replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Turn on -addr-sink-using-gep by default. 

The new codepath has been in the tree for years, and there isn't any
reason to use two codepaths here.

Differential Revision: https://reviews.llvm.org/D30596

llvm-svn: 299723
This commit is contained in:
Eli Friedman 2017-04-06 22:42:18 +00:00
parent 5eb8ea8872
commit 5fba1e53f2
18 changed files with 66 additions and 93 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llvm/lib/CodeGen/CodeGenPrepare.cpp
View File

@ -96,7 +96,7 @@ static cl::opt<bool> DisableSelectToBranch(
cl::desc("Disable select to branch conversion."));
static cl::opt<bool> AddrSinkUsingGEPs(
"addr-sink-using-gep", cl::Hidden, cl::init(false),
"addr-sink-using-gep", cl::Hidden, cl::init(true),
cl::desc("Address sinking in CGP using GEPs."));
static cl::opt<bool> EnableAndCmpSinking(

8
llvm/test/CodeGen/AArch64/aarch64-gep-opt.ll
View File

@ -96,9 +96,13 @@ exit:
; CHECK-NoAA: add i64 [[TMP:%[a-zA-Z0-9]+]], 528
; CHECK-NoAA: add i64 [[TMP]], 532
; CHECK-NoAA: if.true:
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = add i64 [[TMP]], 532
; CHECK-NoAA: inttoptr
; CHECK-NoAA: bitcast
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = getelementptr i8, {{.*}}, i64 532
; CHECK-NoAA: exit:
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = add i64 [[TMP]], 528
; CHECK-NoAA: inttoptr
; CHECK-NoAA: bitcast
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = getelementptr i8, {{.*}}, i64 528
; CHECK-UseAA-LABEL: test_GEP_across_BB(
; CHECK-UseAA: [[PTR0:%[a-zA-Z0-9]+]] = getelementptr

12
llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
View File

@ -36,9 +36,9 @@ done:
; OPT-CI-NOT: getelementptr
; OPT: br i1
; OPT-CI: ptrtoint
; OPT-CI: add
; OPT-CI: inttoptr
; OPT-CI: addrspacecast
; OPT-CI: getelementptr
; OPT-CI: bitcast
; OPT: br label
; GCN-LABEL: {{^}}test_sink_noop_addrspacecast_flat_to_global_i32:
@ -69,9 +69,9 @@ done:
; OPT-CI-NOT: getelementptr
; OPT: br i1
; OPT-CI: ptrtoint
; OPT-CI: add
; OPT-CI: inttoptr
; OPT-CI: addrspacecast
; OPT-CI: getelementptr
; OPT-CI: bitcast
; OPT: br label
; GCN-LABEL: {{^}}test_sink_noop_addrspacecast_flat_to_constant_i32:

43
llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
View File

@ -11,7 +11,7 @@ target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:
; OPT-CI-NOT: getelementptr i32, i32 addrspace(1)* %in
; OPT-VI: getelementptr i32, i32 addrspace(1)* %in
; OPT: br i1
; OPT-CI: ptrtoint
; OPT-CI: getelementptr i8,
; GCN-LABEL: {{^}}test_sink_global_small_offset_i32:
; GCN: {{^}}BB0_2:
@ -124,7 +124,7 @@ done:
; OPT-LABEL: @test_sink_scratch_small_offset_i32(
; OPT-NOT: getelementptr [512 x i32]
; OPT: br i1
; OPT: ptrtoint
; OPT: getelementptr i8,
; GCN-LABEL: {{^}}test_sink_scratch_small_offset_i32:
; GCN: s_and_saveexec_b64
@ -162,7 +162,7 @@ done:
; OPT-LABEL: @test_sink_scratch_small_offset_i32_reserved(
; OPT-NOT: getelementptr [512 x i32]
; OPT: br i1
; OPT: ptrtoint
; OPT: getelementptr i8,
; GCN-LABEL: {{^}}test_sink_scratch_small_offset_i32_reserved:
; GCN: s_and_saveexec_b64
@ -488,7 +488,7 @@ done:
%struct.foo = type { [3 x float], [3 x float] }
; OPT-LABEL: @sink_ds_address(
; OPT: ptrtoint %struct.foo addrspace(3)* %ptr to i32
; OPT: getelementptr i8,
; GCN-LABEL: {{^}}sink_ds_address:
; GCN: s_load_dword [[SREG1:s[0-9]+]],
@ -519,8 +519,7 @@ bb34:
; OPT-LABEL: @test_sink_constant_small_max_mubuf_offset_load_i32_align_1(
; OPT: br i1 %tmp0,
; OPT: if:
; OPT: %sunkaddr = ptrtoint i8 addrspace(2)* %in to i64
; OPT: %sunkaddr1 = add i64 %sunkaddr, 4095
; OPT: getelementptr i8, {{.*}} 4095
define amdgpu_kernel void @test_sink_constant_small_max_mubuf_offset_load_i32_align_1(i32 addrspace(1)* %out, i8 addrspace(2)* %in) {
entry:
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i32 1024
@ -544,10 +543,10 @@ done:
}
; OPT-LABEL: @test_sink_local_small_offset_atomicrmw_i32(
; OPT: %sunkaddr = ptrtoint i32 addrspace(3)* %in to i32
; OPT: %sunkaddr1 = add i32 %sunkaddr, 28
; OPT: %sunkaddr2 = inttoptr i32 %sunkaddr1 to i32 addrspace(3)*
; OPT: %tmp1 = atomicrmw add i32 addrspace(3)* %sunkaddr2, i32 2 seq_cst
; OPT: %0 = bitcast i32 addrspace(3)* %in to i8 addrspace(3)*
; OPT: %sunkaddr = getelementptr i8, i8 addrspace(3)* %0, i32 28
; OPT: %1 = bitcast i8 addrspace(3)* %sunkaddr to i32 addrspace(3)*
; OPT: %tmp1 = atomicrmw add i32 addrspace(3)* %1, i32 2 seq_cst
define amdgpu_kernel void @test_sink_local_small_offset_atomicrmw_i32(i32 addrspace(3)* %out, i32 addrspace(3)* %in) {
entry:
%out.gep = getelementptr i32, i32 addrspace(3)* %out, i32 999999
@ -570,10 +569,10 @@ done:
}
; OPT-LABEL: @test_sink_local_small_offset_cmpxchg_i32(
; OPT: %sunkaddr = ptrtoint i32 addrspace(3)* %in to i32
; OPT: %sunkaddr1 = add i32 %sunkaddr, 28
; OPT: %sunkaddr2 = inttoptr i32 %sunkaddr1 to i32 addrspace(3)*
; OPT: %tmp1.struct = cmpxchg i32 addrspace(3)* %sunkaddr2, i32 undef, i32 2 seq_cst monotonic
; OPT: %0 = bitcast i32 addrspace(3)* %in to i8 addrspace(3)*
; OPT: %sunkaddr = getelementptr i8, i8 addrspace(3)* %0, i32 28
; OPT: %1 = bitcast i8 addrspace(3)* %sunkaddr to i32 addrspace(3)*
; OPT: %tmp1.struct = cmpxchg i32 addrspace(3)* %1, i32 undef, i32 2 seq_cst monotonic
define amdgpu_kernel void @test_sink_local_small_offset_cmpxchg_i32(i32 addrspace(3)* %out, i32 addrspace(3)* %in) {
entry:
%out.gep = getelementptr i32, i32 addrspace(3)* %out, i32 999999
@ -623,10 +622,10 @@ done:
}
; OPT-LABEL: @test_sink_local_small_offset_atomic_inc_i32(
; OPT: %sunkaddr = ptrtoint i32 addrspace(3)* %in to i32
; OPT: %sunkaddr1 = add i32 %sunkaddr, 28
; OPT: %sunkaddr2 = inttoptr i32 %sunkaddr1 to i32 addrspace(3)*
; OPT: %tmp1 = call i32 @llvm.amdgcn.atomic.inc.i32.p3i32(i32 addrspace(3)* %sunkaddr2, i32 2, i32 0, i32 0, i1 false)
; OPT: %0 = bitcast i32 addrspace(3)* %in to i8 addrspace(3)*
; OPT: %sunkaddr = getelementptr i8, i8 addrspace(3)* %0, i32 28
; OPT: %1 = bitcast i8 addrspace(3)* %sunkaddr to i32 addrspace(3)*
; OPT: %tmp1 = call i32 @llvm.amdgcn.atomic.inc.i32.p3i32(i32 addrspace(3)* %1, i32 2, i32 0, i32 0, i1 false)
define amdgpu_kernel void @test_sink_local_small_offset_atomic_inc_i32(i32 addrspace(3)* %out, i32 addrspace(3)* %in) {
entry:
%out.gep = getelementptr i32, i32 addrspace(3)* %out, i32 999999
@ -649,10 +648,10 @@ done:
}
; OPT-LABEL: @test_sink_local_small_offset_atomic_dec_i32(
; OPT: %sunkaddr = ptrtoint i32 addrspace(3)* %in to i32
; OPT: %sunkaddr1 = add i32 %sunkaddr, 28
; OPT: %sunkaddr2 = inttoptr i32 %sunkaddr1 to i32 addrspace(3)*
; OPT: %tmp1 = call i32 @llvm.amdgcn.atomic.dec.i32.p3i32(i32 addrspace(3)* %sunkaddr2, i32 2, i32 0, i32 0, i1 false)
; OPT: %0 = bitcast i32 addrspace(3)* %in to i8 addrspace(3)*
; OPT: %sunkaddr = getelementptr i8, i8 addrspace(3)* %0, i32 28
; OPT: %1 = bitcast i8 addrspace(3)* %sunkaddr to i32 addrspace(3)*
; OPT: %tmp1 = call i32 @llvm.amdgcn.atomic.dec.i32.p3i32(i32 addrspace(3)* %1, i32 2, i32 0, i32 0, i1 false)
define amdgpu_kernel void @test_sink_local_small_offset_atomic_dec_i32(i32 addrspace(3)* %out, i32 addrspace(3)* %in) {
entry:
%out.gep = getelementptr i32, i32 addrspace(3)* %out, i32 999999

