrap())
+ //panic!("PANIC: Some math on dynamic constants is unimplemented")
}
// InterpreterVal::ThreadID(_) => InterpreterVal::Boolean(v),
_ => panic!("PANIC: Some math on dynamic constants is unimplemented"),
@@ -241,7 +270,8 @@ impl<'a> InterpreterVal {
InterpreterVal::UnsignedInteger64(v.try_into().unwrap())
}
InterpreterVal::DynamicConstant(_) => {
- panic!("PANIC: Some math on dynamic constants is unimplemented")
+ InterpreterVal::UnsignedInteger64(v.try_into().unwrap())
+ //panic!("PANIC: Some math on dynamic constants is unimplemented")
}
_ => panic!("PANIC: Some math on dynamic constants is unimplemented"),
},
@@ -772,7 +802,7 @@ impl<'a> InterpreterVal {
}
}
- pub fn unary_op(op: UnaryOperator, val: InterpreterVal) -> Self {
+ pub fn unary_op(types: &Vec<Type>, op: UnaryOperator, val: InterpreterVal) -> Self {
match (op, val) {
(UnaryOperator::Not, Self::Boolean(val)) => Self::Boolean(!val),
(UnaryOperator::Not, Self::Integer8(val)) => Self::Integer8(!val),
@@ -789,7 +819,28 @@ impl<'a> InterpreterVal {
(UnaryOperator::Neg, Self::Integer64(val)) => Self::Integer64(-val),
(UnaryOperator::Neg, Self::Float32(val)) => Self::Float32(-val),
(UnaryOperator::Neg, Self::Float64(val)) => Self::Float64(-val),
- (UnaryOperator::Cast(_), _) => todo!("Write cast impl"),
+ (UnaryOperator::Cast(type_id), val) => {
+ // FIXME: This probably doesn't work.
+ let val = val.as_i128();
+ match types[type_id.idx()] {
+ Type::Control => todo!(),
+ Type::Boolean => todo!(),
+ Type::Integer8 => todo!(),
+ Type::Integer16 => todo!(),
+ Type::Integer32 => Self::Integer32(val.try_into().unwrap()),
+ Type::Integer64 => todo!(),
+ Type::UnsignedInteger8 => todo!(),
+ Type::UnsignedInteger16 => todo!(),
+ Type::UnsignedInteger32 => todo!(),
+ Type::UnsignedInteger64 => Self::UnsignedInteger64(val.try_into().unwrap()),
+ Type::Float32 => todo!(),
+ Type::Float64 => todo!(),
+ Type::Product(_) => todo!(),
+ Type::Summation(_) => todo!(),
+ Type::Array(type_id, _) => todo!(),
+ }
+ }
+ (_, Self::Undef(v)) => InterpreterVal::Undef(v),
_ => panic!("Unsupported combination of unary operation and constant value. Did typechecking succeed?")
}
}
@@ -811,6 +862,23 @@ impl<'a> InterpreterVal {
}
}
+ pub fn as_i128(&self) -> i128 {
+ match *self {
+ InterpreterVal::Boolean(v) => v.try_into().unwrap(),
+ InterpreterVal::Integer8(v) => v.try_into().unwrap(),
+ InterpreterVal::Integer16(v) => v.try_into().unwrap(),
+ InterpreterVal::Integer32(v) => v.try_into().unwrap(),
+ InterpreterVal::Integer64(v) => v.try_into().unwrap(),
+ InterpreterVal::UnsignedInteger8(v) => v.try_into().unwrap(),
+ InterpreterVal::UnsignedInteger16(v) => v.try_into().unwrap(),
+ InterpreterVal::UnsignedInteger32(v) => v.try_into().unwrap(),
+ InterpreterVal::UnsignedInteger64(v) => v.try_into().unwrap(),
+ InterpreterVal::DynamicConstant(v) => v.try_into().unwrap(),
+ InterpreterVal::ThreadID(v) => v.try_into().unwrap(),
+ _ => panic!("PANIC: Value not castable to usize"),
+ }
+ }
+
// Defines row major / how we layout our arrays
pub fn array_idx(extents: &[usize], indices: &[usize]) -> usize {
let a = extents
diff --git a/hercules_test/hercules_tests/Cargo.toml b/hercules_test/hercules_tests/Cargo.toml
index 9bd6fe7b..8c140e75 100644
--- a/hercules_test/hercules_tests/Cargo.toml
+++ b/hercules_test/hercules_tests/Cargo.toml
@@ -9,6 +9,7 @@ clap = { version = "*", features = ["derive"] }
rand = "*"
hercules_ir = { path = "../../hercules_ir" }
hercules_opt = { path = "../../hercules_opt" }
+juno_scheduler = { path = "../../juno_scheduler" }
hercules_interpreter = { path = "../hercules_interpreter" }
itertools = "*"
ordered-float = "*"
diff --git a/hercules_test/hercules_tests/tests/fork_transform_tests.rs b/hercules_test/hercules_tests/tests/fork_transform_tests.rs
new file mode 100644
index 00000000..3799ca0a
--- /dev/null
+++ b/hercules_test/hercules_tests/tests/fork_transform_tests.rs
@@ -0,0 +1,99 @@
+use std::{env, fs::File, io::Read, path::Path};
+
+use hercules_interpreter::*;
+use hercules_ir::ID;
+use juno_scheduler::ir::*;
+
+use juno_scheduler::pass;
+use juno_scheduler::{default_schedule, run_schedule_on_hercules};
+use rand::Rng;
+
+#[test]
+fn fission_simple1() {
+ let module = parse_file("../test_inputs/fork_transforms/fork_fission/simple1.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched = Some(default_schedule![
+ Verify, //Xdot,
+ Unforkify, //Xdot,
+ DCE, Verify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_2);
+ assert_eq!(result_1, result_2)
+}
+
+// #[test]
+// fn fission_simple2() {
+// let module = parse_file("../test_inputs/fork_transforms/fork_fission/simple2.hir");
+// let dyn_consts = [10];
+// let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+// let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+// println!("result: {:?}", result_1);
+
+// let sched: Option<ScheduleStmt> = Some(default_schedule![
+// Verify,
+// ForkFission,
+// DCE,
+// Verify,
+// ]);
+
+// let module = run_schedule_on_hercules(module, sched).unwrap();
+// let result_2 = interp_module!(module, 0, dyn_consts, 2);
+// println!("result: {:?}", result_2);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[ignore] // Wait
+// #[test]
+// fn fission_tricky() {
+// // This either crashes or gives wrong result depending on the order which reduces are observed in.
+// let module = parse_file("../test_inputs/fork_transforms/fork_fission/tricky.hir");
+// let dyn_consts = [10];
+// let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+// let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+// println!("result: {:?}", result_1);
+
+// let sched: Option<ScheduleStmt> = Some(default_schedule![
+// Verify,
+// ForkFission,
+// DCE,
+// Verify,
+// ]);
+
+// let module = run_schedule_on_hercules(module, sched).unwrap();
+// let result_2 = interp_module!(module, 0, dyn_consts, 2);
+// println!("result: {:?}", result_2);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[ignore]
+// #[test]
+// fn inner_loop() {
+// let module = parse_file("../test_inputs/fork_transforms/fork_fission/inner_loop.hir");
+// let dyn_consts = [10, 20];
+// let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+// let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+// println!("result: {:?}", result_1);
+
+// let sched: Option<ScheduleStmt> = Some(default_schedule![
+// Verify,
+// ForkFission,
+// DCE,
+// Verify,
+// ]);
+
+// let module = run_schedule_on_hercules(module, sched).unwrap();
+// let result_2 = interp_module!(module, 0, dyn_consts, 2);
+// println!("result: {:?}", result_2);
+// assert_eq!(result_1, result_2)
+// }
diff --git a/hercules_test/hercules_tests/tests/forkify_tests.rs b/hercules_test/hercules_tests/tests/forkify_tests.rs
new file mode 100644
index 00000000..8ba8e135
--- /dev/null
+++ b/hercules_test/hercules_tests/tests/forkify_tests.rs
@@ -0,0 +1,294 @@
+use std::{env, fs::File, io::Read, path::Path};
+
+use hercules_interpreter::*;
+use hercules_ir::ID;
+
+use hercules_interpreter::*;
+use juno_scheduler::ir::*;
+use juno_scheduler::pass;
+
+use juno_scheduler::{default_schedule, run_schedule_on_hercules};
+use rand::Rng;
+
+#[test]
+#[ignore]
+fn inner_fork_chain() {
+ let module = parse_file("../test_inputs/forkify/inner_fork_chain.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ // let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, PhiElim, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_2);
+ //assert_eq!(result_1, result_2)
+}
+
+#[test]
+fn loop_simple_iv() {
+ let module = parse_file("../test_inputs/forkify/loop_simple_iv.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_2);
+ assert_eq!(result_1, result_2)
+}
+
+#[test]
+#[ignore]
+fn merged_phi_cycle() {
+ let module = parse_file("../test_inputs/forkify/merged_phi_cycle.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_2);
+ assert_eq!(result_1, result_2)
+}
+
+#[test]
+fn split_phi_cycle() {
+ let module = parse_file("../test_inputs/forkify/split_phi_cycle.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_2);
+ assert_eq!(result_1, result_2)
+}
+
+#[test]
+fn loop_sum() {
+ let module = parse_file("../test_inputs/forkify/loop_sum.hir");
+ let dyn_consts = [20];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+ println!("{:?}, {:?}", result_1, result_2);
+}
+
+#[test]
+fn loop_tid_sum() {
+ let module = parse_file("../test_inputs/forkify/loop_tid_sum.hir");
+ let dyn_consts = [20];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+ println!("{:?}, {:?}", result_1, result_2);
+}
+
+#[test]
+fn loop_array_sum() {
+ let module = parse_file("../test_inputs/forkify/loop_array_sum.hir");
+ let len = 5;
+ let dyn_consts = [len];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, params.clone());
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, params);
+ assert_eq!(result_1, result_2);
+ println!("{:?}, {:?}", result_1, result_2);
+}
+
+/** Nested loop 2 is 2 nested loops with different dyn var parameter dimensions.
+ * It is a add of 1 for each iteration, so the result should be dim1 x dim2
+ * The loop PHIs are structured such that on every outer iteration, inner loop increment is set to the running sum,
+ * Notice how there is no outer_var_inc.
+ *
+ * The alternative, seen in nested_loop1, is to intiailize the inner loop to 0 every time, and track
+ * the outer sum more separaetly.
+ *
+ * Idk what im yapping about.
