Auto merge of #28221 - huonw:simd, r=alexcrichton

The ARM equivalents of the AArch64 are annoyingly more complicated (and some of the AArch64 ones are too).

I think I've got exposed all the x86 intrinsics from SSE to AVX2 now (at least, the ones that LLVM implements as callable intrinsics).

Author: bors

src/etc/platform-intrinsics/aarch64.json

@@ -336,6 +336,48 @@
             "ret": "i8",
             "args": ["0"]
         },
+        {
+            "intrinsic": "ld2{0[0].width}_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld2.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);2]","[f(32-64);2]"],
+            "args": ["0.0SPc/0.0"]
+        },
+        {
+            "intrinsic": "ld3{0[0].width}_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld3.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);3]","[f(32-64);3]"],
+            "args": ["0.0SPc/0.0"]
+        },
+        {
+            "intrinsic": "ld4{0[0].width}_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld4.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);4]","[f(32-64);4]"],
+            "args": ["0.0SPc/0.0"]
+        },
+        {
+            "intrinsic": "ld2{0[0].width}_dup_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld2.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);2]","[f(32-64);2]"],
+            "args": ["0.0SPc"]
+        },
+        {
+            "intrinsic": "ld3{0[0].width}_dup_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld3.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);3]","[f(32-64);3]"],
+            "args": ["0.0SPc"]
+        },
+        {
+            "intrinsic": "ld4{0[0].width}_dup_{0[0].data_type}",
+            "width": [64, 128],
+            "llvm": "ld4.{0[0].llvm_name}.{1.llvm_name}",
+            "ret": ["[i(8-64);4]","[f(32-64);4]"],
+            "args": ["0.0SPc"]
+        },
         {
             "intrinsic": "padd{0.width}_{0.data_type}",
             "width": [64, 128],