[PATCH] D159256: [NFC][Clang] Remove redundant function definitions

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. In D159256#4630915 , @jmmartinez wrote: > In D159256#4630876 , @jhuber6 wrote: > >> In D159256#4630410

[PATCH] D159256: [NFC][Clang] Remove redundant function definitions

jhuber6 added a comment. In D159256#4630410 , @jmmartinez wrote: > @jhuber6 I was wondering if there is a reason you kept 3 versions of > `mergeDefaultFunctionDefinitionAttributes` in > https://reviews.llvm.org/D152391 ? I believe it's because one

[PATCH] D159118: [libc] Implement the 'clock()' function on the GPU

This revision was automatically updated to reflect the committed changes. Closed by commit rG30307a7bb795: [libc] Implement the clock() function on the GPU (authored by jhuber6). Changed prior to commit: https://reviews.llvm.org/D159118?vs=554797=554831#toc Repository: rG LLVM Github

[PATCH] D159118: [libc] Implement the 'clock()' function on the GPU

jhuber6 updated this revision to Diff 554797. jhuber6 added a comment. Move the header portion into a common utility header. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D159118/new/ https://reviews.llvm.org/D159118 Files:

[PATCH] D158778: [CUDA] Propagate __float128 support from the host.

jhuber6 added a comment. In D158778#4625892 , @tra wrote: > In D158778#4624408 , @ABataev wrote: > >> Just checks removal should be fine > > Looks like OpenMP handles long double and __float128 differently -- it

[PATCH] D159118: [libc] Implement the 'clock()' function on the GPU

jhuber6 updated this revision to Diff 554434. jhuber6 added a comment. Address nits and add static check for size of `clock_t` type. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D159118/new/ https://reviews.llvm.org/D159118 Files:

[PATCH] D159118: [libc] Implement the 'clock()' function on the GPU

jhuber6 created this revision. jhuber6 added reviewers: arsenm, tra, JonChesterfield, jdoerfert, sivachandra, lntue, michaelrj. Herald added subscribers: libc-commits, tpr. Herald added projects: libc-project, All. jhuber6 requested review of this revision. Herald added subscribers: cfe-commits,

[PATCH] D112921: [clang] Enable sized deallocation by default in C++14 onwards

jhuber6 added a subscriber: sivachandra. jhuber6 added a comment. In D112921#4624580 , @wangpc wrote: > It seems that the linker can't find sized deallocation (no support in the > environment or AMDGPU libraries?). We should have some implementations

[PATCH] D112921: [clang] Enable sized deallocation by default in C++14 onwards

jhuber6 added a comment. This caused some linking errors with the GPU libc test suite, see https://lab.llvm.org/staging/#/builders/247/builds/5659. clang++: error: ld.lld command failed with exit code 1 (use -v to see invocation) [331/473] Linking CXX executable

[PATCH] D153924: [OpenMP] Allow exceptions in target regions when offloading to GPUs

jhuber6 added a comment. The libcxx tests are always broken randomly in my experience. I wouldn't worry about it. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D153924/new/ https://reviews.llvm.org/D153924

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. I think it's fine now given that it's passing tests. Others feel free to comment. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D139730/new/

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added a comment. Just a few more nits. I think it's looking fine but I haven't tested it. Anyone else? Comment at: clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp:406 + // pass on -mllvm options to the clang + for (const opt::Arg *Arg :

[PATCH] D157738: [OpenMP] Emit offloading entries for indirect target variables

This revision was automatically updated to reflect the committed changes. Closed by commit rG9da61aed751e: [OpenMP] Emit offloading entries for indirect target variables (authored by jhuber6). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added inline comments. Comment at: openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h:49 -static_assert(sizeof(AMDGPUImplicitArgsTy) == 56, - "Unexpected size of implicit arguments"); +enum IMPLICITARGS : uint32_t { + COV4_SIZE = 56,

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added inline comments. Comment at: openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h:49 -static_assert(sizeof(AMDGPUImplicitArgsTy) == 56, - "Unexpected size of implicit arguments"); +enum IMPLICITARGS : uint32_t { + COV4_SIZE = 56,

