RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
Code Issues Projects Releases Wiki Activity

[SLP] Add a new test for tree vectorization starting from insertelement

Browse Source 
instruction.

llvm-svn: 288148
This commit is contained in:
Alexey Bataev 2016-11-29 15:37:52 +00:00
parent f1dcbdb14d
commit a4bfd1d3ea
1 changed files with 508 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

541
test/Transforms/SLPVectorizer/X86/insert-element-build-vector.ll
View File

@ -1,13 +1,37 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -S -slp-vectorizer -slp-threshold=-10000 < %s | FileCheck %s
; RUN: opt -S -slp-vectorizer -slp-threshold=0 < %s | FileCheck %s -check-prefix=ZEROTHRESH
; RUN: opt -S -slp-vectorizer -slp-threshold=0 < %s | FileCheck %s --check-prefix=ZEROTHRESH
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0"
define <4 x float> @simple_select(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select(
; CHECK-NEXT: %1 = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: select <4 x i1> %1, <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 0
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 2
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; CHECK-NEXT: ret <4 x float> [[RD]]
;
; ZEROTHRESH-LABEL: @simple_select(
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 0
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 2
; ZEROTHRESH-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; ZEROTHRESH-NEXT: ret <4 x float> [[RD]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -40,8 +64,77 @@ declare void @llvm.assume(i1) nounwind
; This entire tree is ephemeral, don't vectorize any of it.
define <4 x float> @simple_select_eph(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_eph(
; CHECK-NOT: icmp ne <4 x i32>
; CHECK-NOT: select <4 x i1>
; CHECK-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; CHECK-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; CHECK-NEXT: [[C2:%.*]] = extractelement <4 x i32> %c, i32 2
; CHECK-NEXT: [[C3:%.*]] = extractelement <4 x i32> %c, i32 3
; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x float> %a, i32 2
; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x float> %a, i32 3
; CHECK-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; CHECK-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; CHECK-NEXT: [[B2:%.*]] = extractelement <4 x float> %b, i32 2
; CHECK-NEXT: [[B3:%.*]] = extractelement <4 x float> %b, i32 3
; CHECK-NEXT: [[CMP0:%.*]] = icmp ne i32 [[C0]], 0
; CHECK-NEXT: [[CMP1:%.*]] = icmp ne i32 [[C1]], 0
; CHECK-NEXT: [[CMP2:%.*]] = icmp ne i32 [[C2]], 0
; CHECK-NEXT: [[CMP3:%.*]] = icmp ne i32 [[C3]], 0
; CHECK-NEXT: [[S0:%.*]] = select i1 [[CMP0]], float [[A0]], float [[B0]]
; CHECK-NEXT: [[S1:%.*]] = select i1 [[CMP1]], float [[A1]], float [[B1]]
; CHECK-NEXT: [[S2:%.*]] = select i1 [[CMP2]], float [[A2]], float [[B2]]
; CHECK-NEXT: [[S3:%.*]] = select i1 [[CMP3]], float [[A3]], float [[B3]]
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[S0]], i32 0
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[S1]], i32 1
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[S2]], i32 2
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[S3]], i32 3
; CHECK-NEXT: [[Q0:%.*]] = extractelement <4 x float> [[RD]], i32 0
; CHECK-NEXT: [[Q1:%.*]] = extractelement <4 x float> [[RD]], i32 1
; CHECK-NEXT: [[Q2:%.*]] = extractelement <4 x float> [[RD]], i32 2
; CHECK-NEXT: [[Q3:%.*]] = extractelement <4 x float> [[RD]], i32 3
; CHECK-NEXT: [[Q4:%.*]] = fadd float [[Q0]], [[Q1]]
; CHECK-NEXT: [[Q5:%.*]] = fadd float [[Q2]], [[Q3]]
; CHECK-NEXT: [[Q6:%.*]] = fadd float [[Q4]], [[Q5]]
; CHECK-NEXT: [[QI:%.*]] = fcmp olt float [[Q6]], [[Q5]]
; CHECK-NEXT: call void @llvm.assume(i1 [[QI]])
; CHECK-NEXT: ret <4 x float> undef
;
; ZEROTHRESH-LABEL: @simple_select_eph(
; ZEROTHRESH-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; ZEROTHRESH-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; ZEROTHRESH-NEXT: [[C2:%.*]] = extractelement <4 x i32> %c, i32 2
; ZEROTHRESH-NEXT: [[C3:%.*]] = extractelement <4 x i32> %c, i32 3
; ZEROTHRESH-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; ZEROTHRESH-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; ZEROTHRESH-NEXT: [[A2:%.*]] = extractelement <4 x float> %a, i32 2
; ZEROTHRESH-NEXT: [[A3:%.*]] = extractelement <4 x float> %a, i32 3
; ZEROTHRESH-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; ZEROTHRESH-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; ZEROTHRESH-NEXT: [[B2:%.*]] = extractelement <4 x float> %b, i32 2
; ZEROTHRESH-NEXT: [[B3:%.*]] = extractelement <4 x float> %b, i32 3
; ZEROTHRESH-NEXT: [[CMP0:%.*]] = icmp ne i32 [[C0]], 0
; ZEROTHRESH-NEXT: [[CMP1:%.*]] = icmp ne i32 [[C1]], 0
; ZEROTHRESH-NEXT: [[CMP2:%.*]] = icmp ne i32 [[C2]], 0
; ZEROTHRESH-NEXT: [[CMP3:%.*]] = icmp ne i32 [[C3]], 0
; ZEROTHRESH-NEXT: [[S0:%.*]] = select i1 [[CMP0]], float [[A0]], float [[B0]]
; ZEROTHRESH-NEXT: [[S1:%.*]] = select i1 [[CMP1]], float [[A1]], float [[B1]]
; ZEROTHRESH-NEXT: [[S2:%.*]] = select i1 [[CMP2]], float [[A2]], float [[B2]]
; ZEROTHRESH-NEXT: [[S3:%.*]] = select i1 [[CMP3]], float [[A3]], float [[B3]]
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[S0]], i32 0
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[S1]], i32 1
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[S2]], i32 2
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[S3]], i32 3
; ZEROTHRESH-NEXT: [[Q0:%.*]] = extractelement <4 x float> [[RD]], i32 0
; ZEROTHRESH-NEXT: [[Q1:%.*]] = extractelement <4 x float> [[RD]], i32 1
; ZEROTHRESH-NEXT: [[Q2:%.*]] = extractelement <4 x float> [[RD]], i32 2
; ZEROTHRESH-NEXT: [[Q3:%.*]] = extractelement <4 x float> [[RD]], i32 3
; ZEROTHRESH-NEXT: [[Q4:%.*]] = fadd float [[Q0]], [[Q1]]
; ZEROTHRESH-NEXT: [[Q5:%.*]] = fadd float [[Q2]], [[Q3]]
; ZEROTHRESH-NEXT: [[Q6:%.*]] = fadd float [[Q4]], [[Q5]]
; ZEROTHRESH-NEXT: [[QI:%.*]] = fcmp olt float [[Q6]], [[Q5]]
; ZEROTHRESH-NEXT: call void @llvm.assume(i1 [[QI]])
; ZEROTHRESH-NEXT: ret <4 x float> undef
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -82,8 +175,31 @@ define <4 x float> @simple_select_eph(<4 x float> %a, <4 x float> %b, <4 x i32>
; doesn't matter
define <4 x float> @simple_select_insert_out_of_order(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_insert_out_of_order(
; CHECK-NEXT: %1 = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: select <4 x i1> %1, <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 2
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 0
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; CHECK-NEXT: ret <4 x float> [[RD]]
;
; ZEROTHRESH-LABEL: @simple_select_insert_out_of_order(
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 2
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 0
; ZEROTHRESH-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; ZEROTHRESH-NEXT: ret <4 x float> [[RD]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -117,8 +233,33 @@ declare void @f32_user(float) #0
; Multiple users of the final constructed vector
define <4 x float> @simple_select_users(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_users(
; CHECK-NEXT: %1 = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: select <4 x i1> %1, <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 0
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 2
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; CHECK-NEXT: call void @v4f32_user(<4 x float> [[RD]]) #0
; CHECK-NEXT: ret <4 x float> [[RD]]
;
; ZEROTHRESH-LABEL: @simple_select_users(
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = icmp ne <4 x i32> %c, zeroinitializer
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> %a, <4 x float> %b
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP3]], i32 0
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP2]], i32 1
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP4]], i32 1
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP2]], i32 2
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x float> [[RB]], float [[TMP5]], i32 2
; ZEROTHRESH-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP6]], i32 3
; ZEROTHRESH-NEXT: call void @v4f32_user(<4 x float> [[RD]]) #0
; ZEROTHRESH-NEXT: ret <4 x float> [[RD]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -150,8 +291,65 @@ define <4 x float> @simple_select_users(<4 x float> %a, <4 x float> %b, <4 x i32
; Unused insertelement
define <4 x float> @simple_select_no_users(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_no_users(
; CHECK-NOT: icmp ne <4 x i32>
; CHECK-NOT: select <4 x i1>
; CHECK-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; CHECK-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; CHECK-NEXT: [[C2:%.*]] = extractelement <4 x i32> %c, i32 2
; CHECK-NEXT: [[C3:%.*]] = extractelement <4 x i32> %c, i32 3
; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x float> %a, i32 2
; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x float> %a, i32 3
; CHECK-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; CHECK-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; CHECK-NEXT: [[B2:%.*]] = extractelement <4 x float> %b, i32 2
; CHECK-NEXT: [[B3:%.*]] = extractelement <4 x float> %b, i32 3
; CHECK-NEXT: [[CMP0:%.*]] = icmp ne i32 [[C0]], 0
; CHECK-NEXT: [[CMP1:%.*]] = icmp ne i32 [[C1]], 0
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x i32> undef, i32 [[C2]], i32 0
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[C3]], i32 1
; CHECK-NEXT: [[TMP3:%.*]] = icmp ne <2 x i32> [[TMP2]], zeroinitializer
; CHECK-NEXT: [[S0:%.*]] = select i1 [[CMP0]], float [[A0]], float [[B0]]
; CHECK-NEXT: [[S1:%.*]] = select i1 [[CMP1]], float [[A1]], float [[B1]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x float> undef, float [[A2]], i32 0
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x float> [[TMP4]], float [[A3]], i32 1
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x float> undef, float [[B2]], i32 0
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[B3]], i32 1
; CHECK-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP3]], <2 x float> [[TMP5]], <2 x float> [[TMP7]]
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[S0]], i32 0
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[S1]], i32 1
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x float> [[TMP8]], i32 0
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x float> undef, float [[TMP9]], i32 2
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x float> [[TMP8]], i32 1
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[TMP10]], i32 3
; CHECK-NEXT: ret <4 x float> [[RD]]
;
; ZEROTHRESH-LABEL: @simple_select_no_users(
; ZEROTHRESH-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; ZEROTHRESH-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; ZEROTHRESH-NEXT: [[C2:%.*]] = extractelement <4 x i32> %c, i32 2
; ZEROTHRESH-NEXT: [[C3:%.*]] = extractelement <4 x i32> %c, i32 3
; ZEROTHRESH-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; ZEROTHRESH-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; ZEROTHRESH-NEXT: [[A2:%.*]] = extractelement <4 x float> %a, i32 2
; ZEROTHRESH-NEXT: [[A3:%.*]] = extractelement <4 x float> %a, i32 3
; ZEROTHRESH-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; ZEROTHRESH-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; ZEROTHRESH-NEXT: [[B2:%.*]] = extractelement <4 x float> %b, i32 2
; ZEROTHRESH-NEXT: [[B3:%.*]] = extractelement <4 x float> %b, i32 3
; ZEROTHRESH-NEXT: [[CMP0:%.*]] = icmp ne i32 [[C0]], 0
; ZEROTHRESH-NEXT: [[CMP1:%.*]] = icmp ne i32 [[C1]], 0
; ZEROTHRESH-NEXT: [[CMP2:%.*]] = icmp ne i32 [[C2]], 0
; ZEROTHRESH-NEXT: [[CMP3:%.*]] = icmp ne i32 [[C3]], 0
; ZEROTHRESH-NEXT: [[S0:%.*]] = select i1 [[CMP0]], float [[A0]], float [[B0]]
; ZEROTHRESH-NEXT: [[S1:%.*]] = select i1 [[CMP1]], float [[A1]], float [[B1]]
; ZEROTHRESH-NEXT: [[S2:%.*]] = select i1 [[CMP2]], float [[A2]], float [[B2]]
; ZEROTHRESH-NEXT: [[S3:%.*]] = select i1 [[CMP3]], float [[A3]], float [[B3]]
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[S0]], i32 0
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[S1]], i32 1
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x float> undef, float [[S2]], i32 2
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x float> [[RC]], float [[S3]], i32 3
; ZEROTHRESH-NEXT: ret <4 x float> [[RD]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -183,6 +381,27 @@ define <4 x float> @simple_select_no_users(<4 x float> %a, <4 x float> %b, <4 x
; to do this backwards this backwards
define <4 x i32> @reconstruct(<4 x i32> %c) #0 {
; CHECK-LABEL: @reconstruct(
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <4 x i32> %c, i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x i32> undef, i32 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x i32> %c, i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x i32> [[RA]], i32 [[TMP2]], i32 1
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x i32> %c, i32 2
; CHECK-NEXT: [[RC:%.*]] = insertelement <4 x i32> [[RB]], i32 [[TMP3]], i32 2
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x i32> %c, i32 3
; CHECK-NEXT: [[RD:%.*]] = insertelement <4 x i32> [[RC]], i32 [[TMP4]], i32 3
; CHECK-NEXT: ret <4 x i32> [[RD]]
;
; ZEROTHRESH-LABEL: @reconstruct(
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = extractelement <4 x i32> %c, i32 0
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x i32> undef, i32 [[TMP1]], i32 0
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = extractelement <4 x i32> %c, i32 1
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x i32> [[RA]], i32 [[TMP2]], i32 1
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x i32> %c, i32 2
; ZEROTHRESH-NEXT: [[RC:%.*]] = insertelement <4 x i32> [[RB]], i32 [[TMP3]], i32 2
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x i32> %c, i32 3
; ZEROTHRESH-NEXT: [[RD:%.*]] = insertelement <4 x i32> [[RC]], i32 [[TMP4]], i32 3
; ZEROTHRESH-NEXT: ret <4 x i32> [[RD]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%c2 = extractelement <4 x i32> %c, i32 2
@ -196,8 +415,29 @@ define <4 x i32> @reconstruct(<4 x i32> %c) #0 {
define <2 x float> @simple_select_v2(<2 x float> %a, <2 x float> %b, <2 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_v2(
; CHECK: icmp ne <2 x i32>
; CHECK: select <2 x i1>
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne <2 x i32> %c, zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = select <2 x i1> [[TMP1]], <2 x float> %a, <2 x float> %b
