2011-01-01 19:54:22 +00:00
|
|
|
; RUN: opt -basicaa -loop-idiom < %s -S | FileCheck %s
|
implement enough of the memset inference algorithm to recognize and insert
memsets. This is still missing one important validity check, but this is enough
to compile stuff like this:
void test0(std::vector<char> &X) {
for (std::vector<char>::iterator I = X.begin(), E = X.end(); I != E; ++I)
*I = 0;
}
void test1(std::vector<int> &X) {
for (long i = 0, e = X.size(); i != e; ++i)
X[i] = 0x01010101;
}
With:
$ clang t.cpp -S -o - -O2 -emit-llvm | opt -loop-idiom | opt -O3 | llc
to:
__Z5test0RSt6vectorIcSaIcEE: ## @_Z5test0RSt6vectorIcSaIcEE
## BB#0: ## %entry
subq $8, %rsp
movq (%rdi), %rax
movq 8(%rdi), %rsi
cmpq %rsi, %rax
je LBB0_2
## BB#1: ## %bb.nph
subq %rax, %rsi
movq %rax, %rdi
callq ___bzero
LBB0_2: ## %for.end
addq $8, %rsp
ret
...
__Z5test1RSt6vectorIiSaIiEE: ## @_Z5test1RSt6vectorIiSaIiEE
## BB#0: ## %entry
subq $8, %rsp
movq (%rdi), %rax
movq 8(%rdi), %rdx
subq %rax, %rdx
cmpq $4, %rdx
jb LBB1_2
## BB#1: ## %for.body.preheader
andq $-4, %rdx
movl $1, %esi
movq %rax, %rdi
callq _memset
LBB1_2: ## %for.end
addq $8, %rsp
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@122573 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-26 23:42:51 +00:00
|
|
|
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-n8:16:32:64"
|
|
|
|
|
target triple = "x86_64-apple-darwin10.0.0"
|
|
|
|
|
|
|
|
|
|
define void @test1(i8* %Base, i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
store i8 0, i8* %I.0.014, align 1
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test1
|
|
|
|
|
; CHECK: call void @llvm.memset.p0i8.i64(i8* %Base, i8 0, i64 %Size, i32 1, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-02 07:58:36 +00:00
|
|
|
; This is a loop that was rotated but where the blocks weren't merged. This
|
|
|
|
|
; shouldn't perturb us.
|
|
|
|
|
define void @test1a(i8* %Base, i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body.cont ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
store i8 0, i8* %I.0.014, align 1
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
br label %for.body.cont
|
|
|
|
|
for.body.cont:
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test1a
|
|
|
|
|
; CHECK: call void @llvm.memset.p0i8.i64(i8* %Base, i8 0, i64 %Size, i32 1, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
implement enough of the memset inference algorithm to recognize and insert
memsets. This is still missing one important validity check, but this is enough
to compile stuff like this:
void test0(std::vector<char> &X) {
for (std::vector<char>::iterator I = X.begin(), E = X.end(); I != E; ++I)
*I = 0;
}
void test1(std::vector<int> &X) {
for (long i = 0, e = X.size(); i != e; ++i)
X[i] = 0x01010101;
}
With:
$ clang t.cpp -S -o - -O2 -emit-llvm | opt -loop-idiom | opt -O3 | llc
to:
__Z5test0RSt6vectorIcSaIcEE: ## @_Z5test0RSt6vectorIcSaIcEE
## BB#0: ## %entry
subq $8, %rsp
movq (%rdi), %rax
movq 8(%rdi), %rsi
cmpq %rsi, %rax
je LBB0_2
## BB#1: ## %bb.nph
subq %rax, %rsi
movq %rax, %rdi
callq ___bzero
LBB0_2: ## %for.end
addq $8, %rsp
ret
...
__Z5test1RSt6vectorIiSaIiEE: ## @_Z5test1RSt6vectorIiSaIiEE
## BB#0: ## %entry
subq $8, %rsp
movq (%rdi), %rax
movq 8(%rdi), %rdx
subq %rax, %rdx
cmpq $4, %rdx
jb LBB1_2
## BB#1: ## %for.body.preheader
andq $-4, %rdx
movl $1, %esi
movq %rax, %rdi
callq _memset
LBB1_2: ## %for.end
addq $8, %rsp
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@122573 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-26 23:42:51 +00:00
|
|
|
define void @test2(i32* %Base, i64 %Size) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%cmp10 = icmp eq i64 %Size, 0
|
|
|
|
|
br i1 %cmp10, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%i.011 = phi i64 [ %inc, %for.body ], [ 0, %entry ]
|
|
|
|
|
%add.ptr.i = getelementptr i32* %Base, i64 %i.011
|
|
|
|
|
store i32 16843009, i32* %add.ptr.i, align 4
|
|
|
|
|
%inc = add nsw i64 %i.011, 1
|
|
|
|
|
%exitcond = icmp eq i64 %inc, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test2
|
|
|
|
|
; CHECK: br i1 %cmp10,
|
|
|
|
|
; CHECK: %tmp = mul i64 %Size, 4
|
|
|
|
|
; CHECK: call void @llvm.memset.p0i8.i64(i8* %Base1, i8 1, i64 %tmp, i32 4, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
}
|
2011-01-01 19:39:01 +00:00
|
|
|
|
|
|
|
|
; This is a case where there is an extra may-aliased store in the loop, we can't
|
|
|
|
|
; promote the memset.