[PATCH] D158582: [AMDGPU] Respect unresolved symbol option if forwarded to linker

jhuber6 added a comment. In D158582#4610023 , @yaxunl wrote: > The `-Wl` and `-Xlinker` options are intended for the host linker and we > intentionally do not pass them to the device linker. > > If users want to pass options to the device linker, they

[PATCH] D158582: [AMDGPU] Respect unresolved symbol option if forwarded to linker

jhuber6 added a comment. Remember to clang format. Comment at: clang/lib/Driver/ToolChains/AMDGPU.cpp:564-575 + // If the user has manually passed -Wl,--unresolved-symbols=* as a linker + // option, we should not add --no-undefined + bool UnresolvedOpt = false; + for (auto

[PATCH] D153924: [OpenMP] Allow exceptions in target regions when offloading to GPUs

jhuber6 added inline comments. Comment at: clang/test/OpenMP/amdgpu_exceptions.cpp:11 +// RUN: %clang_cc1 -fopenmp -triple amdgcn-amd-amdhsa -fopenmp-is-target-device -fexceptions %s -emit-llvm -S -verify=with -Wopenmp-target-exception -o - &> /dev/null +// RUN: %clang_cc1

[PATCH] D158298: [OpenMP] Always pass the optimization level to the linker wrapper

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG4ab4e40fa294: [OpenMP] Always pass the optimization level to the linker wrapper (authored by jhuber6). Repository: rG LLVM Github Monorepo

[PATCH] D158298: [OpenMP] Always pass the optimization level to the linker wrapper

jhuber6 created this revision. jhuber6 added reviewers: jdoerfert, tianshilei1992, JonChesterfield, ye-luo. Herald added subscribers: kerbowa, guansong, tpr, yaxunl, jvesely. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers: cfe-commits, jplehr,

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added a comment. Some nits. I'm assuming we're getting the code object in the backend now? We'll need to make sure that `-Wl,--amdhsa-code-object-version` is passed to the clang invocation inside of the `clang-linker-wrapper` to handle `-save-temps` mode. Comment

[PATCH] D158131: HIP: Directly use f32 sqrt intrinsic

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. I'm hoping to write some brute force tests for the `fp32` math functions in `libc` soonish. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D158131/new/

[PATCH] D157738: [OpenMP] Emit offloading entries for indirect target variables

jhuber6 updated this revision to Diff 550140. jhuber6 added a comment. Address comments Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D157738/new/ https://reviews.llvm.org/D157738 Files: clang/lib/CodeGen/CGOpenMPRuntime.cpp

[PATCH] D157738: [OpenMP] Emit offloading entries for indirect target variables

jhuber6 marked an inline comment as done. jhuber6 added inline comments. Comment at: clang/lib/CodeGen/CGOpenMPRuntime.cpp:1996-1997 +llvm::GlobalValue *GV) { + std::optional ActiveAttr = +

[PATCH] D157738: [OpenMP] Emit offloading entries for indirect target variables

jhuber6 added a comment. In D157738#4586465 , @JonChesterfield wrote: >> calling device functions via their associated host pointer > > What does this mean? Defining a function foo such that the host and each > individual target each have their own

[PATCH] D157738: [OpenMP] Emit offloading entries for indirect target variables

jhuber6 created this revision. jhuber6 added reviewers: jdoerfert, tianshilei1992, ye-luo, ABataev, RaviNarayanaswamy. Herald added subscribers: guansong, hiraditya, yaxunl. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers: llvm-commits,

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 added a comment. In D156816#4574190 , @arsenm wrote: > Probably should just wrap uses in macros for now In clang? Or just have users deal with `opencl_` on everything and rename it. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST

[PATCH] D157438: [OpenMP] Ensure wrapper headers are included on both host and device

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG61709bbae37a: [OpenMP] Ensure wrapper headers are included on both host and device (authored by jhuber6). Repository: rG LLVM Github Monorepo

[PATCH] D157438: [OpenMP] Ensure wrapper headers are included on both host and device

jhuber6 added inline comments. Comment at: clang/lib/Driver/ToolChains/Clang.cpp:1190-1191 // the resource directory at clang/lib/Headers/llvm_libc_wrappers. -if (C.getActiveOffloadKinds() == Action::OFK_None) { +if ((getToolChain().getTriple().isNVPTX() || +