+*/
+#[test]
+fn nested_loop2() {
+ let module = parse_file("../test_inputs/forkify/nested_loop2.hir");
+ let len = 5;
+ let dyn_consts = [5, 6];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn super_nested_loop() {
+ let module = parse_file("../test_inputs/forkify/super_nested_loop.hir");
+ let len = 5;
+ let dyn_consts = [5, 10, 15];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+}
+
+/**
+ * Tests forkify on a loop where there is control in between the continue projection
+ * and the header. aka control *after* the `loop condition / guard`. This should forkify.
+ */
+#[test]
+fn control_after_condition() {
+ let module = parse_file("../test_inputs/forkify/control_after_condition.hir");
+
+ let size = 10;
+ let dyn_consts = [size];
+ let mut vec = vec![0; size];
+ let mut rng = rand::thread_rng();
+
+ for x in vec.iter_mut() {
+ *x = rng.gen::<i32>() / 100;
+ }
+
+ let result_1 = interp_module!(module, 0, dyn_consts, vec.clone());
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, vec);
+ assert_eq!(result_1, result_2);
+}
+
+/**
+ * Tests forkify on a loop where there is control before the loop condition, so in between the header
+ * and the loop condition. This should not forkify.
+ *
+ * This example is bugged, it reads out of bounds even before forkify.
+ */
+#[ignore]
+#[test]
+fn control_before_condition() {
+ let module = parse_file("../test_inputs/forkify/control_before_condition.hir");
+
+ let size = 11;
+ let dyn_consts = [size - 1];
+ let mut vec = vec![0; size];
+ let mut rng = rand::thread_rng();
+
+ for x in vec.iter_mut() {
+ *x = rng.gen::<i32>() / 100;
+ }
+
+ let result_1 = interp_module!(module, 0, dyn_consts, vec.clone());
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, vec);
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn nested_tid_sum() {
+ let module = parse_file("../test_inputs/forkify/nested_tid_sum.hir");
+ let len = 5;
+ let dyn_consts = [5, 6];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_3 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("{:?}, {:?}, {:?}", result_1, result_2, result_3);
+}
+
+#[test]
+fn nested_tid_sum_2() {
+ let module = parse_file("../test_inputs/forkify/nested_tid_sum_2.hir");
+ let len = 5;
+ let dyn_consts = [5, 6];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+ assert_eq!(result_1, result_2);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_3 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("{:?}, {:?}, {:?}", result_1, result_2, result_3);
+}
+
+/** Tests weird control in outer loop for possible 2d fork-join pair. */
+#[test]
+fn inner_fork_complex() {
+ let module = parse_file("../test_inputs/forkify/inner_fork_complex.hir");
+ let dyn_consts = [5, 6];
+ let params = vec![1, 2, 3, 4, 5]; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 10);
+
+ println!("result: {:?}", result_1);
+
+ let sched: Option<ScheduleStmt> = Some(default_schedule![Verify, Forkify, DCE, Verify,]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 10);
+ assert_eq!(result_1, result_2);
+ println!("{:?}, {:?}", result_1, result_2);
+}
diff --git a/hercules_test/hercules_tests/tests/interpreter_tests.rs b/hercules_test/hercules_tests/tests/interpreter_tests.rs
new file mode 100644
index 00000000..a779c70b
--- /dev/null
+++ b/hercules_test/hercules_tests/tests/interpreter_tests.rs
@@ -0,0 +1,63 @@
+use std::env;
+
+use hercules_interpreter::*;
+use hercules_interpreter::*;
+use hercules_ir::ID;
+use juno_scheduler::ir::*;
+use juno_scheduler::pass;
+
+use juno_scheduler::{default_schedule, run_schedule_on_hercules};
+use rand::Rng;
+
+#[test]
+fn twodeefork() {
+ let module = parse_file("../test_inputs/2d_fork.hir");
+ let d1 = 2;
+ let d2 = 3;
+ let dyn_consts = [d1, d2];
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ let sched = Some(default_schedule![
+ Verify, ForkSplit, //Xdot,
+ Unforkify, //Xdot,
+ DCE, Verify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+
+ let res = (d1 as i32 * d2 as i32);
+ let result_2: InterpreterWrapper = res.into();
+ println!("result: {:?}", result_1); // Should be d1 * d2.
+}
+
+#[test]
+fn threedee() {
+ let module = parse_file("../test_inputs/3d_fork.hir");
+ let d1 = 2;
+ let d2 = 3;
+ let d3 = 5;
+ let dyn_consts = [d1, d2, 5];
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ let sched = Some(default_schedule![
+ Verify, ForkSplit, //Xdot,
+ Unforkify, //Xdot,
+ DCE, Verify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, sched).unwrap();
+ let result_2 = interp_module!(module, 0, dyn_consts, 2);
+
+ let res = (d1 as i32 * d2 as i32 * d3 as i32);
+ let result_2: InterpreterWrapper = res.into();
+ println!("result: {:?}", result_1); // Should be d1 * d2.
+}
+
+#[test]
+fn fivedeefork() {
+ let module = parse_file("../test_inputs/5d_fork.hir");
+ let dyn_consts = [1, 2, 3, 4, 5];
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+ println!("result: {:?}", result_1); // Should be 1 * 2 * 3 * 4 * 5;
+}
diff --git a/hercules_test/hercules_tests/tests/loop_tests.rs b/hercules_test/hercules_tests/tests/loop_tests.rs
new file mode 100644
index 00000000..5832a161
--- /dev/null
+++ b/hercules_test/hercules_tests/tests/loop_tests.rs
@@ -0,0 +1,436 @@
+use std::{env, fs::File, io::Read, path::Path};
+
+use hercules_interpreter::*;
+use hercules_ir::ID;
+use juno_scheduler::ir::*;
+use juno_scheduler::pass;
+
+use juno_scheduler::{default_schedule, run_schedule_on_hercules};
+use rand::random;
+use rand::Rng;
+
+// Tests canonicalization
+
+#[ignore]
+#[test]
+fn loop_trip_count() {
+ let module = parse_file("../test_inputs/loop_analysis/loop_trip_count.hir");
+ let dyn_consts = [10];
+ let params = 2; // TODO: (@xrouth) fix macro to take no params as an option.
+ let result_1 = interp_module!(module, 0, dyn_consts, 2);
+
+ println!("result: {:?}", result_1);
+}
+
+// Test canonicalization
+#[test]
+#[ignore]
+fn alternate_bounds_use_after_loop_no_tid() {
+ let len = 1;
+ let dyn_consts = [len];
+
+ let module =
+ parse_file("../test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let schedule = default_schedule![
+ Forkify,
+ ];
+
+ let module = run_schedule_on_hercules(module, Some(schedule)).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+// Test canonicalization
+#[test]
+#[ignore]
+fn alternate_bounds_use_after_loop() {
+ let len = 4;
+ let dyn_consts = [len];
+
+ let a = vec![3, 4, 5, 6];
+ let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, a.clone());
+
+ println!("result: {:?}", result_1);
+
+ let schedule = Some(default_schedule![
+ Forkify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, a.clone());
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+// Test canonicalization
+#[test]
+#[ignore]
+fn alternate_bounds_use_after_loop2() {
+ let len = 4;
+ let dyn_consts = [len];
+
+ let a = vec![3, 4, 5, 6];
+ let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_use_after_loop2.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, a.clone());
+
+ println!("result: {:?}", result_1);
+
+ let schedule = Some(default_schedule![]);
+
+ let module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, a.clone());
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+// Test canonicalization
+#[test]
+fn do_while_separate_body() {
+ let len = 2;
+ let dyn_consts = [len];
+
+ let a = vec![3, 4, 5, 6];
+ let module = parse_file("../test_inputs/loop_analysis/do_while_separate_body2.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 2i32);
+
+ println!("result: {:?}", result_1);
+
+ let schedule = Some(default_schedule![
+ PhiElim,
+ Forkify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2i32);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn alternate_bounds_internal_control() {
+ let len = 4;
+ let dyn_consts = [len];
+
+ let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_internal_control.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let schedule = Some(default_schedule![
+ PhiElim,
+ Forkify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn alternate_bounds_internal_control2() {
+ let len = 2;
+ let dyn_consts = [len];
+
+ let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_internal_control2.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let schedule = Some(default_schedule![
+ PhiElim,
+ Forkify,
+ ]);
+
+ let module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn alternate_bounds_nested_do_loop() {
+ let len = 1;
+ let dyn_consts = [10, 5];
+
+ let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_nested_do_loop.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn alternate_bounds_nested_do_loop_array() {
+ let len = 1;
+ let dyn_consts = [10, 5];
+
+ let a = vec![4u64, 4, 4, 4, 4];
+ let module =
+ parse_file("../test_inputs/loop_analysis/alternate_bounds_nested_do_loop_array.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, a.clone());
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, a);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+fn alternate_bounds_nested_do_loop_guarded() {
+ let len = 1;
+ let dyn_consts = [3, 2];
+
+ let module =
+ parse_file("../test_inputs/loop_analysis/alternate_bounds_nested_do_loop_guarded.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+
+ let mut pm = PassManager::new(module.clone());
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ println!("{:?}", result_1);
+ println!("{:?}", result_2);
+
+ assert_eq!(result_1, result_2);
+}
+
+// Tests a do while loop that only iterates once,
+// canonicalization *should not* transform this to a while loop, as there is no
+// guard that replicates the loop condition.
