Align LA64 targetpacks with RV64 #102628

am11 · 2024-05-23T20:22:17Z

Also consolidates <NativeAotSupported definitions.

shushanhf

LGTM for LoongArch64.

Thanks very much!

eng/Subsets.props

src/coreclr/tools/aot/crossgen2/crossgen2_publish.csproj

gbalykov · 2024-05-24T07:25:59Z

cc @dotnet/samsung

gbalykov

Looks good to me

am11 · 2024-05-29T16:08:52Z

@MichalStrehovsky, @filipnavara, with #102760 merged the build error is fixed here. This is now publishing linux-arm crossgen2 as an AOT app (instead of R2R), bringing it to the same arm64,x64 plan.

jkotas · 2024-06-02T02:18:11Z

eng/Subsets.props

@@ -121,7 +121,8 @@
    <!-- CLR NativeAot only builds in a subset of the matrix -->
    <_NativeAotSupportedOS Condition="'$(TargetOS)' == 'windows' or '$(TargetOS)' == 'linux' or '$(TargetOS)' == 'osx' or '$(TargetOS)' == 'maccatalyst' or '$(TargetOS)' == 'iossimulator' or '$(TargetOS)' == 'ios' or '$(TargetOS)' == 'tvossimulator' or '$(TargetOS)' == 'tvos' or '$(TargetOS)' == 'freebsd'">true</_NativeAotSupportedOS>
    <_NativeAotSupportedArch Condition="'$(TargetArchitecture)' == 'x64' or '$(TargetArchitecture)' == 'arm64' or '$(TargetArchitecture)' == 'arm' or ('$(TargetOS)' == 'windows' and '$(TargetArchitecture)' == 'x86')">true</_NativeAotSupportedArch>
-    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and $(_NativeAotSupportedArch) == 'true'">true</NativeAotSupported>
+    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and '$(_NativeAotSupportedArch)' == 'true'">true</NativeAotSupported>
+    <NativeAotCanExecuteOnHost Condition="'$(NativeAotSupported)' == 'true' and ('$(CrossBuild)' != 'true' or '$(TargetOS)' == '$(HostOS)')">true</NativeAotCanExecuteOnHost>


Why do we need '$(CrossBuild)' != 'true' in the condition? I would expect '$(TargetOS)' == '$(HostOS)' to be enough.

This part of the condition was adapted from main

runtime/src/coreclr/tools/aot/crossgen2/crossgen2_publish.csproj

Lines 13 to 14 in cb71b60



<NativeAotSupported Condition="'$(CrossBuild)' == 'true' and '$(TargetOS)' != '$(HostOS)'">false</NativeAotSupported>

e.g. x64 binary running on arm64 OS is supported (as long as the host and target OS were same).

x64 binary running on arm64 OS is supported (as long as the host and target OS were same).

This is situation should be covered by '$(TargetOS)' != '$(HostOS)' condition. I do not understand why we also check for CrossBuild in the condition.

jkotas · 2024-06-02T02:25:20Z

eng/Subsets.props

@@ -121,7 +121,8 @@
    <!-- CLR NativeAot only builds in a subset of the matrix -->
    <_NativeAotSupportedOS Condition="'$(TargetOS)' == 'windows' or '$(TargetOS)' == 'linux' or '$(TargetOS)' == 'osx' or '$(TargetOS)' == 'maccatalyst' or '$(TargetOS)' == 'iossimulator' or '$(TargetOS)' == 'ios' or '$(TargetOS)' == 'tvossimulator' or '$(TargetOS)' == 'tvos' or '$(TargetOS)' == 'freebsd'">true</_NativeAotSupportedOS>
    <_NativeAotSupportedArch Condition="'$(TargetArchitecture)' == 'x64' or '$(TargetArchitecture)' == 'arm64' or '$(TargetArchitecture)' == 'arm' or ('$(TargetOS)' == 'windows' and '$(TargetArchitecture)' == 'x86')">true</_NativeAotSupportedArch>
-    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and $(_NativeAotSupportedArch) == 'true'">true</NativeAotSupported>
+    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and '$(_NativeAotSupportedArch)' == 'true'">true</NativeAotSupported>


https://github.com/dotnet/runtime/blob/main/src/native/managed/compile-native.proj#L18-L28 is another place that checks for whether it is possible to use native AOT to produce binaries compiler by the repo. Can we use this property there as well?