[PATCH] D157438: [OpenMP] Ensure wrapper headers are included on both host and device

jhuber6 added inline comments. Comment at: clang/lib/Driver/ToolChains/Clang.cpp:1190-1191 // the resource directory at clang/lib/Headers/llvm_libc_wrappers. -if (C.getActiveOffloadKinds() == Action::OFK_None) { +if ((getToolChain().getTriple().isNVPTX() || +

[PATCH] D157438: [OpenMP] Ensure wrapper headers are included on both host and device

jhuber6 created this revision. jhuber6 added reviewers: jdoerfert, tianshilei1992, tra, JonChesterfield, yaxunl, sivachandra. Herald added a subscriber: guansong. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers: cfe-commits, jplehr, sstefan1,

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 added a comment. In D156014#4567363 , @steven_wu wrote: > This breaks macOS bot: > https://green.lab.llvm.org/green/job/clang-stage1-cmake-RA-incremental/36900/testReport/junit/Clang/SemaCUDA/alias_cu/ I should've fixed that already. Is it

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG0ba9aec38faa: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets (authored by jhuber6). Repository: rG LLVM Github Monorepo

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 updated this revision to Diff 547871. jhuber6 added a comment. Update to check the SDK version. Permit this if there is not passed in SDK version so that freestanding targets can still target CUDA and assume it's supported. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 added a comment. ping Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156014/new/ https://reviews.llvm.org/D156014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:17138 +///and use its value for COV_4 or COV_5 approach. It is used for +///compiling device libraries in ABI-agnostic way. +/// Comment at:

[PATCH] D156928: [Clang][AMDGPU] Fix handling of -mcode-object-version=none arg

jhuber6 added a comment. In D156928#4562239 , @JonChesterfield wrote: > Or, the front end could define those objects directly, without importing IR > files that define the objects with the content clang used to choose the > object file. E.g. instead

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added a comment. In D139730#4561573 , @arsenm wrote: > In D139730#4561540 , @jhuber6 wrote: > >> Could you explain briefly what the approach here is? I'm confused as to >> what's actually changed and how

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added inline comments. Comment at: openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h:36 -// The implicit arguments of AMDGPU kernels. -struct AMDGPUImplicitArgsTy { - uint64_t OffsetX; - uint64_t OffsetY; - uint64_t OffsetZ; - uint64_t HostcallPtr; -

[PATCH] D139730: [OpenMP][DeviceRTL][AMDGPU] Support code object version 5

jhuber6 added a comment. Could you explain briefly what the approach here is? I'm confused as to what's actually changed and how we're handling this difference. I thought if this was just the definition of some builtin function we could just rely on the backend to figure it out. Why do we need

[PATCH] D156936: [Clang] Increase default architecture from sm_35 to sm_52

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGab202aa7004a: [Clang] Increase default architecture from sm_35 to sm_52 (authored by jhuber6). Repository: rG LLVM Github Monorepo CHANGES

[PATCH] D156936: [Clang] Increase default architecture from sm_35 to sm_52

jhuber6 created this revision. jhuber6 added reviewers: tra, jdoerfert, yaxunl, jlebar. Herald added a project: All. jhuber6 requested review of this revision. Herald added a project: clang. Herald added a subscriber: cfe-commits. We previously defaulted to `sm_35` for the purpose of unspecified

[PATCH] D156930: [Clang] Fix Offloading related tests after D156363

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGbc080221b3a2: [Clang] Fix Offloading related tests after D156363 (authored by jhuber6). Changed prior to commit:

[PATCH] D156928: [Clang][AMDGPU] Fix handling of -mcode-object-version=none arg

jhuber6 added a comment. In D156928#4555121 , @yaxunl wrote: > `-mcode-object-version=none` was intentionally designed to work with `clang > -cc1` only, since it does not work with clang driver if users link with > device library. Device library can

