Enable JitDisasm in Release by EgorBo · Pull Request #73365 · dotnet/runtime

EgorBo · 2022-08-04T11:02:23Z

This PR enables DOTNET_JitDisasm for Release config.

It doesn't print lcl tables, EH and gc info (for now).

Adds ~30kb to the binary size of jit.

The only noticeable impact on Release is that instrDesc struct is now 8bytes bigger (_idDebugOnlyInfo) - but I might fix it with a standalone hash-table which will be populated only when JitDisasm variable is not empty.

Example:

publicclassTests{staticintstaticField;stringinstanceField;publicvirtualvoidVirtualCall(){}publicvoidTest(string[]args,IDisposabled){// control flowif(args?.Length==0){Console.WriteLine("field: "+staticField);Console.WriteLine(instanceField);}else{// SIMD (data section)Console.WriteLine(Vector128.Create(1,2,3,4));}// Virtual callVirtualCall();// Interface calld?.Dispose();// Jump tableswitch(args.Length){case1:Console.WriteLine("1");break;case2:Console.WriteLine("2");break;case3:Console.WriteLine("3");break;case4:Console.WriteLine("5");break;case6:Console.WriteLine("7");break;}Console.ReadKey();}}

DOTNET_JitDisasm=Test: (DOTNET_JitDiffableDasm is 0 by default, hence, raw bytes, addresses)

; Assembly listing for method Tests:Test(ref,IDisposable):this; Emitting BLENDED_CODE for X64 CPU with AVX - Windows; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO dataG_M000_IG01: ;; offset=0000H57pushrdi56pushrsi53pushrbx 4883EC30 subrsp,48 C5F877 vzeroupper 488BF1 movrsi,rcx 488BFA movrdi,rdx 498BD8 movrbx,r8G_M000_IG02: ;; offset=0013H 4885FF testrdi,rdi 743D je SHORT G_M000_IG04G_M000_IG03: ;; offset=0018H 837F0800 cmp dword ptr [rdi+8],07537jne SHORT G_M000_IG04 8B0DF8B40C00 movecx, dword ptr [(reloc 0x7ffd9745c06c)] FF15560A1100 call[Number:Int32ToDecStr(int):String] 488BD0 movrdx,rax 48B9D820804A0C020000 movrcx,0x20C4A8020D8 488B09 movrcx, gword ptr [rcx] FF15D0100900 call[String:Concat(String,String):String] 488BC8 movrcx,rax FF15FFAE0F00 call[Console:WriteLine(String)] 488B4E08 movrcx, gword ptr [rsi+8] FF15F5AE0F00 call[Console:WriteLine(String)] EB27 jmp SHORT G_M000_IG05G_M000_IG04: ;; offset=0055H 48B978D64797FD7F0000 