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <2 x float> undef, float [[TMP3]], i32 0
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <2 x float> [[RA]], float [[TMP4]], i32 1
; CHECK-NEXT: ret <2 x float> [[RB]]
;
; ZEROTHRESH-LABEL: @simple_select_v2(
; ZEROTHRESH-NEXT: [[C0:%.*]] = extractelement <2 x i32> %c, i32 0
; ZEROTHRESH-NEXT: [[C1:%.*]] = extractelement <2 x i32> %c, i32 1
; ZEROTHRESH-NEXT: [[A0:%.*]] = extractelement <2 x float> %a, i32 0
; ZEROTHRESH-NEXT: [[A1:%.*]] = extractelement <2 x float> %a, i32 1
; ZEROTHRESH-NEXT: [[B0:%.*]] = extractelement <2 x float> %b, i32 0
; ZEROTHRESH-NEXT: [[B1:%.*]] = extractelement <2 x float> %b, i32 1
; ZEROTHRESH-NEXT: [[CMP0:%.*]] = icmp ne i32 [[C0]], 0
; ZEROTHRESH-NEXT: [[CMP1:%.*]] = icmp ne i32 [[C1]], 0
; ZEROTHRESH-NEXT: [[S0:%.*]] = select i1 [[CMP0]], float [[A0]], float [[B0]]
; ZEROTHRESH-NEXT: [[S1:%.*]] = select i1 [[CMP1]], float [[A1]], float [[B1]]
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <2 x float> undef, float [[S0]], i32 0
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <2 x float> [[RA]], float [[S1]], i32 1
; ZEROTHRESH-NEXT: ret <2 x float> [[RB]]
;
%c0 = extractelement <2 x i32> %c, i32 0
%c1 = extractelement <2 x i32> %c, i32 1
%a0 = extractelement <2 x float> %a, i32 0
@ -217,6 +457,48 @@ define <2 x float> @simple_select_v2(<2 x float> %a, <2 x float> %b, <2 x i32> %
; re-visiting the insertelement chains starting with undef
; (low cost threshold needed to force this to happen)
define <4 x float> @simple_select_partial_vector(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
; CHECK-LABEL: @simple_select_partial_vector(
; CHECK-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; CHECK-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; CHECK-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; CHECK-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x i32> undef, i32 [[C0]], i32 0
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[C1]], i32 1
; CHECK-NEXT: [[TMP3:%.*]] = icmp ne <2 x i32> [[TMP2]], zeroinitializer
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x float> undef, float [[A0]], i32 0
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x float> [[TMP4]], float [[A1]], i32 1
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x float> undef, float [[B0]], i32 0
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[B1]], i32 1
; CHECK-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP3]], <2 x float> [[TMP5]], <2 x float> [[TMP7]]
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x float> [[TMP8]], i32 0
; CHECK-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP9]], i32 0
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x float> [[TMP8]], i32 1
; CHECK-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP10]], i32 1
; CHECK-NEXT: ret <4 x float> [[RB]]
;
; ZEROTHRESH-LABEL: @simple_select_partial_vector(
; ZEROTHRESH-NEXT: [[C0:%.*]] = extractelement <4 x i32> %c, i32 0
; ZEROTHRESH-NEXT: [[C1:%.*]] = extractelement <4 x i32> %c, i32 1
; ZEROTHRESH-NEXT: [[A0:%.*]] = extractelement <4 x float> %a, i32 0
; ZEROTHRESH-NEXT: [[A1:%.*]] = extractelement <4 x float> %a, i32 1
; ZEROTHRESH-NEXT: [[B0:%.*]] = extractelement <4 x float> %b, i32 0
; ZEROTHRESH-NEXT: [[B1:%.*]] = extractelement <4 x float> %b, i32 1
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = insertelement <2 x i32> undef, i32 [[C0]], i32 0
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[C1]], i32 1
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = icmp ne <2 x i32> [[TMP2]], zeroinitializer
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = insertelement <2 x float> undef, float [[A0]], i32 0
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = insertelement <2 x float> [[TMP4]], float [[A1]], i32 1
; ZEROTHRESH-NEXT: [[TMP6:%.*]] = insertelement <2 x float> undef, float [[B0]], i32 0
; ZEROTHRESH-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[B1]], i32 1
; ZEROTHRESH-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP3]], <2 x float> [[TMP5]], <2 x float> [[TMP7]]
; ZEROTHRESH-NEXT: [[TMP9:%.*]] = extractelement <2 x float> [[TMP8]], i32 0
; ZEROTHRESH-NEXT: [[RA:%.*]] = insertelement <4 x float> undef, float [[TMP9]], i32 0
; ZEROTHRESH-NEXT: [[TMP10:%.*]] = extractelement <2 x float> [[TMP8]], i32 1
; ZEROTHRESH-NEXT: [[RB:%.*]] = insertelement <4 x float> [[RA]], float [[TMP10]], i32 1
; ZEROTHRESH-NEXT: ret <4 x float> [[RB]]
;
%c0 = extractelement <4 x i32> %c, i32 0
%c1 = extractelement <4 x i32> %c, i32 1
%a0 = extractelement <4 x float> %a, i32 0
@ -242,7 +524,29 @@ define <4 x float> @simple_select_partial_vector(<4 x float> %a, <4 x float> %b,
; must be rescheduled. The case here is from compiling Julia.