[PATCH] D156928: [Clang][AMDGPU] Fix handling of -mcode-object-version=none arg

jhuber6 added inline comments. Comment at: clang/include/clang/Basic/TargetOptions.h:90 +COV_Default = 400, +COV_MAX = 500 }; Typically we just put a `COV_LAST` to indicate that it's over the accepted enumerations. Comment at:

[PATCH] D156930: [Clang] Fix Offloading related tests after D156363

jhuber6 created this revision. jhuber6 added reviewers: MaskRay, yaxunl, JonChesterfield, tra, jdoerfert, ronlieb, jplehr. Herald added subscribers: mattd, asavonic, ormris, kerbowa, steven_wu, hiraditya, jvesely. Herald added a project: All. jhuber6 requested review of this revision. Herald

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. In D156363#4554913 , @yaxunl wrote: > Thanks. I will wait for your patch. You can leave the tricky ones to me if > you would like. e.g. the rocm-detect.hip You can go ahead and fix that one then. Repository: rG LLVM Github

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. In D156363#4554790 , @ro wrote: > In D156363#4553043 , @ro wrote: > >> It seems the latest commit of this patch has (re-)introduced two failures on >> the Solaris/amd64 buildbot >>

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. In D156363#4554687 , @yaxunl wrote: > In D156363#4554435 , @jhuber6 wrote: > >> Clang :: Driver/amdgpu-hip-system-arch.c >> Clang :: Driver/cuda-bad-arch.cu >> Clang ::

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. Pretty sure most of this is `hip` returning an error if it can't find `ROCm`. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156363/new/ https://reviews.llvm.org/D156363 ___

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. Clang :: Driver/amdgpu-hip-system-arch.c Clang :: Driver/cuda-bad-arch.cu Clang :: Driver/hip-autolink.hip Clang :: Driver/hip-binding.hip Clang :: Driver/hip-cuid-hash.hip Clang :: Driver/hip-cuid.hip Clang :: Driver/hip-default-gpu-arch.hip Clang ::

[PATCH] D156886: [CUDA][HIP] Reorganize options for documentation

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. I think this is good, there's been some confusion around which ones goes to what. Comment at: clang/include/clang/Driver/Options.td:1021 +def offload_host_device :

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 added a comment. Also seems to have impacted one of the AMDGPU bots but not the other, not sure why https://lab.llvm.org/staging/#/builders/247/builds/4145. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156363/new/

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 added a comment. In D156816#4551500 , @jdoerfert wrote: > Macros seems to be good enough. If we really need clang attributes, we need > new docs, and naming convention etc. Yeah an alternative would be a new set of attributes, then we make the

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 added a comment. In D156816#4551409 , @Anastasia wrote: > Why not to just use target address space and define it to some macro with > desirable spelling? > > I don't think renaming OpenCL address space to something else makes sense. It > might

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 added a comment. In D156816#4551338 , @yaxunl wrote: >> FFI isn't the reason you'd use these, it's for generic access to the actual >> backend. E.g. an `addrspace(3)` global is local memory, if it's external >> it's dynamic. Having these named

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 added a comment. In D156816#4551299 , @arsenm wrote: > I don't really see the point of doing this. These introduce ambiguous > terminology. The reason you need the attributes is basically for FFI to > opencl code, so might as well make the

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 updated this revision to Diff 546149. jhuber6 added a comment. Fix typo Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156816/new/ https://reviews.llvm.org/D156816 Files: clang/include/clang/Basic/Attr.td

[PATCH] D156816: [Clang] Make generic aliases to OpenCL address spaces

jhuber6 created this revision. jhuber6 added reviewers: ebevhan, arsenm, JonChesterfield, jdoerfert, tianshilei1992, tra, yaxunl, rjmccall. Herald added subscribers: jeroen.dobbelaere, Naghasan, ldrumm, arichardson, Anastasia. Herald added a reviewer: aaron.ballman. Herald added a project: All.

