diff options
Diffstat (limited to 'test/Transforms/Inline')
22 files changed, 609 insertions, 31 deletions
diff --git a/test/Transforms/Inline/2003-09-22-PHINodeInlineFail.ll b/test/Transforms/Inline/2003-09-22-PHINodeInlineFail.ll index 5ced3b8..b8ca560 100644 --- a/test/Transforms/Inline/2003-09-22-PHINodeInlineFail.ll +++ b/test/Transforms/Inline/2003-09-22-PHINodeInlineFail.ll @@ -3,10 +3,15 @@ define i32 @main() { entry: invoke void @__main( ) - to label %LongJmpBlkPre unwind label %LongJmpBlkPre + to label %LongJmpBlkPost unwind label %LongJmpBlkPre -LongJmpBlkPre: ; preds = %entry, %entry +LongJmpBlkPost: + ret i32 0 + +LongJmpBlkPre: %i.3 = phi i32 [ 0, %entry ], [ 0, %entry ] ; <i32> [#uses=0] + %exn = landingpad {i8*, i32} personality i32 (...)* @__gxx_personality_v0 + cleanup ret i32 0 } @@ -14,3 +19,4 @@ define void @__main() { ret void } +declare i32 @__gxx_personality_v0(...) diff --git a/test/Transforms/Inline/2003-09-22-PHINodesInNormalInvokeDest.ll b/test/Transforms/Inline/2003-09-22-PHINodesInNormalInvokeDest.ll index 1bd5529..43bdd30 100644 --- a/test/Transforms/Inline/2003-09-22-PHINodesInNormalInvokeDest.ll +++ b/test/Transforms/Inline/2003-09-22-PHINodesInNormalInvokeDest.ll @@ -13,6 +13,8 @@ LJDecisionBB: ; preds = %else br label %else RethrowExcept: ; preds = %entry + %exn = landingpad {i8*, i32} personality i32 (...)* @__gxx_personality_v0 + cleanup ret i32 0 } @@ -20,4 +22,4 @@ define void @__main() { ret void } - +declare i32 @__gxx_personality_v0(...) diff --git a/test/Transforms/Inline/2006-11-09-InlineCGUpdate-2.ll b/test/Transforms/Inline/2006-11-09-InlineCGUpdate-2.ll index b4380d0..ee5a378 100644 --- a/test/Transforms/Inline/2006-11-09-InlineCGUpdate-2.ll +++ b/test/Transforms/Inline/2006-11-09-InlineCGUpdate-2.ll @@ -2,7 +2,6 @@ ; PR993 target datalayout = "e-p:32:32" target triple = "i386-unknown-openbsd3.9" -deplibs = [ "stdc++", "c", "crtend" ] %"struct.__gnu_cxx::__normal_iterator<char*,std::basic_string<char, std::char_traits<char>, std::allocator<char> > >" = type { i8* } %"struct.__gnu_cxx::char_producer<char>" = type { i32 (...)** } %struct.__sFILE = type { i8*, i32, i32, i16, i16, %struct.__sbuf, i32, i8*, i32 (i8*)*, i32 (i8*, i8*, i32)*, i64 (i8*, i64, i32)*, i32 (i8*, i8*, i32)*, %struct.__sbuf, i8*, i32, [3 x i8], [1 x i8], %struct.__sbuf, i32, i64 } diff --git a/test/Transforms/Inline/2006-11-09-InlineCGUpdate.ll b/test/Transforms/Inline/2006-11-09-InlineCGUpdate.ll index b754d9f..fb5a4b5 100644 --- a/test/Transforms/Inline/2006-11-09-InlineCGUpdate.ll +++ b/test/Transforms/Inline/2006-11-09-InlineCGUpdate.ll @@ -2,7 +2,6 @@ ; PR992 target datalayout = "e-p:32:32" target triple = "i686-pc-linux-gnu" -deplibs = [ "stdc++", "c", "crtend" ] %struct._IO_FILE = type { i32, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, %struct._IO_marker*, %struct._IO_FILE*, i32, i32, i32, i16, i8, [1 x i8], i8*, i64, i8*, i8*, i32, [52 x i8] } %struct._IO_marker = type { %struct._IO_marker*, %struct._IO_FILE*, i32 } %"struct.__cxxabiv1::__array_type_info" = type { %"struct.std::type_info" } diff --git a/test/Transforms/Inline/2010-05-12-ValueMap.ll b/test/Transforms/Inline/2010-05-12-ValueMap.ll index f9cc13f..f452907 100644 --- a/test/Transforms/Inline/2010-05-12-ValueMap.ll +++ b/test/Transforms/Inline/2010-05-12-ValueMap.ll @@ -1,4 +1,4 @@ -; RUN: opt %s -inline -mergefunc -disable-output +; RUN: opt -inline -mergefunc -disable-output < %s ; This tests for a bug where the inliner kept the functions in a ValueMap after ; it had completed and a ModulePass started to run. LLVM would crash deleting diff --git a/test/Transforms/Inline/alloca_test.ll b/test/Transforms/Inline/alloca_test.ll