[llvm] r271553 - [PowerPC] Run reg2mem on tests to simplify them.
Geoff Berry via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 2 11:02:50 PDT 2016
Author: gberry
Date: Thu Jun 2 13:02:50 2016
New Revision: 271553
URL: http://llvm.org/viewvc/llvm-project?rev=271553&view=rev
Log:
[PowerPC] Run reg2mem on tests to simplify them.
Summary:
Also convert test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll to use
FileCheck instead of two grep and count runs.
This change is needed to avoid spurious diffs in these tests when
EarlyCSE is improved to use MemorySSA and can do more load elimination.
Reviewers: hfinkel
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D20238
Modified:
llvm/trunk/test/CodeGen/PowerPC/builtins-ppc-p8vector.ll
llvm/trunk/test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll
Modified: llvm/trunk/test/CodeGen/PowerPC/builtins-ppc-p8vector.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/builtins-ppc-p8vector.ll?rev=271553&r1=271552&r2=271553&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/builtins-ppc-p8vector.ll (original)
+++ llvm/trunk/test/CodeGen/PowerPC/builtins-ppc-p8vector.ll Thu Jun 2 13:02:50 2016
@@ -4,7 +4,9 @@
; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 < %s | FileCheck %s -check-prefix=CHECK-VSX
@vsc = global <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, align 16
+ at vsc2 = global <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, align 16
@vuc = global <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, align 16
+ at vuc2 = global <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5>, align 16
@res_vll = common global <2 x i64> zeroinitializer, align 16
@res_vull = common global <2 x i64> zeroinitializer, align 16
@res_vsc = common global <16 x i8> zeroinitializer, align 16
@@ -13,54 +15,39 @@
; Function Attrs: nounwind
define void @test1() {
entry:
- %__a.addr.i = alloca <16 x i8>, align 16
- %__b.addr.i = alloca <16 x i8>, align 16
%0 = load <16 x i8>, <16 x i8>* @vsc, align 16
- %1 = load <16 x i8>, <16 x i8>* @vsc, align 16
- store <16 x i8> %0, <16 x i8>* %__a.addr.i, align 16
- store <16 x i8> %1, <16 x i8>* %__b.addr.i, align 16
- %2 = load <16 x i8>, <16 x i8>* %__a.addr.i, align 16
- %3 = load <16 x i8>, <16 x i8>* %__b.addr.i, align 16
- %4 = call <2 x i64> @llvm.ppc.altivec.vbpermq(<16 x i8> %2, <16 x i8> %3)
- store <2 x i64> %4, <2 x i64>* @res_vll, align 16
+ %1 = load <16 x i8>, <16 x i8>* @vsc2, align 16
+ %2 = call <2 x i64> @llvm.ppc.altivec.vbpermq(<16 x i8> %0, <16 x i8> %1)
+ store <2 x i64> %2, <2 x i64>* @res_vll, align 16
ret void
; CHECK-LABEL: @test1
-; CHECK: lvx [[REG1:[0-9]+]],
-; CHECK: lvx [[REG2:[0-9]+]],
-; CHECK: vbpermq {{[0-9]+}}, [[REG2]], [[REG1]]
+; CHECK: lvx [[REG1:[0-9]+]], 0, 3
+; CHECK: lvx [[REG2:[0-9]+]], 0, 4
+; CHECK: vbpermq {{[0-9]+}}, [[REG1]], [[REG2]]
; CHECK-VSX: vbpermq {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
}
; Function Attrs: nounwind
define void @test2() {
entry:
- %__a.addr.i = alloca <16 x i8>, align 16
- %__b.addr.i = alloca <16 x i8>, align 16
%0 = load <16 x i8>, <16 x i8>* @vuc, align 16
