summaryrefslogtreecommitdiffstats
path: root/llvm/lib
diff options
context:
space:
mode:
authorAdam Nemet <anemet@apple.com>2014-06-27 00:43:38 +0000
committerAdam Nemet <anemet@apple.com>2014-06-27 00:43:38 +0000
commit73f72e15ac73a1a84bb5e36b5434815cc5273006 (patch)
tree4f8729259cc421365446824822205aa40f98ef9f /llvm/lib
parentdc13b21d639ad075d164a2ad04a81d672f971729 (diff)
downloadbcm5719-llvm-73f72e15ac73a1a84bb5e36b5434815cc5273006.tar.gz
bcm5719-llvm-73f72e15ac73a1a84bb5e36b5434815cc5273006.zip
[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
Diffstat (limited to 'llvm/lib')
-rw-r--r--llvm/lib/Target/X86/X86InstrAVX512.td22
1 files changed, 22 insertions, 0 deletions
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index 110dba9c9ac..75656f3cdf0 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))),
OpenPOWER on IntegriCloud