[PATCH] D156368: [OpenMP] Do not always emit unused extern variables

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG141c4e7a9403: [OpenMP] Do not always emit unused extern variables (authored by jhuber6). Herald added a project: OpenMP. Herald added a subscriber:

[PATCH] D156368: [OpenMP] Do not always emit unused extern variables

jhuber6 added inline comments. Comment at: clang/test/OpenMP/declare_target_codegen.cpp:264 -// CHECK-DAG: !{i32 1, !"aaa", i32 0, i32 {{[0-9]+}}} -// CHECK-DAG: !{i32 1, !"ccc", i32 0, i32 {{[0-9]+}}} // CHECK-DAG: !{{{.+}}virtual_foo tianshilei1992 wrote: >

[PATCH] D156368: [OpenMP] Do not always emit unused extern variables

jhuber6 updated this revision to Diff 545168. jhuber6 added a comment. Add OpenMP runtime test Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156368/new/ https://reviews.llvm.org/D156368 Files: clang/lib/CodeGen/CGOpenMPRuntime.cpp

[PATCH] D156363: [Driver] -###: exit with code 1 if hasErrorOccurred

jhuber6 accepted this revision. jhuber6 added a comment. I'd wager a lot of the tests that return non-zero aren't even intentional, so it's probably good to enforce this. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156363/new/

[PATCH] D156426: [HIP] link HIP runtime library without --hip-link

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. LG, thanks. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156426/new/ https://reviews.llvm.org/D156426 ___ cfe-commits mailing list

[PATCH] D156426: [HIP] link HIP runtime library without --hip-link

jhuber6 added inline comments. Comment at: clang/include/clang/Driver/Driver.h:712 + /// Whether there are HIP input files. + bool hasHIPInputs() const { return HasHIPInputs; } + Shouldn't we have access to the compilation? I figured we could check

[PATCH] D156426: [HIP] link HIP runtime library without --hip-link

jhuber6 added a comment. So this is equivalent to `nvcc` implicitly calling `-lcudart`? I've had thoughts about the `clang-linker-wrapper` adding known runtime flags to the link job if it's not found. E.g. if we find a CUDA image we pass `-lcudart`. CHANGES SINCE LAST ACTION

[PATCH] D156368: [OpenMP] Do not always emit unused extern variables

jhuber6 created this revision. jhuber6 added reviewers: JonChesterfield, jdoerfert, tianshilei1992, ye-luo, RaviNarayanaswamy, ABataev. Herald added subscribers: sunshaoce, guansong, yaxunl. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers:

[PATCH] D156366: HIP: Use __builtin_sqrt instead of routing through ocml sqrt for f64

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. Thanks CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156366/new/ https://reviews.llvm.org/D156366 ___ cfe-commits mailing list

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 added inline comments. Comment at: clang/lib/Sema/SemaDeclAttr.cpp:1995 } - if (S.Context.getTargetInfo().getTriple().isNVPTX()) { -S.Diag(AL.getLoc(), diag::err_alias_not_supported_on_nvptx); jhuber6 wrote: > tra wrote: > > tra wrote: > > >

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 added inline comments. Comment at: clang/lib/Sema/SemaDeclAttr.cpp:1995 } - if (S.Context.getTargetInfo().getTriple().isNVPTX()) { -S.Diag(AL.getLoc(), diag::err_alias_not_supported_on_nvptx); tra wrote: > tra wrote: > > Allowing or not

[PATCH] D156014: [Clang][NVPTX] Permit use of the alias attribute for NVPTX targets

jhuber6 created this revision. jhuber6 added reviewers: tra, arsenm, jlebar, kushanam, aaron.ballman, yaxunl, jdoerfert. Herald added subscribers: mattd, gchakrabarti, asavonic, jeroen.dobbelaere. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers:

[PATCH] D155211: [NVPTX] Add initial support for '.alias' in PTX

This revision was automatically updated to reflect the committed changes. Closed by commit rGf4381d464457: [NVPTX] Add initial support for .alias in PTX (authored by jhuber6). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155211/new/

[PATCH] D155727: [OpenMP][Docs] Add some things to the OpenMP support

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGa37d74722254: [OpenMP][Docs] Add some things to the OpenMP support (authored by jhuber6). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST

[PATCH] D155727: [OpenMP][Docs] Add some things to the OpenMP support

jhuber6 created this revision. jhuber6 added reviewers: jdoerfert, tianshilei1992, JonChesterfield, kevinsala, jplehr. Herald added subscribers: sunshaoce, guansong, yaxunl. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers: cfe-commits, sstefan1.

