[graalvm-users] How to extend or create a new LLVM Bitcode Visitor
Roland Schatz
roland.schatz at oracle.com
Wed Feb 12 07:25:27 PST 2020
Hi Alberto,
On 2/12/20 2:52 PM, Alberto Barbaro wrote:
> Hi Roland,
> I think that's exactly what I need. I was able to build simpletool and
> to use Truffle so I'll adapt it to my need using LLVMTruffleInstrument
> instead TruffleInsturment. Does it sound correct to you?
I think either way is fine. It's probably a good idea to do your own
instrument, but you can start from the LLVMTracerInstrument, because it
already does something very similar to what you want to do.
>
> In addition I tried lli with the experimental options but I had the
> following errors:
>
> [...]
> ERROR: java.lang.IllegalStateException: The registerService method can
> only be called during the execution of the Env.createContext method.
> org.graalvm.polyglot.PolyglotException:
> java.lang.IllegalStateException: The registerService method can only
> be called during the execution of the Env.createContext method.
> at
> com.oracle.truffle.api.TruffleLanguage$Env.registerService(TruffleLanguage.java:2382)
> at
> com.oracle.truffle.llvm.instruments.trace.LLVMTracerInstrument.initialize(LLVMTracerInstrument.java:58)
> at
> com.oracle.truffle.llvm.runtime.LLVMContext.initialize(LLVMContext.java:276)
> [...]
From the stack trace, looks like you're on version 19.3, right? Then
that is a known bug, it's already fixed on current master and on the
20.0 release branch. You can use one of our dev builds
(https://github.com/graalvm/graalvm-ce-dev-builds/releases) in the
meantime. It will also be fixed in the upcoming 20.0 release.
- Roland
>
> Il giorno mer 12 feb 2020 alle ore 11:24 Roland Schatz
> <roland.schatz at oracle.com <mailto:roland.schatz at oracle.com>> ha scritto:
>
> Hi Alberto,
>
> On 2/11/20 6:40 PM, Alberto Barbaro wrote:
>> Hi Roland,
>> Yes you are right, that's the proper class name.
>>
>> Atm the goal for me is just to print all the instructions that
>> are executed In the LLVM IR firm if possible.
>
> The LLVMBitcodeInstructionVisitor is used just for parsing the
> bitcode. For runtime instrumentation, there is the Truffle
> instrumentation framework that might be better suited for what
> you're trying to do:
> https://www.graalvm.org/docs/graalvm-as-a-platform/implement-instrument/
> <https://urldefense.com/v3/__https://www.graalvm.org/docs/graalvm-as-a-platform/implement-instrument/__;!!GqivPVa7Brio!M_VQB1VYM2oG0tUtVXzAs5zr9rDHleDIiNnN14WgTynkSubDpWG5C3J0oIZzg4DLWps$>
>
> We already have a few instruments that you might find useful,
> either for using directly, or for building on top of them.
>
> You can already get a trace of all executed bitcode instructions:
>> $ llvm-dis hello.bc
>> $ lli --experimental-options --llvm.llDebug --llvm.traceIR hello.bc
>> [lli] >> Entering function @main at hello.ll:9:1 with
>> arguments:[StackPointer 0x7f62b3fff010 (Bounds: 0x7f62aefff010 -
>> 0x7f62b3fff010), 1, 0x55ad10a8ba48, 0x55ad10a8ba58]
>> [lli] >> hello.ll:10:1 -> %1 = alloca i32, align 4
>> [lli] >> hello.ll:11:1 -> store i32 0, i32* %1, align 4
>> [lli] >> hello.ll:12:1 -> %2 = call i32 (i8*, ...) @printf(i8*
>> getelementptr inbounds ([15 x i8], [15 x i8]* @.str, i32 0, i32
>> 0)), !dbg !13
>> Hello, World!
>> [lli] >> hello.ll:13:1 -> ret i32 0, !dbg !14
>> [lli] >> Leaving @main
>
> The code for that is mainly in the classes LLVMTracerInstrument
> and LLVMTraceNodeFactory. That might be an easier entry point for
> what you're trying to do.