jkotas · 2024-06-02T02:42:36Z

eng/Subsets.props

@@ -121,7 +121,8 @@
    <!-- CLR NativeAot only builds in a subset of the matrix -->
    <_NativeAotSupportedOS Condition="'$(TargetOS)' == 'windows' or '$(TargetOS)' == 'linux' or '$(TargetOS)' == 'osx' or '$(TargetOS)' == 'maccatalyst' or '$(TargetOS)' == 'iossimulator' or '$(TargetOS)' == 'ios' or '$(TargetOS)' == 'tvossimulator' or '$(TargetOS)' == 'tvos' or '$(TargetOS)' == 'freebsd'">true</_NativeAotSupportedOS>
    <_NativeAotSupportedArch Condition="'$(TargetArchitecture)' == 'x64' or '$(TargetArchitecture)' == 'arm64' or '$(TargetArchitecture)' == 'arm' or ('$(TargetOS)' == 'windows' and '$(TargetArchitecture)' == 'x86')">true</_NativeAotSupportedArch>
-    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and $(_NativeAotSupportedArch) == 'true'">true</NativeAotSupported>
+    <NativeAotSupported Condition="'$(_NativeAotSupportedOS)' == 'true' and '$(_NativeAotSupportedArch)' == 'true'">true</NativeAotSupported>
+    <NativeAotCanExecuteOnHost Condition="'$(NativeAotSupported)' == 'true' and ('$(CrossBuild)' != 'true' or '$(TargetOS)' == '$(HostOS)')">true</NativeAotCanExecuteOnHost>


I am not sure whether NativeAotCanExecuteOnHost is the right name for this property. We always have ilc that can execute on host for all host OSes. The problem is that we do not always have a working host/target build environment.

Would something like UseNativeAotForComponents be a better name?

am11 · 2024-06-02T10:23:25Z

src/native/managed/compile-native.proj

-        <SupportsNativeAotComponents Condition="'$(SupportsNativeAotComponents)' == '' and ('$(TargetsWindows)' == 'true' or '$(TargetsOSX)' == 'true' or ('$(TargetsLinux)' == 'true' and '$(TargetsAndroid)' != 'true' and '$(TargetsLinuxMusl)' != 'true'))">true</SupportsNativeAotComponents>
-        <SupportsNativeAotComponents Condition="'$(SupportsNativeAotComponents)' == ''">false</SupportsNativeAotComponents>
-    </PropertyGroup>
-
    <!-- some special kinds of runtime builds need extra NativeAOT flags -->
    <PropertyGroup>
        <SysRoot Condition="'$(CrossBuild)' == 'true' and '$(HostOS)' != 'windows'">$(ROOTFS_DIR)</SysRoot>


These SysRoot setup can also be brought to the same plan as ILCompiler.csproj and crossgen2.csproj (better deduplicate all three using init-compiler.sh thingy). I have a feeling it will take some digging which is why @lambdageek chose this approach, but worth a try (in a follow-up PR). 🙂

MichalStrehovsky · 2024-06-03T05:28:02Z

am11 requested review from jkoritzinsky, jkotas and MichalStrehovsky 3 days ago

I think it would help reviewing if you could split off the native AOT change to a separate PR. I don't know about target packs or RV64 and cannot review or sign off on that.

I (and probably others) often skim commit histories and PRs to see what's going on and skip stuff with titles that are not related to my areas - a PR named "Align LA64 targetpacks" that spends most lines of the diff on something unrelated breaks such workflows.

am11 · 2024-06-03T08:56:59Z

@MichalStrehovsky, the PR description has the line and if PR name was the issue, you or @jkotas could have told a bit earlier without going through the cycles of feedback and it would have saved both some time.

MichalStrehovsky · 2024-06-03T09:06:06Z

@MichalStrehovsky, the PR description has the line and if PR name was the issue, you or @jkotas could have told a bit earlier without going through the cycles of feedback and it would have saved both some time.

Sorry, it was not clear who would be responsible for getting this merged since it touched multiple things owned/worked on by different teams. I often comment on PRs without an intention of taking the full responsibility of merging it - whoever merged it is the one who's ultimately responsible even if it was authored by you. It was not clear to me who could merge a PR in this shape (I guess Jan often gets to volunteer himself for these, but I don't like dumping all work on Jan). Since you asked me specifically to take a look, my feedback is really that I won't take responsibility for targeting pack changes and we're going to be blocked on finding whoever is qualified (I don't know who), unless the native AOT change is separated out.

jkotas · 2024-06-03T21:07:45Z

I know that it is tempting to make multiple somewhat unrelated change in the same PR. I have done it myself many times, and sometimes had to revert it or split into multiple PRs to make review and approval easier.

