Define AVX broadcast intrinsics

This defines `_mm256_broadcast_ps` and `_mm256_broadcast_pd`. The `_ss`
and `_sd` variants are not supported by LLVM. In Clang these intrinsics
are implemented as inline functions in C++.

Intel reference: https://software.intel.com/en-us/node/514144.

Note: the argument type should really be "0hPc" (a pointer to a vector
of half the width), but internally the LLVM intrinsic takes a pointer to
a signed integer, and for any other type LLVM will complain. This means
that a transmute is required to call these intrinsics.

The AVX2 broadcast intrinsics `_mm256_broadcastss_ps` and
`_mm256_broadcastsd_pd` are not available as LLVM intrinsics. In Clang
they are implemented using the shufflevector builtin.

1 files changed, 7 insertions, 0 deletions

diff --git a/src/etc/platform-intrinsics/x86/avx.json b/src/etc/platform-intrinsics/x86/avx.json
index 2c1492c2954..981838536b2 100644
--- a/src/etc/platform-intrinsics/x86/avx.json
+++ b/src/etc/platform-intrinsics/x86/avx.json
@@ -9,6 +9,13 @@
             "args": ["0", "0"]
         },
         {
+            "intrinsic": "256_broadcast_{0.data_type}",
+            "width": [256],
+            "llvm": "vbroadcastf128.{0.data_type}.256",
+            "ret": "f(32-64)",
+            "args": ["s8SPc"]
+        },
+        {
             "intrinsic": "256_dp_ps",
             "width": [256],
             "llvm": "dp.ps.256",