>
>
>
> We also have a code coverage tool:
> https://www.graalvm.org/docs/reference-manual/tools/#code-coverage
> <https://urldefense.com/v3/__https://www.graalvm.org/docs/reference-manual/tools/*code-coverage__;Iw!!GqivPVa7Brio!M_VQB1VYM2oG0tUtVXzAs5zr9rDHleDIiNnN14WgTynkSubDpWG5C3J0oIZzo_rD3k0$>
> https://github.com/oracle/graal/tree/master/tools/src/com.oracle.truffle.tools.coverage/src/com/oracle/truffle/tools/coverage
> <https://urldefense.com/v3/__https://github.com/oracle/graal/tree/master/tools/src/com.oracle.truffle.tools.coverage/src/com/oracle/truffle/tools/coverage__;!!GqivPVa7Brio!M_VQB1VYM2oG0tUtVXzAs5zr9rDHleDIiNnN14WgTynkSubDpWG5C3J0oIZzHgM1A5U$>
>
> If you run it with regular "lli --coverage ...", you'll get C code
> covarage, but you can combine that with the "--llvm.llDebug"
> option to get bitcode coverage, too (see below for example output).
>
>
> I hope this helps,
> Roland
>
>
>
>> $ lli --coverage --coverage.Output=detailed
>> --experimental-options --llvm.llDebug hello.bc
>> Hello, World!
>> ------------------------------------------------------------------------------------------
>> Code coverage per line of code and what percent of each element
>> was covered during execution (per source)
>> + indicates the line is covered during execution
>> - indicates the line is not covered during execution
>> p indicates the line is part of a statement that was
>> incidentally covered during execution
>> e.g. a not-taken branch of a covered if statement
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> /home/roland/test/hello/hello.ll | 100.00% | 100.00% | 100.00%
>>
>> ; ModuleID = 'hello.bc'
>> source_filename = "hello.c"
>> target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
>> target triple = "x86_64-pc-linux-gnu"
>>
>> @.str = private unnamed_addr constant [15 x i8] c"Hello,
>> World!\0A\00", align 1
>>
>> ; Function Attrs: noinline nounwind optnone sspstrong uwtable
>> define dso_local i32 @main() #0 !dbg !9 {
>> + %1 = alloca i32, align 4
>> + store i32 0, i32* %1, align 4
>> + %2 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds
>> ([15 x i8], [15 x i8]* @.str, i32 0, i32 0)), !dbg !13
>> + ret i32 0, !dbg !14
>> }
>>
>> declare i32 @printf(i8*, ...) #1
>>
>> attributes #0 = { noinline nounwind optnone sspstrong uwtable
>> "correctly-rounded-divide-sqrt-fp-math"="false"
>> "disable-tail-calls"="false" "less-precise-fpmad"="false"
>> "min-legal-vector-width"="0" "no-frame-pointer-elim"="true"
>> "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false"
>> "no-jump-tables"="false" "no-nans-fp-math"="false"
>> "no-signed-zeros-fp-math"="false" "no-trapping-math"="false"
>> "stack-protector-buffer-size"="8" "target-cpu"="x86-64"
>> "target-features"="+fxsr,+mmx,+sse,+sse2,+x87"
>> "unsafe-fp-math"="false" "use-soft-float"="false" }
>> attributes #1 = {
>> "correctly-rounded-divide-sqrt-fp-math"="false"
>> "disable-tail-calls"="false" "less-precise-fpmad"="false"
>> "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"
>> "no-infs-fp-math"="false" "no-nans-fp-math"="false"
>> "no-signed-zeros-fp-math"="false" "no-trapping-math"="false"
>> "stack-protector-buffer-size"="8" "target-cpu"="x86-64"
>> "target-features"="+fxsr,+mmx,+sse,+sse2,+x87"
>> "unsafe-fp-math"="false" "use-soft-float"="false" }
>>
>> !llvm.dbg.cu
>> <https://urldefense.com/v3/__http://llvm.dbg.cu__;!!GqivPVa7Brio!M_VQB1VYM2oG0tUtVXzAs5zr9rDHleDIiNnN14WgTynkSubDpWG5C3J0oIZzO4ag0HM$>
>> = !{!0}
>> !llvm.module.flags = !{!3, !4, !5, !6, !7}
>> !llvm.ident = !{!8}
>>
>> !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1,
>> producer: "clang version 8.0.1 (tags/RELEASE_801/final)",
>> isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug,
>> enums: !2, nameTableKind: None)
>> !1 = !DIFile(filename: "hello.c", directory:
>> "/home/roland/test/hello")
>> !2 = !{}
>> !3 = !{i32 2, !"Dwarf Version", i32 4}
>> !4 = !{i32 2, !"Debug Info Version", i32 3}
>> !5 = !{i32 1, !"wchar_size", i32 4}
>> !6 = !{i32 7, !"PIC Level", i32 2}
>> !7 = !{i32 7, !"PIE Level", i32 2}
>> !8 = !{!"clang version 8.0.1 (tags/RELEASE_801/final)"}
>> !9 = distinct !DISubprogram(name: "main", scope: !1, file: !1,
>> line: 3, type: !10, scopeLine: 3, spFlags: DISPFlagDefinition,
>> unit: !0, retainedNodes: !2)
>> !10 = !DISubroutineType(types: !11)
>> !11 = !{!12}
>> !12 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
>> !13 = !DILocation(line: 4, column: 5, scope: !9)
>> !14 = !DILocation(line: 5, column: 5, scope: !9)
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> com.oracle.truffle.llvm.libraries.bitcode/src/crt0.c |
>> 100.00% | | 100.00%
>>
>> NO CONTENT AVAILABLE
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> com.oracle.truffle.llvm.libraries.bitcode/src/exit.c |
>> 87.50% | | 100.00%
>>
>> NO CONTENT AVAILABLE
>> ------------------------------------------------------------------------------------------
>
>> $ lli --coverage --coverage.Output=detailed hello.bc
>> Hello, World!