+#[ignore]
+#[test]
+fn do_loop_not_continued() {
+ // let len = 1;
+ // let dyn_consts = [len];
+ // let params = vec![1, 2, 3, 4, 5];
+
+ // let module = parse_file("../test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir");
+ // let result_1 = interp_module!(module, 0,dyn_consts, params);
+
+ // println!("result: {:?}", result_1);
+}
+
+// Tests a do while loop that is guarded, so should be canonicalized
+// It also has
+#[test]
+fn do_loop_complex_immediate_guarded() {
+ let len = 1;
+ let dyn_consts = [len];
+
+ let module = parse_file("../test_inputs/loop_analysis/do_loop_immediate_guard.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, 3);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 3);
+ assert_eq!(result_1, result_2);
+}
+
+#[ignore]
+#[test]
+fn loop_canonical_sum() {
+ let len = 1;
+ let dyn_consts = [len];
+ let params = vec![1, 2, 3, 4, 5];
+
+ let module = parse_file("../test_inputs/loop_analysis/loop_array_sum.hir");
+ let result_1 = interp_module!(module, 0, dyn_consts, params);
+
+ println!("result: {:?}", result_1);
+}
+
+#[test]
+#[ignore]
+fn antideps_pipeline() {
+ let len = 1;
+ let dyn_consts = [2, 2, 2];
+
+ // FIXME: This path should not leave the crate
+ let module = parse_module_from_hbin("../../juno_samples/antideps/antideps.hbin");
+ let result_1 = interp_module!(module, 0, dyn_consts, 9i32);
+
+ println!("result: {:?}", result_1);
+
+ let module = run_schedule_on_hercules(module, None).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 9i32);
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+#[ignore]
+fn implicit_clone_pipeline() {
+ let len = 1;
+ let dyn_consts = [2, 2, 2];
+
+ // FIXME: This path should not leave the crate
+ let module = parse_module_from_hbin("../../juno_samples/implicit_clone/out.hbin");
+ let result_1 = interp_module!(module, 0, dyn_consts, 2u64, 2u64);
+
+ println!("result: {:?}", result_1);
+ let schedule = default_schedule![
+ Forkify,
+ ForkGuardElim,
+ Forkify,
+ ForkGuardElim,
+ Forkify,
+ ForkGuardElim,
+ DCE,
+ ForkSplit,
+ Unforkify,
+ GVN,
+ DCE,
+ DCE,
+ AutoOutline,
+ InterproceduralSROA,
+ SROA,
+ InferSchedules,
+ DCE,
+ GCM,
+ DCE,
+ FloatCollections,
+ GCM,
+ ];
+ let module = run_schedule_on_hercules(module, Some(schedule)).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, 2u64, 2u64);
+ assert_eq!(result_1, result_2);
+}
+
+#[test]
+#[ignore]
+fn look_at_local() {
+ const I: usize = 4;
+ const J: usize = 4;
+ const K: usize = 4;
+ let a: Vec<i32> = (0..I * J).map(|_| random::<i32>() % 100).collect();
+ let b: Vec<i32> = (0..J * K).map(|_| random::<i32>() % 100).collect();
+ let dyn_consts = [I, J, K];
+ let mut correct_c: Box<[i32]> = (0..I * K).map(|_| 0).collect();
+ for i in 0..I {
+ for k in 0..K {
+ for j in 0..J {
+ correct_c[i * K + k] += a[i * J + j] * b[j * K + k];
+ }
+ }
+ }
+
+ let module = parse_module_from_hbin(
+ "/home/xavierrouth/dev/hercules/hercules_test/hercules_tests/save_me.hbin",
+ );
+
+ let schedule = Some(default_schedule![
+ ]);
+
+ let result_1 = interp_module!(module, 0, dyn_consts, a.clone(), b.clone());
+
+ let module = run_schedule_on_hercules(module.clone(), schedule).unwrap();
+
+ let schedule = Some(default_schedule![
+ Unforkify, Verify,
+ ]);
+
+ let module = run_schedule_on_hercules(module.clone(), schedule).unwrap();
+
+ let result_2 = interp_module!(module, 0, dyn_consts, a.clone(), b.clone());
+
+ println!("golden: {:?}", correct_c);
+ println!("result: {:?}", result_2);
+}
+#[test]
+#[ignore]
+fn matmul_pipeline() {
+ let len = 1;
+
+ const I: usize = 4;
+ const J: usize = 4;
+ const K: usize = 4;
+ let a: Vec<i32> = (0i32..(I * J) as i32).map(|v| v + 1).collect();
+ let b: Vec<i32> = ((I * J) as i32..(J * K) as i32 + (I * J) as i32)
+ .map(|v| v + 1)
+ .collect();
+ let a: Vec<i32> = (0..I * J).map(|_| random::<i32>() % 100).collect();
+ let b: Vec<i32> = (0..J * K).map(|_| random::<i32>() % 100).collect();
+ let dyn_consts = [I, J, K];
+
+ // FIXME: This path should not leave the crate
+ let mut module = parse_module_from_hbin("../../juno_samples/matmul/out.hbin");
+ //
+ let mut correct_c: Box<[i32]> = (0..I * K).map(|_| 0).collect();
+ for i in 0..I {
+ for k in 0..K {
+ for j in 0..J {
+ correct_c[i * K + k] += a[i * J + j] * b[j * K + k];
+ }
+ }
+ }
+
+ let result_1 = interp_module!(module, 1, dyn_consts, a.clone(), b.clone());
+
+ println!("golden: {:?}", correct_c);
+ println!("result: {:?}", result_1);
+
+ let InterpreterVal::Array(_, d) = result_1.clone() else {
+ panic!()
+ };
+ let InterpreterVal::Integer32(value) = d[0] else {
+ panic!()
+ };
+ assert_eq!(correct_c[0], value);
+
+ let schedule = Some(default_schedule![Xdot, ForkSplit, Unforkify, Xdot,]);
+
+ module = run_schedule_on_hercules(module, schedule).unwrap();
+
+ let result_2 = interp_module!(module, 1, dyn_consts, a.clone(), b.clone());
+
+ println!("result: {:?}", result_2);
+ assert_eq!(result_1, result_2);
+}
diff --git a/hercules_test/hercules_tests/tests/opt_tests.rs b/hercules_test/hercules_tests/tests/opt_tests.rs
index 388dfedd..2f85b78b 100644
--- a/hercules_test/hercules_tests/tests/opt_tests.rs
+++ b/hercules_test/hercules_tests/tests/opt_tests.rs
@@ -3,198 +3,205 @@ use std::env;
use rand::Rng;
use hercules_interpreter::*;
-use hercules_opt::pass::Pass;
-
-#[test]
-fn matmul_int() {
- let module = parse_file("../test_inputs/matmul_int.hir");
- let dyn_consts = [2, 2, 2];
- let m1 = vec![3, 4, 5, 6];
- let m2 = vec![7, 8, 9, 10];
- let result_1 = interp_module!(module, dyn_consts, m1.clone(), m2.clone());
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
-
- let passes = vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ];
-
- for pass in passes {
- pm.add_pass(pass);
- }
- pm.run_passes();
-
- let module = pm.get_module();
- let result_2 = interp_module!(module, dyn_consts, m1, m2);
- assert_eq!(result_1, result_2)
-}
-
-#[test]
-fn ccp_example() {
- let module = parse_file("../test_inputs/ccp_example.hir");
- let dyn_consts = [];
- let x = 34;
- let result_1 = interp_module!(module, dyn_consts, x);
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
-
- let passes = vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ];
-
- for pass in passes {
- pm.add_pass(pass);
- }
- pm.run_passes();
-
- let module = pm.get_module();
- let result_2 = interp_module!(module, dyn_consts, x);
- assert_eq!(result_1, result_2)
-}
-
-#[test]
-fn gvn_example() {
- let module = parse_file("../test_inputs/gvn_example.hir");
-
- let dyn_consts = [];
- let x: i32 = rand::random();
- let x = x / 32;
- let y: i32 = rand::random();
- let y = y / 32; // prevent overflow,
- let result_1 = interp_module!(module, dyn_consts, x, y);
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
-
- let passes = vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ];
-
- for pass in passes {
- pm.add_pass(pass);
- }
- pm.run_passes();
-
- let module = pm.get_module();
- let result_2 = interp_module!(module, dyn_consts, x, y);
- assert_eq!(result_1, result_2)
-}
-
-#[test]
-fn sum_int() {
- let module = parse_file("../test_inputs/sum_int1.hir");
-
- let size = 100;
- let dyn_consts = [size];
- let mut vec = vec![0; size];
- let mut rng = rand::thread_rng();
-
- for x in vec.iter_mut() {
- *x = rng.gen::<i32>() / 100;
- }
-
- let result_1 = interp_module!(module, dyn_consts, vec.clone());
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
-
- let passes = vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ];
-
- for pass in passes {
- pm.add_pass(pass);
- }
- pm.run_passes();
-
- let module = pm.get_module();
- let result_2 = interp_module!(module, dyn_consts, vec);
- assert_eq!(result_1, result_2)
-}
-
-#[test]
-fn sum_int2() {
- let module = parse_file("../test_inputs/sum_int2.hir");
-
- let size = 100;
- let dyn_consts = [size];
- let mut vec = vec![0; size];
- let mut rng = rand::thread_rng();
-
- for x in vec.iter_mut() {
- *x = rng.gen::<i32>() / 100;
- }
-
- let result_1 = interp_module!(module, dyn_consts, vec.clone());
-
- let mut pm = hercules_opt::pass::PassManager::new(module.clone());
-
- let passes = vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ];
-
- for pass in passes {
- pm.add_pass(pass);
- }
- pm.run_passes();
-
- let module = pm.get_module();
- let result_2 = interp_module!(module, dyn_consts, vec);
- assert_eq!(result_1, result_2)
-}
-
-#[test]
-fn sum_int2_smaller() {
- interp_file_with_passes!("../test_inputs/sum_int2.hir",
- [100],
- vec![
- Pass::Verify,
- Pass::CCP,
- Pass::DCE,
- Pass::GVN,
- Pass::DCE,
- Pass::Forkify,
- Pass::DCE,
- Pass::Predication,
- Pass::DCE,
- ],
- vec![1; 100]);
-}
+use hercules_ir::ID;
+use juno_scheduler::*;
+
+// #[test]
+// fn matmul_int() {
+// let module = parse_file("../test_inputs/matmul_int.hir");
+// let dyn_consts = [2, 2, 2];
+// let m1 = vec![3, 4, 5, 6];
+// let m2 = vec![7, 8, 9, 10];
+// let result_1 = interp_module!(module, 0, dyn_consts, m1.clone(), m2.clone());
+
+// let mut pm = hercules_opt::pass::PassManager::new(module.clone());
+
+// let passes = vec![
+// // Pass::Verify,
+// // Pass::CCP,
+// // Pass::DCE,
+// // Pass::GVN,
+// // Pass::DCE,
+// // Pass::Forkify,
+// // Pass::DCE,
+// // Pass::Predication,
+// // Pass::DCE,
+// ];
+
+// for pass in passes {
+// pm.add_pass(pass);
+// }
+// pm.run_passes();
+
+// let module = pm.get_module();
+// let result_2 = interp_module!(module, 0, dyn_consts, m1, m2);
+// // println!("result: {:?}", result_1);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[test]
+// fn ccp_example() {
+// let module = parse_file("../test_inputs/ccp_example.hir");
+// let dyn_consts = [];
+// let x = 34;
+// let result_1 = interp_module!(module, 0, dyn_consts, x);
+
+// let mut pm = hercules_opt::pass::PassManager::new(module.clone());
+
+// let passes = vec![
+// Pass::Verify,
+// Pass::CCP,
+// Pass::DCE,
+// Pass::GVN,
+// Pass::DCE,
+// Pass::Forkify,
+// Pass::DCE,
+// Pass::Predication,
+// Pass::DCE,
+// ];
+
+// for pass in passes {
+// pm.add_pass(pass);
+// }
+// pm.run_passes();
+
+// let module = pm.get_module();
+// let result_2 = interp_module!(module, 0, dyn_consts, x);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[test]
+// fn gvn_example() {
+// let module = parse_file("../test_inputs/gvn_example.hir");
+
+// let dyn_consts = [];
+// let x: i32 = rand::random();
+// let x = x / 32;
+// let y: i32 = rand::random();
