From a50d4dd9b0378cce9400139c7808204540a1f5a9 Mon Sep 17 00:00:00 2001
From: Adam Nemet <anemet@apple.com>
Date: Fri, 27 Jun 2014 00:43:38 +0000
Subject: [X86] AVX512: Add vbroadcasti*

For now I used a separate template for these sub-vector/tuple broadcasts
rather than sharing the mem variants with avx512_int_broadcast_rm.

<rdar://problem/17402869>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211828 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86InstrAVX512.td | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

(limited to 'lib')

diff --git a/lib/Target/X86/X86InstrAVX512.td b/lib/Target/X86/X86InstrAVX512.td
index 110dba9c9a..75656f3cdf 100644
--- a/lib/Target/X86/X86InstrAVX512.td
+++ b/lib/Target/X86/X86InstrAVX512.td
@@ -476,6 +476,28 @@ defm VPBROADCASTQZ  : avx512_int_broadcast_rm<0x59, "vpbroadcastq", i64mem,
                       loadi64, VR512, v8i64, v2i64, VK8WM>,  EVEX_V512, VEX_W,
                       EVEX_CD8<64, CD8VT1>;
 
+multiclass avx512_int_subvec_broadcast_rm<bits<8> opc, string OpcodeStr,
+                          X86MemOperand x86memop, PatFrag ld_frag,
+                          RegisterClass KRC> {
+  let mayLoad = 1 in {
+  def rm : AVX5128I<opc, MRMSrcMem, (outs VR512:$dst), (ins x86memop:$src),
+                  !strconcat(OpcodeStr, " \t{$src, $dst|$dst, $src}"),
+                  []>, EVEX;
+  def krm : AVX5128I<opc, MRMSrcMem, (outs VR512:$dst), (ins KRC:$mask,
+                                                         x86memop:$src),
+                  !strconcat(OpcodeStr,
+                      " \t{$src, ${dst} {${mask}} {z}|${dst} {${mask}} {z}, $src}"),
+                  []>, EVEX, EVEX_KZ;
+  }
+}
+
+defm VBROADCASTI32X4 : avx512_int_subvec_broadcast_rm<0x5a, "vbroadcasti32x4",
+                       i128mem, loadv2i64, VK16WM>,
+                       EVEX_V512, EVEX_CD8<32, CD8VT4>;
+defm VBROADCASTI64X4 : avx512_int_subvec_broadcast_rm<0x5b, "vbroadcasti64x4",
+                       i256mem, loadv4i64, VK16WM>, VEX_W,
+                       EVEX_V512, EVEX_CD8<64, CD8VT4>;
+
 def : Pat<(v16i32 (int_x86_avx512_pbroadcastd_512 (v4i32 VR128X:$src))),
           (VPBROADCASTDZrr VR128X:$src)>;
 def : Pat<(v8i64 (int_x86_avx512_pbroadcastq_512 (v2i64 VR128X:$src))),
-- 
cgit v1.2.3