mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 19:23:23 +01:00
75fa314f27
Summary: If the first insertelement instruction has multiple users and inserts at position 0, we can re-use this instruction when folding a chain of insertelement instructions. As we need to generate the first insertelement instruction anyways, this should be a strict improvement. We could get rid of the restriction of inserting at position 0 by creating a different shufflemask, but it is probably worth to keep the first insertelement instruction with position 0, as this is easier to do efficiently than at other positions I think. Reviewers: grosser, mkuper, fpetrogalli, efriedma Reviewed By: fpetrogalli Subscribers: gareevroman, llvm-commits Differential Revision: https://reviews.llvm.org/D37064 llvm-svn: 312110
138 lines
5.4 KiB
LLVM
138 lines
5.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt -instcombine -S < %s | FileCheck %s
|
|
|
|
; CHECK-LABEL: good1
|
|
; CHECK: %[[INS:.*]] = insertelement <4 x float> undef, float %arg, i32 0
|
|
; CHECK-NEXT: %[[BCAST:.*]] = shufflevector <4 x float> %[[INS]], <4 x float> undef, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: ret <4 x float> %[[BCAST]]
|
|
define <4 x float> @good1(float %arg) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: good2
|
|
; CHECK: %[[INS:.*]] = insertelement <4 x float> undef, float %arg, i32 0
|
|
; CHECK-NEXT: %[[BCAST:.*]] = shufflevector <4 x float> %[[INS]], <4 x float> undef, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: ret <4 x float> %[[BCAST]]
|
|
define <4 x float> @good2(float %arg) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 1
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 2
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 0
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: good3
|
|
; CHECK: %[[INS:.*]] = insertelement <4 x float> undef, float %arg, i32 0
|
|
; CHECK-NEXT: %[[BCAST:.*]] = shufflevector <4 x float> %[[INS]], <4 x float> undef, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: ret <4 x float> %[[BCAST]]
|
|
define <4 x float> @good3(float %arg) {
|
|
%tmp = insertelement <4 x float> zeroinitializer, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: good4
|
|
; CHECK: %[[INS:.*]] = insertelement <4 x float> undef, float %arg, i32 0
|
|
; CHECK-NEXT: %[[ADD:.*]] = fadd <4 x float> %[[INS]], %[[INS]]
|
|
; CHECK-NEXT: %[[BCAST:.*]] = shufflevector <4 x float> %[[ADD]], <4 x float> undef, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: ret <4 x float> %[[BCAST]]
|
|
define <4 x float> @good4(float %arg) {
|
|
%tmp = insertelement <4 x float> zeroinitializer, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
%tmp7 = fadd <4 x float> %tmp6, %tmp6
|
|
ret <4 x float> %tmp7
|
|
}
|
|
|
|
; CHECK-LABEL: @good5(
|
|
; CHECK-NEXT: %ins1 = insertelement <4 x float> undef, float %v, i32 0
|
|
; CHECK-NEXT: %a1 = fadd <4 x float> %ins1, %ins1
|
|
; CHECK-NEXT: %ins4 = shufflevector <4 x float> %ins1, <4 x float> undef, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: %res = fadd <4 x float> %a1, %ins4
|
|
; CHECK-NEXT: ret <4 x float> %res
|
|
define <4 x float> @good5(float %v) {
|
|
%ins1 = insertelement <4 x float> undef, float %v, i32 0
|
|
%a1 = fadd <4 x float> %ins1, %ins1
|
|
%ins2 = insertelement<4 x float> %ins1, float %v, i32 1
|
|
%ins3 = insertelement<4 x float> %ins2, float %v, i32 2
|
|
%ins4 = insertelement<4 x float> %ins3, float %v, i32 3
|
|
%res = fadd <4 x float> %a1, %ins4
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
; CHECK-LABEL: bad1
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad1(float %arg) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 1
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: bad2
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad2(float %arg) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 0
|
|
%tmp5 = insertelement <4 x float> %tmp, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: bad3
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad3(float %arg, float %arg2) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg2, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: bad4
|
|
; CHECK-NOT: shufflevector
|
|
define <1 x float> @bad4(float %arg) {
|
|
%tmp = insertelement <1 x float> undef, float %arg, i32 0
|
|
ret <1 x float> %tmp
|
|
}
|
|
|
|
; CHECK-LABEL: bad5
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad5(float %arg) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 2
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
%tmp7 = fadd <4 x float> %tmp6, %tmp4
|
|
ret <4 x float> %tmp7
|
|
}
|
|
|
|
; CHECK-LABEL: bad6
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad6(float %arg, i32 %k) {
|
|
%tmp = insertelement <4 x float> undef, float %arg, i32 0
|
|
%tmp4 = insertelement <4 x float> %tmp, float %arg, i32 1
|
|
%tmp5 = insertelement <4 x float> %tmp4, float %arg, i32 %k
|
|
%tmp6 = insertelement <4 x float> %tmp5, float %arg, i32 3
|
|
ret <4 x float> %tmp6
|
|
}
|
|
|
|
; CHECK-LABEL: @bad7(
|
|
; CHECK-NOT: shufflevector
|
|
define <4 x float> @bad7(float %v) {
|
|
%ins1 = insertelement <4 x float> undef, float %v, i32 1
|
|
%a1 = fadd <4 x float> %ins1, %ins1
|
|
%ins2 = insertelement<4 x float> %ins1, float %v, i32 2
|
|
%ins3 = insertelement<4 x float> %ins2, float %v, i32 3
|
|
%ins4 = insertelement<4 x float> %ins3, float %v, i32 0
|
|
%res = fadd <4 x float> %a1, %ins4
|
|
ret <4 x float> %res
|
|
}
|