1
llvm/test/CodeGen/ARM/phi.ll
View File

@ -1,5 +1,4 @@
; RUN: llc -mtriple=arm-eabi -mattr=+v4t %s -o - | FileCheck %s
; RUN: llc -mtriple=arm-eabi -mattr=+v4t -addr-sink-using-gep=1 %s -o - | FileCheck %s
; <rdar://problem/8686347>

1
llvm/test/CodeGen/PowerPC/2007-11-16-landingpad-split.ll
View File

@ -1,5 +1,4 @@
; RUN: llc -mcpu=g5 < %s | FileCheck %s
; RUN: llc -mcpu=g5 -addr-sink-using-gep=1 < %s | FileCheck %s
;; Formerly crashed, see PR 1508
target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f128:64:128"
target triple = "powerpc64-apple-darwin8"

4
llvm/test/CodeGen/PowerPC/ppc64-gep-opt.ll
View File

@ -84,9 +84,9 @@ exit:
; CHECK-NoAA: add i64 [[TMP:%[a-zA-Z0-9]+]], 528
; CHECK-NoAA: add i64 [[TMP]], 532
; CHECK-NoAA: if.true:
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = add i64 [[TMP]], 532
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = getelementptr i8, i8* {{.*}}, i64 532
; CHECK-NoAA: exit:
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = add i64 [[TMP]], 528
; CHECK-NoAA: {{%sunk[a-zA-Z0-9]+}} = getelementptr i8, i8* {{.*}}, i64 528
; CHECK-UseAA-LABEL: test_GEP_across_BB(
; CHECK-UseAA: [[PTR0:%[a-zA-Z0-9]+]] = getelementptr

1
llvm/test/CodeGen/X86/2007-03-15-GEP-Idx-Sink.ll
View File

@ -1,5 +1,4 @@
; RUN: llc < %s -march=x86 -mtriple=i686-darwin | FileCheck %s
; RUN: llc < %s -march=x86 -mtriple=i686-darwin -addr-sink-using-gep=1 | FileCheck %s
define void @foo(i8** %buf, i32 %size, i32 %col, i8* %p) nounwind {
entry:

1
llvm/test/CodeGen/X86/MergeConsecutiveStores.ll
View File

@ -1,6 +1,5 @@
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx -fixup-byte-word-insts=1 < %s | FileCheck -check-prefix=CHECK -check-prefix=BWON %s
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx -fixup-byte-word-insts=0 < %s | FileCheck -check-prefix=CHECK -check-prefix=BWOFF %s
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx -addr-sink-using-gep=1 < %s | FileCheck -check-prefix=CHECK -check-prefix=BWON %s
%struct.A = type { i8, i8, i8, i8, i8, i8, i8, i8 }
%struct.B = type { i32, i32, i32, i32, i32, i32, i32, i32 }