+// let y = y / 32; // prevent overflow,
+// let result_1 = interp_module!(module, 0, dyn_consts, x, y);
+
+// let mut pm = hercules_opt::pass::PassManager::new(module.clone());
+
+// let passes = vec![
+// Pass::Verify,
+// Pass::CCP,
+// Pass::DCE,
+// Pass::GVN,
+// Pass::DCE,
+// Pass::Forkify,
+// Pass::DCE,
+// Pass::Predication,
+// Pass::DCE,
+// ];
+
+// for pass in passes {
+// pm.add_pass(pass);
+// }
+// pm.run_passes();
+
+// let module = pm.get_module();
+// let result_2 = interp_module!(module, 0, dyn_consts, x, y);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[test]
+// fn sum_int() {
+// let module = parse_file("../test_inputs/sum_int1.hir");
+
+// let size = 2;
+// let dyn_consts = [size];
+// let mut vec = vec![0; size];
+// let mut rng = rand::thread_rng();
+
+// for x in vec.iter_mut() {
+// *x = rng.gen::<i32>() / 100;
+// }
+
+// println!("{:?}", vec);
+
+// let result_1 = interp_module!(module, 0, dyn_consts, vec.clone());
+
+// println!("{:?}", result_1);
+
+// let mut pm = hercules_opt::pass::PassManager::new(module.clone());
+
+// let passes = vec![
+// Pass::Verify,
+// Pass::CCP,
+// Pass::DCE,
+// Pass::GVN,
+// Pass::DCE,
+// Pass::Forkify,
+// Pass::DCE,
+// Pass::Predication,
+// Pass::DCE,
+// ];
+
+// for pass in passes {
+// pm.add_pass(pass);
+// }
+// pm.run_passes();
+
+// let module = pm.get_module();
+// let result_2 = interp_module!(module, 0, dyn_consts, vec);
+
+// assert_eq!(result_1, result_2)
+// }
+
+// #[test]
+// fn sum_int2() {
+// let module = parse_file("../test_inputs/sum_int2.hir");
+
+// let size = 10;
+// let dyn_consts = [size];
+// let mut vec = vec![0; size];
+// let mut rng = rand::thread_rng();
+
+// for x in vec.iter_mut() {
+// *x = rng.gen::<i32>() / 100;
+// }
+
+// let result_1 = interp_module!(module, 0, dyn_consts, vec.clone());
+
+// let mut pm = hercules_opt::pass::PassManager::new(module.clone());
+
+// let passes = vec![
+// Pass::Verify,
+// Pass::CCP,
+// Pass::DCE,
+// Pass::GVN,
+// Pass::DCE,
+// Pass::Forkify,
+// Pass::DCE,
+// Pass::Predication,
+// Pass::DCE,
+// ];
+
+// for pass in passes {
+// pm.add_pass(pass);
+// }
+// pm.run_passes();
+
+// let module = pm.get_module();
+// let result_2 = interp_module!(module, 0, dyn_consts, vec);
+// assert_eq!(result_1, result_2)
+// }
+
+// #[test]
+// fn sum_int2_smaller() {
+// interp_file_with_passes!("../test_inputs/sum_int2.hir",
+// [100],
+// vec![
+// Pass::Verify,
+// Pass::CCP,
+// Pass::DCE,
+// Pass::GVN,
+// Pass::DCE,
+// Pass::Forkify,
+// Pass::DCE,
+// Pass::Predication,
+// Pass::DCE,
+// ],
+// vec![1; 100]);
+// }
diff --git a/hercules_test/test_inputs/2d_fork.hir b/hercules_test/test_inputs/2d_fork.hir
new file mode 100644
index 00000000..e784c1db
--- /dev/null
+++ b/hercules_test/test_inputs/2d_fork.hir
@@ -0,0 +1,8 @@
+fn twodeefork<2>(x: i32) -> i32
+ zero = constant(i32, 0)
+ one = constant(i32, 1)
+ f = fork(start, #1, #0)
+ j = join(f)
+ add = add(r, one)
+ r = reduce(j, zero, add)
+ z = return(j, r)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/3d_fork.hir b/hercules_test/test_inputs/3d_fork.hir
new file mode 100644
index 00000000..746fd902
--- /dev/null
+++ b/hercules_test/test_inputs/3d_fork.hir
@@ -0,0 +1,8 @@
+fn twodeefork<3>(x: i32) -> i32
+ zero = constant(i32, 0)
+ one = constant(i32, 1)
+ f = fork(start, #2, #1, #0)
+ j = join(f)
+ add = add(r, one)
+ r = reduce(j, zero, add)
+ z = return(j, r)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/5d_fork.hir b/hercules_test/test_inputs/5d_fork.hir
new file mode 100644
index 00000000..94299601
--- /dev/null
+++ b/hercules_test/test_inputs/5d_fork.hir
@@ -0,0 +1,8 @@
+fn fivedeefork<5>(x: i32) -> i32
+ zero = constant(i32, 0)
+ one = constant(i32, 1)
+ f = fork(start, #4, #3, #2, #1, #0)
+ j = join(f)
+ add = add(r, one)
+ r = reduce(j, zero, add)
+ z = return(j, r)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/inner_control.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/inner_control.hir
new file mode 100644
index 00000000..052bbdb8
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/inner_control.hir
@@ -0,0 +1,15 @@
+fn fun<2>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ f2 = fork(f, #1)
+ j2 = join(f2)
+ j = join(j2)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, one)
+ reduce1 = reduce(j, zero, add1)
+ add2 = add(reduce2, two)
+ reduce2 = reduce(j, zero, add2)
+ out1 = add(reduce1, reduce2)
+ z = return(j, out1)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/inner_loop.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/inner_loop.hir
new file mode 100644
index 00000000..0cc13b2f
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/inner_loop.hir
@@ -0,0 +1,23 @@
+fn fun<2>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#1)
+ loop = region(f, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ j = join(if_false)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, idx)
+ reduce1 = reduce(j, zero, add1)
+ add2 = add(reduce2, idx_inc)
+ reduce2 = reduce(j, zero, add2)
+ out1 = add(reduce1, reduce2)
+ z = return(j, out1)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/intermediate_buffer_simple.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/intermediate_buffer_simple.hir
new file mode 100644
index 00000000..75e0f157
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/intermediate_buffer_simple.hir
@@ -0,0 +1,10 @@
+fn fun<1>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ j = join(f)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, two)
+ reduce1 = reduce(j, zero, add1)
+ z = return(j, reduce1)
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/simple1.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/simple1.hir
new file mode 100644
index 00000000..aaed60d9
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/simple1.hir
@@ -0,0 +1,13 @@
+fn fun<1>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ j = join(f)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, one)
+ reduce1 = reduce(j, zero, add1)
+ add2 = add(reduce2, two)
+ reduce2 = reduce(j, zero, add2)
+ out1 = add(reduce1, reduce2)
+ z = return(j, out1)
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/simple2.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/simple2.hir
new file mode 100644
index 00000000..6be6d2c7
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/simple2.hir
@@ -0,0 +1,19 @@
+fn fun<1>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ j = join(f)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, one)
+ reduce1 = reduce(j, zero, add1)
+ add2 = add(reduce2, two)
+ reduce2 = reduce(j, zero, add2)
+ add3 = add(reduce3, tid)
+ reduce3 = reduce(j, zero, add3)
+ add4 = mul(reduce4, tid)
+ reduce4 = reduce(j, zero, add4)
+ out1 = add(reduce1, reduce2)
+ out2 = add(reduce3, reduce4)
+ out3 = add(out1, out2)
+ z = return(j, out3)
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fission/tricky.hir b/hercules_test/test_inputs/fork_transforms/fork_fission/tricky.hir
new file mode 100644
index 00000000..6fb895c4
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/fork_fission/tricky.hir
@@ -0,0 +1,13 @@
+fn fun<1>(x: u64) -> u64
+ zero = constant(u64, 0)
+ one = constant(u64, 1)
+ two = constant(u64, 2)
+ f = fork(start, #0)
+ j = join(f)
+ tid = thread_id(f, 0)
+ add1 = add(reduce1, one)
+ reduce1 = reduce(j, zero, add1)
+ add2 = add(reduce2, reduce1)
+ reduce2 = reduce(j, zero, add2)
+ out1 = add(reduce1, reduce2)
+ z = return(j, out1)
diff --git a/hercules_test/test_inputs/fork_transforms/fork_fusion.hir b/hercules_test/test_inputs/fork_transforms/fork_fusion.hir
new file mode 100644
index 00000000..e69de29b
diff --git a/hercules_test/test_inputs/fork_transforms/fork_interchange.hir b/hercules_test/test_inputs/fork_transforms/fork_interchange.hir
new file mode 100644
index 00000000..e69de29b
diff --git a/hercules_test/test_inputs/fork_transforms/matmul_int.hir b/hercules_test/test_inputs/fork_transforms/matmul_int.hir
new file mode 100644
index 00000000..ab0f384a
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/matmul_int.hir
@@ -0,0 +1,18 @@
+fn matmul<3>(a: array(i32, #0, #1), b: array(i32, #1, #2)) -> array(i32, #0, #2)
+ c = constant(array(i32, #0, #2), [])
+ i_j_ctrl = fork(start, #0, #2)
+ i_idx = thread_id(i_j_ctrl, 0)
+ j_idx = thread_id(i_j_ctrl, 1)
+ k_ctrl = fork(i_j_ctrl, #1)
+ k_idx = thread_id(k_ctrl, 0)
+ k_join_ctrl = join(k_ctrl)
+ i_j_join_ctrl = join(k_join_ctrl)
+ r = return(i_j_join_ctrl, update_i_j_c)
+ zero = constant(i32, 0)
+ a_val = read(a, position(i_idx, k_idx))
+ b_val = read(b, position(k_idx, j_idx))
+ mul = mul(a_val, b_val)
+ add = add(mul, dot)
+ dot = reduce(k_join_ctrl, zero, add)
+ update_c = write(update_i_j_c, dot, position(i_idx, j_idx))
+ update_i_j_c = reduce(i_j_join_ctrl, c, update_c)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/fork_transforms/tiled_matmul_int.hir b/hercules_test/test_inputs/fork_transforms/tiled_matmul_int.hir
new file mode 100644
index 00000000..ab0f384a
--- /dev/null
+++ b/hercules_test/test_inputs/fork_transforms/tiled_matmul_int.hir
@@ -0,0 +1,18 @@
+fn matmul<3>(a: array(i32, #0, #1), b: array(i32, #1, #2)) -> array(i32, #0, #2)
+ c = constant(array(i32, #0, #2), [])
+ i_j_ctrl = fork(start, #0, #2)
+ i_idx = thread_id(i_j_ctrl, 0)
+ j_idx = thread_id(i_j_ctrl, 1)
+ k_ctrl = fork(i_j_ctrl, #1)
+ k_idx = thread_id(k_ctrl, 0)
+ k_join_ctrl = join(k_ctrl)
+ i_j_join_ctrl = join(k_join_ctrl)
+ r = return(i_j_join_ctrl, update_i_j_c)
+ zero = constant(i32, 0)
+ a_val = read(a, position(i_idx, k_idx))
+ b_val = read(b, position(k_idx, j_idx))
+ mul = mul(a_val, b_val)
+ add = add(mul, dot)
+ dot = reduce(k_join_ctrl, zero, add)
+ update_c = write(update_i_j_c, dot, position(i_idx, j_idx))
+ update_i_j_c = reduce(i_j_join_ctrl, c, update_c)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/alternate_bounds.hir b/hercules_test/test_inputs/forkify/alternate_bounds.hir
new file mode 100644
index 00000000..4a9ba015
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/alternate_bounds.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/broken_sum.hir b/hercules_test/test_inputs/forkify/broken_sum.hir
new file mode 100644
index 00000000..d15ef561
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/broken_sum.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/control_after_condition.hir b/hercules_test/test_inputs/forkify/control_after_condition.hir