@am11 Please take @MichalStrehovsky comment only as a suggestion. If you prefer to keep both changes in one PR, that's ok - it is still manageable. We appreciate your contributions.

dotnet-issue-labeler bot added the area-crossgen2-coreclr label May 23, 2024

dotnet-policy-service bot added the community-contribution Indicates that the PR has been added by a community member label May 23, 2024

am11 added the arch-loongarch64 label May 23, 2024

am11 force-pushed the feature/crossgen2/LA64 branch from 459b546 to dfa88cb Compare May 23, 2024 21:36

am11 added the arch-riscv Related to the RISC-V architecture label May 23, 2024

am11 requested review from MichalStrehovsky, gbalykov and shushanhf May 24, 2024 00:12

am11 marked this pull request as ready for review May 24, 2024 00:12

shushanhf approved these changes May 24, 2024

View reviewed changes

build-analysis bot mentioned this pull request May 24, 2024

NativeAOT legs timing out in CI #102239

Closed

MichalStrehovsky reviewed May 24, 2024

View reviewed changes

eng/Subsets.props Outdated Show resolved Hide resolved

src/coreclr/tools/aot/crossgen2/crossgen2_publish.csproj Outdated Show resolved Hide resolved

gbalykov approved these changes May 24, 2024

View reviewed changes

tomeksowi approved these changes May 24, 2024

View reviewed changes

Bajtazar approved these changes May 24, 2024

View reviewed changes

am11 force-pushed the feature/crossgen2/LA64 branch from 32d98b7 to 8479358 Compare May 24, 2024 10:00

clamp03 approved these changes May 29, 2024

View reviewed changes

am11 added 4 commits May 29, 2024 19:04

Align LA64 targetpacks with RV64

07eb8d7

Block x86 during publish (until preview 5)

32df8eb

Separate CanExecute and Supported concerns

cda5ba9

Revert x86 workaround

7eec988

am11 force-pushed the feature/crossgen2/LA64 branch from b40bca0 to 7eec988 Compare May 29, 2024 16:04

am11 requested a review from jkoritzinsky May 31, 2024 08:04

am11 requested review from jkotas and MichalStrehovsky May 31, 2024 08:04

jkotas reviewed Jun 2, 2024

View reviewed changes

am11 added 4 commits June 2, 2024 10:03

Address CR feedback

34d83f4

Merge dotnet/main into feature/crossgen2/LA64

4be199b

Change to explicit boolean condition

52fd95c

Record the original CrossBuild condition for tools

d68366e

am11 commented Jun 2, 2024

View reviewed changes

am11 added 5 commits June 2, 2024 16:50

Fix windows build

3aedb2b

Use public props

83c0fab

Update Subsets.props

41e9a48

Remve duplicate prop

b736756

Sort props

ea7d6f1

build-analysis bot mentioned this pull request Jun 2, 2024

System.Net.Tests.HttpWebRequestTest_Async.GetResponseAsync_ParametersAreNotCachable_CreateNewClient test fails #100912

Closed

am11 closed this Jun 3, 2024

LuckyXu-HF mentioned this pull request Jun 11, 2024

[LoongArch64] Add linux-loongarch64 RID for Net90AppHostRids and Net90RuntimePackRids. dotnet/sdk#41515

Closed

github-actions bot locked and limited conversation to collaborators Jul 4, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Align LA64 targetpacks with RV64 #102628

Align LA64 targetpacks with RV64 #102628

am11 commented May 23, 2024

shushanhf left a comment

gbalykov commented May 24, 2024

gbalykov left a comment

am11 commented May 29, 2024

jkotas Jun 2, 2024

am11 Jun 2, 2024

jkotas Jun 2, 2024 •

edited

Loading

jkotas Jun 2, 2024

jkotas Jun 2, 2024

am11 Jun 2, 2024

MichalStrehovsky commented Jun 3, 2024

am11 commented Jun 3, 2024

MichalStrehovsky commented Jun 3, 2024

jkotas commented Jun 3, 2024

	<!-- Publish crossgen2 as a single-file app on native-OS builds. Cross-OS NativeAOT compilation is not supported yet -->
	<NativeAotSupported Condition="'$(CrossBuild)' == 'true' and '$(TargetOS)' != '$(HostOS)'">false</NativeAotSupported>

Align LA64 targetpacks with RV64 #102628

Align LA64 targetpacks with RV64 #102628

Conversation

am11 commented May 23, 2024

shushanhf left a comment

Choose a reason for hiding this comment

gbalykov commented May 24, 2024

gbalykov left a comment

Choose a reason for hiding this comment

am11 commented May 29, 2024

jkotas Jun 2, 2024

Choose a reason for hiding this comment

am11 Jun 2, 2024

Choose a reason for hiding this comment

jkotas Jun 2, 2024 • edited Loading

Choose a reason for hiding this comment

jkotas Jun 2, 2024

Choose a reason for hiding this comment

jkotas Jun 2, 2024

Choose a reason for hiding this comment

am11 Jun 2, 2024

Choose a reason for hiding this comment

MichalStrehovsky commented Jun 3, 2024

am11 commented Jun 3, 2024

MichalStrehovsky commented Jun 3, 2024

jkotas commented Jun 3, 2024

jkotas Jun 2, 2024 •

edited

Loading