Magic-Mirror/llvm-capstone
1
0
Fork 0
mirror of https://github.com/capstone-engine/llvm-capstone.git synced 2025-01-09 09:32:20 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[X86] Combine (store (v1i1 (scalar_to_vector (i8 X)))) -> (store (i8 X)).

Browse Source 
llvm-svn: 326670
This commit is contained in:
Craig Topper 2018-03-04 01:48:02 +00:00
parent be31585be8
commit a476026f70
2 changed files with 24 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -35021,6 +35021,16 @@ static SDValue combineStore(SDNode *N, SelectionDAG &DAG,
SDValue StoredVal = St->getOperand(1);
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
// If this is a store of a scalar_to_vector to v1i1, just use a scalar store.
// This will avoid a copy to k-register.
if (VT == MVT::v1i1 && VT == StVT && Subtarget.hasAVX512() &&
StoredVal.getOpcode() == ISD::SCALAR_TO_VECTOR &&
StoredVal.getOperand(0).getValueType() == MVT::i8) {
return DAG.getStore(St->getChain(), dl, StoredVal.getOperand(0),
St->getBasePtr(), St->getPointerInfo(),
St->getAlignment(), St->getMemOperand()->getFlags());
}
// If we are saving a concatenation of two XMM registers and 32-byte stores
// are slow, such as on Sandy Bridge, perform two 16-byte stores.
bool Fast;

28
llvm/test/CodeGen/X86/avx512-load-trunc-store-i1.ll
View File

@ -5,8 +5,8 @@
define void @load_v1i2_trunc_v1i1_store(<1 x i2>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i2_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i2_trunc_v1i1_store:
@ -22,8 +22,8 @@ define void @load_v1i2_trunc_v1i1_store(<1 x i2>* %a0,<1 x i1>* %a1) {
define void @load_v1i3_trunc_v1i1_store(<1 x i3>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i3_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i3_trunc_v1i1_store:
@ -39,8 +39,8 @@ define void @load_v1i3_trunc_v1i1_store(<1 x i3>* %a0,<1 x i1>* %a1) {
define void @load_v1i4_trunc_v1i1_store(<1 x i4>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i4_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i4_trunc_v1i1_store:
@ -56,8 +56,8 @@ define void @load_v1i4_trunc_v1i1_store(<1 x i4>* %a0,<1 x i1>* %a1) {
define void @load_v1i8_trunc_v1i1_store(<1 x i8>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i8_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i8_trunc_v1i1_store:
@ -73,8 +73,8 @@ define void @load_v1i8_trunc_v1i1_store(<1 x i8>* %a0,<1 x i1>* %a1) {
define void @load_v1i16_trunc_v1i1_store(<1 x i16>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i16_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i16_trunc_v1i1_store:
@ -90,8 +90,8 @@ define void @load_v1i16_trunc_v1i1_store(<1 x i16>* %a0,<1 x i1>* %a1) {
define void @load_v1i32_trunc_v1i1_store(<1 x i32>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i32_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i32_trunc_v1i1_store:
@ -107,8 +107,8 @@ define void @load_v1i32_trunc_v1i1_store(<1 x i32>* %a0,<1 x i1>* %a1) {
define void @load_v1i64_trunc_v1i1_store(<1 x i64>* %a0,<1 x i1>* %a1) {
; AVX512-ALL-LABEL: load_v1i64_trunc_v1i1_store:
; AVX512-ALL: # %bb.0:
; AVX512-ALL-NEXT: kmovb (%rdi), %k0
; AVX512-ALL-NEXT: kmovb %k0, (%rsi)
; AVX512-ALL-NEXT: movb (%rdi), %al
; AVX512-ALL-NEXT: movb %al, (%rsi)
; AVX512-ALL-NEXT: retq
;
; AVX512-ONLY-LABEL: load_v1i64_trunc_v1i1_store: