maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/catchpad-regmask.ll

145 lines
4.9 KiB
LLVM
Raw Normal View History

[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
; RUN: llc < %s | FileCheck %s
; Based on this code:
;
; extern "C" int array[4];
; extern "C" void global_array(int idx1, int idx2, int idx3) {
; try {
; array[idx1] = 111;
; throw;
; } catch (...) {
; array[idx2] = 222;
; }
; array[idx3] = 333;
; }
; extern "C" __declspec(dllimport) int imported;
; extern "C" void access_imported() {
; try {
; imported = 111;
; throw;
; } catch (...) {
; imported = 222;
; }
; imported = 333;
; }
target datalayout = "e-m:w-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-pc-windows-msvc18.0.0"
%eh.ThrowInfo = type { i32, i32, i32, i32 }
@array = external global [4 x i32], align 16
@imported = external dllimport global i32, align 4
; Function Attrs: uwtable
define void @global_array(i32 %idx1, i32 %idx2, i32 %idx3) #0 personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
entry:
%idxprom = sext i32 %idx1 to i64
%arrayidx = getelementptr inbounds [4 x i32], [4 x i32]* @array, i64 0, i64 %idxprom
store i32 111, i32* %arrayidx, align 4, !tbaa !2
invoke void @_CxxThrowException(i8* null, %eh.ThrowInfo* null) #1
to label %unreachable unwind label %catch.dispatch
catch.dispatch: ; preds = %entry
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
%cs1 = catchswitch within none [label %catch] unwind to caller
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
catch: ; preds = %catch.dispatch
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
%0 = catchpad within %cs1 [i8* null, i32 64, i8* null]
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
%idxprom1 = sext i32 %idx2 to i64
%arrayidx2 = getelementptr inbounds [4 x i32], [4 x i32]* @array, i64 0, i64 %idxprom1
store i32 222, i32* %arrayidx2, align 4, !tbaa !2
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
catchret from %0 to label %try.cont
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
try.cont: ; preds = %catch
%idxprom3 = sext i32 %idx3 to i64
%arrayidx4 = getelementptr inbounds [4 x i32], [4 x i32]* @array, i64 0, i64 %idxprom3
store i32 333, i32* %arrayidx4, align 4, !tbaa !2
ret void
unreachable: ; preds = %entry
unreachable
}
; CHECK-LABEL: global_array: # @global_array
; CHECK: pushq %rbp
; First array access
; CHECK: movslq %ecx, %[[idx:[^ ]*]]
; CHECK: leaq array(%rip), %[[base:[^ ]*]]
; CHECK: movl $111, (%[[base]],%[[idx]],4)
; Might throw an exception and return to below...
; CHECK: callq _CxxThrowException
; Third array access must remat the address of array
; CHECK: movslq {{.*}}, %[[idx:[^ ]*]]
; CHECK: leaq array(%rip), %[[base:[^ ]*]]
; CHECK: movl $333, (%[[base]],%[[idx]],4)
; CHECK: popq %rbp
; CHECK: retq
; CHECK: "?catch$2@?0?global_array@4HA":
; CHECK: pushq %rbp
; CHECK: movslq {{.*}}, %[[idx:[^ ]*]]
; CHECK: leaq array(%rip), %[[base:[^ ]*]]
; CHECK: movl $222, (%[[base]],%[[idx]],4)
; CHECK: popq %rbp
; CHECK: retq # CATCHRET
declare void @_CxxThrowException(i8*, %eh.ThrowInfo*)
declare i32 @__CxxFrameHandler3(...)
; Function Attrs: uwtable
define void @access_imported() #0 personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
entry:
store i32 111, i32* @imported, align 4, !tbaa !2
invoke void @_CxxThrowException(i8* null, %eh.ThrowInfo* null) #1
to label %unreachable unwind label %catch.dispatch
catch.dispatch: ; preds = %entry
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
%cs1 = catchswitch within none [label %catch] unwind to caller
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
catch: ; preds = %catch.dispatch
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
%0 = catchpad within %cs1 [i8* null, i32 64, i8* null]
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
store i32 222, i32* @imported, align 4, !tbaa !2
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
catchret from %0 to label %try.cont
[WinEH] Mark funclet entries and exits as clobbering all registers Summary: In this implementation, LiveIntervalAnalysis invents a few register masks on basic block boundaries that preserve no registers. The nice thing about this is that it prevents the prologue inserter from thinking it needs to spill all XMM CSRs, because it doesn't see any explicit physreg defs in the MI. Reviewers: MatzeB, qcolombet, JosephTremoulet, majnemer Subscribers: MatzeB, llvm-commits Differential Revision: http://reviews.llvm.org/D14407 llvm-svn: 252318
2015-11-07 01:06:38 +08:00
try.cont: ; preds = %catch
store i32 333, i32* @imported, align 4, !tbaa !2
ret void
unreachable: ; preds = %entry
unreachable
}
; CHECK-LABEL: access_imported: # @access_imported
; CHECK: pushq %rbp
; CHECK: movq __imp_imported(%rip), %[[base:[^ ]*]]
; CHECK: movl $111, (%[[base]])
; Might throw an exception and return to below...
; CHECK: callq _CxxThrowException
; Third access must reload the address of imported
; CHECK: movq __imp_imported(%rip), %[[base:[^ ]*]]
; CHECK: movl $333, (%[[base]])
; CHECK: popq %rbp
; CHECK: retq
; CHECK: "?catch$2@?0?access_imported@4HA":
; CHECK: pushq %rbp
; CHECK: movq __imp_imported(%rip), %[[base:[^ ]*]]
; CHECK: movl $222, (%[[base]])
; CHECK: popq %rbp
; CHECK: retq # CATCHRET
attributes #0 = { uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { noreturn }
!llvm.module.flags = !{!0}
!llvm.ident = !{!1}
!0 = !{i32 1, !"PIC Level", i32 2}
!1 = !{!"clang version 3.8.0 "}
!2 = !{!3, !3, i64 0}
!3 = !{!"int", !4, i64 0}
!4 = !{!"omnipotent char", !5, i64 0}
!5 = !{!"Simple C/C++ TBAA"}