movrcx,0x7FFD9747D678 E8FCAEAE5F call CORINFO_HELP_NEWSFAST C5F91005E4000000 vmovupdxmm0, xmmword ptr [reloc @RWD00] C5F9114008 vmovupd xmmword ptr [rax+8],xmm0 488BC8 movrcx,rax FF15B6AE0F00 call[Console:WriteLine(Object)]G_M000_IG05: ;; offset=007AH 488BCE movrcx,rsi 488B06 movrax, qword ptr [rsi] 488B4040 movrax, qword ptr [rax+64] FF5020 call[rax+32]Tests:VirtualCall():this 4885DB testrbx,rbx7410je SHORT G_M000_IG07G_M000_IG06: ;; offset=008CH 488BCB movrcx,rbx 49BB08001E97FD7F0000 movr11,0x7FFD971E0008 41FF13 call[r11]IDisposable:Dispose():thisG_M000_IG07: ;; offset=009CH 8B4F08 movecx, dword ptr [rdi+8] FFC9 dececx 83F905 cmpecx,5 0F877F000000 ja G_M000_IG13 8BC9 movecx,ecx 488D05AD000000 learax,[reloc @RWD16] 8B0488 moveax, dword ptr [rax+4*rcx] 488D1556FFFFFF leardx, G_M000_IG02 4803C2 addrax,rdx FFE0 jmpraxG_M000_IG08: ;; offset=00C2H 48B9E020804A0C020000 movrcx,0x20C4A8020E0 488B09 movrcx, gword ptr [rcx] FF1573AE0F00 call[Console:WriteLine(String)] EB52 jmp SHORT G_M000_IG13G_M000_IG09: ;; offset=00D7H 48B9E820804A0C020000 movrcx,0x20C4A8020E8 488B09 movrcx, gword ptr [rcx] FF155EAE0F00 call[Console:WriteLine(String)] EB3D jmp SHORT G_M000_IG13G_M000_IG10: ;; offset=00ECH 48B9F020804A0C020000 movrcx,0x20C4A8020F0 488B09 movrcx, gword ptr [rcx] FF1549AE0F00 call[Console:WriteLine(String)] EB28 jmp SHORT G_M000_IG13G_M000_IG11: ;; offset=0101H 48B9F820804A0C020000 movrcx,0x20C4A8020F8 488B09 movrcx, gword ptr [rcx] FF1534AE0F00 call[Console:WriteLine(String)] EB13 jmp SHORT G_M000_IG13G_M000_IG12: ;; offset=0116H 48B90021804A0C020000 movrcx,0x20C4A802100 488B09 movrcx, gword ptr [rcx] FF151FAE0F00 call[Console:WriteLine(String)]G_M000_IG13: ;; offset=0129H 488D4C2420 learcx, bword ptr 33D2 xoredx,edx FF158AEE0F00 call[ConsolePal:ReadKey(bool):ConsoleKeyInfo]90nopG_M000_IG14: ;; offset=0137H 4883C430 addrsp,48 5B poprbx 5E poprsi 5F poprdi C3 retRWD00 dq 0000000200000001h,0000000400000003hRWD16 dd 000000AFh ; case G_M000_IG08dd000000C4h ; case G_M000_IG09dd000000D9h ; case G_M000_IG10dd000000EEh ; case G_M000_IG11dd00000116h ; case G_M000_IG13dd00000103h ; case G_M000_IG12; Total bytes of code 319