index e5791d5..8464259 100644 --- a/test/Transforms/Inline/alloca_test.ll +++ b/test/Transforms/Inline/alloca_test.ll @@ -1,7 +1,7 @@ ; This test ensures that alloca instructions in the entry block for an inlined ; function are moved to the top of the function they are inlined into. ; -; RUN: opt -S -inline %s | FileCheck %s +; RUN: opt -S -inline < %s | FileCheck %s define i32 @func(i32 %i) { %X = alloca i32 ; <i32*> [#uses=1] diff --git a/test/Transforms/Inline/basictest.ll b/test/Transforms/Inline/basictest.ll index 609a3d4..39e25cb 100644 --- a/test/Transforms/Inline/basictest.ll +++ b/test/Transforms/Inline/basictest.ll @@ -45,3 +45,48 @@ define i32 @test2(i1 %cond) { ; CHECK-NOT: = alloca ; CHECK: ret i32 } + +declare void @barrier() noduplicate + +define internal i32 @f() { + call void @barrier() noduplicate + ret i32 1 +} + +define i32 @g() { + call void @barrier() noduplicate + ret i32 2 +} + +define internal i32 @h() { + call void @barrier() noduplicate + ret i32 3 +} + +define i32 @test3() { + %b = call i32 @f() + ret i32 %b +} + +; The call to @f cannot be inlined as there is another callsite +; calling @f, and @f contains a noduplicate call. +; +; The call to @g cannot be inlined as it has external linkage. +; +; The call to @h *can* be inlined. + +; CHECK: @test +define i32 @test() { +; CHECK: call i32 @f() + %a = call i32 @f() +; CHECK: call i32 @g() + %b = call i32 @g() +; CHECK-NOT: call i32 @h() + %c = call i32 @h() + + %d = add i32 %a, %b + %e = add i32 %d, %c + + ret i32 %e +; CHECK: } +} diff --git a/test/Transforms/Inline/crash2.ll b/test/Transforms/Inline/crash2.ll index cb1f44d..be634f6 100644 --- a/test/Transforms/Inline/crash2.ll +++ b/test/Transforms/Inline/crash2.ll @@ -1,4 +1,4 @@ -; RUN: opt -inline -scalarrepl -max-cg-scc-iterations=1 %s -disable-output +; RUN: opt -inline -scalarrepl -max-cg-scc-iterations=1 -disable-output < %s 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-darwin10.3" diff --git a/test/Transforms/Inline/delete-call.ll b/test/Transforms/Inline/delete-call.ll index 7716d6a..97c52af 100644 --- a/test/Transforms/Inline/delete-call.ll +++ b/test/Transforms/Inline/delete-call.ll @@ -1,5 +1,9 @@ -; RUN: opt %s -S -inline -functionattrs -stats 2>&1 | grep "Number of call sites deleted, not inlined" -; RUN: opt %s -S -inline -stats 2>&1 | grep "Number of functions inlined" +; REQUIRES: asserts +; RUN: opt -S -inline -stats < %s 2>&1 | FileCheck %s +; CHECK: Number of functions inlined + +; RUN: opt -S -inline -functionattrs -stats < %s 2>&1 | FileCheck -check-prefix=FUNCTIONATTRS %s +; CHECK-FUNCTIONATTRS: Number of call sites deleted, not inlined target datalayout = "e-p:32:32:32-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-f80:128:128-n8:16:32" target triple = "i386-apple-darwin9.8" diff --git a/test/Transforms/Inline/devirtualize-3.ll b/test/Transforms/Inline/devirtualize-3.ll index c32be4e..3f01967 100644 --- a/test/Transforms/Inline/devirtualize-3.ll +++ b/test/Transforms/Inline/devirtualize-3.ll @@ -1,4 +1,4 @@ -; RUN: opt -basicaa -inline -S -scalarrepl -gvn -instcombine %s | FileCheck %s +; RUN: opt -basicaa -inline -S -scalarrepl -gvn -instcombine < %s | FileCheck %s ; PR5009 ; CHECK: define i32 @main() diff --git a/test/Transforms/Inline/devirtualize.ll b/test/Transforms/Inline/devirtualize.ll index 51ea4ba..d46154e 100644 --- a/test/Transforms/Inline/devirtualize.ll +++ b/test/Transforms/Inline/devirtualize.ll @@ -1,4 +1,4 @@ -; RUN: opt -S -basicaa -inline -scalarrepl -instcombine -simplifycfg -instcombine -gvn -globaldce %s | FileCheck %s +; RUN: opt -S -Os < %s | FileCheck %s 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" diff --git