maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopIdiom/basic.ll

243 lines
8.2 KiB
LLVM
Raw Normal View History

improve validity check to handle constant-trip-count loops more aggressively. In practice, this doesn't help anything though, see the todo. llvm-svn: 122660
2011-01-02 03:54:22 +08:00
; 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 llvm-svn: 122573
2010-12-27 07:42:51 +08:00
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
}
Allow loop-idiom to run on multiple BB loops, but still only scan the loop header for now for memset/memcpy opportunities. It turns out that loop-rotate is successfully rotating loops, but *DOESN'T MERGE THE BLOCKS*, turning "for loops" into 2 basic block loops that loop-idiom was ignoring. With this fix, we form many *many* more memcpy and memsets than before, including on the "history" loops in the viterbi benchmark, which look like this: for (j=0; j<MAX_history; ++j) { history_new[i][j+1] = history[2*i][j]; } Transforming these loops into memcpy's speeds up the viterbi benchmark from 11.98s to 3.55s on my machine. Woo. llvm-svn: 122685
2011-01-02 15:58:36 +08:00
; 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 llvm-svn: 122573
2010-12-27 07:42:51 +08:00
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
}
implement the "no aliasing accesses in loop" safety check. This pass should be correct now. llvm-svn: 122659
2011-01-02 03:39:01 +08:00
; 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
}
improve validity check to handle constant-trip-count loops more aggressively. In practice, this doesn't help anything though, see the todo. llvm-svn: 122660
2011-01-02 03:54:22 +08:00
;; 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
}
add a validity check that was missed, fixing a crash on the new testcase. llvm-svn: 122662
2011-01-02 04:12:04 +08:00
; 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
}
teach loop idiom recognition to form memcpy's from simple loops. llvm-svn: 122678
2011-01-02 11:37:56 +08:00
;; 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
}
enhance loop idiom recognition to scan *all* unconditionally executed blocks in a loop, instead of just the header block. This makes it more aggressive, able to handle Duncan's Ada examples. llvm-svn: 122704
2011-01-03 03:01:03 +08:00
; 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
}
If a loop iterates exactly once (has backedge count = 0) then don't mess with it. We'd rather peel/unroll it than convert all of its stores into memsets. llvm-svn: 122711
2011-01-03 04:24:21 +08:00
; 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
}
teach loop idiom recognition to form memcpy's from simple loops. llvm-svn: 122678
2011-01-02 11:37:56 +08:00
fix a miscompilation of tramp3d-v4: when forming a memcpy, we have to make sure that the loop we're promoting into a memcpy doesn't mutate the input of the memcpy. Before we were just checking that the dest of the memcpy wasn't mod/ref'd by the loop. llvm-svn: 122712
2011-01-03 05:14:18 +08:00
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
}