new file mode 100644
index 00000000..db40225b
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/control_after_condition.hir
@@ -0,0 +1,25 @@
+fn alt_sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two_idx = constant(u64, 2)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, negate_bottom)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ rem = rem(idx, two_idx)
+ odd = eq(rem, one_idx)
+ negate_if = if(loop_continue, odd)
+ negate_if_false = projection(negate_if, 0)
+ negate_if_true = projection(negate_if, 1)
+ negate_bottom = region(negate_if_false, negate_if_true)
+ read = read(a, position(idx))
+ read_neg = neg(read)
+ read_phi = phi(negate_bottom, read, read_neg)
+ red_add = add(red, read_phi)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ loop_exit = projection(if, 0)
+ loop_continue = projection(if, 1)
+ r = return(loop_exit, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/control_before_condition.hir b/hercules_test/test_inputs/forkify/control_before_condition.hir
new file mode 100644
index 00000000..f24b565a
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/control_before_condition.hir
@@ -0,0 +1,25 @@
+fn alt_sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two_idx = constant(u64, 2)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ rem = rem(idx, two_idx)
+ odd = eq(rem, one_idx)
+ negate_if = if(loop, odd)
+ negate_if_false = projection(negate_if, 0)
+ negate_if_true = projection(negate_if, 1)
+ negate_bottom = region(negate_if_false, negate_if_true)
+ read = read(a, position(idx))
+ read_neg = neg(read)
+ read_phi = phi(negate_bottom, read, read_neg)
+ red_add = add(red, read_phi)
+ in_bounds = lt(idx, bound)
+ if = if(negate_bottom, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/expected_fails.hir/bad_3nest_return.hir b/hercules_test/test_inputs/forkify/expected_fails.hir/bad_3nest_return.hir
new file mode 100644
index 00000000..f5ec4370
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/expected_fails.hir/bad_3nest_return.hir
@@ -0,0 +1,35 @@
+fn loop<3>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_loop = region(outer_outer_if_true, inner_if_false)
+ inner_loop = region(outer_if_true, inner_if_true)
+ outer_var = phi(outer_loop, outer_outer_var, inner_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, one_var)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ outer_bound = dynamic_constant(#1)
+ outer_outer_bound = dynamic_constant(#2)
+ outer_outer_loop = region(start, outer_if_false)
+ outer_outer_var = phi(outer_outer_loop, zero_var, outer_var)
+ outer_outer_if = if(outer_outer_loop, outer_outer_in_bounds)
+ outer_outer_if_false = projection(outer_outer_if, 0)
+ outer_outer_if_true = projection(outer_outer_if, 1)
+ outer_outer_idx = phi(outer_outer_loop, zero_idx, outer_outer_idx_inc, outer_outer_idx)
+ outer_outer_idx_inc = add(outer_outer_idx, one_idx)
+ outer_outer_in_bounds = lt(outer_outer_idx, outer_outer_bound)
+ r = return(outer_outer_if_false, inner_var)
+
diff --git a/hercules_test/test_inputs/forkify/expected_fails.hir/bad_loop_tid_sum.hir b/hercules_test/test_inputs/forkify/expected_fails.hir/bad_loop_tid_sum.hir
new file mode 100644
index 00000000..8dda179b
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/expected_fails.hir/bad_loop_tid_sum.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, idx)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var_inc)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/inner_fork.hir b/hercules_test/test_inputs/forkify/inner_fork.hir
new file mode 100644
index 00000000..e2c96a68
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/inner_fork.hir
@@ -0,0 +1,22 @@
+fn loop<2>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, inner_join)
+ outer_if_true = projection(outer_if, 1)
+ inner_fork = fork(outer_if_true, #0)
+ inner_join = join(inner_fork)
+ outer_var = phi(outer_loop, zero_var, inner_var)
+ inner_var = reduce(inner_fork, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, inner_idx)
+ inner_idx = thread_id(inner_fork, 0)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ r = return(outer_if_false, outer_var)
+
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/inner_fork_complex.hir b/hercules_test/test_inputs/forkify/inner_fork_complex.hir
new file mode 100644
index 00000000..91eb00fa
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/inner_fork_complex.hir
@@ -0,0 +1,32 @@
+fn loop<2>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ ten = constant(u64, 10)
+ two = constant(u64, 2)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, inner_condition_true_projection, inner_condition_false_projection )
+ outer_if_true = projection(outer_if, 1)
+ other_phi_weird = phi(outer_loop, zero_var, inner_var, other_phi_weird)
+ inner_fork = fork(outer_if_true, #0)
+ inner_join = join(inner_fork)
+ inner_condition_eq = eq(outer_idx, two)
+ inner_condition_if = if(inner_join, inner_condition_eq)
+ inner_condition_true_projection = projection(inner_condition_if, 1)
+ inner_condition_false_projection = projection(inner_condition_if, 0)
+ outer_var = phi(outer_loop, zero_var, inner_var, inner_var)
+ inner_var = reduce(inner_join, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, inner_var_inc_3)
+ inner_var_inc_2 = mul(ten, outer_idx)
+ inner_var_inc_3 = add(inner_var_inc_2, inner_idx)
+ inner_idx = thread_id(inner_fork, 0)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx_inc)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ ret_val = add(outer_var, other_phi_weird)
+ r = return(outer_if_false, ret_val)
+
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/loop_array_sum.hir b/hercules_test/test_inputs/forkify/loop_array_sum.hir
new file mode 100644
index 00000000..f9972b59
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/loop_array_sum.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/loop_simple_iv.hir b/hercules_test/test_inputs/forkify/loop_simple_iv.hir
new file mode 100644
index 00000000..c25b9a2c
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/loop_simple_iv.hir
@@ -0,0 +1,12 @@
+fn loop<1>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, idx)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/loop_sum.hir b/hercules_test/test_inputs/forkify/loop_sum.hir
new file mode 100644
index 00000000..fd9c4deb
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/loop_sum.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/loop_tid_sum.hir b/hercules_test/test_inputs/forkify/loop_tid_sum.hir
new file mode 100644
index 00000000..2d3ca34d
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/loop_tid_sum.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, idx)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/merged_phi_cycle.hir b/hercules_test/test_inputs/forkify/merged_phi_cycle.hir
new file mode 100644
index 00000000..cee473a0
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/merged_phi_cycle.hir
@@ -0,0 +1,18 @@
+fn sum<1>(a: i32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ first_red = phi(loop, zero_idx, first_red_add)
+ second_red = phi(loop, zero_idx, first_red_add_2)
+ first_red_add = add(first_red, idx)
+ second_red_add_1 = add(first_red, idx)
+ second_red_add_2 = add(first_red_add, two)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, first_red_add_2)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/nested_loop2.hir b/hercules_test/test_inputs/forkify/nested_loop2.hir
new file mode 100644
index 00000000..0f29ec74
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/nested_loop2.hir
@@ -0,0 +1,25 @@
+fn loop<2>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, inner_if_false)
+ inner_loop = region(outer_if_true, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, one_var)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, outer_var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/nested_tid_sum.hir b/hercules_test/test_inputs/forkify/nested_tid_sum.hir
new file mode 100644
index 00000000..5539202d
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/nested_tid_sum.hir
@@ -0,0 +1,25 @@
+fn loop<2>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, inner_if_false)
+ inner_loop = region(outer_if_true, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, inner_idx)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, outer_var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/nested_tid_sum_2.hir b/hercules_test/test_inputs/forkify/nested_tid_sum_2.hir
new file mode 100644
index 00000000..9221fd47
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/nested_tid_sum_2.hir
@@ -0,0 +1,26 @@
+fn loop<2>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, inner_if_false)
+ inner_loop = region(outer_if_true, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ iv_mul = mul(inner_idx, outer_idx)
+ inner_var_inc = add(inner_var, iv_mul)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, outer_var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/phi_loop4.hir b/hercules_test/test_inputs/forkify/phi_loop4.hir
new file mode 100644
index 00000000..e69ecc3d
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/phi_loop4.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var_inc)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/split_phi_cycle.hir b/hercules_test/test_inputs/forkify/split_phi_cycle.hir
new file mode 100644
index 00000000..96de73c8
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/split_phi_cycle.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: i32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ first_red = phi(loop, zero_idx, first_red_add_2)
+ first_red_add = add(first_red, idx)
+ first_red_add_2 = add(first_red_add, two)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, first_red_add_2)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/forkify/super_nested_loop.hir b/hercules_test/test_inputs/forkify/super_nested_loop.hir
new file mode 100644
index 00000000..6853efbf
--- /dev/null
+++ b/hercules_test/test_inputs/forkify/super_nested_loop.hir
@@ -0,0 +1,35 @@
+fn loop<3>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_loop = region(outer_outer_if_true, inner_if_false)
+ inner_loop = region(outer_if_true, inner_if_true)