ghost · 2022-08-04T11:02:40Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Issue Details

This PR enables DOTNT_JitDisasm, DOTNET_NgenDisasm and DOTNET_JitDiffableDasm for Release config.

It doesn't print lcl tables and gc info (for now).

Adds ~30kb to the binary size of jit.

The only noticeable impact on Release is that instrDesc struct is now 8bytes bigger (_idDebugOnlyInfo) - but I might fix it with a standalone hash-table which will be populated only when JitDisasm variable is not empty.

Example:

publicclassTests{staticintstaticField;stringinstanceField;publicvirtualvoidVirtualCall(){}publicvoidTest(string[]args,IDisposabled){// control flowif(args?.Length==0){Console.WriteLine("field: "+staticField);Console.WriteLine(instanceField);}else{// SIMD (data section)Console.WriteLine(Vector128.Create(1,2,3,4));}// Virtual callVirtualCall();// Interface calld?.Dispose();// Jump tableswitch(args.Length){case1:Console.WriteLine("1");break;case2:Console.WriteLine("2");break;case3:Console.WriteLine("3");break;case4:Console.WriteLine("5");break;case6:Console.WriteLine("7");break;}Console.ReadKey();}}

; Emitting BLENDED_CODE for X64 CPU with AVX - Windows; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO dataIG_01:57pushrdi56pushrsi53pushrbx 4883EC30 subrsp,48 C5F877 vzeroupper 488BF1 movrsi,rcx 488BFA movrdi,rdx 498BD8 movrbx,r8IG_02: 4885FF testrdi,rdi 743D je SHORT G_M000_IG04IG_03: 837F0800 cmp dword ptr [rdi+8],07537jne SHORT G_M000_IG04 8B0DF8B40C00 movecx, dword ptr [(reloc 0x7ffd2d2dc06c)] FF15560A1100 call[Number:Int32ToDecStr(int):String] 488BD0 movrdx,rax 48B9D820C0EF74020000 movrcx,0x274EFC020D8 488B09 movrcx, gword ptr [rcx] FF15D0100900 call[String:Concat(String,String):String] 488BC8 movrcx,rax FF15FFAE0F00 call[Console:WriteLine(String)] 488B4E08 movrcx, gword ptr [rsi+8] FF15F5AE0F00 call[Console:WriteLine(String)] EB27 jmp SHORT G_M000_IG05IG_04: 48B978D62F2DFD7F0000 movrcx,0x7FFD2D2FD678 E8FCAEAF5F call CORINFO_HELP_NEWSFAST C5F91005E4000000 vmovupdxmm0, xmmword ptr [reloc @RWD00] C5F9114008 vmovupd xmmword ptr [rax+8],xmm0 488BC8 movrcx,rax FF15B6AE0F00 call[Console:WriteLine(Object)]IG_05: 488BCE movrcx,rsi 488B06 movrax, qword ptr [rsi] 488B4040 movrax, qword ptr [rax+64] FF5020 call[rax+32]Tests:VirtualCall():this 4885DB testrbx,rbx7410je SHORT G_M000_IG07IG_06: 488BCB movrcx,rbx 49BB0800062DFD7F0000 movr11,0x7FFD2D060008 41FF13 call[r11]IDisposable:Dispose():thisIG_07: 8B4F08 movecx, dword ptr [rdi+8] FFC9 dececx 83F905 cmpecx,5 0F877F000000 ja G_M000_IG13 8BC9 movecx,ecx 488D05AD000000 learax,[reloc @RWD16] 8B0488 moveax, dword ptr [rax+4*rcx] 488D1556FFFFFF leardx, G_M000_IG02 4803C2 addrax,rdx FFE0 jmpraxIG_08: 48B9E020C0EF74020000 movrcx,0x274EFC020E0 488B09 movrcx, gword ptr [rcx] FF1573AE0F00 call[Console:WriteLine(String)] EB52 jmp SHORT G_M000_IG13IG_09: 48B9E820C0EF74020000 movrcx,0x274EFC020E8 488B09 movrcx, gword ptr [rcx] FF155EAE0F00 call[Console:WriteLine(String)] EB3D jmp SHORT G_M000_IG13IG_10: 48B9F020C0EF74020000 movrcx,0x274EFC020F0 488B09 movrcx, gword ptr [rcx] FF1549AE0F00 call[Console:WriteLine(String)] EB28 jmp SHORT G_M000_IG13IG_11: 48B9F820C0EF74020000 movrcx,0x274EFC020F8 488B09 movrcx, gword ptr [rcx] FF1534AE0F00 call[Console:WriteLine(String)] EB13 jmp SHORT G_M000_IG13IG_12: 48B90021C0EF74020000 movrcx,0x274EFC02100 488B09 movrcx, gword ptr [rcx] FF151FAE0F00 call[Console:WriteLine(String)]IG_13: 488D4C2420 learcx, bword ptr 33D2 xoredx,edx FF158AEE0F00 call[ConsolePal:ReadKey(bool):ConsoleKeyInfo]90nopIG_14: 4883C430 addrsp,48 5B poprbx 5E poprsi 5F poprdi C3 retRWD00 dq 0000000200000001h,0000000400000003hRWD16 dd 000000AFh ; case IG_08dd000000C4h ; case IG_09dd000000D9h ; case IG_10dd000000EEh ; case IG_11dd00000116h ; case IG_13dd00000103h ; case IG_12; Total bytes of code 319

Author:	EgorBo
Assignees:	-
Labels:	`area-CodeGen-coreclr`
Milestone:	-

stephentoub · 2022-08-04T11:09:03Z

This would be super helpful.

hez2010 · 2022-08-04T11:12:10Z

Does this PR also cover crossgen2?

EgorBo · 2022-08-04T11:19:32Z

Does this PR also cover crossgen2?