a/test/Transforms/Inline/gvn-inline-iteration.ll b/test/Transforms/Inline/gvn-inline-iteration.ll index e502fd5..526ed79 100644 --- a/test/Transforms/Inline/gvn-inline-iteration.ll +++ b/test/Transforms/Inline/gvn-inline-iteration.ll @@ -1,4 +1,4 @@ -; RUN: opt -basicaa -inline -gvn %s -S -max-cg-scc-iterations=1 | FileCheck %s +; RUN: opt -basicaa -inline -gvn -S -max-cg-scc-iterations=1 < %s | FileCheck %s ; rdar://6295824 and PR6724 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" diff --git a/test/Transforms/Inline/inline-optsize.ll b/test/Transforms/Inline/inline-optsize.ll index 20d7426..3ad573a 100644 --- a/test/Transforms/Inline/inline-optsize.ll +++ b/test/Transforms/Inline/inline-optsize.ll @@ -1,5 +1,5 @@ -; RUN: opt -S -Oz %s | FileCheck %s -check-prefix=OZ -; RUN: opt -S -O2 %s | FileCheck %s -check-prefix=O2 +; RUN: opt -S -Oz < %s | FileCheck %s -check-prefix=OZ +; RUN: opt -S -O2 < %s | FileCheck %s -check-prefix=O2 ; The inline threshold for a function with the optsize attribute is currently ; the same as the global inline threshold for -Os. Check that the optsize diff --git a/test/Transforms/Inline/inline_constprop.ll b/test/Transforms/Inline/inline_constprop.ll index 0b48a72..77bc378 100644 --- a/test/Transforms/Inline/inline_constprop.ll +++ b/test/Transforms/Inline/inline_constprop.ll @@ -111,6 +111,82 @@ bb.false: ret i32 %sub } +declare {i8, i1} @llvm.uadd.with.overflow.i8(i8 %a, i8 %b) + +define i8 @caller4(i8 %z) { +; Check that we can constant fold through intrinsics such as the +; overflow-detecting arithmetic instrinsics. These are particularly important +; as they are used heavily in standard library code and generic C++ code where +; the arguments are oftent constant but complete generality is required. +; +; CHECK: @caller4 +; CHECK-NOT: call +; CHECK: ret i8 -1 + +entry: + %x = call i8 @callee4(i8 254, i8 14, i8 %z) + ret i8 %x +} + +define i8 @callee4(i8 %x, i8 %y, i8 %z) { + %uadd = call {i8, i1} @llvm.uadd.with.overflow.i8(i8 %x, i8 %y) + %o = extractvalue {i8, i1} %uadd, 1 + br i1 %o, label %bb.true, label %bb.false + +bb.true: + ret i8 -1 + +bb.false: + ; This block musn't be counted in the inline cost. + %z1 = add i8 %z, 1 + %z2 = add i8 %z1, 1 + %z3 = add i8 %z2, 1 + %z4 = add i8 %z3, 1 + %z5 = add i8 %z4, 1 + %z6 = add i8 %z5, 1 + %z7 = add i8 %z6, 1 + %z8 = add i8 %z7, 1 + ret i8 %z8 +} + +define i64 @caller5(i64 %y) { +; Check that we can round trip constants through various kinds of casts etc w/o +; losing track of the constant prop in the inline cost analysis. +; +; CHECK: @caller5 +; CHECK-NOT: call +; CHECK: ret i64 -1 + +entry: + %x = call i64 @callee5(i64 42, i64 %y) + ret i64 %x +} + +define i64 @callee5(i64 %x, i64 %y) { + %inttoptr = inttoptr i64 %x to i8* + %bitcast = bitcast i8* %inttoptr to i32* + %ptrtoint = ptrtoint i32* %bitcast to i64 + %trunc = trunc i64 %ptrtoint to i32 + %zext = zext i32 %trunc to i64 + %cmp = icmp eq i64 %zext, 42 + br i1 %cmp, label %bb.true, label %bb.false + +bb.true: + ret i64 -1 + +bb.false: + ; This block musn't be counted in the inline cost. + %y1 = add i64 %y, 1 + %y2 = add i64 %y1, 1 + %y3 = add i64 %y2, 1 + %y4 = add i64 %y3, 1 + %y5 = add i64 %y4, 1 + %y6 = add i64 %y5, 1 + %y7 = add i64 %y6, 1 + %y8 = add i64 %y7, 1 + ret i64 %y8 +} + define i32 @PR13412.main() { ; This is a somewhat complicated three layer subprogram that was reported to diff --git a/test/Transforms/Inline/inline_invoke.ll b/test/Transforms/Inline/inline_invoke.ll index 9f5f670..c394138 100644 --- a/test/Transforms/Inline/inline_invoke.ll +++ b/test/Transforms/Inline/inline_invoke.ll @@ -96,6 +96,7 @@ eh.resume: ; CHECK: landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 ; CHECK-NEXT: cleanup ; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) +; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) ; CHECK-NEXT: invoke void @_ZN1AD1Ev(%struct.A* [[A]]) ; CHECK-NEXT: to label %[[LBL:[^\s]+]] unwind ; CHECK: [[LBL]]: @@ -166,6 +167,7 @@ eh.resume: ; CHECK-NEXT: [[LPADVAL1:%.*]] = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 ; CHECK-NEXT: cleanup ; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) +; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) ; CHECK-NEXT: invoke void @_ZN1AD1Ev(%struct.A* [[A1]]) ; CHECK-NEXT: to label %[[RESUME1:[^\s]+]] unwind ; CHECK: [[RESUME1]]: @@ -185,6 +187,7 @@ eh.resume: ; CHECK-NEXT: [[LPADVAL2:%.*]] = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 ; CHECK-NEXT: cleanup ; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) +; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) ; CHECK-NEXT: invoke void @_ZN1AD1Ev(%struct.A* [[A2]]) ; CHECK-NEXT: to label %[[RESUME2:[^\s]+]] unwind ; CHECK: [[RESUME2]]: @@ -272,6 +275,7 @@ lpad.cont: ; CHECK: landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 ; CHECK-NEXT: cleanup ; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) +; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) ; CHECK-NEXT: invoke void @_ZN1AD1Ev( ; CHECK-NEXT: to label %[[L:[^\s]+]] unwind ; CHECK: [[L]]: @@ -318,6 +322,7 @@ terminate: ; CHECK: landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 ; CHECK-NEXT: cleanup ; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) +; CHECK-NEXT: catch i8* bitcast (i8** @_ZTIi to i8*) ; CHECK-NEXT: invoke void @_ZN1AD1Ev( ; CHECK-NEXT: to label %[[L:[^\s]+]] unwind ; CHECK: [[L]]: @@ -330,7 +335,7 @@ terminate: ; CHECK-NEXT: br label %[[JOIN]] ; CHECK: [[JOIN]]: ; CHECK-NEXT: phi { i8*, i32 } -; CHECK-NEXT: call void @opaque() nounwind +; CHECK-NEXT: call void @opaque() [[NUW:#[0-9]+]] ; CHECK-NEXT: br label %[[FIX:[^\s]+]] ; CHECK: lpad: ; CHECK-NEXT: landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 @@ -340,3 +345,8 @@ terminate: ; CHECK-NEXT: [[T1:%.*]] = phi i32 [ 0, %[[JOIN]] ], [ 1, %lpad ] ; CHECK-NEXT: call void @use(i32 [[T1]]) ; CHECK-NEXT: call void @_ZSt9terminatev() + +; CHECK: attributes [[NUW]] = { nounwind } +; CHECK: attributes #1 = { nounwind readnone } +; CHECK: attributes #2 = { ssp uwtable } +; CHECK: attributes #3 = { noreturn nounwind } diff --git a/test/Transforms/Inline/inline_minisize.ll b/test/Transforms/Inline/inline_minisize.ll new file mode 100644 index 0000000..3dddbcf --- /dev/null +++ b/test/Transforms/Inline/inline_minisize.ll @@ -0,0 +1,232 @@ +; RUN: opt -O2 -S < %s | FileCheck %s + +@data = common global i32* null, align 8 + +define i32 @fct1(i32 %a) nounwind uwtable ssp { +entry: + %a.addr = alloca i32, align 4 + %res = alloca i32, align 4 + %i = alloca i32, align 4 + store i32 %a, i32* %a.addr, align 4 + %tmp = load i32* %a.addr, align 4 + %idxprom = sext i32 %tmp to i64 + %tmp1 = load i32** @data, align 8 + %arrayidx = getelementptr inbounds i32* %tmp1, i64 %idxprom + %tmp2 = load i32* %arrayidx, align 4 + %tmp3 = load i32* %a.addr, align 4 + %add = add nsw i32 %tmp3, 1 + %idxprom1 = sext i32 %add to i64 + %tmp4 = load i32** @data, align 8 + %arrayidx2 = getelementptr inbounds i32* %tmp4, i64 %idxprom1 + %tmp5 = load i32* %arrayidx2, align 4 + %mul = mul nsw i32 %tmp2, %tmp5 + store i32 %mul, i32* %res, align 4 + store i32 0, i32* %i, align 4 + store i32 0, i32* %i, align 4 + br label %for.cond + +for.cond: ; preds = %for.inc, %entry + %tmp6 = load i32* %i, align 4 + %tmp7 = load i32* %res, align 4 + %cmp = icmp slt i32 %tmp6, %tmp7 + br i1 %cmp, label %for.body, label %for.end + +for.body: ; preds = %for.cond + %tmp8 = load i32* %i, align 4 + %idxprom3 = sext i32 %tmp8 to i64 + %tmp9 = load i32** @data, align 8 + %arrayidx4 = getelementptr inbounds i32* %tmp9, i64 %idxprom3 + call void @fct0(i32* %arrayidx4) + br label %for.inc + +for.inc: ; preds = %for.body + %tmp10 = load i32* %i, align 4 + %inc = add nsw i32 %tmp10, 1 + store i32 %inc, i32* %i, align 4 + br label %for.cond + +for.end: ; preds = %for.cond + store i32 0, i32* %i, align 4 + br label %for.cond5 + +for.cond5: ; preds = %for.inc10, %for.end + %tmp11 = load i32* %i, align 4 + %tmp12 = load i32* %res, align 4 + %cmp6 = icmp slt i32 %tmp11, %tmp12 + br i1 %cmp6, label %for.body7, label %for.end12 + +for.body7: ; preds = %for.cond5 + %tmp13 = load i32* %i, align 4 + %idxprom8 = sext i32 %tmp13 to i64 + %tmp14 = load i32** @data, align 8 + %arrayidx9 = getelementptr inbounds i32* %tmp14, i64 %idxprom8 + call void @fct0(i32* %arrayidx9) + br label %for.inc10 + +for.inc10: ; preds = %for.body7 + %tmp15 = load i32* %i, align 4 + %inc11 = add nsw i32 %tmp15, 1 + store i32 %inc11, i32* %i, align 4 + br label %for.cond5 + +for.end12: ; preds = %for.cond5 + store i32 0, i32* %i, align 4 + br label %for.cond13 + +for.cond13: ; preds = %for.inc18, %for.end12 + %tmp16 = load i32* %i, align 4 + %tmp17 = load i32* %res, align 4 + %cmp14 = icmp slt i32 %tmp16, %tmp17 + br i1 %cmp14, label %for.body15, label %for.end20 + +for.body15: ; preds = %for.cond13 + %tmp18 = load i32* %i, align 4 + %idxprom16 = sext i32 %tmp18 to i64 + %tmp19 = load i32** @data, align 8 + %arrayidx17 = getelementptr inbounds i32* %tmp19, i64 %idxprom16 + call void @fct0(i32* %arrayidx17) + br label %for.inc18 + +for.inc18: ; preds = %for.body15 + %tmp20 = load i32* %i, align 4 + %inc19 = add nsw i32 %tmp20, 1 + store i32 %inc19, i32* %i, align 4 + br label %for.cond13 + +for.end20: ; preds = %for.cond13 + %tmp21 = load i32* %res, align 4 + ret i32 %tmp21 +} + +declare void @fct0(i32*) + +define i32 @fct2(i32 %a) nounwind uwtable inlinehint ssp { +entry: + %a.addr = alloca i32, align 4 + %res = alloca i32, align 4 + %i = alloca i32, align 4 + store i32 %a, i32* %a.addr, align 4 + %tmp = load i32* %a.addr, align 4 + %shl = shl i32 %tmp, 1 + %idxprom = sext i32 %shl to i64 + %tmp1 = load i32** @data, align 8 + %arrayidx = getelementptr inbounds i32* %tmp1, i64 %idxprom + %tmp2 = load i32* %arrayidx, align 4 + %tmp3 = load i32* %a.addr, align 4 + %shl1 = shl i32 %tmp3, 1 + %add = add nsw i32 %shl1, 13 + %idxprom2 = sext i32 %add to i64 + %tmp4 = load i32** @data, align 8 + %arrayidx3 = getelementptr inbounds i32* %tmp4, i64 %idxprom2 + %tmp5 = load i32* %arrayidx3, align 4 + %mul = mul nsw i32 %tmp2, %tmp5 + store i32 %mul, i32* %res, align 4 + store i32 0, i32* %i, align 4 + store i32 0, i32* %i, align 4 + br label %for.cond + +for.cond: ; preds = %for.inc, %entry + %tmp6 = load i32* %i, align 4 + %tmp7 = load i32* %res, align 4 + %cmp = icmp slt i32 %tmp6, %tmp7 + br i1 %cmp, label %for.body, label %for.end + +for.body: ; preds = %for.cond + %tmp8 = load i32* %i, align 4 + %idxprom4 = sext i32 %tmp8 to i64 + %tmp9 = load i32** @data, align 8 + %arrayidx5 = getelementptr inbounds i32* %tmp9, i64 %idxprom4 + call void @fct0(i32* %arrayidx5) + br label %for.inc + +for.inc: ; preds = %for.body + %tmp10 = load i32* %i, align 4 + %inc = add nsw i32 %tmp10, 1 + store i32 %inc, i32* %i, align 4 + br label %for.cond + +for.end: ; preds = %for.cond + store i32 0, i32* %i, align 4 + br label %for.cond6 + +for.cond6: ; preds = %for.inc11, %for.end + %tmp11 = load i32* %i, align 4 + %tmp12 = load i32* %res, align 4 + %cmp7 = icmp slt i32 %tmp11, %tmp12 + br i1 %cmp7, label %for.body8, label %for.end13 + +for.body8: ; preds = %for.cond6 + %tmp13 = load i32* %i, align 4 + %idxprom9 = sext i32 %tmp13 to i64 + %tmp14 = load i32** @data, align 8 + %arrayidx10 = getelementptr inbounds i32* %tmp14, i64 %idxprom9 + call void @fct0(i32* %arrayidx10) + br label %for.inc11 + +for.inc11: ; preds = %for.body8 + %tmp15 = load i32* %i, align 4 + %inc12 = add nsw i32 %tmp15, 1 + store i32 %inc12, i32* %i, align 4 + br label %for.cond6 + +for.end13: ; preds = %for.cond6 + store i32 0, i32* %i, align 4 + br label %for.cond14 + +for.cond14: ; preds = %for.inc19, %for.end13 + %tmp16 = load i32* %i, align 4 + %tmp17 = load i32* %res, align 4 + %cmp15 = icmp slt i32 %tmp16, %tmp17 + br i1 %cmp15, label %for.body16, label %for.end21 + +for.body16: ; preds = %for.cond14 + %tmp18 = load i32* %i, align 4 + %idxprom17 = sext i32 %tmp18 to i64 + %tmp19 = load i32** @data, align 8 + %arrayidx18 = getelementptr inbounds i32* %tmp19, i64 %idxprom17 + call void @fct0(i32* %arrayidx18) + br label %for.inc19 + +for.inc19: ; preds = %for.body16 + %tmp20 = load i32* %i, align 4 + %inc20 = add nsw i32 %tmp20, 1 + store i32 %inc20, i32* %i, align 4 + br label %for.cond14 + +for.end21: ; preds = %for.cond14 + %tmp21 = load i32* %res, align 4 + ret i32 %tmp21 +} + +define i32 @fct3(i32 %c) nounwind uwtable ssp { +entry: + ;CHECK: @fct3 + ;CHECK: call i32 @fct1 + ; The inline keyword gives a sufficient benefits to inline fct2 + ;CHECK-NOT: call i32 @fct2 + %c.addr = alloca i32, align 4 + store i32 %c, i32* %c.addr, align 4 + %tmp = load i32* %c.addr, align 4 + %call = call i32 @fct1(i32 %tmp) + %tmp1 = load i32* %c.addr, align 4 + %call1 = call i32 @fct2(i32 %tmp1) + %add = add nsw i32 %call, %call1 + ret i32 %add +} + +define i32 @fct4(i32 %c) minsize nounwind uwtable ssp { +entry: + ;CHECK: @fct4 + ;CHECK: call i32 @fct1 + ; With Oz (minsize attribute), the benefit of inlining fct2 + ; is the same as fct1, thus no inlining for fct2 + ;CHECK: call i32 @fct2 + %c.addr = alloca i32, align 4 + store i32 %c, i32* %c.addr, align 4 + %tmp = load i32* %c.addr, align 4 + %call = call i32 @fct1(i32 %tmp) + %tmp1 = load i32* %c.addr, align 4 + %call1 = call i32 @fct2(i32 %tmp1) + %add = add nsw i32 %call, %call1 + ret i32 %add +} diff --git a/test/Transforms/Inline/inline_ssp.ll b/test/Transforms/Inline/inline_ssp.ll new file mode 100644 index 0000000..a4b43a7 --- /dev/null +++ b/test/Transforms/Inline/inline_ssp.ll @@ -0,0 +1,160 @@ +; RUN: opt -inline %s -S | FileCheck %s +; Ensure SSP attributes are propagated correctly when inlining. + +@.str = private unnamed_addr constant [11 x i8] c"fun_nossp\0A\00", align 1 +@.str1 = private unnamed_addr constant [9 x i8] c"fun_ssp\0A\00", align 1 +@.str2 = private unnamed_addr constant [15 x i8] c"fun_sspstrong\0A\00", align 1 +@.str3 = private unnamed_addr constant [12 x i8] c"fun_sspreq\0A\00", align 1 + +; These first four functions (@fun_sspreq, @fun_sspstrong, @fun_ssp, @fun_nossp) +; are used by the remaining functions to ensure that the SSP attributes are +; propagated correctly. The caller should have its SSP attribute set as: +; strictest(caller-ssp-attr, callee-ssp-attr), where strictness is ordered as: +; sspreq > sspstrong > ssp > [no ssp] +define internal void @fun_sspreq() nounwind sspreq uwtable { +entry: + %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([12 x i8]* @.str3, i32 0, i32 0)) + ret void +} + +define internal void @fun_sspstrong() nounwind sspstrong uwtable { +entry: + %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([15 x i8]* @.str2, i32 0, i32 0)) + ret void +} + +define internal void @fun_ssp() nounwind ssp uwtable { +entry: + %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([9 