- %1 = load <16 x i8>, <16 x i8>* @vuc, align 16
- store <16 x i8> %0, <16 x i8>* %__a.addr.i, align 16
- store <16 x i8> %1, <16 x i8>* %__b.addr.i, align 16
- %2 = load <16 x i8>, <16 x i8>* %__a.addr.i, align 16
- %3 = load <16 x i8>, <16 x i8>* %__b.addr.i, align 16
- %4 = call <2 x i64> @llvm.ppc.altivec.vbpermq(<16 x i8> %2, <16 x i8> %3)
- store <2 x i64> %4, <2 x i64>* @res_vull, align 16
+ %1 = load <16 x i8>, <16 x i8>* @vuc2, align 16
+ %2 = call <2 x i64> @llvm.ppc.altivec.vbpermq(<16 x i8> %0, <16 x i8> %1)
+ store <2 x i64> %2, <2 x i64>* @res_vull, align 16
ret void
; CHECK-LABEL: @test2
-; CHECK: lvx [[REG1:[0-9]+]],
-; CHECK: lvx [[REG2:[0-9]+]],
-; CHECK: vbpermq {{[0-9]+}}, [[REG2]], [[REG1]]
+; CHECK: lvx [[REG1:[0-9]+]], 0, 3
+; CHECK: lvx [[REG2:[0-9]+]], 0, 4
+; CHECK: vbpermq {{[0-9]+}}, [[REG1]], [[REG2]]
; CHECK-VSX: vbpermq {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
}
; Function Attrs: nounwind
define void @test3() {
entry:
- %__a.addr.i = alloca <16 x i8>, align 16
%0 = load <16 x i8>, <16 x i8>* @vsc, align 16
- store <16 x i8> %0, <16 x i8>* %__a.addr.i, align 16
- %1 = load <16 x i8>, <16 x i8>* %__a.addr.i, align 16
- %2 = call <16 x i8> @llvm.ppc.altivec.vgbbd(<16 x i8> %1)
- store <16 x i8> %2, <16 x i8>* @res_vsc, align 16
+ %1 = call <16 x i8> @llvm.ppc.altivec.vgbbd(<16 x i8> %0)
+ store <16 x i8> %1, <16 x i8>* @res_vsc, align 16
ret void
; CHECK-LABEL: @test3
; CHECK: lvx [[REG1:[0-9]+]],
@@ -71,12 +58,9 @@ entry:
; Function Attrs: nounwind
define void @test4() {
entry:
- %__a.addr.i = alloca <16 x i8>, align 16
%0 = load <16 x i8>, <16 x i8>* @vuc, align 16
- store <16 x i8> %0, <16 x i8>* %__a.addr.i, align 16
- %1 = load <16 x i8>, <16 x i8>* %__a.addr.i, align 16
- %2 = call <16 x i8> @llvm.ppc.altivec.vgbbd(<16 x i8> %1)
- store <16 x i8> %2, <16 x i8>* @res_vuc, align 16
+ %1 = call <16 x i8> @llvm.ppc.altivec.vgbbd(<16 x i8> %0)
+ store <16 x i8> %1, <16 x i8>* @res_vuc, align 16
ret void
; CHECK-LABEL: @test4
; CHECK: lvx [[REG1:[0-9]+]],
Modified: llvm/trunk/test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll?rev=271553&r1=271552&r2=271553&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll (original)
+++ llvm/trunk/test/CodeGen/PowerPC/vsx-ldst-builtin-le.ll Thu Jun 2 13:02:50 2016
@@ -1,6 +1,4 @@
-; RUN: llc -mcpu=pwr8 -mattr=+vsx -O2 -mtriple=powerpc64le-unknown-linux-gnu < %s > %t
-; RUN: grep lxvd2x < %t | count 18
-; RUN: grep stxvd2x < %t | count 18
+; RUN: llc -mcpu=pwr8 -mattr=+vsx -O2 -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
@vf = global <4 x float> <float -1.500000e+00, float 2.500000e+00, float -3.500000e+00, float 4.500000e+00>, align 16
@vd = global <2 x double> <double 3.500000e+00, double -7.500000e+00>, align 16
@@ -18,150 +16,60 @@
define void @test1() {
entry:
; CHECK-LABEL: test1
- %__a.addr.i31 = alloca i32, align 4
- %__b.addr.i32 = alloca <4 x i32>*, align 8
- %__a.addr.i29 = alloca i32, align 4
- %__b.addr.i30 = alloca <4 x float>*, align 8
- %__a.addr.i27 = alloca i32, align 4
- %__b.addr.i28 = alloca <2 x i64>*, align 8
- %__a.addr.i25 = alloca i32, align 4
