Make Memory{T} respect the max alignment of the heap (#53748)

Fixes #53683

(cherry picked from commit 62b5159)

Author: gbaraldi

src/codegen.cpp

@@ -3793,6 +3793,8 @@ static bool emit_f_opmemory(jl_codectx_t &ctx, jl_cgval_t *ret, jl_value_t *f,
     bool needlock = isatomic && layout->size > MAX_ATOMIC_SIZE;
     size_t elsz = layout->size;
     size_t al = layout->alignment;
+    if (al > JL_HEAP_ALIGNMENT)
+        al = JL_HEAP_ALIGNMENT;
     if (isatomic == (order == jl_memory_order_notatomic)) {
         emit_atomic_error(ctx,
                 issetmemory ?
@@ -4120,6 +4122,8 @@ static bool emit_builtin_call(jl_codectx_t &ctx, jl_cgval_t *ret, jl_value_t *f,
             bool isunion = layout->flags.arrayelem_isunion;
             size_t elsz = layout->size;
             size_t al = layout->alignment;
+            if (al > JL_HEAP_ALIGNMENT)
+                al = JL_HEAP_ALIGNMENT;
             bool needlock = isatomic && !isboxed && elsz > MAX_ATOMIC_SIZE;
             AtomicOrdering Order = (needlock || order <= jl_memory_order_notatomic)
                                     ? (isboxed ? AtomicOrdering::Unordered : AtomicOrdering::NotAtomic)

test/vecelement.jl

@@ -1,5 +1,5 @@
 # This file is a part of Julia. License is MIT: https://julialang.org/license
-
+using InteractiveUtils
 make_value(::Type{T}, i::Integer) where {T<:Integer} = 3*i%T
 make_value(::Type{T},i::Integer) where {T<:AbstractFloat} = T(3*i)
 
@@ -120,3 +120,9 @@ for T in (Float64, Float32, Int64, Int32)
         @test b == result
     end
 end
+@testset "vecelement overalignment" begin
+    io = IOBuffer()
+    code_llvm(io,getindex, (Array{NTuple{5, VecElement{Float64}}, 1}, Int64), optimize=false)
+    ir = String(take!(io))
+    @test match(r"align 64", ir) === nothing
+end