From 73f72e15ac73a1a84bb5e36b5434815cc5273006 Mon Sep 17 00:00:00 2001
From: Adam Nemet <anemet@apple.com>
Date: Fri, 27 Jun 2014 00:43:38 +0000
Subject: [PATCH] [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>

llvm-svn: 211828
---
 llvm/lib/Target/X86/X86InstrAVX512.td | 22 ++++++++++++++++++++++
 llvm/test/MC/X86/avx512-encodings.s   |  8 ++++++++
 2 files changed, 30 insertions(+)

diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index 110dba9..75656f3 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/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))),
diff --git a/llvm/test/MC/X86/avx512-encodings.s b/llvm/test/MC/X86/avx512-encodings.s
index e37c8b6..97841fa 100644
--- a/llvm/test/MC/X86/avx512-encodings.s
+++ b/llvm/test/MC/X86/avx512-encodings.s
@@ -3128,6 +3128,14 @@ vpsrad 512(%rdi, %rsi, 4), %zmm12, %zmm25
 // CHECK: encoding: [0x62,0xf2,0x7d,0xc9,0x58,0xc8]
 vpbroadcastd  %xmm0, %zmm1 {%k1} {z}
 
+// CHECK: vbroadcasti32x4 {{.*}} {%k7} {z}
+// CHECK: encoding: [0x67,0x62,0xf2,0x7d,0xcf,0x5a,0x52,0x02]
+vbroadcasti32x4 0x20(%edx), %zmm2 {%k7} {z}
+
+// CHECK: vbroadcasti64x4 {{.*}} %zmm22
+// CHECK: encoding: [0x62,0xe2,0xfd,0x48,0x5b,0x72,0x02]
+vbroadcasti64x4 0x40(%rdx), %zmm22
+
 // CHECK: vmovdqu64 {{.*}} {%k3}
 // CHECK: encoding: [0x62,0xf1,0xfe,0x4b,0x7f,0x07]
 vmovdqu64 %zmm0, (%rdi) {%k3}
-- 
2.7.4