[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 created this revision. gandhi21299 added reviewers: arsenm, yaxunl, rampitec. gandhi21299 created this object with edit policy "Members of Project: AMDGPU". Herald added subscribers: foad, kerbowa, jfb, hiraditya, t-tye, tpr, dstuttard, nhaehnle, jvesely, kzhuravl. gandhi21299 requeste

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362093. gandhi21299 added a comment. - replaced `check-prefixes` to `check-prefix` in the test - Function ffp1 emits a `flat_atomic_cmpswap` instead of `global_atomic_add`, I am not sure if that is considered an unsafe hardware instruction. @arsenm ? Re

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 created this revision. gandhi21299 added reviewers: arsenm, yaxunl, rampitec. Herald added subscribers: kerbowa, jfb, tpr, nhaehnle, jvesely. gandhi21299 requested review of this revision. Herald added subscribers: cfe-commits, wdng. Herald added a project: clang. Implement target buil

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362157. gandhi21299 added a comment. - corrected test by replacing the hardware instruction check by `v_add_f32_e64` Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D10689

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362208. gandhi21299 added a comment. Instruction selector is unable to select unsafe hw atomic instruction for add. Adding relevant diagnostic in the test case. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 362433. gandhi21299 added a comment. - added more tests with supported combinations of address spaces - minor nits Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362470. gandhi21299 added a comment. - added an OpenCL test to check for unsafe HW instructions - CUDA forces CAS loop over using unsafe HW instructions even with `-munsafe-fp-atomics`, eliminated that test Repository: rG LLVM Github Monorepo CHANGES

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 362487. gandhi21299 added a comment. - added scope argument - removed irrelevant tests and updated them - replaced __constant by __generic qualifier for two of the test functions Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362518. gandhi21299 added a comment. - code format Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu clan

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 362522. gandhi21299 added a comment. Herald added subscribers: llvm-commits, foad, hiraditya. Herald added a project: LLVM. - removed `kernel` from functions taking in `__generic` qualified `addr` Repository: rG LLVM Github Monorepo CHANGES SINCE LAST

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 362555. gandhi21299 added a comment. - eliminated scope argument, seemed irrelevant Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 Files: clang/include/clang/B

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362795. gandhi21299 added a comment. - refreshing patch Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 362804. gandhi21299 added a comment. fixed tests by replacing checks for the functions with generic-qualified arguments. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 362807. gandhi21299 added a comment. applied clang-format Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. ping Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org

[PATCH] D92956: Fix range-loop-analysis checks for trivial copyability

gandhi21299 added a comment. Herald added a project: All. @Quuxplusone I can commit this patch on behalf of @fanfuqiang. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D92956/new/ https://reviews.llvm.org/D92956 _

[PATCH] D92956: Fix range-loop-analysis checks for trivial copyability

gandhi21299 added a comment. Perhaps, a description could be added to this patch before committing. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D92956/new/ https://reviews.llvm.org/D92956 ___ cfe-commit

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. @yaxunl Under what criteria should an alias not be removed? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 ___ cfe-commits mailing

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. Internal linkage detection works great for our purposes but it causes a failure in llvm/test/CodeGen/AMDGPU/inline-calls.ll due to `@func_alias` unable to be casted into a `Function`. If we pass through that, the `@kernel3` causes the error: `scalar registers (98) e

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. @yaxunl I think we have two ways to go from here: 1. If appropriate, reset the maximum number of scalar registers allowed in `@kernel3` (inline-calls.ll) to fix the test. 2. Determine a stronger condition for inlining. Repository: rG LLVM Github Monorepo CHANGES

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 373257. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - Prevent removing alias if the GlobalAlias does not have internal linkage Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 374354. gandhi21299 added a comment. - replaced a `cast` with a `dyn_cast` since the return value from `getCalleeFunction()` is not always a Function - `RUN on line 2` was causing 2 more scalar registers to be used on tonga due to @func_alias not being i

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 374369. gandhi21299 added a comment. - refreshing patch Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 Files: clang/lib/Driver/ToolChains/Clang.cpp clang/tes

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp:65 - return cast(Op.getGlobal()); + return dyn_cast(Op.getGlobal()); } arsenm wrote: > I think this is not the right place for this. If we can determine the