define <4 x float> @reschedule_extract(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: @reschedule_extract(
; CHECK: %1 = fadd <4 x float> %a, %b
; CHECK-NEXT: [[TMP1:%.*]] = fadd <4 x float> %a, %b
; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
; CHECK-NEXT: [[V0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
; CHECK-NEXT: [[V1:%.*]] = insertelement <4 x float> [[V0]], float [[TMP3]], i32 1
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
; CHECK-NEXT: [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[TMP4]], i32 2
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
; CHECK-NEXT: [[V3:%.*]] = insertelement <4 x float> [[V2]], float [[TMP5]], i32 3
; CHECK-NEXT: ret <4 x float> [[V3]]
;
; ZEROTHRESH-LABEL: @reschedule_extract(
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = fadd <4 x float> %a, %b
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
; ZEROTHRESH-NEXT: [[V0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
; ZEROTHRESH-NEXT: [[V1:%.*]] = insertelement <4 x float> [[V0]], float [[TMP3]], i32 1
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
; ZEROTHRESH-NEXT: [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[TMP4]], i32 2
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
; ZEROTHRESH-NEXT: [[V3:%.*]] = insertelement <4 x float> [[V2]], float [[TMP5]], i32 3
; ZEROTHRESH-NEXT: ret <4 x float> [[V3]]
;
%a0 = extractelement <4 x float> %a, i32 0
%b0 = extractelement <4 x float> %b, i32 0
%c0 = fadd float %a0, %b0
@ -265,8 +569,30 @@ define <4 x float> @reschedule_extract(<4 x float> %a, <4 x float> %b) {
; Check that cost model for vectorization takes credit for
; instructions that are erased.
define <4 x float> @take_credit(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: @take_credit(
; CHECK-NEXT: [[TMP1:%.*]] = fadd <4 x float> %a, %b
; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
; CHECK-NEXT: [[V0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
; CHECK-NEXT: [[V1:%.*]] = insertelement <4 x float> [[V0]], float [[TMP3]], i32 1
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
; CHECK-NEXT: [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[TMP4]], i32 2
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
; CHECK-NEXT: [[V3:%.*]] = insertelement <4 x float> [[V2]], float [[TMP5]], i32 3
; CHECK-NEXT: ret <4 x float> [[V3]]
;
; ZEROTHRESH-LABEL: @take_credit(
; ZEROTHRESH: %1 = fadd <4 x float> %a, %b
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = fadd <4 x float> %a, %b
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
; ZEROTHRESH-NEXT: [[V0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
; ZEROTHRESH-NEXT: [[V1:%.*]] = insertelement <4 x float> [[V0]], float [[TMP3]], i32 1
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP1]], i32 2
; ZEROTHRESH-NEXT: [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[TMP4]], i32 2
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
; ZEROTHRESH-NEXT: [[V3:%.*]] = insertelement <4 x float> [[V2]], float [[TMP5]], i32 3
; ZEROTHRESH-NEXT: ret <4 x float> [[V3]]
;
%a0 = extractelement <4 x float> %a, i32 0
%b0 = extractelement <4 x float> %b, i32 0
%c0 = fadd float %a0, %b0
@ -288,6 +614,46 @@ define <4 x float> @take_credit(<4 x float> %a, <4 x float> %b) {
; Make sure we handle multiple trees that feed one build vector correctly.
define <4 x double> @multi_tree(double %w, double %x, double %y, double %z) {
; CHECK-LABEL: @multi_tree(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double %w, i32 0
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double %x, i32 1
; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[TMP1]], <double 0.000000e+00, double 1.000000e+00>
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x double> undef, double %y, i32 0
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP3]], double %z, i32 1
; CHECK-NEXT: [[TMP5:%.*]] = fadd <2 x double> [[TMP4]], <double 2.000000e+00, double 3.000000e+00>
; CHECK-NEXT: [[TMP6:%.*]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, [[TMP2]]
; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x double> [[TMP6]], i32 0
; CHECK-NEXT: [[I1:%.*]] = insertelement <4 x double> undef, double [[TMP7]], i32 3
; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP6]], i32 1
; CHECK-NEXT: [[I2:%.*]] = insertelement <4 x double> [[I1]], double [[TMP8]], i32 2
; CHECK-NEXT: [[TMP9:%.*]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, [[TMP5]]
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP9]], i32 0
; CHECK-NEXT: [[I3:%.*]] = insertelement <4 x double> [[I2]], double [[TMP10]], i32 1
; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x double> [[TMP9]], i32 1
; CHECK-NEXT: [[I4:%.*]] = insertelement <4 x double> [[I3]], double [[TMP11]], i32 0
; CHECK-NEXT: ret <4 x double> [[I4]]
;
; ZEROTHRESH-LABEL: @multi_tree(
; ZEROTHRESH-NEXT: entry:
; ZEROTHRESH-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double %w, i32 0
; ZEROTHRESH-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double %x, i32 1
; ZEROTHRESH-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[TMP1]], <double 0.000000e+00, double 1.000000e+00>
; ZEROTHRESH-NEXT: [[TMP3:%.*]] = insertelement <2 x double> undef, double %y, i32 0
; ZEROTHRESH-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP3]], double %z, i32 1
; ZEROTHRESH-NEXT: [[TMP5:%.*]] = fadd <2 x double> [[TMP4]], <double 2.000000e+00, double 3.000000e+00>
; ZEROTHRESH-NEXT: [[TMP6:%.*]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, [[TMP2]]
; ZEROTHRESH-NEXT: [[TMP7:%.*]] = extractelement <2 x double> [[TMP6]], i32 0
; ZEROTHRESH-NEXT: [[I1:%.*]] = insertelement <4 x double> undef, double [[TMP7]], i32 3
; ZEROTHRESH-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP6]], i32 1
; ZEROTHRESH-NEXT: [[I2:%.*]] = insertelement <4 x double> [[I1]], double [[TMP8]], i32 2
; ZEROTHRESH-NEXT: [[TMP9:%.*]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, [[TMP5]]
; ZEROTHRESH-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP9]], i32 0
; ZEROTHRESH-NEXT: [[I3:%.*]] = insertelement <4 x double> [[I2]], double [[TMP10]], i32 1
; ZEROTHRESH-NEXT: [[TMP11:%.*]] = extractelement <2 x double> [[TMP9]], i32 1
; ZEROTHRESH-NEXT: [[I4:%.*]] = insertelement <4 x double> [[I3]], double [[TMP11]], i32 0
; ZEROTHRESH-NEXT: ret <4 x double> [[I4]]
;
entry:
%t0 = fadd double %w , 0.000000e+00
%t1 = fadd double %x , 1.000000e+00
@ -303,23 +669,132 @@ entry:
%i4 = insertelement <4 x double> %i3, double %t7, i32 0
ret <4 x double> %i4
}
; CHECK-LABEL: @multi_tree
; CHECK-DAG: %[[V0:.+]] = insertelement <2 x double> undef, double %w, i32 0
; CHECK-DAG: %[[V1:.+]] = insertelement <2 x double> %[[V0]], double %x, i32 1
; CHECK-DAG: %[[V2:.+]] = fadd <2 x double> %[[V1]], <double 0.000000e+00, double 1.000000e+00>
; CHECK-DAG: %[[V3:.+]] = insertelement <2 x double> undef, double %y, i32 0
; CHECK-DAG: %[[V4:.+]] = insertelement <2 x double> %[[V3]], double %z, i32 1
; CHECK-DAG: %[[V5:.+]] = fadd <2 x double> %[[V4]], <double 2.000000e+00, double 3.000000e+00>
; CHECK-DAG: %[[V6:.+]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, %[[V2]]
; CHECK-DAG: %[[V7:.+]] = extractelement <2 x double> %[[V6]], i32 0
; CHECK-DAG: %[[I1:.+]] = insertelement <4 x double> undef, double %[[V7]], i32 3
; CHECK-DAG: %[[V8:.+]] = extractelement <2 x double> %[[V6]], i32 1
; CHECK-DAG: %[[I2:.+]] = insertelement <4 x double> %[[I1]], double %[[V8]], i32 2
; CHECK-DAG: %[[V9:.+]] = fmul <2 x double> <double 1.000000e+00, double 1.000000e+00>, %[[V5]]
; CHECK-DAG: %[[V10:.+]] = extractelement <2 x double> %[[V9]], i32 0
; CHECK-DAG: %[[I3:.+]] = insertelement <4 x double> %i2, double %[[V10]], i32 1
; CHECK-DAG: %[[V11:.+]] = extractelement <2 x double> %[[V9]], i32 1
; CHECK-DAG: %[[I4:.+]] = insertelement <4 x double> %i3, double %[[V11]], i32 0
; CHECK: ret <4 x double> %[[I4]]
define <8 x float> @_vadd256(<8 x float> %a, <8 x float> %b) local_unnamed_addr #0 {
; CHECK-LABEL: @_vadd256(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[VECEXT:%.*]] = extractelement <8 x float> %a, i32 0
; CHECK-NEXT: [[VECEXT1:%.*]] = extractelement <8 x float> %b, i32 0
; CHECK-NEXT: [[VECEXT2:%.*]] = extractelement <8 x float> %a, i32 1
; CHECK-NEXT: [[VECEXT3:%.*]] = extractelement <8 x float> %b, i32 1
; CHECK-NEXT: [[VECEXT5:%.*]] = extractelement <8 x float> %a, i32 2
