maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[AArch64] Fix i64 nontemporal high-half extraction. 

Since we only have pair - not single - nontemporal store instructions,
we have to extract the high part into a separate register to be able
to use them.

When the initial nontemporal codegen support was added, I wrote the
extract using the nonsensical UBFX [0,32[.
Use the correct LSR form instead.

llvm-svn: 259134
This commit is contained in:
Ahmed Bougacha 2016-01-29 01:08:41 +00:00
parent bcf27523f5
commit 53010a0d5b
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llvm/lib/Target/AArch64/AArch64InstrInfo.td
View File

@ -5982,7 +5982,7 @@ def : NTStore64Pat<v8i8>;
def : Pat<(nontemporalstore GPR64:$Rt,
(am_indexed7s32 GPR64sp:$Rn, simm7s4:$offset)),
(STNPWi (EXTRACT_SUBREG GPR64:$Rt, sub_32),
(EXTRACT_SUBREG (UBFMXri GPR64:$Rt, 0, 31), sub_32),
(EXTRACT_SUBREG (UBFMXri GPR64:$Rt, 32, 63), sub_32),
GPR64sp:$Rn, simm7s4:$offset)>;
} // AddedComplexity=10
} // Predicates = [IsLE]

6
llvm/test/CodeGen/AArch64/nontemporal.ll
View File

@ -112,7 +112,7 @@ define void @test_stnp_v1i64(<1 x i64>* %p, <1 x i64> %v) #0 {
define void @test_stnp_i64(i64* %p, i64 %v) #0 {
; CHECK-LABEL: test_stnp_i64:
; CHECK-NEXT: ubfx x[[HI:[0-9]+]], x1, #0, #32
; CHECK-NEXT: lsr x[[HI:[0-9]+]], x1, #32
; CHECK-NEXT: stnp w1, w[[HI]], [x0]
; CHECK-NEXT: ret
store i64 %v, i64* %p, align 1, !nontemporal !0
@ -162,7 +162,7 @@ define void @test_stnp_v2f32_offset_neg(<2 x float>* %p, <2 x float> %v) #0 {
define void @test_stnp_i64_offset(i64* %p, i64 %v) #0 {
; CHECK-LABEL: test_stnp_i64_offset:
; CHECK-NEXT: ubfx x[[HI:[0-9]+]], x1, #0, #32
; CHECK-NEXT: lsr x[[HI:[0-9]+]], x1, #32
; CHECK-NEXT: stnp w1, w[[HI]], [x0, #8]
; CHECK-NEXT: ret
%tmp0 = getelementptr i64, i64* %p, i32 1
@ -172,7 +172,7 @@ define void @test_stnp_i64_offset(i64* %p, i64 %v) #0 {
define void @test_stnp_i64_offset_neg(i64* %p, i64 %v) #0 {
; CHECK-LABEL: test_stnp_i64_offset_neg:
; CHECK-NEXT: ubfx x[[HI:[0-9]+]], x1, #0, #32
; CHECK-NEXT: lsr x[[HI:[0-9]+]], x1, #32
; CHECK-NEXT: stnp w1, w[[HI]], [x0, #-8]
; CHECK-NEXT: ret
%tmp0 = getelementptr i64, i64* %p, i32 -1