41
llvm/test/CodeGen/X86/codegen-prepare-addrmode-sext.ll
View File

@ -1,5 +1,4 @@
; RUN: opt -S -codegenprepare %s -o - | FileCheck %s
; RUN: opt -S -codegenprepare -addr-sink-using-gep=1 %s -o - | FileCheck -check-prefix=CHECK-GEP %s
; This file tests the different cases what are involved when codegen prepare
; tries to get sign/zero extension out of the way of addressing mode.
; This tests require an actual target as addressing mode decisions depends
@ -309,33 +308,18 @@ define i8 @twoArgsNoPromotionRemove(i1 %arg1, i8 %arg2, i8* %base) {
; CHECK: [[ADD:%[a-zA-Z_0-9-]+]] = add nsw i32 [[SHL]], %arg2
; CHECK: [[SEXTADD:%[a-zA-Z_0-9-]+]] = sext i32 [[ADD]] to i64
; BB then
; CHECK: [[BASE1:%[a-zA-Z_0-9-]+]] = add i64 [[SEXTADD]], 48
; CHECK: [[ADDR1:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[BASE1]] to i32*
; CHECK: [[BASE1:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[SEXTADD]] to i32*
; CHECK: [[BCC1:%[a-zA-Z_0-9-]+]] = bitcast i32* [[BASE1]] to i8*
; CHECK: [[FULL1:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[BCC1]], i64 48
; CHECK: [[ADDR1:%[a-zA-Z_0-9-]+]] = bitcast i8* [[FULL1]] to i32*
; CHECK: load i32, i32* [[ADDR1]]
; BB else
; CHECK: [[BASE2:%[a-zA-Z_0-9-]+]] = add i64 [[SEXTADD]], 48
; CHECK: [[ADDR2:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[BASE2]] to i32*
; CHECK: [[BASE2:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[SEXTADD]] to i32*
; CHECK: [[BCC2:%[a-zA-Z_0-9-]+]] = bitcast i32* [[BASE2]] to i8*
; CHECK: [[FULL2:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[BCC2]], i64 48
; CHECK: [[ADDR2:%[a-zA-Z_0-9-]+]] = bitcast i8* [[FULL2]] to i32*
; CHECK: load i32, i32* [[ADDR2]]
; CHECK: ret
; CHECK-GEP-LABEL: @checkProfitability
; CHECK-GEP-NOT: {{%[a-zA-Z_0-9-]+}} = sext i32 %arg1 to i64
; CHECK-GEP-NOT: {{%[a-zA-Z_0-9-]+}} = sext i32 %arg2 to i64
; CHECK-GEP: [[SHL:%[a-zA-Z_0-9-]+]] = shl nsw i32 %arg1, 1
; CHECK-GEP: [[ADD:%[a-zA-Z_0-9-]+]] = add nsw i32 [[SHL]], %arg2
; CHECK-GEP: [[SEXTADD:%[a-zA-Z_0-9-]+]] = sext i32 [[ADD]] to i64
; BB then
; CHECK-GEP: [[BASE1:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[SEXTADD]] to i32*
; CHECK-GEP: [[BCC1:%[a-zA-Z_0-9-]+]] = bitcast i32* [[BASE1]] to i8*
; CHECK-GEP: [[FULL1:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[BCC1]], i64 48
; CHECK-GEP: [[ADDR1:%[a-zA-Z_0-9-]+]] = bitcast i8* [[FULL1]] to i32*
; CHECK-GEP: load i32, i32* [[ADDR1]]
; BB else
; CHECK-GEP: [[BASE2:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[SEXTADD]] to i32*
; CHECK-GEP: [[BCC2:%[a-zA-Z_0-9-]+]] = bitcast i32* [[BASE2]] to i8*
; CHECK-GEP: [[FULL2:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[BCC2]], i64 48
; CHECK-GEP: [[ADDR2:%[a-zA-Z_0-9-]+]] = bitcast i8* [[FULL2]] to i32*
; CHECK-GEP: load i32, i32* [[ADDR2]]
; CHECK-GEP: ret
define i32 @checkProfitability(i32 %arg1, i32 %arg2, i1 %test) {
%shl = shl nsw i32 %arg1, 1
%add1 = add nsw i32 %shl, %arg2
@ -371,11 +355,10 @@ end:
; Use it at the starting point for the matching.
; CHECK: %conv.i = zext i16 [[PLAIN_OPND:%[.a-zA-Z_0-9-]+]] to i32
; CHECK-NEXT: [[PROMOTED_CONV:%[.a-zA-Z_0-9-]+]] = zext i16 [[PLAIN_OPND]] to i64
; CHECK-NEXT: [[BASE:%[a-zA-Z_0-9-]+]] = ptrtoint %struct.dns_packet* %P to i64
; CHECK-NEXT: [[ADD:%[a-zA-Z_0-9-]+]] = add i64 [[BASE]], [[PROMOTED_CONV]]
; CHECK-NEXT: [[ADDR:%[a-zA-Z_0-9-]+]] = add i64 [[ADD]], 7
; CHECK-NEXT: [[CAST:%[a-zA-Z_0-9-]+]] = inttoptr i64 [[ADDR]] to i8*
; CHECK-NEXT: load i8, i8* [[CAST]], align 1
; CHECK-NEXT: [[BASE:%[a-zA-Z_0-9-]+]] = bitcast %struct.dns_packet* %P to i8*
; CHECK-NEXT: [[ADD:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[BASE]], i64 [[PROMOTED_CONV]]
; CHECK-NEXT: [[ADDR:%[a-zA-Z_0-9-]+]] = getelementptr i8, i8* [[ADD]], i64 7
; CHECK-NEXT: load i8, i8* [[ADDR]], align 1
define signext i16 @fn3(%struct.dns_packet* nocapture readonly %P) {
entry:
%tmp = getelementptr inbounds %struct.dns_packet, %struct.dns_packet* %P, i64 0, i32 2

