OSchip
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
13
0
Fork 6
Code Issues Packages Projects Releases Wiki Activity

[X86] Fold (trunc (i32 (zextload i16))) into vbroadcast. 

When matching non-LSB-extracting truncating broadcasts, we now insert
the necessary SRL. If the scalar resulted from a load, the SRL will be
folded into it, creating a narrower, offset, load.

However, i16 loads aren't Desirable, so we get i16->i32 zextloads.
We already catch i16 aextloads; catch these as well.

llvm-svn: 252363
This commit is contained in:
Ahmed Bougacha 2015-11-06 23:16:48 +00:00
parent 05a0514b12
commit b49eb3ab4b
3 changed files with 10 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llvm/lib/Target/X86/X86InstrSSE.td
View File

@ -8365,6 +8365,12 @@ let Predicates = [HasAVX2] in {
(VPBROADCASTWrm addr:$src)>;
def : Pat<(v16i16 (X86VBroadcast (i16 (trunc (i32 (load addr:$src)))))),
(VPBROADCASTWYrm addr:$src)>;
def : Pat<(v8i16 (X86VBroadcast
(i16 (trunc (i32 (zextloadi16 addr:$src)))))),
(VPBROADCASTWrm addr:$src)>;
def : Pat<(v16i16 (X86VBroadcast
(i16 (trunc (i32 (zextloadi16 addr:$src)))))),
(VPBROADCASTWYrm addr:$src)>;
// Provide aliases for broadcast from the same register class that
// automatically does the extract.

8
llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
View File

@ -2258,9 +2258,7 @@ define <8 x i16> @insert_dup_elt1_mem_v8i16_i32(i32* %ptr) {
;
; AVX2-LABEL: insert_dup_elt1_mem_v8i16_i32:
; AVX2: # BB#0:
; AVX2-NEXT: movzwl 2(%rdi), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw 2(%rdi), %xmm0
; AVX2-NEXT: retq
%tmp = load i32, i32* %ptr, align 4
%tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
@ -2298,9 +2296,7 @@ define <8 x i16> @insert_dup_elt3_mem_v8i16_i32(i32* %ptr) {
;
; AVX2-LABEL: insert_dup_elt3_mem_v8i16_i32:
; AVX2: # BB#0:
; AVX2-NEXT: movzwl 2(%rdi), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw 2(%rdi), %xmm0
; AVX2-NEXT: retq
%tmp = load i32, i32* %ptr, align 4
%tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 1

8
llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
View File

@ -3342,9 +3342,7 @@ define <16 x i16> @insert_dup_elt1_mem_v16i16_i32(i32* %ptr) #0 {
;
; AVX2-LABEL: insert_dup_elt1_mem_v16i16_i32:
; AVX2: # BB#0:
; AVX2-NEXT: movzwl 2(%rdi), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %ymm0
; AVX2-NEXT: vpbroadcastw 2(%rdi), %ymm0
; AVX2-NEXT: retq
%tmp = load i32, i32* %ptr, align 4
%tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
@ -3363,9 +3361,7 @@ define <16 x i16> @insert_dup_elt3_mem_v16i16_i32(i32* %ptr) #0 {
;
; AVX2-LABEL: insert_dup_elt3_mem_v16i16_i32:
; AVX2: # BB#0:
; AVX2-NEXT: movzwl 2(%rdi), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %ymm0
; AVX2-NEXT: vpbroadcastw 2(%rdi), %ymm0
; AVX2-NEXT: retq
%tmp = load i32, i32* %ptr, align 4
%tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 1