|
|
|
|
|
define void @test3(i32* %Base, i64 %Size, i8 *%MayAlias) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%i.011 = phi i64 [ %inc, %for.body ], [ 0, %entry ]
|
|
|
|
|
%add.ptr.i = getelementptr i32* %Base, i64 %i.011
|
|
|
|
|
store i32 16843009, i32* %add.ptr.i, align 4
|
|
|
|
|
|
|
|
|
|
store i8 42, i8* %MayAlias
|
|
|
|
|
%inc = add nsw i64 %i.011, 1
|
|
|
|
|
%exitcond = icmp eq i64 %inc, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test3
|
|
|
|
|
; CHECK-NOT: memset
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
2011-01-01 19:54:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
;; TODO: We should be able to promote this memset. Not yet though.
|
|
|
|
|
define void @test4(i8* %Base) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
%Base100 = getelementptr i8* %Base, i64 1000
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
store i8 0, i8* %I.0.014, align 1
|
|
|
|
|
|
|
|
|
|
;; Store beyond the range memset, should be safe to promote.
|
|
|
|
|
store i8 42, i8* %Base100
|
|
|
|
|
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, 100
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK-TODO: @test4
|
|
|
|
|
; CHECK-TODO: call void @llvm.memset.p0i8.i64(i8* %Base, i8 0, i64 100, i32 1, i1 false)
|
|
|
|
|
; CHECK-TODO-NOT: store
|
|
|
|
|
}
|
2011-01-01 20:12:04 +00:00
|
|
|
|
|
|
|
|
; This can't be promoted: the memset is a store of a loop variant value.
|
|
|
|
|
define void @test5(i8* %Base, i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
|
|
|
|
|
%V = trunc i64 %indvar to i8
|
|
|
|
|
store i8 %V, i8* %I.0.014, align 1
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test5
|
|
|
|
|
; CHECK-NOT: memset
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-02 03:37:56 +00:00
|
|
|
|
|
|
|
|
;; memcpy formation
|
|
|
|
|
define void @test6(i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph:
|
|
|
|
|
%Base = alloca i8, i32 10000
|
|
|
|
|
%Dest = alloca i8, i32 10000
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
%DestI = getelementptr i8* %Dest, i64 %indvar
|
|
|
|
|
%V = load i8* %I.0.014, align 1
|
|
|
|
|
store i8 %V, i8* %DestI, align 1
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test6
|
|
|
|
|
; CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %Dest, i8* %Base, i64 %Size, i32 1, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2011-01-02 19:01:03 +00:00
|
|
|
; This is a loop that was rotated but where the blocks weren't merged. This
|
|
|
|
|
; shouldn't perturb us.
|
|
|
|
|
define void @test7(i8* %Base, i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body.cont ]
|
|
|
|
|
br label %for.body.cont
|
|
|
|
|
for.body.cont:
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
store i8 0, i8* %I.0.014, align 1
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test7
|
|
|
|
|
; CHECK: call void @llvm.memset.p0i8.i64(i8* %Base, i8 0, i64 %Size, i32 1, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-02 20:24:21 +00:00
|
|
|
; This is a loop should not be transformed, it only executes one iteration.
|
|
|
|
|
define void @test8(i64* %Ptr, i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph: ; preds = %entry
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%PI = getelementptr i64* %Ptr, i64 %indvar
|
|
|
|
|
store i64 0, i64 *%PI
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, 1
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test8
|
|
|
|
|
; CHECK: store i64 0, i64* %PI
|
|
|
|
|
}
|
2011-01-02 03:37:56 +00:00
|
|
|
|
2011-01-02 21:14:18 +00:00
|
|
|
declare i8* @external(i8*)
|
|
|
|
|
|
|
|
|
|
;; This cannot be transformed into a memcpy, because the read-from location is
|
|
|
|
|
;; mutated by the loop.
