[llvm-dev] [RFC][ThinLTO] llvm-dis ThinLTO summary dump format
Peter Collingbourne via llvm-dev
llvm-dev at lists.llvm.org
Fri Jun 2 14:01:23 PDT 2017
On Fri, Jun 2, 2017 at 1:51 PM, Teresa Johnson <tejohnson at google.com> wrote:
>
>
> On Fri, Jun 2, 2017 at 1:16 PM, Peter Collingbourne <peter at pcc.me.uk>
> wrote:
>
>> On Fri, Jun 2, 2017 at 12:49 PM, Teresa Johnson <tejohnson at google.com>
>> wrote:
>>
>>>
>>>
>>> On Fri, Jun 2, 2017 at 12:39 PM, Peter Collingbourne <peter at pcc.me.uk>
>>> wrote:
>>>
>>>> On Fri, Jun 2, 2017 at 12:29 PM, Teresa Johnson <tejohnson at google.com>
>>>> wrote:
>>>>
>>>>> On Fri, Jun 2, 2017 at 12:18 PM, Peter Collingbourne via llvm-dev <
>>>>> llvm-dev at lists.llvm.org> wrote:
>>>>>
>>>>>> Why do we need a custom dumping format for the summary? Since we
>>>>>> already need the YAML format anyway, wouldn't it be better to extend that
>>>>>> to cover the entire summary?
>>>>>>
>>>>>
>>>>> IMO it is useful/convenient to be able to see the summary in the
>>>>> llvm-dis output.
>>>>>
>>>>
>>>> There are two things:
>>>> 1) the format of the summary
>>>> 2) where to add the dumper
>>>> I'm not too concerned about 2 (I wouldn't add it to llvm-dis, but
>>>> rather as a subcommand to llvm-lto2, but I don't have a strong opinion
>>>> about that).
>>>>
>>>> So that leaves the text based summary format. We need a format that is
>>>> easily parseable so that we can test components that manipulate summaries
>>>> in isolation. That is what the YAML format is. Why add a second format that
>>>> is not parseable?
>>>>
>>>
>>> So in this case it is purely for visual inspection - I suggested that
>>> Charles add this as LLVM assembly comments, to make it clear that we can't
>>> currently reconstruct the summary if the LLVM assembly file is re-assembled
>>> (we currently need to rebuild the summary from scratch). I know you have
>>> added support for inputing the type id summaries in YAML form for testing
>>> purposes, since they were/are not being generated from IR during the
>>> current module summary index builder, but we have the opposite issue with
>>> the rest of the summaries (no serialization, but can build from IR).
>>>
>>
>> Again: we could address that issue by adding the rest of the summary to
>> the YAML serialization, no? I don't want to have to update the tests and
>> the YAML dumper and the new-format dumper when I make a change to the
>> summary format.
>>
>
> What is the option for emitting the summary in YAML (I know it is
> incomplete, but just so we can see what it currently looks like)? Where
> does it get emitted?
>
There is no option currently, but I've attached a quick patch that adds a
dump-summary subcommand to llvm-lto2, so you can see what it looks like.
You can use it like this:
$ ra/bin/llvm-lto2 dump-summary
ra/test/ThinLTO/X86/Output/deadstrip.ll.tmp1.bc
---
GlobalValueMap:
2305701868152283314:
- TypeTests: [ ]
7342339837106705152:
- TypeTests: [ ]
10474011137621744663:
- TypeTests: [ ]
15822663052811949562:
- TypeTests: [ ]
16434608426314478903:
- TypeTests: [ ]
TypeIdMap:
...