- %__b.addr.i26 = alloca <2 x i64>*, align 8
- %__a.addr.i23 = alloca i32, align 4
- %__b.addr.i24 = alloca <2 x double>*, align 8
- %__a.addr.i20 = alloca <4 x i32>, align 16
- %__b.addr.i21 = alloca i32, align 4
- %__c.addr.i22 = alloca <4 x i32>*, align 8
- %__a.addr.i17 = alloca <4 x i32>, align 16
- %__b.addr.i18 = alloca i32, align 4
- %__c.addr.i19 = alloca <4 x i32>*, align 8
- %__a.addr.i14 = alloca <4 x float>, align 16
- %__b.addr.i15 = alloca i32, align 4
- %__c.addr.i16 = alloca <4 x float>*, align 8
- %__a.addr.i11 = alloca <2 x i64>, align 16
- %__b.addr.i12 = alloca i32, align 4
- %__c.addr.i13 = alloca <2 x i64>*, align 8
- %__a.addr.i8 = alloca <2 x i64>, align 16
- %__b.addr.i9 = alloca i32, align 4
- %__c.addr.i10 = alloca <2 x i64>*, align 8
- %__a.addr.i6 = alloca <2 x double>, align 16
- %__b.addr.i7 = alloca i32, align 4
- %__c.addr.i = alloca <2 x double>*, align 8
- %__a.addr.i = alloca i32, align 4
- %__b.addr.i = alloca <4 x i32>*, align 8
- store i32 0, i32* %__a.addr.i, align 4
- store <4 x i32>* @vsi, <4 x i32>** %__b.addr.i, align 8
- %0 = load i32, i32* %__a.addr.i, align 4
- %1 = load <4 x i32>*, <4 x i32>** %__b.addr.i, align 8
- %2 = bitcast <4 x i32>* %1 to i8*
- %3 = getelementptr i8, i8* %2, i32 %0
- %4 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* %3)
- store <4 x i32> %4, <4 x i32>* @res_vsi, align 16
- store i32 0, i32* %__a.addr.i31, align 4
- store <4 x i32>* @vui, <4 x i32>** %__b.addr.i32, align 8
- %5 = load i32, i32* %__a.addr.i31, align 4
- %6 = load <4 x i32>*, <4 x i32>** %__b.addr.i32, align 8
- %7 = bitcast <4 x i32>* %6 to i8*
- %8 = getelementptr i8, i8* %7, i32 %5
- %9 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* %8)
- store <4 x i32> %9, <4 x i32>* @res_vui, align 16
- store i32 0, i32* %__a.addr.i29, align 4
- store <4 x float>* @vf, <4 x float>** %__b.addr.i30, align 8
- %10 = load i32, i32* %__a.addr.i29, align 4
- %11 = load <4 x float>*, <4 x float>** %__b.addr.i30, align 8
- %12 = bitcast <4 x float>* %11 to i8*
- %13 = getelementptr i8, i8* %12, i32 %10
- %14 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* %13)
- %15 = bitcast <4 x i32> %14 to <4 x float>
- store <4 x float> %15, <4 x float>* @res_vf, align 16
- store i32 0, i32* %__a.addr.i27, align 4
- store <2 x i64>* @vsll, <2 x i64>** %__b.addr.i28, align 8
- %16 = load i32, i32* %__a.addr.i27, align 4
- %17 = load <2 x i64>*, <2 x i64>** %__b.addr.i28, align 8
- %18 = bitcast <2 x i64>* %17 to i8*
- %19 = getelementptr i8, i8* %18, i32 %16
- %20 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* %19)
- %21 = bitcast <2 x double> %20 to <2 x i64>
- store <2 x i64> %21, <2 x i64>* @res_vsll, align 16
- store i32 0, i32* %__a.addr.i25, align 4
- store <2 x i64>* @vull, <2 x i64>** %__b.addr.i26, align 8
- %22 = load i32, i32* %__a.addr.i25, align 4
- %23 = load <2 x i64>*, <2 x i64>** %__b.addr.i26, align 8
- %24 = bitcast <2 x i64>* %23 to i8*
- %25 = getelementptr i8, i8* %24, i32 %22
- %26 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* %25)
- %27 = bitcast <2 x double> %26 to <2 x i64>
- store <2 x i64> %27, <2 x i64>* @res_vull, align 16
