RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-19 11:02:59 +02:00
Code Issues Projects Releases Wiki Activity

[X86] Fix some patterns that select VLX instructions, but were incorrectly also checking presence of BWI instructions.

Browse Source 
The EVEX->VEX pass probably obscures this.

llvm-svn: 315365
This commit is contained in:
Craig Topper 2017-10-10 21:07:14 +00:00
parent edf19ef4ee
commit 61bc7dca24
2 changed files with 6 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Target/X86/X86InstrAVX512.td
View File

@ -1339,12 +1339,14 @@ let Predicates = [HasAVX512] in {
(VPBROADCASTQZm addr:$src)>;
}
let Predicates = [HasVLX, HasBWI] in {
let Predicates = [HasVLX] in {
// 32-bit targets will fail to load a i64 directly but can use ZEXT_LOAD.
def : Pat<(v2i64 (X86VBroadcast (v2i64 (X86vzload addr:$src)))),
(VPBROADCASTQZ128m addr:$src)>;
def : Pat<(v4i64 (X86VBroadcast (v4i64 (X86vzload addr:$src)))),
(VPBROADCASTQZ256m addr:$src)>;
}
let Predicates = [HasVLX, HasBWI] in {
// loadi16 is tricky to fold, because !isTypeDesirableForOp, justifiably.
// This means we'll encounter truncated i32 loads; match that here.
def : Pat<(v8i16 (X86VBroadcast (i16 (trunc (i32 (load addr:$src)))))),

4
lib/Target/X86/X86InstrSSE.td
View File

@ -7855,12 +7855,14 @@ defm VPBROADCASTD : avx2_broadcast<0x58, "vpbroadcastd", i32mem, loadi32,
defm VPBROADCASTQ : avx2_broadcast<0x59, "vpbroadcastq", i64mem, loadi64,
v2i64, v4i64, NoVLX>;
let Predicates = [HasAVX2, NoVLX_Or_NoBWI] in {
let Predicates = [HasAVX2, NoVLX] in {
// 32-bit targets will fail to load a i64 directly but can use ZEXT_LOAD.
def : Pat<(v2i64 (X86VBroadcast (v2i64 (X86vzload addr:$src)))),
(VPBROADCASTQrm addr:$src)>;
def : Pat<(v4i64 (X86VBroadcast (v4i64 (X86vzload addr:$src)))),
(VPBROADCASTQYrm addr:$src)>;
}
let Predicates = [HasAVX2, NoVLX_Or_NoBWI] in {
// loadi16 is tricky to fold, because !isTypeDesirableForOp, justifiably.
// This means we'll encounter truncated i32 loads; match that here.
def : Pat<(v8i16 (X86VBroadcast (i16 (trunc (i32 (load addr:$src)))))),