Peter
> Teresa
>
>
>>
>> Peter
>>
>>
>>>
>>> Teresa
>>>
>>>
>>>> Peter
>>>>
>>>> Teresa
>>>>>
>>>>>
>>>>>> Peter
>>>>>>
>>>>>> On Fri, Jun 2, 2017 at 8:46 AM, Charles Saternos via llvm-dev <
>>>>>> llvm-dev at lists.llvm.org> wrote:
>>>>>>
>>>>>>> Hey all,
>>>>>>>
>>>>>>> Below is the proposed format for the dump of the ThinLTO module
>>>>>>> summary in the llvm-dis utility:
>>>>>>>
>>>>>>> > ../build/bin/llvm-dis t.o && cat t.o.ll
>>>>>>> ; ModuleID = '2.o'
>>>>>>> source_filename = "2.ll"
>>>>>>> target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
>>>>>>> target triple = "x86_64-unknown-linux-gnu"
>>>>>>>
>>>>>>> @X = constant i32 42, section "foo", align 4
>>>>>>>
>>>>>>> @a = weak alias i32, i32* @X
>>>>>>>
>>>>>>> define void @afun() {
>>>>>>> %1 = load i32, i32* @a
>>>>>>> ret void
>>>>>>> }
>>>>>>>
>>>>>>> define void @testtest() {
>>>>>>> tail call void @boop()
>>>>>>> ret void
>>>>>>> }
>>>>>>>
>>>>>>> declare void @boop()
>>>>>>>
>>>>>>> ; Module summary:
>>>>>>> ; testtest (External linkage)
>>>>>>> ; Function (2 instructions)
>>>>>>> ; Calls: boop
>>>>>>> ; X (External linkage)
>>>>>>> ; Global Variable
>>>>>>> ; afun (External linkage)
>>>>>>> ; Function (2 instructions)
>>>>>>> ; Refs:
>>>>>>> ; a
>>>>>>> ; a (Weak any linkage)
>>>>>>> ; Alias (aliasee X)
>>>>>>>
>>>>>>> I've implemented the above format in the llvm-dis utility, since
>>>>>>> there currently isn't really a way of getting ThinLTO summaries in a
>>>>>>> human-readable format.
>>>>>>>
>>>>>>> Let me know what you think of this format, and what information you
>>>>>>> think should be added/removed.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Charles
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> LLVM Developers mailing list
>>>>>>> llvm-dev at lists.llvm.org
>>>>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> --
>>>>>> Peter
>>>>>>
>>>>>> _______________________________________________
>>>>>> LLVM Developers mailing list
>>>>>> llvm-dev at lists.llvm.org
>>>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Teresa Johnson | Software Engineer | tejohnson at google.com |
>>>>> 408-460-2413 <(408)%20460-2413>
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> --
>>>> Peter
>>>>
>>>
>>>
>>>
>>> --
>>> Teresa Johnson | Software Engineer | tejohnson at google.com |
>>> 408-460-2413 <(408)%20460-2413>
>>>
>>
>>
>>
>> --
>> --
>> Peter
>>
>
>
>
> --
> Teresa Johnson | Software Engineer | tejohnson at google.com |
> 408-460-2413 <(408)%20460-2413>
>
--
--
Peter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170602/0ef51500/attachment.html>
-------------- next part --------------
diff --git a/llvm/tools/llvm-lto2/llvm-lto2.cpp b/llvm/tools/llvm-lto2/llvm-lto2.cpp
index 3d2643db85bd..395160d1b0af 100644
--- a/llvm/tools/llvm-lto2/llvm-lto2.cpp
+++ b/llvm/tools/llvm-lto2/llvm-lto2.cpp
@@ -17,8 +17,10 @@
//===----------------------------------------------------------------------===//
#include "llvm/LTO/Caching.h"
+#include "llvm/Bitcode/BitcodeReader.h"
#include "llvm/CodeGen/CommandFlags.h"
#include "llvm/IR/DiagnosticPrinter.h"
+#include "llvm/IR/ModuleSummaryIndexYAML.h"
#include "llvm/LTO/LTO.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/FileSystem.h"
@@ -343,6 +345,19 @@ static int dumpSymtab(int argc, char **argv) {
return 0;
}
+static int dumpSummary(int argc, char **argv) {
+ for (StringRef F : make_range(argv + 1, argv + argc)) {
+ std::unique_ptr<MemoryBuffer> MB = check(MemoryBuffer::getFile(F), F);
+ std::unique_ptr<ModuleSummaryIndex> Index =
+ check(getModuleSummaryIndex(*MB), F);
+
+ yaml::Output Out(outs());
+ Out << *Index;
+ }
+
+ return 0;
+}
+
int main(int argc, char **argv) {
InitializeAllTargets();
InitializeAllTargetMCs();
@@ -360,6 +375,8 @@ int main(int argc, char **argv) {
argv[1] = argv[0];
if (Subcommand == "dump-symtab")
return dumpSymtab(argc - 1, argv + 1);
+ if (Subcommand == "dump-summary")
+ return dumpSummary(argc - 1, argv + 1);
if (Subcommand == "run")
return run(argc - 1, argv + 1);
return usage();
More information about the llvm-dev
mailing list