|
|
|
|
|
define void @test9(i64 %Size) nounwind ssp {
|
|
|
|
|
bb.nph:
|
|
|
|
|
%Base = alloca i8, i32 10000
|
|
|
|
|
%Dest = alloca i8, i32 10000
|
|
|
|
|
|
|
|
|
|
%BaseAlias = call i8* @external(i8* %Base)
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %bb.nph, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body ]
|
|
|
|
|
%I.0.014 = getelementptr i8* %Base, i64 %indvar
|
|
|
|
|
%DestI = getelementptr i8* %Dest, i64 %indvar
|
|
|
|
|
%V = load i8* %I.0.014, align 1
|
|
|
|
|
store i8 %V, i8* %DestI, align 1
|
|
|
|
|
|
|
|
|
|
;; This store can clobber the input.
|
|
|
|
|
store i8 4, i8* %BaseAlias
|
|
|
|
|
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, %Size
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test9
|
|
|
|
|
; CHECK-NOT: llvm.memcpy
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-04 07:46:33 +00:00
|
|
|
; Two dimensional nested loop should be promoted to one big memset.
|
|
|
|
|
define void @test10(i8* %X) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
br label %bb.nph
|
|
|
|
|
|
|
|
|
|
bb.nph: ; preds = %entry, %for.inc10
|
|
|
|
|
%i.04 = phi i32 [ 0, %entry ], [ %inc12, %for.inc10 ]
|
|
|
|
|
br label %for.body5
|
|
|
|
|
|
|
|
|
|
for.body5: ; preds = %for.body5, %bb.nph
|
|
|
|
|
%j.02 = phi i32 [ 0, %bb.nph ], [ %inc, %for.body5 ]
|
|
|
|
|
%mul = mul nsw i32 %i.04, 100
|
|
|
|
|
%add = add nsw i32 %j.02, %mul
|
|
|
|
|
%idxprom = sext i32 %add to i64
|
|
|
|
|
%arrayidx = getelementptr inbounds i8* %X, i64 %idxprom
|
|
|
|
|
store i8 0, i8* %arrayidx, align 1
|
|
|
|
|
%inc = add nsw i32 %j.02, 1
|
|
|
|
|
%cmp4 = icmp eq i32 %inc, 100
|
|
|
|
|
br i1 %cmp4, label %for.inc10, label %for.body5
|
|
|
|
|
|
|
|
|
|
for.inc10: ; preds = %for.body5
|
|
|
|
|
%inc12 = add nsw i32 %i.04, 1
|
|
|
|
|
%cmp = icmp eq i32 %inc12, 100
|
|
|
|
|
br i1 %cmp, label %for.end13, label %bb.nph
|
|
|
|
|
|
|
|
|
|
for.end13: ; preds = %for.inc10
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test10
|
|
|
|
|
; CHECK: entry:
|
|
|
|
|
; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %X, i8 0, i64 10000, i32 1, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
Implement rdar://9009151, transforming strided loop stores of
unsplatable values into memset_pattern16 when it is available
(recent darwins). This transforms lots of strided loop stores
of ints for example, like 5 in vpr:
Formed memset: call void @memset_pattern16(i8* %4, i8* getelementptr inbounds ([16 x i8]* @.memset_pattern9, i32 0, i32 0), i64 %tmp25)
from store to: {%3,+,4}<%11> at: store i32 3, i32* %scevgep, align 4, !tbaa !4
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126040 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:31:39 +00:00
|
|
|
|
|
|
|
|
; On darwin10 (which is the triple in this .ll file) this loop can be turned
|
|
|
|
|
; into a memset_pattern call.
|
|
|
|
|
; rdar://9009151
|
rewrite the memset_pattern pattern generation stuff to accept any 2/4/8/16-byte
constant, including globals. This makes us generate much more "pretty" pattern
globals as well because it doesn't break it down to an array of bytes all the
time.
This enables us to handle stores of relocatable globals. This kicks in about
48 times in 254.gap, giving us stuff like this:
@.memset_pattern40 = internal constant [2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*] [%struct.TypHeader* (%struct.TypHeader*, %struct
.TypHeader*)* @IsFalse, %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)* @IsFalse], align 16
...
call void @memset_pattern16(i8* %scevgep5859, i8* bitcast ([2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*]* @.memset_pattern40 to i8*
), i64 %tmp75) nounwind
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126044 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:56:44 +00:00
|
|
|
define void @test11_pattern(i32* nocapture %P) nounwind ssp {
|
Implement rdar://9009151, transforming strided loop stores of
unsplatable values into memset_pattern16 when it is available
(recent darwins). This transforms lots of strided loop stores
of ints for example, like 5 in vpr:
Formed memset: call void @memset_pattern16(i8* %4, i8* getelementptr inbounds ([16 x i8]* @.memset_pattern9, i32 0, i32 0), i64 %tmp25)
from store to: {%3,+,4}<%11> at: store i32 3, i32* %scevgep, align 4, !tbaa !4
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126040 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:31:39 +00:00
|
|
|
entry:
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %entry ], [ %indvar.next, %for.body ]
|
|
|
|
|
%arrayidx = getelementptr i32* %P, i64 %indvar
|
|
|
|
|
store i32 1, i32* %arrayidx, align 4
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, 10000
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body
|
|
|
|
|
ret void
|
rewrite the memset_pattern pattern generation stuff to accept any 2/4/8/16-byte
constant, including globals. This makes us generate much more "pretty" pattern
globals as well because it doesn't break it down to an array of bytes all the
time.