[PATCH] D155606: [Clang] Only emit CUDA version warnings when creating the CUDA toolchain

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGd2ac0069a21b: [Clang] Only emit CUDA version warnings when creating the CUDA toolchain (authored by jhuber6). Repository: rG LLVM Github Monorepo

[PATCH] D155606: [Clang] Only emit CUDA version warnings when creating the CUDA toolchain

jhuber6 updated this revision to Diff 541594. jhuber6 added a comment. Add test Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155606/new/ https://reviews.llvm.org/D155606 Files: clang/lib/Driver/Driver.cpp clang/lib/Driver/ToolChains/Cuda.cpp

[PATCH] D155606: [Clang] Only emit CUDA version warnings when creating the CUDA toolchain

jhuber6 updated this revision to Diff 541562. jhuber6 added a comment. Fix wrong comment Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D155606/new/ https://reviews.llvm.org/D155606 Files: clang/lib/Driver/Driver.cpp

[PATCH] D155606: [Clang] Only emit CUDA version warnings when creating the CUDA toolchain

jhuber6 created this revision. jhuber6 added reviewers: tra, yaxunl, jdoerfert, tianshilei1992, JonChesterfield. Herald added a subscriber: mattd. Herald added a project: All. jhuber6 requested review of this revision. Herald added subscribers: cfe-commits, wangpc, jplehr, sstefan1, MaskRay.

[PATCH] D155211: [NVPTX] Add initial support for '.alias' in PTX

jhuber6 updated this revision to Diff 540062. jhuber6 added a comment. Remove changes in `clang` that I forgot to remove to keep this restricted to the codegen. Afterwards we can remove the sema in clang and test it there. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D155211: [NVPTX] Add initial support for '.alias' in PTX

jhuber6 created this revision. jhuber6 added reviewers: tra, arsenm, jlebar, kushanam. Herald added subscribers: mattd, gchakrabarti, asavonic, jeroen.dobbelaere, hiraditya. Herald added a reviewer: aaron.ballman. Herald added a project: All. jhuber6 requested review of this revision. Herald

[PATCH] D154850: [libc] Remove GPU string functions incompatible with C++

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGb454e7aa7ceb: [libc] Remove GPU string functions incompatible with C++ (authored by jhuber6). Repository: rG LLVM Github Monorepo CHANGES SINCE

[PATCH] D154850: [libc] Remove GPU string functions incompatible with C++

jhuber6 created this revision. jhuber6 added reviewers: jdoerfert, JonChesterfield, sivachandra, lntue, michaelrj, ronlieb. Herald added projects: libc-project, All. Herald added a subscriber: libc-commits. jhuber6 requested review of this revision. Herald added a project: clang. Herald added a

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added a comment. In D153725#4485039 , @arsenm wrote: > And the libomptarget build is in fact doing that, but it shouldn't have to. > What it's doing actually seems really unreasonable. It's only building the > locally found targets when it

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added a comment. In D153725#4484966 , @arsenm wrote: > In D153725#4484754 , > @JonChesterfield wrote: > >> - if you open the driver too many times at once it fails to open, so running >> a parallel

[PATCH] D154591: [OpenMP][OMPIRBuilder] Rename IsEmbedded and IsTargetCodegen flags

jhuber6 added a comment. Yeah that happens sometimes, if it's working on your system it's safe to assume it's fine. Worst case scenario you can always revert, it's not a big deal. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D154591/new/

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. LG CHANGES SINCE LAST ACTION https://reviews.llvm.org/D153725/new/ https://reviews.llvm.org/D153725 ___ cfe-commits mailing list

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added inline comments. Comment at: clang/tools/amdgpu-arch/AMDGPUArch.cpp:48-52 + if (!printGPUsByHSA()) +return 0; +#endif + return printGPUsByHSA(); Are we missing something here ? They look the same. CHANGES SINCE LAST ACTION

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added inline comments. Comment at: clang/tools/amdgpu-arch/AMDGPUArch.cpp:50 +#else + return printGPUsByHSA(); +#endif arsenm wrote: > The HIP path should work on linux too. I generally think we should build as > much code as possible on all hosts, so