Yes works for R2R. And, probably, for NativeAOT as well but afair it redirects stdout cc @MichalStrehovsky

EgorBo · 2022-08-04T12:32:09Z

One noticeable advantage over sharplab for end-users is the ability to print names for helper calls, e.g.

publicclassC{objecto;publicvoidM(object[]a){a[0]=42;o=a;}}

Sharplab:

JitDisasm:

; Assembly listing for method C:M(ref):this; Emitting BLENDED_CODE for X64 CPU with AVX - Windows; optimized code; rsp based frame; partially interruptible; No PGO dataG_M000_IG01: ;; offset=0000H57pushrdi56pushrsi 4883EC28 subrsp,40 488BF9 movrdi,rcx 488BF2 movrsi,rdxG_M000_IG02: ;; offset=000CH 48B958E4DC2CFD7F0000 movrcx,0x7FFD2CDCE458 E835AFB15F call CORINFO_HELP_NEWSFAST C740082A000000 mov dword ptr [rax+8],42 4C8BC0 movr8,rax 488BCE movrcx,rsi 33D2 xoredx,edx E891F4FFFF call CORINFO_HELP_ARRADDR_ST 488D4F08 learcx, bword ptr [rdi+8] 488BD6 movrdx,rsi E875F4E2FF call CORINFO_HELP_ASSIGN_REF90nopG_M000_IG03: ;; offset=003CH 4883C428 addrsp,40 5E poprsi 5F poprdi C3 ret; Total bytes of code 67

Perhaps, it will be useful for sharplab too

JulieLeeMSFT · 2022-08-04T13:09:53Z

@EgorBo do you plan to merge this to .NET 7?

EgorBo · 2022-08-04T13:13:18Z

@EgorBo do you plan to merge this to .NET 7?

It probably does not meet the bar but I'd love to merge it in .NET 7.0, but no strong opinion

stephentoub · 2022-08-04T13:14:24Z

It probably does not meet the bar but I'd love to merge it in .NET 7.0, but no strong opinion

FWIW, I would love that, too :) (as long as there's no meaningful perf regressions associated with it and there are no concerns about the data this exposes).

EgorBo · 2022-08-04T13:20:24Z

no concerns about the data this exposes

Yep, we discussed this one too and came to conclusion that there is nothing to be afraid of and our competitors provide this ability too

jkotas · 2022-08-04T13:35:25Z

Does this PR also cover crossgen2?

Yes works for R2R. And, probably, for NativeAOT

NativeAOT compiler and crossgen use exact same infrastructure. If it works for one, it should work for the other one too.

You have to specify the option on the command line instead of env variable. Is that correct?

DOTNET_NgenDisasm

We should not be injecting new life into NGen name. Can we rename this so that it does not have NGen in the name?

I am wondering why we need both NgenDisasm and JitDisasm. Can we have just one that kicks in everywhere?

EgorBo · 2022-08-04T13:38:19Z

You have to specify the option on the command line instead of env variable. Is that correct?

Correct, e.g. --codegenopt NgenDisasm="Test"

We should not be injecting new life into NGen name. Can we rename this so that it does not have NGen in the name?

I fully agree, should I still keep NgenDisasm for backward compatibility?

jkotas · 2022-08-04T13:45:08Z

I fully agree, should I still keep NgenDisasm for backward compatibility?

There is no backward compatibility to be worried about. We should delete all configs that start with NGen and use the Jit configs everywhere.

If we have any infra that use the NGen configs (I do not see any), it needs to be updated.

I see a few places that clear COMPlus_Ngen* in superpmi. They should be deleted. They are not doing anything useful anyway since NGen configs are not read from the environment.

EgorBo · 2022-08-04T13:54:26Z

There is no backward compatibility to be worried about.

https://godbolt.org/ relies on it to print ASM for C#, but I assume it will only benefit from moving to release .NET as it won't have to compile .NET from sources and use public previews/daily builds of Release .NET instead cc @hez2010

hez2010 · 2022-08-04T23:37:14Z

There is no backward compatibility to be worried about.
https://godbolt.org/ relies on it to print ASM for C#.

There is nothing to worry about. We can definitely setup a new net7.0 config for godbolt while keeping net6.0 as is.