+ outer_var = phi(outer_loop, outer_outer_var, inner_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, one_var)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(outer_loop, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ outer_bound = dynamic_constant(#1)
+ outer_outer_bound = dynamic_constant(#2)
+ outer_outer_loop = region(start, outer_if_false)
+ outer_outer_var = phi(outer_outer_loop, zero_var, outer_var)
+ outer_outer_if = if(outer_outer_loop, outer_outer_in_bounds)
+ outer_outer_if_false = projection(outer_outer_if, 0)
+ outer_outer_if_true = projection(outer_outer_if, 1)
+ outer_outer_idx = phi(outer_outer_loop, zero_idx, outer_outer_idx_inc, outer_outer_idx)
+ outer_outer_idx_inc = add(outer_outer_idx, one_idx)
+ outer_outer_in_bounds = lt(outer_outer_idx, outer_outer_bound)
+ r = return(outer_outer_if_false, outer_outer_var)
+
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds.hir
new file mode 100644
index 00000000..4df92a18
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds.hir
@@ -0,0 +1,14 @@
+fn sum<1>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, one_idx)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control.hir
new file mode 100644
index 00000000..8b4431bf
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control.hir
@@ -0,0 +1,23 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ ten = constant(u64, 10)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ inner_ctrl = region(loop)
+ inner_phi = phi(inner_ctrl, idx)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, two)
+ red2 = phi(loop, zero_idx, red_add2)
+ red_add2 = add(red, inner_phi)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_ctrl, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red_add, ten)
+ red_add_2_plus_blah = add(red2, plus_ten)
+ final_add = add(inner_phi, red_add_2_plus_blah)
+ r = return(if_false, final_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control2.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control2.hir
new file mode 100644
index 00000000..f4adf643
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_internal_control2.hir
@@ -0,0 +1,21 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ ten = constant(u64, 10)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ inner_ctrl = region(loop)
+ inner_phi = phi(inner_ctrl, idx)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, two)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_ctrl, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red_add, ten)
+ red_add_2_plus_blah = add(inner_phi, plus_ten)
+ final_add = add(inner_phi, red_add_2_plus_blah)
+ r = return(if_false, final_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop.hir
new file mode 100644
index 00000000..52f70172
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop.hir
@@ -0,0 +1,28 @@
+fn loop<2>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ ten = constant(u64, 10)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, outer_if_true)
+ inner_loop = region(outer_loop, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var_inc)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, blah2)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ blah = mul(outer_idx, ten)
+ blah2 = add(blah, inner_idx)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx_inc, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx_inc, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(inner_if_false, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, inner_var_inc)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop2.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop2.hir
new file mode 100644
index 00000000..f295b391
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop2.hir
@@ -0,0 +1,25 @@
+fn loop<2>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, outer_if_true)
+ inner_loop = region(outer_loop, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var_inc)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, one_var)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx_inc, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx_inc, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(inner_if_false, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, inner_var_inc)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_array.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_array.hir
new file mode 100644
index 00000000..e5401779
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_array.hir
@@ -0,0 +1,28 @@
+fn loop<2>(a: array(u64, #1)) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ ten = constant(u64, 10)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(start, outer_if_true)
+ inner_loop = region(outer_loop, inner_if_true)
+ outer_var = phi(outer_loop, zero_var, inner_var_inc)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ inner_var_inc = add(inner_var, blah2)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ blah = read(a, position(outer_idx))
+ blah2 = add(blah, inner_idx)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx_inc, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx_inc, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(inner_if_false, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ r = return(outer_if_false, inner_var_inc)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_guarded.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_guarded.hir
new file mode 100644
index 00000000..b979ad42
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_nested_do_loop_guarded.hir
@@ -0,0 +1,40 @@
+fn loop<2>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ ten = constant(u64, 10)
+ outer_guard_if = if(start, outer_guard_lt)
+ outer_guard_if_false = projection(outer_guard_if, 0)
+ outer_guard_if_true = projection(outer_guard_if, 1)
+ outer_guard_lt = lt(zero_idx, outer_bound)
+ outer_join = region(outer_guard_if_false, outer_if_false)
+ outer_join_var = phi(outer_join, zero_idx, join_var)
+ inner_bound = dynamic_constant(#0)
+ outer_bound = dynamic_constant(#1)
+ outer_loop = region(outer_guard_if_true, outer_if_true)
+ inner_loop = region(guard_if_true, inner_if_true)
+ guard_lt = lt(zero_idx, inner_bound)
+ guard_if = if(outer_loop, guard_lt)
+ guard_if_true = projection(guard_if, 1)
+ guard_if_false = projection(guard_if, 0)
+ guard_join = region(guard_if_false, inner_if_false)
+ inner_idx = phi(inner_loop, zero_idx, inner_idx_inc)
+ inner_idx_inc = add(inner_idx, one_idx)
+ inner_in_bounds = lt(inner_idx_inc, inner_bound)
+ outer_idx = phi(outer_loop, zero_idx, outer_idx_inc, outer_idx)
+ outer_idx_inc = add(outer_idx, one_idx)
+ outer_in_bounds = lt(outer_idx_inc, outer_bound)
+ inner_if = if(inner_loop, inner_in_bounds)
+ inner_if_false = projection(inner_if, 0)
+ inner_if_true = projection(inner_if, 1)
+ outer_if = if(guard_join, outer_in_bounds)
+ outer_if_false = projection(outer_if, 0)
+ outer_if_true = projection(outer_if, 1)
+ outer_var = phi(outer_loop, zero_var, join_var)
+ inner_var = phi(inner_loop, outer_var, inner_var_inc)
+ blah = mul(outer_idx, ten)
+ blah2 = add(blah, inner_idx)
+ inner_var_inc = add(inner_var, blah2)
+ join_var = phi(guard_join, outer_var, inner_var_inc)
+ r = return(outer_join, outer_join_var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir
new file mode 100644
index 00000000..2fe4ca57
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop.hir
@@ -0,0 +1,21 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ ten = constant(i32, 10)
+ three = constant(i32, 3)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red_add, ten)
+ mult = mul(read, three)
+ final = add(plus_ten, mult)
+ r = return(if_false, final)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop2.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop2.hir
new file mode 100644
index 00000000..760ae5ad
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop2.hir
@@ -0,0 +1,21 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ ten = constant(i32, 10)
+ three = constant(i32, 3)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx_inc))
+ red_add = add(red, read)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red, ten)
+ mult = mul(read, three)
+ final = add(plus_ten, mult)
+ r = return(if_false, final)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid.hir
new file mode 100644
index 00000000..4b937509
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid.hir
@@ -0,0 +1,17 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ ten = constant(u64, 10)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, two)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red_add, ten)
+ r = return(if_false, plus_ten)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid2.hir b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid2.hir
new file mode 100644
index 00000000..fd06eb7d
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/alternate_bounds_use_after_loop_no_tid2.hir
@@ -0,0 +1,19 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 2)
+ ten = constant(u64, 10)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, two)
+ blah = phi(loop, zero_idx, red_add)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ plus_ten = add(red_add, ten)
+ plus_blah = add(blah, red_add)
+ r = return(if_false, plus_blah)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/broken_sum.hir b/hercules_test/test_inputs/loop_analysis/broken_sum.hir
new file mode 100644
index 00000000..d15ef561
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/broken_sum.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/do_loop_far_guard.hir b/hercules_test/test_inputs/loop_analysis/do_loop_far_guard.hir
new file mode 100644
index 00000000..4df92a18
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/do_loop_far_guard.hir
@@ -0,0 +1,14 @@
+fn sum<1>(a: u32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, one_idx)
+ in_bounds = lt(idx_inc, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/do_loop_immediate_guard.hir b/hercules_test/test_inputs/loop_analysis/do_loop_immediate_guard.hir
new file mode 100644
index 00000000..a4732cde
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/do_loop_immediate_guard.hir
@@ -0,0 +1,21 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ guard_lt = lt(zero_idx, bound)
+ guard = if(start, guard_lt)
+ guard_true = projection(guard, 1)
+ guard_false = projection(guard, 0)
+ loop = region(guard_true, if_true)