- store i32 0, i32* %__a.addr.i23, align 4
- store <2 x double>* @vd, <2 x double>** %__b.addr.i24, align 8
- %28 = load i32, i32* %__a.addr.i23, align 4
- %29 = load <2 x double>*, <2 x double>** %__b.addr.i24, align 8
- %30 = bitcast <2 x double>* %29 to i8*
- %31 = getelementptr i8, i8* %30, i32 %28
- %32 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* %31)
- store <2 x double> %32, <2 x double>* @res_vd, align 16
- %33 = load <4 x i32>, <4 x i32>* @vsi, align 16
- store <4 x i32> %33, <4 x i32>* %__a.addr.i20, align 16
- store i32 0, i32* %__b.addr.i21, align 4
- store <4 x i32>* @res_vsi, <4 x i32>** %__c.addr.i22, align 8
- %34 = load <4 x i32>, <4 x i32>* %__a.addr.i20, align 16
- %35 = load i32, i32* %__b.addr.i21, align 4
- %36 = load <4 x i32>*, <4 x i32>** %__c.addr.i22, align 8
- %37 = bitcast <4 x i32>* %36 to i8*
- %38 = getelementptr i8, i8* %37, i32 %35
- call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %34, i8* %38)
- %39 = load <4 x i32>, <4 x i32>* @vui, align 16
- store <4 x i32> %39, <4 x i32>* %__a.addr.i17, align 16
- store i32 0, i32* %__b.addr.i18, align 4
- store <4 x i32>* @res_vui, <4 x i32>** %__c.addr.i19, align 8
- %40 = load <4 x i32>, <4 x i32>* %__a.addr.i17, align 16
- %41 = load i32, i32* %__b.addr.i18, align 4
- %42 = load <4 x i32>*, <4 x i32>** %__c.addr.i19, align 8
- %43 = bitcast <4 x i32>* %42 to i8*
- %44 = getelementptr i8, i8* %43, i32 %41
- call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %40, i8* %44)
- %45 = load <4 x float>, <4 x float>* @vf, align 16
- store <4 x float> %45, <4 x float>* %__a.addr.i14, align 16
- store i32 0, i32* %__b.addr.i15, align 4
- store <4 x float>* @res_vf, <4 x float>** %__c.addr.i16, align 8
- %46 = load <4 x float>, <4 x float>* %__a.addr.i14, align 16
- %47 = bitcast <4 x float> %46 to <4 x i32>
- %48 = load i32, i32* %__b.addr.i15, align 4
- %49 = load <4 x float>*, <4 x float>** %__c.addr.i16, align 8
- %50 = bitcast <4 x float>* %49 to i8*
- %51 = getelementptr i8, i8* %50, i32 %48
- call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %47, i8* %51) #1
- %52 = load <2 x i64>, <2 x i64>* @vsll, align 16
- store <2 x i64> %52, <2 x i64>* %__a.addr.i11, align 16
- store i32 0, i32* %__b.addr.i12, align 4
- store <2 x i64>* @res_vsll, <2 x i64>** %__c.addr.i13, align 8
- %53 = load <2 x i64>, <2 x i64>* %__a.addr.i11, align 16
- %54 = bitcast <2 x i64> %53 to <2 x double>
- %55 = load i32, i32* %__b.addr.i12, align 4
- %56 = load <2 x i64>*, <2 x i64>** %__c.addr.i13, align 8
- %57 = bitcast <2 x i64>* %56 to i8*
- %58 = getelementptr i8, i8* %57, i32 %55
- call void @llvm.ppc.vsx.stxvd2x(<2 x double> %54, i8* %58)
- %59 = load <2 x i64>, <2 x i64>* @vull, align 16
- store <2 x i64> %59, <2 x i64>* %__a.addr.i8, align 16
- store i32 0, i32* %__b.addr.i9, align 4
- store <2 x i64>* @res_vull, <2 x i64>** %__c.addr.i10, align 8
- %60 = load <2 x i64>, <2 x i64>* %__a.addr.i8, align 16
- %61 = bitcast <2 x i64> %60 to <2 x double>
- %62 = load i32, i32* %__b.addr.i9, align 4
- %63 = load <2 x i64>*, <2 x i64>** %__c.addr.i10, align 8
- %64 = bitcast <2 x i64>* %63 to i8*
- %65 = getelementptr i8, i8* %64, i32 %62
- call void @llvm.ppc.vsx.stxvd2x(<2 x double> %61, i8* %65)
- %66 = load <2 x double>, <2 x double>* @vd, align 16
- store <2 x double> %66, <2 x double>* %__a.addr.i6, align 16
- store i32 0, i32* %__b.addr.i7, align 4
- store <2 x double>* @res_vd, <2 x double>** %__c.addr.i, align 8
- %67 = load <2 x double>, <2 x double>* %__a.addr.i6, align 16
- %68 = load i32, i32* %__b.addr.i7, align 4
- %69 = load <2 x double>*, <2 x double>** %__c.addr.i, align 8
- %70 = bitcast <2 x double>* %69 to i8*
- %71 = getelementptr i8, i8* %70, i32 %68
- call void @llvm.ppc.vsx.stxvd2x(<2 x double> %67, i8* %71)
+; CHECK: lxvd2x
+ %0 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* bitcast (<4 x i32>* @vsi to i8*))
+; CHECK: stxvd2x
+ store <4 x i32> %0, <4 x i32>* @res_vsi, align 16
+; CHECK: lxvd2x
+ %1 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* bitcast (<4 x i32>* @vui to i8*))
+; CHECK: stxvd2x
+ store <4 x i32> %1, <4 x i32>* @res_vui, align 16
+; CHECK: lxvd2x
+ %2 = call <4 x i32> @llvm.ppc.vsx.lxvw4x(i8* bitcast (<4 x float>* @vf to i8*))
+ %3 = bitcast <4 x i32> %2 to <4 x float>
+; CHECK: stxvd2x
+ store <4 x float> %3, <4 x float>* @res_vf, align 16
+; CHECK: lxvd2x
+ %4 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* bitcast (<2 x i64>* @vsll to i8*))
+ %5 = bitcast <2 x double> %4 to <2 x i64>
+; CHECK: stxvd2x
+ store <2 x i64> %5, <2 x i64>* @res_vsll, align 16
+; CHECK: lxvd2x
+ %6 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* bitcast (<2 x i64>* @vull to i8*))
+ %7 = bitcast <2 x double> %6 to <2 x i64>
+; CHECK: stxvd2x
+ store <2 x i64> %7, <2 x i64>* @res_vull, align 16
+; CHECK: lxvd2x
+ %8 = call <2 x double> @llvm.ppc.vsx.lxvd2x(i8* bitcast (<2 x double>* @vd to i8*))
+; CHECK: stxvd2x
+ store <2 x double> %8, <2 x double>* @res_vd, align 16
+; CHECK: lxvd2x
+ %9 = load <4 x i32>, <4 x i32>* @vsi, align 16
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %9, i8* bitcast (<4 x i32>* @res_vsi to i8*))
+; CHECK: lxvd2x
+ %10 = load <4 x i32>, <4 x i32>* @vui, align 16
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %10, i8* bitcast (<4 x i32>* @res_vui to i8*))
+; CHECK: lxvd2x
+ %11 = load <4 x float>, <4 x float>* @vf, align 16
+ %12 = bitcast <4 x float> %11 to <4 x i32>
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvw4x(<4 x i32> %12, i8* bitcast (<4 x float>* @res_vf to i8*))
+; CHECK: lxvd2x
+ %13 = load <2 x i64>, <2 x i64>* @vsll, align 16
+ %14 = bitcast <2 x i64> %13 to <2 x double>
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvd2x(<2 x double> %14, i8* bitcast (<2 x i64>* @res_vsll to i8*))
+; CHECK: lxvd2x
+ %15 = load <2 x i64>, <2 x i64>* @vull, align 16
+ %16 = bitcast <2 x i64> %15 to <2 x double>
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvd2x(<2 x double> %16, i8* bitcast (<2 x i64>* @res_vull to i8*))
+; CHECK: lxvd2x
+ %17 = load <2 x double>, <2 x double>* @vd, align 16
+; CHECK: stxvd2x
+ call void @llvm.ppc.vsx.stxvd2x(<2 x double> %17, i8* bitcast (<2 x double>* @res_vd to i8*))
ret void
}
More information about the llvm-commits
mailing list