[Mlir-commits] [llvm] [mlir] [llvm][OpenMP] Allow Chunk Size on SIMD Guided (PR #178853)

[Jack Styles]

https://github.com/Stylie777 created https://github.com/llvm/llvm-project/pull/178853

As per the OpenMP Spec, Chunk Size is allowed when using the guided kind-type with the schedule clause. However, when being used in cases such as `!$omp do simd schedule (simd:guided,4)`, this was not allowed as the base type, BaseGuidedSimd, would hit an assert not allowing ChunkSizes.

By making this change, we can allow the use of the Guided type, with a ChunkSize and the schedule clause when using OMPIRBuidler.

Fixes #82106 

>From e1cc109ae0afb393c319aefc9ddb13c166228f38 Mon Sep 17 00:00:00 2001
From: Jack Styles <jack.styles at arm.com>
Date: Fri, 30 Jan 2026 09:17:25 +0000
Subject: [PATCH] [llvm][OpenMP] Allow Chunk Size on SIMD Guided

As per the OpenMP Spec, Chunk Size is allowed when using the guided
kind-type with the schedule clause. However, when being used in cases
such as `!$omp do simd schedule (simd:guided,4)`, this was not allowed
as the base type, BaseGuidedSimd, would hit an assert not allowing
ChunkSizes.

By making this change, we can allow the use of the Guided type, with
a ChunkSize and the schedule clause when using OMPIRBuidler.

Fixes #150747
---
 llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp     |  2 +-
 .../Target/LLVMIR/openmp-simd-guided.mlir     | 23 +++++++++++++++++++
 2 files changed, 24 insertions(+), 1 deletion(-)
 create mode 100644 mlir/test/Target/LLVMIR/openmp-simd-guided.mlir

diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 8d7a207a91f5a..719b0177c0198 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -5956,11 +5956,11 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::applyWorkshareLoop(
   case OMPScheduleType::BaseGreedy:
   case OMPScheduleType::BaseBalanced:
   case OMPScheduleType::BaseSteal:
-  case OMPScheduleType::BaseGuidedSimd:
   case OMPScheduleType::BaseRuntimeSimd:
     assert(!ChunkSize &&
            "schedule type does not support user-defined chunk sizes");
     [[fallthrough]];
+  case OMPScheduleType::BaseGuidedSimd:
   case OMPScheduleType::BaseDynamicChunked:
   case OMPScheduleType::BaseGuidedChunked:
   case OMPScheduleType::BaseGuidedIterativeChunked:
diff --git a/mlir/test/Target/LLVMIR/openmp-simd-guided.mlir b/mlir/test/Target/LLVMIR/openmp-simd-guided.mlir
new file mode 100644
index 0000000000000..ac46b4a56ba32
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-simd-guided.mlir
@@ -0,0 +1,23 @@
+// Ensure that schedule can be used with the guided kind-type and simd construct
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
+
+omp.private {type = private} @_QFEi_private_i32 : i32
+llvm.func @test_simd_guided() {
+  %0 = llvm.mlir.constant (1 : i64) : i64
+  %c1_i32 = llvm.mlir.constant (1 : i32) : i32
+  %c64_i32 = llvm.mlir.constant (64 : i32) : i32
+  %c0_i32 = llvm.mlir.constant (0 : i32) : i32
+  %c4_i32 = llvm.mlir.constant (4 : i32) : i32
+  %1 = llvm.alloca %0 x i32 {bindc_name = "i"} : (i64) -> !llvm.ptr
+  omp.wsloop schedule(guided = %c4_i32 : i32, simd) {
+    omp.simd linear(%1 = %c1_i32 : !llvm.ptr) private(@_QFEi_private_i32 %1 -> %arg0 : !llvm.ptr) {
+      omp.loop_nest (%arg1) : i32 = (%c0_i32) to (%c64_i32) inclusive step (%c1_i32) {
+        omp.yield
+      }
+    } {linear_var_types = [i32], omp.composite}
+  } {omp.composite}
+  llvm.return
+}
+
+// CHECK:   %[[omp_global_thread_num:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
+// CHECK:   call void @__kmpc_dispatch_init_4u(ptr @1, i32 %[[omp_global_thread_num]], i32 1073741870, i32 1, i32 65, i32 1, i32 4)