; CHECK-NEXT: [[VECEXT6:%.*]] = extractelement <8 x float> %b, i32 2
; CHECK-NEXT: [[VECEXT8:%.*]] = extractelement <8 x float> %a, i32 3
; CHECK-NEXT: [[VECEXT9:%.*]] = extractelement <8 x float> %b, i32 3
; CHECK-NEXT: [[TMP0:%.*]] = insertelement <4 x float> undef, float [[VECEXT]], i32 0
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x float> [[TMP0]], float [[VECEXT2]], i32 1
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x float> [[TMP1]], float [[VECEXT5]], i32 2
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x float> [[TMP2]], float [[VECEXT8]], i32 3
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x float> undef, float [[VECEXT1]], i32 0
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x float> [[TMP4]], float [[VECEXT3]], i32 1
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x float> [[TMP5]], float [[VECEXT6]], i32 2
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x float> [[TMP6]], float [[VECEXT9]], i32 3
; CHECK-NEXT: [[TMP8:%.*]] = fadd <4 x float> [[TMP3]], [[TMP7]]
; CHECK-NEXT: [[VECEXT11:%.*]] = extractelement <8 x float> %a, i32 4
; CHECK-NEXT: [[VECEXT12:%.*]] = extractelement <8 x float> %b, i32 4
; CHECK-NEXT: [[VECEXT14:%.*]] = extractelement <8 x float> %a, i32 5
; CHECK-NEXT: [[VECEXT15:%.*]] = extractelement <8 x float> %b, i32 5
; CHECK-NEXT: [[VECEXT17:%.*]] = extractelement <8 x float> %a, i32 6
; CHECK-NEXT: [[VECEXT18:%.*]] = extractelement <8 x float> %b, i32 6
; CHECK-NEXT: [[VECEXT20:%.*]] = extractelement <8 x float> %a, i32 7
; CHECK-NEXT: [[VECEXT21:%.*]] = extractelement <8 x float> %b, i32 7
; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x float> undef, float [[VECEXT11]], i32 0
; CHECK-NEXT: [[TMP10:%.*]] = insertelement <4 x float> [[TMP9]], float [[VECEXT14]], i32 1
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x float> [[TMP10]], float [[VECEXT17]], i32 2
; CHECK-NEXT: [[TMP12:%.*]] = insertelement <4 x float> [[TMP11]], float [[VECEXT20]], i32 3
; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x float> undef, float [[VECEXT12]], i32 0
; CHECK-NEXT: [[TMP14:%.*]] = insertelement <4 x float> [[TMP13]], float [[VECEXT15]], i32 1
; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x float> [[TMP14]], float [[VECEXT18]], i32 2
; CHECK-NEXT: [[TMP16:%.*]] = insertelement <4 x float> [[TMP15]], float [[VECEXT21]], i32 3
; CHECK-NEXT: [[TMP17:%.*]] = fadd <4 x float> [[TMP12]], [[TMP16]]
; CHECK-NEXT: [[TMP18:%.*]] = extractelement <4 x float> [[TMP8]], i32 0
; CHECK-NEXT: [[VECINIT_I:%.*]] = insertelement <8 x float> undef, float [[TMP18]], i32 0
; CHECK-NEXT: [[TMP19:%.*]] = extractelement <4 x float> [[TMP8]], i32 1
; CHECK-NEXT: [[VECINIT1_I:%.*]] = insertelement <8 x float> [[VECINIT_I]], float [[TMP19]], i32 1
; CHECK-NEXT: [[TMP20:%.*]] = extractelement <4 x float> [[TMP8]], i32 2
; CHECK-NEXT: [[VECINIT2_I:%.*]] = insertelement <8 x float> [[VECINIT1_I]], float [[TMP20]], i32 2
; CHECK-NEXT: [[TMP21:%.*]] = extractelement <4 x float> [[TMP8]], i32 3
; CHECK-NEXT: [[VECINIT3_I:%.*]] = insertelement <8 x float> [[VECINIT2_I]], float [[TMP21]], i32 3
; CHECK-NEXT: [[TMP22:%.*]] = extractelement <4 x float> [[TMP17]], i32 0
; CHECK-NEXT: [[VECINIT4_I:%.*]] = insertelement <8 x float> [[VECINIT3_I]], float [[TMP22]], i32 4
; CHECK-NEXT: [[TMP23:%.*]] = extractelement <4 x float> [[TMP17]], i32 1
; CHECK-NEXT: [[VECINIT5_I:%.*]] = insertelement <8 x float> [[VECINIT4_I]], float [[TMP23]], i32 5
; CHECK-NEXT: [[TMP24:%.*]] = extractelement <4 x float> [[TMP17]], i32 2
; CHECK-NEXT: [[VECINIT6_I:%.*]] = insertelement <8 x float> [[VECINIT5_I]], float [[TMP24]], i32 6
; CHECK-NEXT: [[TMP25:%.*]] = extractelement <4 x float> [[TMP17]], i32 3
; CHECK-NEXT: [[VECINIT7_I:%.*]] = insertelement <8 x float> [[VECINIT6_I]], float [[TMP25]], i32 7
; CHECK-NEXT: ret <8 x float> [[VECINIT7_I]]
;
; ZEROTHRESH-LABEL: @_vadd256(
; ZEROTHRESH-NEXT: entry:
; ZEROTHRESH-NEXT: [[VECEXT:%.*]] = extractelement <8 x float> %a, i32 0
; ZEROTHRESH-NEXT: [[VECEXT1:%.*]] = extractelement <8 x float> %b, i32 0
; ZEROTHRESH-NEXT: [[ADD:%.*]] = fadd float [[VECEXT]], [[VECEXT1]]
; ZEROTHRESH-NEXT: [[VECEXT2:%.*]] = extractelement <8 x float> %a, i32 1