EgorBo · 2022-08-08T09:34:01Z

Outerloop pipelines failures seem unrelated:

Fuzzlyn: same here, e.g. x86 failure is x86: Assertion failed '!"Too many unreachable block removal loops"' during 'Global local var liveness' #70786
coreclr Pri0 Runtime Tests Run Linux arm checked failure is GC/API/GC/GetGCMemoryInfo/GetGCMemoryInfo.sh test failing intermittently on CoreCLR Linux ARM32 #73247
runtime-coreclr crossgen2 outerloop failures are infra-related e.g. error : A call to an Azure DevOps api returned 401, which may indicate a bad 'System.AccessToken' value.
runtime-coreclr gcstress0x3-gcstress0xc are known issues, e.g. Test failure JIT/Methodical/Arrays/lcs/lcs2_r/lcs2_r.cmd #69092 (comment)
runtime-coreclr outerloop failures seem to be infra-related '"C:\h\w\A18C08B3\p\dotnet.exe"' is not recognized as an internal or external command

jakobbotsch · 2022-08-08T09:47:07Z

I have started a non-PR run of runtimehere, which will get us some more coverage for release JIT with the libraries tests.

The change is somewhat low risk for release since when m_debugInfoSize = 0, the layout just collapses back to what it was before. So even if we forgot to account for it somewhere, it is unlikely that will cause problems for release non-disasm builds.

My main concern would be that we have an idDebugOnlyInfo() use somewhere that is not conditioned on the debug info being present, but I've audited the uses and wasn't able to find any way that could happen.

EgorBo · 2022-08-08T09:53:32Z

My main concern would be that we have an idDebugOnlyInfo() use somewhere that is not conditioned on the debug info being present, but I've audited the uses and wasn't able to find any way that could happen.

Right, I audited those too (and inserted a failfast if we entered them with .disAsm = false) locally

EgorBo · 2022-08-08T15:22:10Z

@dotnet/jit-contrib @jakobbotsch @BruceForstall PTAL

AndyAyersMS · 2022-08-08T16:08:55Z

Seems like we should allow users to specify JitDiffableDasm too?

If you are live debugging it would be nice to see the actual addresses and such.

runtime/src/coreclr/jit/disasm.cpp

Lines 1461 to 1467 in e71a958

 #ifdef DEBUG 
// Should we make it diffable?
 disDiffable = disComp->opts.dspDiffable; 
 #else// !DEBUG
// NOTE: non-debug builds are always diffable!
 disDiffable = true; 
 #endif// !DEBUG

Might also be nice to be able to control JitDasmWithAddress and JitDasmWithAlignmentBoundaries and perhaps even JitDisasmWithGC and JitDisasmWithDebugInfo.

There are a whole bunch more like this: unwind, eh, ... but they might take a lot more work.

omariom · 2022-08-08T16:21:39Z

@EgorBo
Can we assume that it will always start with "; Assembly listing for method " and end with "; Total bytes of code "?

EgorBo · 2022-08-08T17:45:05Z

Seems like we should allow users to specify JitDiffableDasm too?
If you are live debugging it would be nice to see the actual addresses and such.
runtime/src/coreclr/jit/disasm.cpp
Lines 1461 to 1467 in e71a958
#ifdef DEBUG
// Should we make it diffable?
disDiffable = disComp->opts.dspDiffable;
#else// !DEBUG
// NOTE: non-debug builds are always diffable!
disDiffable = true;
#endif// !DEBUG
Might also be nice to be able to control JitDasmWithAddress and JitDasmWithAlignmentBoundaries and perhaps even JitDisasmWithGC and JitDisasmWithDebugInfo.
There are a whole bunch more like this: unwind, eh, ... but they might take a lot more work.

Yes, I had to draw the line somewhere for the initial impl since, obviously, supporting all of them is a lot more work, I am going to take a look at JitDiffableDasm separately if you don't mind. JitDasmWithAlignmentBoundaries is also nice to have - it draws nice lines 😄

EgorBo · 2022-08-08T17:46:28Z

@EgorBo Can we assume that it will always start with "; Assembly listing for method " and end with "; Total bytes of code "?