+ inner_side_effect = region(loop)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, one_idx)
+ join_phi = phi(final, zero_idx, red_add)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_side_effect, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ final = region(guard_false, if_false)
+ r = return(final, join_phi)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/do_loop_no_guard.hir b/hercules_test/test_inputs/loop_analysis/do_loop_no_guard.hir
new file mode 100644
index 00000000..9e22e14b
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/do_loop_no_guard.hir
@@ -0,0 +1,15 @@
+fn sum<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ inner_side_effect = region(loop)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_idx, red_add)
+ red_add = add(red, one_idx)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_side_effect, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red_add)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/do_while_separate_body.hir b/hercules_test/test_inputs/loop_analysis/do_while_separate_body.hir
new file mode 100644
index 00000000..42269040
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/do_while_separate_body.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: i32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ inner_region = region(loop)
+ inner_red = phi(inner_region, red_add)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ outer_red = phi(loop, zero_idx, inner_red)
+ red_add = add(outer_red, idx)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_region, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, inner_red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/do_while_separate_body2.hir b/hercules_test/test_inputs/loop_analysis/do_while_separate_body2.hir
new file mode 100644
index 00000000..a751952d
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/do_while_separate_body2.hir
@@ -0,0 +1,18 @@
+fn sum<1>(a: i32) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ two = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ inner_region = region(loop)
+ inner_red = phi(inner_region, red_mul)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ outer_red = phi(loop, zero_idx, inner_red)
+ red_add = add(outer_red, idx)
+ red_mul = mul(red_add, idx)
+ in_bounds = lt(idx_inc, bound)
+ if = if(inner_region, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, inner_red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/loop_array_sum.hir b/hercules_test/test_inputs/loop_analysis/loop_array_sum.hir
new file mode 100644
index 00000000..f9972b59
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/loop_array_sum.hir
@@ -0,0 +1,16 @@
+fn sum<1>(a: array(i32, #0)) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_inc = constant(i32, 0)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ red = phi(loop, zero_inc, red_add)
+ read = read(a, position(idx))
+ red_add = add(red, read)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, red)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/loop_body_count.hir b/hercules_test/test_inputs/loop_analysis/loop_body_count.hir
new file mode 100644
index 00000000..c6f3cbf6
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/loop_body_count.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u64) -> u64
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/loop_sum.hir b/hercules_test/test_inputs/loop_analysis/loop_sum.hir
new file mode 100644
index 00000000..fd9c4deb
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/loop_sum.hir
@@ -0,0 +1,16 @@
+fn loop<1>(a: u32) -> i32
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(i32, 0)
+ one_var = constant(i32, 1)
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ r = return(if_false, var)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/loop_trip_count.hir b/hercules_test/test_inputs/loop_analysis/loop_trip_count.hir
new file mode 100644
index 00000000..b756f090
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/loop_trip_count.hir
@@ -0,0 +1,19 @@
+fn loop<1>(b: prod(u64, u64)) -> prod(u64, u64)
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ c = constant(prod(u64, u64), (0, 0))
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ tuple1 = write(c, var, field(0))
+ tuple2 = write(tuple1, idx, field(1))
+ r = return(if_false, tuple2)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/loop_analysis/loop_trip_count_tuple.hir b/hercules_test/test_inputs/loop_analysis/loop_trip_count_tuple.hir
new file mode 100644
index 00000000..b756f090
--- /dev/null
+++ b/hercules_test/test_inputs/loop_analysis/loop_trip_count_tuple.hir
@@ -0,0 +1,19 @@
+fn loop<1>(b: prod(u64, u64)) -> prod(u64, u64)
+ zero_idx = constant(u64, 0)
+ one_idx = constant(u64, 1)
+ zero_var = constant(u64, 0)
+ one_var = constant(u64, 1)
+ c = constant(prod(u64, u64), (0, 0))
+ bound = dynamic_constant(#0)
+ loop = region(start, if_true)
+ var = phi(loop, zero_var, var_inc)
+ var_inc = add(var, one_var)
+ idx = phi(loop, zero_idx, idx_inc)
+ idx_inc = add(idx, one_idx)
+ in_bounds = lt(idx, bound)
+ if = if(loop, in_bounds)
+ if_false = projection(if, 0)
+ if_true = projection(if, 1)
+ tuple1 = write(c, var, field(0))
+ tuple2 = write(tuple1, idx, field(1))
+ r = return(if_false, tuple2)
\ No newline at end of file
diff --git a/hercules_test/test_inputs/matmul_int.hir b/hercules_test/test_inputs/matmul_int.hir
index 1e496bab..ab0f384a 100644
--- a/hercules_test/test_inputs/matmul_int.hir
+++ b/hercules_test/test_inputs/matmul_int.hir
@@ -1,21 +1,18 @@
fn matmul<3>(a: array(i32, #0, #1), b: array(i32, #1, #2)) -> array(i32, #0, #2)
c = constant(array(i32, #0, #2), [])
- i_ctrl = fork(start, #0)
- i_idx = thread_id(i_ctrl)
- j_ctrl = fork(i_ctrl, #2)
- j_idx = thread_id(j_ctrl)
- k_ctrl = fork(j_ctrl, #1)
- k_idx = thread_id(k_ctrl)
+ i_j_ctrl = fork(start, #0, #2)
+ i_idx = thread_id(i_j_ctrl, 0)
+ j_idx = thread_id(i_j_ctrl, 1)
+ k_ctrl = fork(i_j_ctrl, #1)
+ k_idx = thread_id(k_ctrl, 0)
k_join_ctrl = join(k_ctrl)
- j_join_ctrl = join(k_join_ctrl)
- i_join_ctrl = join(j_join_ctrl)
- r = return(i_join_ctrl, update_i_c)
+ i_j_join_ctrl = join(k_join_ctrl)
+ r = return(i_j_join_ctrl, update_i_j_c)
zero = constant(i32, 0)
a_val = read(a, position(i_idx, k_idx))
b_val = read(b, position(k_idx, j_idx))
mul = mul(a_val, b_val)
add = add(mul, dot)
dot = reduce(k_join_ctrl, zero, add)
- updated_c = write(update_j_c, dot, position(i_idx, j_idx))
- update_j_c = reduce(j_join_ctrl, update_i_c, updated_c)
- update_i_c = reduce(i_join_ctrl, c, update_j_c)
+ update_c = write(update_i_j_c, dot, position(i_idx, j_idx))
+ update_i_j_c = reduce(i_j_join_ctrl, c, update_c)
\ No newline at end of file
diff --git a/juno_samples/cava/build.rs b/juno_samples/cava/build.rs
index 929d3eba..7f60f801 100644
--- a/juno_samples/cava/build.rs
+++ b/juno_samples/cava/build.rs
@@ -1,4 +1,3 @@
-extern crate juno_build;
use juno_build::JunoCompiler;
fn main() {
diff --git a/juno_samples/cava/src/main.rs b/juno_samples/cava/src/main.rs
index 73a75a94..8ad6824f 100644
--- a/juno_samples/cava/src/main.rs
+++ b/juno_samples/cava/src/main.rs
@@ -59,7 +59,10 @@ fn run_cava(
tonemap,
)
.await
- }).as_slice::<u8>().to_vec().into_boxed_slice()
+ })
+ .as_slice::<u8>()
+ .to_vec()
+ .into_boxed_slice()
}
enum Error {
diff --git a/juno_samples/matmul/build.rs b/juno_samples/matmul/build.rs
index 926fbc33..511bf483 100644
--- a/juno_samples/matmul/build.rs
+++ b/juno_samples/matmul/build.rs
@@ -4,6 +4,8 @@ fn main() {
JunoCompiler::new()
.file_in_src("matmul.jn")
.unwrap()
+ //.schedule_in_src("sched.sch")
+ //.unwrap()
.build()
.unwrap();
}
diff --git a/juno_samples/matmul/src/main.rs b/juno_samples/matmul/src/main.rs
index fa5d1f04..624ee565 100644
--- a/juno_samples/matmul/src/main.rs
+++ b/juno_samples/matmul/src/main.rs
@@ -24,10 +24,14 @@ fn main() {
let a = HerculesCPURef::from_slice(&a);
let b = HerculesCPURef::from_slice(&b);
let mut r = runner!(matmul);
- let c = r.run(I as u64, J as u64, K as u64, a.clone(), b.clone()).await;
+ let c = r
+ .run(I as u64, J as u64, K as u64, a.clone(), b.clone())
+ .await;
assert_eq!(c.as_slice::<i32>(), &*correct_c);
let mut r = runner!(tiled_64_matmul);
- let tiled_c = r.run(I as u64, J as u64, K as u64, a.clone(), b.clone()).await;
+ let tiled_c = r
+ .run(I as u64, J as u64, K as u64, a.clone(), b.clone())
+ .await;
assert_eq!(tiled_c.as_slice::<i32>(), &*correct_c);
});
}
@@ -36,4 +40,3 @@ fn main() {
fn matmul_test() {
main();
}
-
diff --git a/juno_samples/matmul/src/sched.sch b/juno_samples/matmul/src/sched.sch
new file mode 100644
index 00000000..3999f923
--- /dev/null
+++ b/juno_samples/matmul/src/sched.sch
@@ -0,0 +1,76 @@
+macro juno-setup!(X) {
+ gvn(X);
+ dce(X);
+ phi-elim(X);
+}
+
+macro default!(X) {
+ dce(X);
+ crc(X);
+ dce(X);
+ slf(X);
+ dce(X);
+ inline(X);
+ ip-sroa(X);
+ sroa(X);
+ phi-elim(X);
+ dce(X);
+ ccp(X);
+ dce(X);
+ gvn(X);
+ dce(X);
+ write-predication(X);
+ phi-elim(X);
+ dce(X);
+ crc(X);
+ dce(X);
+ slf(X);
+ dce(X);
+ predication(X);
+ dce(X);
+ ccp(X);
+ dce(X);
+ gvn(X);
+ dce(X);
+ lift-dc-math(X);
+ dce(X);
+ gvn(X);
+ dce(X);
+}
+
+macro codegen-prep!(X) {
+ verify(*);
+ ip-sroa(*);
+ sroa(*);
+ infer-schedules(X);
+ dce(X);
+ gcm(X);
+ dce(X);
+ phi-elim(X);
+ float-collections(X);
+ gcm(X);
+}
+
+juno-setup!(*);
+default!(*);
+// your stuff here.
+
+fixpoint stop after 13 {
+ forkify(*);
+ fork-guard-elim(*);
+ fork-coalesce(*);
+ phi-elim(*);
+ dce(*);
+}
+
+xdot[true](*);
+// serialize(*);
+
+fork-split(*);
+unforkify(*);
+
+gvn(*);
+dce(*);
+
+auto-outline(*);
+codegen-prep!(*);
diff --git a/juno_scheduler/Cargo.toml b/juno_scheduler/Cargo.toml
index 1c837d4a..04ab156c 100644
--- a/juno_scheduler/Cargo.toml
+++ b/juno_scheduler/Cargo.toml
@@ -18,3 +18,5 @@ hercules_cg = { path = "../hercules_cg" }
hercules_ir = { path = "../hercules_ir" }
hercules_opt = { path = "../hercules_opt" }
juno_utils = { path = "../juno_utils" }
+postcard = { version = "*", features = ["alloc"] }
+serde = { version = "*", features = ["derive"] }
\ No newline at end of file
diff --git a/juno_scheduler/src/compile.rs b/juno_scheduler/src/compile.rs
index fc3c1279..11a8ec53 100644
--- a/juno_scheduler/src/compile.rs
+++ b/juno_scheduler/src/compile.rs
@@ -4,8 +4,7 @@ use crate::parser;
use juno_utils::env::Env;
use juno_utils::stringtab::StringTable;
-extern crate hercules_ir;