; ZEROTHRESH-NEXT: [[VECEXT3:%.*]] = extractelement <8 x float> %b, i32 1
; ZEROTHRESH-NEXT: [[ADD4:%.*]] = fadd float [[VECEXT2]], [[VECEXT3]]
; ZEROTHRESH-NEXT: [[VECEXT5:%.*]] = extractelement <8 x float> %a, i32 2
; ZEROTHRESH-NEXT: [[VECEXT6:%.*]] = extractelement <8 x float> %b, i32 2
; ZEROTHRESH-NEXT: [[ADD7:%.*]] = fadd float [[VECEXT5]], [[VECEXT6]]
; ZEROTHRESH-NEXT: [[VECEXT8:%.*]] = extractelement <8 x float> %a, i32 3
; ZEROTHRESH-NEXT: [[VECEXT9:%.*]] = extractelement <8 x float> %b, i32 3
; ZEROTHRESH-NEXT: [[ADD10:%.*]] = fadd float [[VECEXT8]], [[VECEXT9]]
; ZEROTHRESH-NEXT: [[VECEXT11:%.*]] = extractelement <8 x float> %a, i32 4
; ZEROTHRESH-NEXT: [[VECEXT12:%.*]] = extractelement <8 x float> %b, i32 4
; ZEROTHRESH-NEXT: [[ADD13:%.*]] = fadd float [[VECEXT11]], [[VECEXT12]]
; ZEROTHRESH-NEXT: [[VECEXT14:%.*]] = extractelement <8 x float> %a, i32 5
; ZEROTHRESH-NEXT: [[VECEXT15:%.*]] = extractelement <8 x float> %b, i32 5
; ZEROTHRESH-NEXT: [[ADD16:%.*]] = fadd float [[VECEXT14]], [[VECEXT15]]
; ZEROTHRESH-NEXT: [[VECEXT17:%.*]] = extractelement <8 x float> %a, i32 6
; ZEROTHRESH-NEXT: [[VECEXT18:%.*]] = extractelement <8 x float> %b, i32 6
; ZEROTHRESH-NEXT: [[ADD19:%.*]] = fadd float [[VECEXT17]], [[VECEXT18]]
; ZEROTHRESH-NEXT: [[VECEXT20:%.*]] = extractelement <8 x float> %a, i32 7
; ZEROTHRESH-NEXT: [[VECEXT21:%.*]] = extractelement <8 x float> %b, i32 7
; ZEROTHRESH-NEXT: [[ADD22:%.*]] = fadd float [[VECEXT20]], [[VECEXT21]]
; ZEROTHRESH-NEXT: [[VECINIT_I:%.*]] = insertelement <8 x float> undef, float [[ADD]], i32 0
; ZEROTHRESH-NEXT: [[VECINIT1_I:%.*]] = insertelement <8 x float> [[VECINIT_I]], float [[ADD4]], i32 1
; ZEROTHRESH-NEXT: [[VECINIT2_I:%.*]] = insertelement <8 x float> [[VECINIT1_I]], float [[ADD7]], i32 2
; ZEROTHRESH-NEXT: [[VECINIT3_I:%.*]] = insertelement <8 x float> [[VECINIT2_I]], float [[ADD10]], i32 3
; ZEROTHRESH-NEXT: [[VECINIT4_I:%.*]] = insertelement <8 x float> [[VECINIT3_I]], float [[ADD13]], i32 4
; ZEROTHRESH-NEXT: [[VECINIT5_I:%.*]] = insertelement <8 x float> [[VECINIT4_I]], float [[ADD16]], i32 5
; ZEROTHRESH-NEXT: [[VECINIT6_I:%.*]] = insertelement <8 x float> [[VECINIT5_I]], float [[ADD19]], i32 6
; ZEROTHRESH-NEXT: [[VECINIT7_I:%.*]] = insertelement <8 x float> [[VECINIT6_I]], float [[ADD22]], i32 7
; ZEROTHRESH-NEXT: ret <8 x float> [[VECINIT7_I]]
;
entry:
%vecext = extractelement <8 x float> %a, i32 0
%vecext1 = extractelement <8 x float> %b, i32 0
%add = fadd float %vecext, %vecext1
%vecext2 = extractelement <8 x float> %a, i32 1
%vecext3 = extractelement <8 x float> %b, i32 1
%add4 = fadd float %vecext2, %vecext3
%vecext5 = extractelement <8 x float> %a, i32 2
%vecext6 = extractelement <8 x float> %b, i32 2
%add7 = fadd float %vecext5, %vecext6
%vecext8 = extractelement <8 x float> %a, i32 3
%vecext9 = extractelement <8 x float> %b, i32 3
%add10 = fadd float %vecext8, %vecext9
%vecext11 = extractelement <8 x float> %a, i32 4
%vecext12 = extractelement <8 x float> %b, i32 4
%add13 = fadd float %vecext11, %vecext12
%vecext14 = extractelement <8 x float> %a, i32 5
%vecext15 = extractelement <8 x float> %b, i32 5
%add16 = fadd float %vecext14, %vecext15
%vecext17 = extractelement <8 x float> %a, i32 6
%vecext18 = extractelement <8 x float> %b, i32 6
%add19 = fadd float %vecext17, %vecext18
%vecext20 = extractelement <8 x float> %a, i32 7
%vecext21 = extractelement <8 x float> %b, i32 7
%add22 = fadd float %vecext20, %vecext21
%vecinit.i = insertelement <8 x float> undef, float %add, i32 0
%vecinit1.i = insertelement <8 x float> %vecinit.i, float %add4, i32 1
%vecinit2.i = insertelement <8 x float> %vecinit1.i, float %add7, i32 2
%vecinit3.i = insertelement <8 x float> %vecinit2.i, float %add10, i32 3
%vecinit4.i = insertelement <8 x float> %vecinit3.i, float %add13, i32 4
%vecinit5.i = insertelement <8 x float> %vecinit4.i, float %add16, i32 5
%vecinit6.i = insertelement <8 x float> %vecinit5.i, float %add19, i32 6
%vecinit7.i = insertelement <8 x float> %vecinit6.i, float %add22, i32 7
ret <8 x float> %vecinit7.i
}
attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"="true" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }