[flang-commits] [flang] [Flang]Fix propagation of loop collapse number for target-based directives (PR #162707)

[Alexey Bataev via flang-commits]

https://github.com/alexey-bataev created https://github.com/llvm/llvm-project/pull/162707

Currently, loops collapse number for target-based directives is not
propagated, which leads to the fact that the loops are not collapsed
anymore and perf regressions.


>From e004f9a268a28f40760fcb7dfff46dec21436f88 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev at outlook.com>
Date: Thu, 9 Oct 2025 10:32:02 -0700
Subject: [PATCH] =?UTF-8?q?[=F0=9D=98=80=F0=9D=97=BD=F0=9D=97=BF]=20initia?=
 =?UTF-8?q?l=20version?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Created using spr 1.3.7
---
 flang/lib/Lower/OpenMP/OpenMP.cpp                       | 1 +
 flang/test/Lower/OpenMP/distribute-parallel-do-simd.f90 | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 9e56c2bfb7e25..bd94651919960 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -153,6 +153,7 @@ class HostEvalInfo {
     clauseOps.loopLowerBounds = ops.loopLowerBounds;
     clauseOps.loopUpperBounds = ops.loopUpperBounds;
     clauseOps.loopSteps = ops.loopSteps;
+    clauseOps.collapseNumLoops = ops.collapseNumLoops;
     ivOut.append(iv);
     return true;
   }
diff --git a/flang/test/Lower/OpenMP/distribute-parallel-do-simd.f90 b/flang/test/Lower/OpenMP/distribute-parallel-do-simd.f90
index 142bc02ae8c1d..c769152318663 100644
--- a/flang/test/Lower/OpenMP/distribute-parallel-do-simd.f90
+++ b/flang/test/Lower/OpenMP/distribute-parallel-do-simd.f90
@@ -112,7 +112,7 @@ subroutine lastprivate_cond_in_composite_construct(x_min, x_max, y_min, y_max)
 ! CHECK:                 omp.distribute {
 ! CHECK:                   omp.wsloop {
 ! CHECK:                     omp.simd private({{.*}}) {
-! CHECK:                       omp.loop_nest (%[[I_IV:.*]], %[[J_IV:.*]]) : i32 = ({{.*}}) to ({{.*}}) inclusive step ({{.*}}) {
+! CHECK:                       omp.loop_nest (%[[I_IV:.*]], %[[J_IV:.*]]) : i32 = ({{.*}}) to ({{.*}}) inclusive step ({{.*}}) collapse(2) {
 ! CHECK:                         %[[Y_MAX_PRIV:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "{{.*}}y_max"}
 
 ! CHECK:                         %[[I_UB:.*]] = fir.load %[[X_MAX_MAPPED]]#0 : !fir.ref<i32>