1
llvm/test/CodeGen/X86/codegen-prepare.ll
View File

@ -1,5 +1,4 @@
; RUN: llc < %s -mtriple=x86_64-pc-linux | FileCheck %s
; RUN: llc < %s -mtriple=x86_64-pc-linux -addr-sink-using-gep=1 | FileCheck %s
; Check that the CodeGenPrepare Pass
; does not wrongly rewrite the address computed by Instruction %4

1
llvm/test/CodeGen/X86/isel-sink.ll
View File

@ -1,5 +1,4 @@
; RUN: llc < %s -march=x86 | FileCheck %s
; RUN: llc < %s -march=x86 -addr-sink-using-gep=1 | FileCheck %s
define i32 @test(i32* %X, i32 %B) {
; CHECK-LABEL: test:

1
llvm/test/CodeGen/X86/merge_store.ll
View File

@ -1,5 +1,4 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -addr-sink-using-gep=1 | FileCheck %s
define void @merge_store(i32* nocapture %a) {
; CHECK-LABEL: merge_store:

8
llvm/test/Transforms/CodeGenPrepare/X86/computedgoto.ll
View File

@ -218,10 +218,10 @@ define void @nophi(i32* %p) {
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: br label [[INDIRECTGOTO]]
; CHECK: indirectgoto:
; CHECK-NEXT: [[SUNKADDR:%.*]] = ptrtoint i32* [[P]] to i64
; CHECK-NEXT: [[SUNKADDR1:%.*]] = add i64 [[SUNKADDR]], 4
; CHECK-NEXT: [[SUNKADDR2:%.*]] = inttoptr i64 [[SUNKADDR1]] to i32*
; CHECK-NEXT: [[NEWP:%.*]] = load i32, i32* [[SUNKADDR2]], align 4
; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[P]] to i8*
; CHECK-NEXT: [[SUNKADDR:%.*]] = getelementptr i8, i8* [[TMP0]], i64 4
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i8* [[SUNKADDR]] to i32*
; CHECK-NEXT: [[NEWP:%.*]] = load i32, i32* [[TMP1]], align 4
; CHECK-NEXT: [[IDX:%.*]] = sext i32 [[NEWP]] to i64
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* @nophi.targets, i64 0, i64 [[IDX]]
; CHECK-NEXT: [[NEWOP:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8

