Rename `tests/codegen` into `tests/codegen-llvm`

1 files changed, 35 insertions, 0 deletions

diff --git a/tests/codegen-llvm/array-map.rs b/tests/codegen-llvm/array-map.rs
new file mode 100644
index 00000000000..f49dddcfc20
--- /dev/null
+++ b/tests/codegen-llvm/array-map.rs
@@ -0,0 +1,35 @@
+//@ compile-flags: -C opt-level=3 -C target-cpu=x86-64-v3
+//@ only-x86_64
+
+#![crate_type = "lib"]
+
+// CHECK-LABEL: @short_integer_map
+#[no_mangle]
+pub fn short_integer_map(x: [u32; 8]) -> [u32; 8] {
+    // CHECK: load <8 x i32>
+    // CHECK: shl <8 x i32>
+    // CHECK: or{{( disjoint)?}} <8 x i32>
+    // CHECK: store <8 x i32>
+    x.map(|x| 2 * x + 1)
+}
+
+// This test is checking that LLVM can SRoA away a bunch of the overhead,
+// like fully moving the iterators to registers.  Notably, previous implementations
+// of `map` ended up `alloca`ing the whole `array::IntoIterator`, meaning both a
+// hard-to-eliminate `memcpy` and that the iteration counts needed to be written
+// out to stack every iteration, even for infallible operations on `Copy` types.
+//
+// This is still imperfect, as there's more copies than would be ideal,
+// but hopefully work like #103830 will improve that in future,
+// and update this test to be stricter.
+//
+// CHECK-LABEL: @long_integer_map
+#[no_mangle]
+pub fn long_integer_map(x: [u32; 512]) -> [u32; 512] {
+    // CHECK: start:
+    // CHECK-NEXT: alloca [2048 x i8]
+    // CHECK-NOT: alloca
+    // CHECK: mul <{{[0-9]+}} x i32>
+    // CHECK: add <{{[0-9]+}} x i32>
+    x.map(|x| 13 * x + 7)
+}