Good question, I assume some 3rd party tools are going to depend on it (parse) and we should not ever change it
So I personally is fine with the current format, if anyone has better ideas - feel free to change now

MichalPetryka · 2022-08-08T17:50:31Z

@EgorBo Can we assume that it will always start with "; Assembly listing for method " and end with "; Total bytes of code "?
Good question, I assume some 3rd party tools are going to depend on it (parse) and we should not ever change it So I personally is fine with the current format, if anyone has better ideas - feel free to change now

There could be a separate env var that'd enable a special stable parsable output, something like git does for tools invoking it.

BruceForstall

Looks good. It would be nice if the number of places doing + m_debugInfoSize were more minimized, via wrapper functions or the like.

It would also be nice if some of the additional controls were available, as previously suggested, like JitStdOutFile, JitDiffableDasm, etc.

BruceForstall · 2022-08-09T16:45:08Z

 #endif // MULTIREG_HAS_SECOND_GC_RET
 };

+ // TODO-Cleanup: Uses of stack-allocated instrDescs should be refactored to be unnecessary.


Why should we eliminate stack-allocated instrDescs? It seems perfectly reasonable to allow them.

We are creating these temporary instrDesc simply to be able to call existing methods to emit code/do printing. That seems like a code smell that the codegen/printing methods themselves should be refactored so that we can use the specific parts we want directly instead of having to indirect through an instrDesc structure. BTW, there's just three uses of this pattern:
One in emitxarch.cpp that wants to call an emit method. It already has this same cleanup comment.
One in emitarm.cpp and emitarm64.cpp for printing LARGEJMP instructions.

BruceForstall · 2022-08-09T16:48:03Z

+ {
+ private:
+ instrDescDebugInfo* idDebugInfo;
+ alignas(alignof(T)) char idStorage[sizeof(T)];


Given the static_assert_no_msg below, isn't this alignas not useful? Or equivalent to alignas(sizeof(instrDescDebugInfo*))?

In theory, some future instrDesc could have larger-than-pointer alignment requirement. That would certainly require some work to support and I just want it to be noisy that this is a place that would need to be fixed.

kunalspathak · 2022-08-09T17:03:05Z

Started to review, but it seems we are not printing PerfScore yet?

kunalspathak · 2022-08-09T17:21:18Z

It would be nice if the number of places doing + m_debugInfoSize were more minimized, via wrapper functions or the like.

I agree with @BruceForstall . There are too many places where we are doing + m_debugInfoSize and it is easy to forget doing that in future. A wrapper function will be helpful.

kunalspathak

LGTM.

Consolidate most uses of m_debugInfoSize into these functions.

jakobbotsch · 2022-08-09T18:01:48Z

I pushed a commit that adds emitFirstInstrDesc and emitAdvanceInstrDesc and uses these in place of the manual castto/buffer accesses, which consolidates most uses of m_debugInfoSize.

kunalspathak

LGTM

BruceForstall · 2022-08-09T18:09:46Z

I pushed a commit that adds emitFirstInstrDesc and emitAdvanceInstrDesc and uses these in place of the manual castto/buffer accesses, which consolidates most uses of m_debugInfoSize.

The functions are a nice improvement.

jakobbotsch · 2022-08-09T19:02:10Z

Does the JIT dll import table change? (I.e., are there any new dependencies in the Release build?)

We get two new IAT entries on Windows: __stdio_common_vsprintf_s and strnlen.

EgorBo · 2022-08-10T14:58:57Z

Merging now. We're going to keep eyes on TP-related benchmarks (e.g. "Crossgen2 Throughput Trends" page) in coming weeks.

Kudos to @jakobbotsch for mitigating all the initial TP/Memory regressions 🎉

EgorBo added 6 commits August 2, 2022 01:04

Use liberal vn for tls indirects
bf756bc

Merge branch 'main' of github.com:dotnet/runtime into main
c33bd4b

Merge branch 'main' of github.com:dotnet/runtime into main
e9164de

Initial impl
24fc8a6

Enable JitDisasm in Release
75db099

Merge branch 'main' of github.com:dotnet/runtime into jit-disasm-release
0dfb332

ghost added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Aug 4, 2022

ghost assigned EgorBo Aug 4, 2022

jakobbotsch reviewed Aug 4, 2022
View reviewed changes

Comment threadsrc/coreclr/jit/emitxarch.cpp Outdated

Fix compilation errors
98a2a9d

EgorBo force-pushed the jit-disasm-release branch from 2849219 to 98a2a9dCompare August 4, 2022 13:57