26
llvm/test/Transforms/CodeGenPrepare/X86/sink-addrmode.ll
View File

@ -7,7 +7,7 @@ target triple = "x86_64-unknown-linux-gnu"
; Can we sink single addressing mode computation to use?
define void @test1(i1 %cond, i64* %base) {
; CHECK-LABEL: @test1
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
entry:
%addr = getelementptr inbounds i64, i64* %base, i64 5
%casted = bitcast i64* %addr to i32*
@ -33,7 +33,7 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
%cmp = icmp eq i32 %v1, 0
@ -41,7 +41,7 @@ if.then:
next:
; CHECK-LABEL: next:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v2 = load i32, i32* %casted, align 4
call void @foo(i32 %v2)
br label %fallthrough
@ -61,10 +61,10 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
; CHECK-NOT: add i64 {{.+}}, 40
; CHECK-NOT: getelementptr i8, {{.+}}, 40
%v2 = load i32, i32* %casted, align 4
call void @foo(i32 %v2)
br label %fallthrough
@ -84,7 +84,7 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
%cmp = icmp eq i32 %v1, 0
@ -95,7 +95,7 @@ fallthrough:
rare.1:
; CHECK-LABEL: rare.1:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
call void @slowpath(i32 %v1, i32* %casted) cold
br label %fallthrough
}
@ -111,7 +111,7 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK-NOT: add i64 {{.+}}, 40
; CHECK-NOT: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
%cmp = icmp eq i32 %v1, 0
@ -136,7 +136,7 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK-NOT: add i64 {{.+}}, 40
; CHECK-NOT: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
%cmp = icmp eq i32 %v1, 0
@ -162,7 +162,7 @@ entry:
if.then:
; CHECK-LABEL: if.then:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v1 = load i32, i32* %casted, align 4
call void @foo(i32 %v1)
%cmp = icmp eq i32 %v1, 0
@ -170,7 +170,7 @@ if.then:
next:
; CHECK-LABEL: next:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
%v2 = load i32, i32* %casted, align 4
call void @foo(i32 %v2)
%cmp2 = icmp eq i32 %v2, 0
@ -181,13 +181,13 @@ fallthrough:
rare.1:
; CHECK-LABEL: rare.1:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
call void @slowpath(i32 %v1, i32* %casted) cold
br label %next
rare.2:
; CHECK-LABEL: rare.2:
; CHECK: add i64 {{.+}}, 40
; CHECK: getelementptr i8, {{.+}} 40
call void @slowpath(i32 %v2, i32* %casted) cold
br label %fallthrough
}

5
llvm/test/Transforms/CodeGenPrepare/X86/sink-addrspacecast.ll
View File

@ -1,12 +1,10 @@
; RUN: opt -S -codegenprepare < %s | FileCheck %s -check-prefix=CHECK -check-prefix=INT
; RUN: opt -S -codegenprepare -addr-sink-using-gep=true < %s | FileCheck %s -check-prefix=CHECK -check-prefix=GEP
; RUN: opt -S -codegenprepare < %s | FileCheck %s -check-prefix=CHECK -check-prefix=GEP
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-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"
; CHECK-LABEL: @load_cast_gep
; INT: add i64 %sunkaddr, 40
; GEP: [[CAST:%[0-9]+]] = addrspacecast i64* %base to i8 addrspace(1)*
; GEP: getelementptr i8, i8 addrspace(1)* [[CAST]], i64 40
define void @load_cast_gep(i1 %cond, i64* %base) {
@ -24,7 +22,6 @@ fallthrough:
}
; CHECK-LABEL: @store_gep_cast
; INT: add i64 %sunkaddr, 20
; GEP: [[CAST:%[0-9]+]] = addrspacecast i64* %base to i8 addrspace(1)*
; GEP: getelementptr i8, i8 addrspace(1)* [[CAST]], i64 20
define void @store_gep_cast(i1 %cond, i64* %base) {

1
llvm/test/Transforms/LoopStrengthReduce/ARM/ivchain-ARM.ll
View File

@ -1,5 +1,4 @@
; RUN: llc -O3 -mtriple=thumb-eabi -mcpu=cortex-a9 %s -o - | FileCheck %s -check-prefix=A9
; RUN: llc -O3 -mtriple=thumb-eabi -mcpu=cortex-a9 -addr-sink-using-gep=1 %s -o - | FileCheck %s -check-prefix=A9
; @simple is the most basic chain of address induction variables. Chaining
; saves at least one register and avoids complex addressing and setup

2
llvm/test/Transforms/LoopStrengthReduce/X86/ivchain-X86.ll
View File

@ -1,7 +1,5 @@
; RUN: llc < %s -O3 -march=x86-64 -mcpu=core2 | FileCheck %s -check-prefix=X64
; RUN: llc < %s -O3 -march=x86 -mcpu=core2 | FileCheck %s -check-prefix=X32
; RUN: llc < %s -O3 -march=x86-64 -mcpu=core2 -addr-sink-using-gep=1 | FileCheck %s -check-prefix=X64
; RUN: llc < %s -O3 -march=x86 -mcpu=core2 -addr-sink-using-gep=1 | FileCheck %s -check-prefix=X32
; @simple is the most basic chain of address induction variables. Chaining
; saves at least one register and avoids complex addressing and setup