>> ------------------------------------------------------------------------------------------
>> Code coverage per line of code and what percent of each element
>> was covered during execution (per source)
>> + indicates the line is covered during execution
>> - indicates the line is not covered during execution
>> p indicates the line is part of a statement that was
>> incidentally covered during execution
>> e.g. a not-taken branch of a covered if statement
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> /home/roland/test/hello/hello.c | 100.00% | 100.00% | 100.00%
>>
>> #include <stdio.h>
>>
>> int main() {
>> + printf("Hello, World!\n");
>> + return 0;
>> }
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> com.oracle.truffle.llvm.libraries.bitcode/src/crt0.c |
>> 100.00% | | 100.00%
>>
>> NO CONTENT AVAILABLE
>> ------------------------------------------------------------------------------------------
>> Path |
>> Statements | Lines | Roots
>> com.oracle.truffle.llvm.libraries.bitcode/src/exit.c |
>> 87.50% | | 100.00%
>>
>> NO CONTENT AVAILABLE
>> ------------------------------------------------------------------------------------------
>
>
>>
>> If all goes well I'd like to work on a code coverage tool for the
>> bitcode.
>>
>> Thanks
>> Alberto
>>
>> On Tue, Feb 11, 2020, 16:02 Roland Schatz
>> <roland.schatz at oracle.com <mailto:roland.schatz at oracle.com>> wrote:
>>
>> Hi Alberto!
>>
>> I assume you mean the LLVMBitcodeInstructionVisitor class,
>> right? This class is part of the internal implementation of
>> the LLVM runtime in GraalVM, and not meant to be extended or
>> used outside of the LLVM runtime. This is not meant as API,
>> we might change that class without notice, even in minor
>> releases.
>>
>> What are you trying to achieve by subclassing it?
>>
>> - Roland
>>
>> On 2/11/20 4:31 PM, Alberto Barbaro wrote:
>>> Hi all,
>>> I've just started to use graalvm and I've noticed that the
>>> LLVMBitcodeVisitor class is market as final. I'd like to
>>> create my own visitor extending it... So I was wondering if
>>> there is a better approach rather then modifying a bit the
>>> source code. How would you recommend to do it?
>>>
>>> Thanks
>>> Alberto
>>>
>>> _______________________________________________
>>> GraalVM-Users mailing list
>>> GraalVM-Users at oss.oracle.com <mailto:GraalVM-Users at oss.oracle.com>
>>> https://oss.oracle.com/mailman/listinfo/graalvm-users
>>
>>
>> _______________________________________________
>> GraalVM-Users mailing list
>> GraalVM-Users at oss.oracle.com
>> <mailto:GraalVM-Users at oss.oracle.com>
>> https://oss.oracle.com/mailman/listinfo/graalvm-users
>>
>
> _______________________________________________
> GraalVM-Users mailing list
> GraalVM-Users at oss.oracle.com <mailto:GraalVM-Users at oss.oracle.com>
> https://oss.oracle.com/mailman/listinfo/graalvm-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://oss.oracle.com/pipermail/graalvm-users/attachments/20200212/cd660c31/attachment-0001.html
More information about the GraalVM-Users
mailing list