diff --git a/src/pystencils/backend/platforms/sycl.py b/src/pystencils/backend/platforms/sycl.py
index fdc19a0c4ad126e9a9e3622cf781e31125292348..52953115ab3a50139c6bfdf047cc8f9fa53714d8 100644
--- a/src/pystencils/backend/platforms/sycl.py
+++ b/src/pystencils/backend/platforms/sycl.py
@@ -101,7 +101,13 @@ class SyclPlatform(GenericGpu):
 
         indexing_decls = [id_decl]
         conds = []
-        for i, dim in enumerate(dimensions[::-1]):
+
+        #   Other than in CUDA, SYCL ids are linearized in C order
+        #   The leftmost entry of an ID varies slowest, and the rightmost entry varies fastest
+        #   See https://registry.khronos.org/SYCL/specs/sycl-2020/html/sycl-2020.html#sec:multi-dim-linearization
+
+        for i, dim in enumerate(dimensions):
+            #   Slowest to fastest
             coord = PsExpression.make(PsConstant(i, self._ctx.index_dtype))
             work_item_idx = PsSubscript(id_symbol, coord)