[PATCH] D154036: [libc] Add support for creating wrapper headers for offloading in clang

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGa4a26374aa11: [libc] Add support for creating wrapper headers for offloading in clang (authored by jhuber6). Repository: rG LLVM Github Monorepo

[PATCH] D154591: [OpenMP][OMPIRBuilder] Rename IsEmbedded and IsTargetCodegen flags

jhuber6 accepted this revision. jhuber6 added a comment. This revision is now accepted and ready to land. Noisy change but shouldn't really affect much. Names area little overloaded here so it's good to be more specific. Thanks. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D154036: [libc] Add support for creating wrapper headers for offloading in clang

jhuber6 updated this revision to Diff 537787. jhuber6 added a comment. Changing this to only apply to OpenMP for now. It breaks CUDA / HIP builds because they already have forward declarations of things like `malloc` or `memcpy` on the GPU that conflict. We'll need to clean those up later.

[PATCH] D154036: [libc] Add support for creating wrapper headers for offloading in clang

jhuber6 added inline comments. Comment at: libc/include/CMakeLists.txt:8 +if(LIBC_TARGET_ARCHITECTURE_IS_GPU) + include(GetClangResourceDir) +endif() sivachandra wrote: > Where does this come from? It's a global CMake module that LLVM provides in

[PATCH] D154036: [libc] Add support for creating wrapper headers for offloading in clang

jhuber6 updated this revision to Diff 537535. jhuber6 added a comment. Fix guard on the headers for offloading languages Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D154036/new/ https://reviews.llvm.org/D154036 Files:

[PATCH] D154378: [LinkerWrapper] Set the GPU LTO job to be freestanding

jhuber6 created this revision. jhuber6 added reviewers: JonChesterfield, jdoerfert, yaxunl, tianshilei1992. Herald added a subscriber: inglorion. Herald added a project: All. jhuber6 requested review of this revision. Herald added a project: clang. Herald added a subscriber: cfe-commits. The LTO

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added inline comments. Comment at: clang/tools/amdgpu-arch/AMDGPUArch.cpp:47 - // Attempt to load the HSA runtime. - if (llvm::Error Err = loadHSA()) { -logAllUnhandledErrors(std::move(Err), llvm::errs()); -return 1; - } - - hsa_status_t Status =

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added inline comments. Comment at: clang/tools/amdgpu-arch/AMDGPUArch.cpp:47 - // Attempt to load the HSA runtime. - if (llvm::Error Err = loadHSA()) { -logAllUnhandledErrors(std::move(Err), llvm::errs()); -return 1; - } - - hsa_status_t Status =

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added a comment. Also w.r.t. target-id, I'm wondering what a good solution would be. Right now the main usage of `amdgpu-arch` is both to detect the `-mcpu / -march` in CMake and to fill in the architecture via `--offload-arch=native` or `-fopenmp-target=amdgcn-amd-amdhsa`. We may want

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added inline comments. Comment at: clang/tools/amdgpu-arch/AMDGPUArch.cpp:47 - // Attempt to load the HSA runtime. - if (llvm::Error Err = loadHSA()) { -logAllUnhandledErrors(std::move(Err), llvm::errs()); -return 1; - } - - hsa_status_t Status =

[PATCH] D153725: [clang] Make amdgpu-arch tool work on Windows

jhuber6 added a comment. In D153725#4463589 , @arsenm wrote: > Unrelated but can we get this to start reporting xnack and ecc? A lot of CMake relies on this just being an ordered list of architectures, so we'd probably need to make that an opt-in

[PATCH] D154145: [HIP] Fix -mllvm option for device lld linker

jhuber6 added inline comments. Comment at: clang/lib/Driver/ToolChains/HIPAMD.cpp:164 +StringRef ArgVal = Arg->getValue(1); +if (ArgVal.startswith("-mllvm=")) { + ArgVal = ArgVal.substr(strlen("-mllvm=")); arsenm wrote: > StringRef Prefix("-mllvm=")

  1   2   3   4   5   6   7   8   9   10   >