[llvm] [SLPVectorizer] Widen strided loads. (PR #153074)

[Mikhail Gudim via llvm-commits]

================
@@ -37,6 +37,11 @@ static cl::opt<unsigned> SLPMaxVF(
         "exclusively by SLP vectorizer."),
     cl::Hidden);
 
+static cl::opt<bool> SLPPreferAltOpcVectorization(
+    "riscv-v-slp-prefer-alt-opc-vectorization",
+    cl::desc("Controls preferAlternateOpcodeVectorization"), cl::init(false),
+    cl::Hidden);
+
----------------
mgudim wrote:

this is just to show what it does on x264. This option can be removed when we merge. But we need to figure out the right solution to this. Without this option, satd in x264 won't generate strided load.

https://github.com/llvm/llvm-project/pull/153074