This enables us to handle stores of relocatable globals. This kicks in about
48 times in 254.gap, giving us stuff like this:
@.memset_pattern40 = internal constant [2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*] [%struct.TypHeader* (%struct.TypHeader*, %struct
.TypHeader*)* @IsFalse, %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)* @IsFalse], align 16
...
call void @memset_pattern16(i8* %scevgep5859, i8* bitcast ([2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*]* @.memset_pattern40 to i8*
), i64 %tmp75) nounwind
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126044 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:56:44 +00:00
|
|
|
; CHECK: @test11_pattern
|
Implement rdar://9009151, transforming strided loop stores of
unsplatable values into memset_pattern16 when it is available
(recent darwins). This transforms lots of strided loop stores
of ints for example, like 5 in vpr:
Formed memset: call void @memset_pattern16(i8* %4, i8* getelementptr inbounds ([16 x i8]* @.memset_pattern9, i32 0, i32 0), i64 %tmp25)
from store to: {%3,+,4}<%11> at: store i32 3, i32* %scevgep, align 4, !tbaa !4
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126040 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:31:39 +00:00
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
|
; CHECK-NEXT: bitcast
|
|
|
|
|
; CHECK-NEXT: memset_pattern
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
|
|
|
|
|
2011-02-19 19:35:49 +00:00
|
|
|
; Store of null should turn into memset of zero.
|
|
|
|
|
define void @test12(i32** nocapture %P) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %entry ], [ %indvar.next, %for.body ]
|
|
|
|
|
%arrayidx = getelementptr i32** %P, i64 %indvar
|
|
|
|
|
store i32* null, i32** %arrayidx, align 4
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, 10000
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test12
|
|
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
|
; CHECK-NEXT: bitcast
|
|
|
|
|
; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %P1, i8 0, i64 80000, i32 4, i1 false)
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|
Implement rdar://9009151, transforming strided loop stores of
unsplatable values into memset_pattern16 when it is available
(recent darwins). This transforms lots of strided loop stores
of ints for example, like 5 in vpr:
Formed memset: call void @memset_pattern16(i8* %4, i8* getelementptr inbounds ([16 x i8]* @.memset_pattern9, i32 0, i32 0), i64 %tmp25)
from store to: {%3,+,4}<%11> at: store i32 3, i32* %scevgep, align 4, !tbaa !4
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126040 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:31:39 +00:00
|
|
|
|
rewrite the memset_pattern pattern generation stuff to accept any 2/4/8/16-byte
constant, including globals. This makes us generate much more "pretty" pattern
globals as well because it doesn't break it down to an array of bytes all the
time.
This enables us to handle stores of relocatable globals. This kicks in about
48 times in 254.gap, giving us stuff like this:
@.memset_pattern40 = internal constant [2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*] [%struct.TypHeader* (%struct.TypHeader*, %struct
.TypHeader*)* @IsFalse, %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)* @IsFalse], align 16
...
call void @memset_pattern16(i8* %scevgep5859, i8* bitcast ([2 x %struct.TypHeader* (%struct.TypHeader*, %struct.TypHeader*)*]* @.memset_pattern40 to i8*
), i64 %tmp75) nounwind
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@126044 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-19 19:56:44 +00:00
|
|
|
@G = global i32 5
|
|
|
|
|
|
|
|
|
|
; This store-of-address loop can be turned into a memset_pattern call.
|
|
|
|
|
; rdar://9009151
|
|
|
|
|
define void @test13_pattern(i32** nocapture %P) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%indvar = phi i64 [ 0, %entry ], [ %indvar.next, %for.body ]
|
|
|
|
|
%arrayidx = getelementptr i32** %P, i64 %indvar
|
|
|
|
|
store i32* @G, i32** %arrayidx, align 4
|
|
|
|
|
%indvar.next = add i64 %indvar, 1
|
|
|
|
|
%exitcond = icmp eq i64 %indvar.next, 10000
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body
|
|
|
|
|
ret void
|
|
|
|
|
; CHECK: @test13_pattern
|
|
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
|
; CHECK-NEXT: bitcast
|
|
|
|
|
; CHECK-NEXT: memset_pattern
|
|
|
|
|
; CHECK-NOT: store
|
|
|
|
|
; CHECK: ret void
|
|
|
|
|
}
|