Run jit-format
3952560

EgorBo added this to the 7.0.0 milestone Aug 4, 2022

jkotas reviewed Aug 4, 2022
View reviewed changes

Comment threadsrc/coreclr/jit/compiler.cpp

Don't allocate instrDescDebugInfo when JitDisasm is not set
c2d397a

fix build
f760723

EgorBoand others added 3 commits August 8, 2022 11:35

Update codegencommon.cpp
13190b8

Tiny optimization
b075a03

Switch a couple of checks to m_debugInfoSize > 0
d6759c2

Move assert back
00e754b

Update emitarm64.cpp
318f275

BruceForstall approved these changes Aug 9, 2022
View reviewed changes

kunalspathak approved these changes Aug 9, 2022
View reviewed changes

Add emitFirstInstrDesc and emitAdvanceInstrDesc
963596d
Consolidate most uses of m_debugInfoSize into these functions.

kunalspathak approved these changes Aug 9, 2022
View reviewed changes

Fix function header
840f26e

BruceForstall reviewed Aug 9, 2022
View reviewed changes

Comment threadsrc/coreclr/jit/emit.cpp Outdated

Fix build, fix bad refactor
4057d73

EgorBo merged commit 87d5ece into dotnet:mainAug 10, 2022

EgorBo mentioned this pull request Aug 23, 2022
[release/7.0-rc1] Improve DOTNET_JitDisasm and introduce DOTNET_JitDisasmSummary #74392
Merged

ghost locked as resolved and limited conversation to collaborators Sep 9, 2022

Uh oh!

Conversation

EgorBo commented Aug 4, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ghost commented Aug 4, 2022

Uh oh!

stephentoub commented Aug 4, 2022

Uh oh!

Uh oh!

hez2010 commented Aug 4, 2022

Uh oh!

EgorBo commented Aug 4, 2022

Uh oh!

EgorBo commented Aug 4, 2022

Sharplab:

JitDisasm:

Uh oh!

JulieLeeMSFT commented Aug 4, 2022

Uh oh!

EgorBo commented Aug 4, 2022

Uh oh!

stephentoub commented Aug 4, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

EgorBo commented Aug 4, 2022

Uh oh!

jkotas commented Aug 4, 2022

Uh oh!

EgorBo commented Aug 4, 2022

Uh oh!

jkotas commented Aug 4, 2022

Uh oh!

EgorBo commented Aug 4, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

hez2010 commented Aug 4, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

EgorBo commented Aug 8, 2022

Uh oh!

jakobbotsch commented Aug 8, 2022

Uh oh!

EgorBo commented Aug 8, 2022

Uh oh!

EgorBo commented Aug 8, 2022

Uh oh!

AndyAyersMS commented Aug 8, 2022

Uh oh!

omariom commented Aug 8, 2022

Uh oh!

EgorBo commented Aug 8, 2022

Uh oh!

EgorBo commented Aug 8, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MichalPetryka commented Aug 8, 2022• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

BruceForstall left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BruceForstallAug 9, 2022

Choose a reason for hiding this comment

Uh oh!

jakobbotschAug 9, 2022

Choose a reason for hiding this comment

Uh oh!

BruceForstallAug 9, 2022

Choose a reason for hiding this comment

Uh oh!

jakobbotschAug 9, 2022

Choose a reason for hiding this comment

Uh oh!

kunalspathak commented Aug 9, 2022

Uh oh!

kunalspathak commented Aug 9, 2022

Uh oh!

EgorBo commented Aug 4, 2022•
edited
Loading

stephentoub commented Aug 4, 2022•
edited
Loading

EgorBo commented Aug 4, 2022•
edited
Loading

hez2010 commented Aug 4, 2022•
edited
Loading

EgorBo commented Aug 8, 2022•
edited
Loading

MichalPetryka commented Aug 8, 2022•
edited
Loading

BruceForstall left a comment •
edited
Loading