-use self::hercules_ir::ir::{Device, Schedule};
+use hercules_ir::ir::{Device, Schedule};
use lrlex::DefaultLexerTypes;
use lrpar::NonStreamingLexer;
@@ -117,8 +116,11 @@ impl FromStr for Appliable {
"slf" | "store-load-forward" => Ok(Appliable::Pass(ir::Pass::SLF)),
"sroa" => Ok(Appliable::Pass(ir::Pass::SROA)),
"unforkify" => Ok(Appliable::Pass(ir::Pass::Unforkify)),
+ "fork-coalesce" => Ok(Appliable::Pass(ir::Pass::ForkCoalesce)),
"verify" => Ok(Appliable::Pass(ir::Pass::Verify)),
"xdot" => Ok(Appliable::Pass(ir::Pass::Xdot)),
+ "serialize" => Ok(Appliable::Pass(ir::Pass::Serialize)),
+ "write-predication" => Ok(Appliable::Pass(ir::Pass::WritePredication)),
"cpu" | "llvm" => Ok(Appliable::Device(Device::LLVM)),
"gpu" | "cuda" | "nvidia" => Ok(Appliable::Device(Device::CUDA)),
diff --git a/juno_scheduler/src/default.rs b/juno_scheduler/src/default.rs
index 46b51b43..fd45a371 100644
--- a/juno_scheduler/src/default.rs
+++ b/juno_scheduler/src/default.rs
@@ -1,5 +1,6 @@
use crate::ir::*;
+#[macro_export]
macro_rules! pass {
($p:ident) => {
ScheduleStmt::Let {
@@ -13,6 +14,7 @@ macro_rules! pass {
};
}
+#[macro_export]
macro_rules! default_schedule {
() => {
ScheduleStmt::Block {
@@ -64,8 +66,9 @@ pub fn default_schedule() -> ScheduleStmt {
DCE,
GVN,
DCE,
- /*Forkify,*/
- /*ForkGuardElim,*/
+ // Forkify,
+ // ForkGuardElim,
+ // ForkCoalesce,
DCE,
ForkSplit,
Unforkify,
diff --git a/juno_scheduler/src/ir.rs b/juno_scheduler/src/ir.rs
index 381c3475..d6a41baf 100644
--- a/juno_scheduler/src/ir.rs
+++ b/juno_scheduler/src/ir.rs
@@ -1,6 +1,4 @@
-extern crate hercules_ir;
-
-use self::hercules_ir::ir::{Device, Schedule};
+use hercules_ir::ir::{Device, Schedule};
#[derive(Debug, Copy, Clone)]
pub enum Pass {
@@ -12,6 +10,7 @@ pub enum Pass {
FloatCollections,
ForkGuardElim,
ForkSplit,
+ ForkCoalesce,
Forkify,
GCM,
GVN,
@@ -28,6 +27,7 @@ pub enum Pass {
WritePredication,
Verify,
Xdot,
+ Serialize,
}
impl Pass {
diff --git a/juno_scheduler/src/lib.rs b/juno_scheduler/src/lib.rs
index 1caafe4f..571d1fbf 100644
--- a/juno_scheduler/src/lib.rs
+++ b/juno_scheduler/src/lib.rs
@@ -14,7 +14,7 @@ use crate::parser::lexer;
mod compile;
mod default;
-mod ir;
+pub mod ir;
pub mod labels;
mod pm;
@@ -22,7 +22,7 @@ use crate::compile::*;
use crate::default::*;
use crate::ir::*;
use crate::labels::*;
-use crate::pm::*;
+pub use crate::pm::*;
// Given a schedule's filename parse and process the schedule
fn build_schedule(sched_filename: String) -> Result<ScheduleStmt, String> {
@@ -107,6 +107,45 @@ pub fn schedule_juno(
.map_err(|e| format!("Scheduling Error: {}", e))
}
+pub fn run_schedule_on_hercules(
+ module: Module,
+ sched: Option<ScheduleStmt>,
+) -> Result<Module, String> {
+ let sched = if let Some(sched) = sched {
+ sched
+ } else {
+ default_schedule()
+ };
+
+ // Prepare the scheduler's string table and environment
+ // For this, we put all of the Hercules function names into the environment
+ // and string table
+ let mut strings = StringTable::new();
+ let mut env = Env::new();
+
+ env.open_scope();
+
+ for (idx, func) in module.functions.iter().enumerate() {
+ let func_name = strings.lookup_string(func.name.clone());
+ env.insert(
+ func_name,
+ Value::HerculesFunction {
+ func: FunctionID::new(idx),
+ },
+ );
+ }
+
+ env.open_scope();
+ schedule_module(
+ module,
+ sched,
+ strings,
+ env,
+ JunoFunctions { func_ids: vec![] },
+ )
+ .map_err(|e| format!("Scheduling Error: {}", e))
+}
+
pub fn schedule_hercules(
module: Module,
sched_filename: Option<String>,
diff --git a/juno_scheduler/src/pm.rs b/juno_scheduler/src/pm.rs
index c388833c..9888f3d2 100644
--- a/juno_scheduler/src/pm.rs
+++ b/juno_scheduler/src/pm.rs
@@ -4,6 +4,8 @@ use hercules_cg::*;
use hercules_ir::*;
use hercules_opt::*;
+use serde::Deserialize;
+use serde::Serialize;
use tempfile::TempDir;
use juno_utils::env::Env;
@@ -132,7 +134,7 @@ impl Value {
}
}
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize, Deserialize)]
pub enum SchedulerError {
UndefinedVariable(String),
UndefinedField(String),
@@ -159,8 +161,8 @@ impl fmt::Display for SchedulerError {
}
}
-#[derive(Debug)]
-struct PassManager {
+#[derive(Debug, Clone)]
+pub struct PassManager {
functions: Vec<Function>,
types: RefCell<Vec<Type>>,
constants: RefCell<Vec<Constant>>,
@@ -189,7 +191,7 @@ struct PassManager {
}
impl PassManager {
- fn new(module: Module) -> Self {
+ pub fn new(module: Module) -> Self {
let Module {
functions,
types,
@@ -500,6 +502,31 @@ impl PassManager {
res
}
+ pub fn get_module(&self) -> Module {
+ let PassManager {
+ functions,
+ types,
+ constants,
+ dynamic_constants,
+ labels,
+ typing: _,
+ control_subgraphs: _,
+ bbs: _,
+ collection_objects: _,
+ callgraph: _,
+ ..
+ } = self;
+
+ let module = Module {
+ functions: functions.to_vec(),
+ types: types.clone().into_inner(),
+ constants: constants.clone().into_inner(),
+ dynamic_constants: dynamic_constants.clone().into_inner(),
+ labels: labels.clone().into_inner(),
+ };
+ module
+ }
+
fn codegen(mut self, output_dir: String, module_name: String) -> Result<(), SchedulerError> {
self.make_typing();
self.make_control_subgraphs();
@@ -629,6 +656,18 @@ pub fn schedule_codegen(
pm.codegen(output_dir, module_name)
}
+pub fn schedule_module(
+ module: Module,
+ schedule: ScheduleStmt,
+ mut stringtab: StringTable,
+ mut env: Env<usize, Value>,
+ functions: JunoFunctions,
+) -> Result<Module, SchedulerError> {
+ let mut pm = PassManager::new(module);
+ let _ = schedule_interpret(&mut pm, &schedule, &mut stringtab, &mut env, &functions)?;
+ Ok(pm.get_module())
+}
+
// Interpreter for statements and expressions returns a bool indicating whether
// any optimization ran and changed the IR. This is used for implementing
// the fixpoint
@@ -1239,30 +1278,85 @@ fn run_pass(
pm.clear_analyses();
}
Pass::ForkGuardElim => {
- todo!("Fork Guard Elim doesn't use editor")
- }
- Pass::ForkSplit => {
assert!(args.is_empty());
pm.make_fork_join_maps();
- pm.make_reduce_cycles();
let fork_join_maps = pm.fork_join_maps.take().unwrap();
- let reduce_cycles = pm.reduce_cycles.take().unwrap();
- for ((func, fork_join_map), reduce_cycles) in build_selection(pm, selection)
+ for (func, fork_join_map) in build_selection(pm, selection)
.into_iter()
.zip(fork_join_maps.iter())
- .zip(reduce_cycles.iter())
{
let Some(mut func) = func else {
continue;
};
- fork_split(&mut func, fork_join_map, reduce_cycles);
+ fork_guard_elim(&mut func, fork_join_map);
changed |= func.modified();
}
pm.delete_gravestones();
pm.clear_analyses();
}
+ Pass::Serialize => {
+ // FIXME: How to get module name here?
+ let output_file = "out.hbin";
+ let module = pm.clone().get_module().clone();
+ let module_contents: Vec<u8> = postcard::to_allocvec(&module).unwrap();
+ let mut file =
+ File::create(&output_file).expect("PANIC: Unable to open output module file.");
+ file.write_all(&module_contents)
+ .expect("PANIC: Unable to write output module file contents.");
+ }
+ Pass::ForkSplit => {
+ assert!(args.is_empty());
+ loop {
+ let mut inner_changed = false;
+ pm.make_fork_join_maps();
+ pm.make_reduce_cycles();
+ let fork_join_maps = pm.fork_join_maps.take().unwrap();
+ let reduce_cycles = pm.reduce_cycles.take().unwrap();
+ for ((func, fork_join_map), reduce_cycles) in build_selection(pm, selection.clone())
+ .into_iter()
+ .zip(fork_join_maps.iter())
+ .zip(reduce_cycles.iter())
+ {
+ let Some(mut func) = func else {
+ continue;
+ };
+ fork_split(&mut func, fork_join_map, reduce_cycles);
+ changed |= func.modified();
+ inner_changed |= func.modified();
+ }
+ pm.delete_gravestones();
+ pm.clear_analyses();
+
+ if !inner_changed {
+ break;
+ }
+ }
+ }
Pass::Forkify => {
- todo!("Forkify doesn't use editor")
+ assert!(args.is_empty());
+ pm.make_fork_join_maps();
+ pm.make_control_subgraphs();
+ pm.make_loops();
+ let fork_join_maps = pm.fork_join_maps.take().unwrap();
+ let loops = pm.loops.take().unwrap();
+ let control_subgraphs = pm.control_subgraphs.take().unwrap();
+ for (((func, fork_join_map), loop_nest), control_subgraph) in
+ build_selection(pm, selection)
+ .into_iter()
+ .zip(fork_join_maps.iter())
+ .zip(loops.iter())
+ .zip(control_subgraphs.iter())
+ {
+ let Some(mut func) = func else {
+ continue;
+ };
+ // TODO: uses direct return from forkify for now instead of
+ // func.modified, see comment on top of `forkify` for why. Fix
+ // this eventually.
+ changed |= forkify(&mut func, control_subgraph, fork_join_map, loop_nest);
+ }
+ pm.delete_gravestones();
+ pm.clear_analyses();
}
Pass::GCM => {
assert!(args.is_empty());
@@ -1576,21 +1670,50 @@ fn run_pass(
Pass::Unforkify => {
assert!(args.is_empty());
pm.make_fork_join_maps();
+ pm.make_loops();
+
let fork_join_maps = pm.fork_join_maps.take().unwrap();
+ let loops = pm.loops.take().unwrap();
- for (func, fork_join_map) in build_selection(pm, selection)
+ for ((func, fork_join_map), loop_tree) in build_selection(pm, selection)
.into_iter()
.zip(fork_join_maps.iter())
+ .zip(loops.iter())
{
let Some(mut func) = func else {
continue;
};
- unforkify(&mut func, fork_join_map);
+ unforkify(&mut func, fork_join_map, loop_tree);
changed |= func.modified();
}
pm.delete_gravestones();
pm.clear_analyses();
}
+ Pass::ForkCoalesce => {
+ assert!(args.is_empty());
+ pm.make_fork_join_maps();
+ pm.make_control_subgraphs();
+ pm.make_loops();
+ pm.make_reduce_cycles();
+ let fork_join_maps = pm.fork_join_maps.take().unwrap();
+ let loops = pm.loops.take().unwrap();
+ let control_subgraphs = pm.control_subgraphs.take().unwrap();
+ for (((func, fork_join_map), loop_nest), control_subgraph) in
+ build_selection(pm, selection)
+ .into_iter()
+ .zip(fork_join_maps.iter())
+ .zip(loops.iter())
+ .zip(control_subgraphs.iter())
+ {
+ let Some(mut func) = func else {
+ continue;
+ };
+ changed |= fork_coalesce(&mut func, loop_nest, fork_join_map);
+ // func.modified();
+ }
+ pm.delete_gravestones();
+ pm.clear_analyses();
+ }
Pass::WritePredication => {
assert!(args.is_empty());
for func in build_selection(pm, selection) {
diff --git a/juno_utils/src/stringtab.rs b/juno_utils/src/stringtab.rs
index e151b830..45ee0864 100644
--- a/juno_utils/src/stringtab.rs
+++ b/juno_utils/src/stringtab.rs
@@ -1,6 +1,4 @@
-extern crate serde;
-
-use self::serde::{Deserialize, Serialize};
+use serde::{Deserialize, Serialize};
use std::collections::HashMap;
--
GitLab