x i8]* @.str1, i32 0, i32 0)) + ret void +} + +define internal void @fun_nossp() nounwind uwtable { +entry: + %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([11 x i8]* @.str, i32 0, i32 0)) + ret void +} + +; Tests start below + +define void @inline_req_req() nounwind sspreq uwtable { +entry: +; CHECK: @inline_req_req() #0 + call void @fun_sspreq() + ret void +} + +define void @inline_req_strong() nounwind sspstrong uwtable { +entry: +; CHECK: @inline_req_strong() #0 + call void @fun_sspreq() + ret void +} + +define void @inline_req_ssp() nounwind ssp uwtable { +entry: +; CHECK: @inline_req_ssp() #0 + call void @fun_sspreq() + ret void +} + +define void @inline_req_nossp() nounwind uwtable { +entry: +; CHECK: @inline_req_nossp() #0 + call void @fun_sspreq() + ret void +} + +define void @inline_strong_req() nounwind sspreq uwtable { +entry: +; CHECK: @inline_strong_req() #0 + call void @fun_sspstrong() + ret void +} + + +define void @inline_strong_strong() nounwind sspstrong uwtable { +entry: +; CHECK: @inline_strong_strong() #1 + call void @fun_sspstrong() + ret void +} + +define void @inline_strong_ssp() nounwind ssp uwtable { +entry: +; CHECK: @inline_strong_ssp() #1 + call void @fun_sspstrong() + ret void +} + +define void @inline_strong_nossp() nounwind uwtable { +entry: +; CHECK: @inline_strong_nossp() #1 + call void @fun_sspstrong() + ret void +} + +define void @inline_ssp_req() nounwind sspreq uwtable { +entry: +; CHECK: @inline_ssp_req() #0 + call void @fun_ssp() + ret void +} + + +define void @inline_ssp_strong() nounwind sspstrong uwtable { +entry: +; CHECK: @inline_ssp_strong() #1 + call void @fun_ssp() + ret void +} + +define void @inline_ssp_ssp() nounwind ssp uwtable { +entry: +; CHECK: @inline_ssp_ssp() #2 + call void @fun_ssp() + ret void +} + +define void @inline_ssp_nossp() nounwind uwtable { +entry: +; CHECK: @inline_ssp_nossp() #2 + call void @fun_ssp() + ret void +} + +define void @inline_nossp_req() nounwind uwtable sspreq { +entry: +; CHECK: @inline_nossp_req() #0 + call void @fun_nossp() + ret void +} + + +define void @inline_nossp_strong() nounwind sspstrong uwtable { +entry: +; CHECK: @inline_nossp_strong() #1 + call void @fun_nossp() + ret void +} + +define void @inline_nossp_ssp() nounwind ssp uwtable { +entry: +; CHECK: @inline_nossp_ssp() #2 + call void @fun_nossp() + ret void +} + +define void @inline_nossp_nossp() nounwind uwtable { +entry: +; CHECK: @inline_nossp_nossp() #3 + call void @fun_nossp() + ret void +} + +declare i32 @printf(i8*, ...) + +; CHECK: attributes #0 = { nounwind sspreq uwtable } +; CHECK: attributes #1 = { nounwind sspstrong uwtable } +; CHECK: attributes #2 = { nounwind ssp uwtable } +; CHECK: attributes #3 = { nounwind uwtable } diff --git a/test/Transforms/Inline/lifetime-no-datalayout.ll b/test/Transforms/Inline/lifetime-no-datalayout.ll new file mode 100644 index 0000000..f4ffef3 --- /dev/null +++ b/test/Transforms/Inline/lifetime-no-datalayout.ll @@ -0,0 +1,23 @@ +; RUN: opt -inline -S < %s | FileCheck %s + +declare void @use(i8* %a) + +define void @helper() { + %a = alloca i8 + call void @use(i8* %a) + ret void +} + +; Size in llvm.lifetime.X should be -1 (unknown). +define void @test() { +; CHECK: @test +; CHECK-NOT: lifetime +; CHECK: llvm.lifetime.start(i64 -1 +; CHECK-NOT: lifetime +; CHECK: llvm.lifetime.end(i64 -1 + call void @helper() +; CHECK-NOT: lifetime +; CHECK: ret void + ret void +} + diff --git a/test/Transforms/Inline/lifetime.ll b/test/Transforms/Inline/lifetime.ll index a95c836..fc73385 100644 --- a/test/Transforms/Inline/lifetime.ll +++ b/test/Transforms/Inline/lifetime.ll @@ -1,22 +1,25 @@ -; RUN: opt -inline %s -S -o - | FileCheck %s +; RUN: opt -inline -S < %s | FileCheck %s +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-S128" declare void @llvm.lifetime.start(i64, i8*) declare void @llvm.lifetime.end(i64, i8*) define void @helper_both_markers() { %a = alloca i8 - call void @llvm.lifetime.start(i64 1, i8* %a) - call void @llvm.lifetime.end(i64 1, i8* %a) + ; Size in llvm.lifetime.start / llvm.lifetime.end differs from + ; allocation size. We should use the former. + call void @llvm.lifetime.start(i64 2, i8* %a) + call void @llvm.lifetime.end(i64 2, i8* %a) ret void } define void @test_both_markers() { ; CHECK: @test_both_markers -; CHECK: llvm.lifetime.start(i64 1 -; CHECK-NEXT: llvm.lifetime.end(i64 1 +; CHECK: llvm.lifetime.start(i64 2 +; CHECK-NEXT: llvm.lifetime.end(i64 2 call void @helper_both_markers() -; CHECK-NEXT: llvm.lifetime.start(i64 1 -; CHECK-NEXT: llvm.lifetime.end(i64 1 +; CHECK-NEXT: llvm.lifetime.start(i64 2 +; CHECK-NEXT: llvm.lifetime.end(i64 2 call void @helper_both_markers() ; CHECK-NEXT: ret void ret void @@ -27,7 +30,7 @@ define void @test_both_markers() { declare void @use(i8* %a) define void @helper_no_markers() { - %a = alloca i8 + %a = alloca i8 ; Allocation size is 1 byte. call void @use(i8* %a) ret void } @@ -37,14 +40,14 @@ define void @helper_no_markers() { define void @test_no_marker() { ; CHECK: @test_no_marker ; CHECK-NOT: lifetime -; CHECK: llvm.lifetime.start(i64 -1 +; CHECK: llvm.lifetime.start(i64 1 ; CHECK-NOT: lifetime -; CHECK: llvm.lifetime.end(i64 -1 +; CHECK: llvm.lifetime.end(i64 1 call void @helper_no_markers() ; CHECK-NOT: lifetime -; CHECK: llvm.lifetime.start(i64 -1 +; CHECK: llvm.lifetime.start(i64 1 ; CHECK-NOT: lifetime -; CHECK: llvm.lifetime.end(i64 -1 +; CHECK: llvm.lifetime.end(i64 1 call void @helper_no_markers() ; CHECK-NOT: lifetime ; CHECK: ret void @@ -76,3 +79,22 @@ define void @test_two_casts() { ; CHECK: ret void ret void } + +define void @helper_arrays_alloca() { + %a = alloca [10 x i32], align 16 + %1 = bitcast [10 x i32]* %a to i8* + call void @use(i8* %1) + ret void +} + +define void @test_arrays_alloca() { +; CHECK: @test_arrays_alloca +; CHECK-NOT: lifetime +; CHECK: llvm.lifetime.start(i64 40, +; CHECK-NOT: lifetime +; CHECK: llvm.lifetime.end(i64 40, + call void @helper_arrays_alloca() +; CHECK-NOT: lifetime +; CHECK: ret void + ret void +} diff --git a/test/Transforms/Inline/noinline-recursive-fn.ll b/test/Transforms/Inline/noinline-recursive-fn.ll index 6cde0e2..5520093 100644 --- a/test/Transforms/Inline/noinline-recursive-fn.ll +++ b/test/Transforms/Inline/noinline-recursive-fn.ll @@ -2,7 +2,7 @@ ; This effectively is just peeling off the first iteration of a loop, and the ; inliner heuristics are not set up for this. -; RUN: opt -inline %s -S | FileCheck %s +; RUN: opt -inline -S < %s | FileCheck %s 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-darwin10.3" diff --git a/test/Transforms/Inline/noinline.ll b/test/Transforms/Inline/noinline.ll index dc3f6e0..7667114 100644 --- a/test/Transforms/Inline/noinline.ll +++ b/test/Transforms/Inline/noinline.ll @@ -1,4 +1,4 @@ -; RUN: opt %s -inline -S | FileCheck %s +; RUN: opt -inline -S < %s | FileCheck %s ; PR6682 declare void @foo() nounwind diff --git a/test/Transforms/Inline/recursive.ll b/test/Transforms/Inline/recursive.ll index 5fe8d16..fe1c041 100644 --- a/test/Transforms/Inline/recursive.ll +++ b/test/Transforms/Inline/recursive.ll @@ -1,4 +1,4 @@ -; RUN: opt %s -inline -S | FileCheck %s +; RUN: opt -inline -S < %s | FileCheck %s target datalayout = "e-p:32:32:32-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-f80:128:128" target triple = "i386-apple-darwin10.0" |