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 375079. gandhi21299 added a comment. - Declare an unhandled call lowering in SelectionDAG when a callee is encountered which cannot be casted into a Function - I am still investigating the effects on GlobalISel side of things, there seems to be a problem

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. It does not look like function calls are supported yet in AMDGPUCallLowering, is that correct? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 ___

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. In D109707#3016438 , @gandhi21299 wrote: > - replaced a `cast` with a `dyn_cast` since the return value from > `getCalleeFunction()` is not always a Function > - `RUN on line 2` was causing 2 more scalar registers to be used

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 375284. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - added the `REQUIRES` line as requested by Sam Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.l

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. @yaxunl Should inline-calls.ll be converted into an expected failing test or removed? (to avoid cast failure in AMDGPUResourceAnalysis to break the test) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https:

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 375655. gandhi21299 added a comment. - declare failure when lowering an accessor of a callee which is not a function, in GlobalISel Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp:96-97 if (Function* F = dyn_cast(A.getAliasee())) { + if (A.getLinkage() != GlobalValue::InternalLinkage) +continue; A.replaceAllUsesWith(F);

[PATCH] D110772: [clang] Allow pointers from any address space to gfx90a builtins

gandhi21299 created this revision. gandhi21299 added reviewers: yaxunl, rampitec. gandhi21299 requested review of this revision. Herald added a project: clang. Herald added a subscriber: cfe-commits. Explicitly address space cast the pointer if its address space differs from the address space of t

[PATCH] D110772: [clang] Allow pointers from any address space to gfx90a builtins

gandhi21299 updated this revision to Diff 376095. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - corrected test, address space cast should go under `unsafeAtomicAdd(...)` codegen Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llv

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 reclaimed this revision. gandhi21299 added a comment. Sorry, that was a mistake. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 ___ cfe-commits mai

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. inline-calls.ll failed on gfx908 due to the change in SIISelLowering.cpp, line 3015. Without the change, there is a failure in AMDGPUResourceAnalysis.cpp, line 65 because Op.getGlobal() is not a Function. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST AC

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 376559. gandhi21299 added a comment. - Since callees may alias to a function pointer, it makes sense for `getCalleeFunction(...)` to return a `Function` which is a cast of the operand of a `GlobalAlias`. Repository: rG LLVM Github Monorepo CHANGES S

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 updated this revision to Diff 376564. gandhi21299 added a comment. - eliminated changes in SIISelLowering Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 Files: clang/lib/Driver/ToolChains/C

[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

gandhi21299 added a comment. ping Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109707/new/ https://reviews.llvm.org/D109707 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added a comment. Passed PSDB Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.l

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:16114 +Intrinsic::ID IID; +llvm::Type *ArgTy = llvm::Type::getDoubleTy(getLLVMContext()); +switch (BuiltinID) { arsenm wrote: > Initializing this here is strange, sink down

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. Passed PSDB Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.l

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 marked an inline comment as done. gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12086 + TargetLowering::AtomicExpansionKind Kind, bool UnsafeFlag) { + ORE = new OptimizationRemarkEmitter(RMW->getFunction())

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 marked 7 inline comments as done. gandhi21299 added inline comments. Comment at: clang/include/clang/Basic/BuiltinsAMDGPU.def:201 +TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fadd_f32, "ff*1fi", "t", "gfx90a-insts") +TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fa

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 363159. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. addressed reviewers' feedback: - changed a builtin name, - corrected tests, - minor formatting nits Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION htt

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 363196. gandhi21299 added a comment. refreshing patch Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 Files: clang/include/clang/Basic/BuiltinsAMDGPU.def clan

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363324. gandhi21299 added a comment. Herald added a subscriber: pengfei. - transferred code over to AtomicExpandPass to be able to call `getAnalysis().getORE()` and simplify code as `SITargetLowering::shouldExpandAtomicRMW()` is called there. Repositor

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363484. gandhi21299 added a comment. refreshing patch Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu c

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363490. gandhi21299 marked 3 inline comments as done. gandhi21299 added a comment. generalizing the description of the remark for conversion into CAS loop. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D10689

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363498. gandhi21299 added a comment. correction in the CUDA LIT check Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-o

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363590. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. requested changes Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clan

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. @rampitec should the unsafe check go in some pass later in the pipeline then? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ c

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 marked 3 inline comments as done. gandhi21299 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:16212 + case AMDGPU::BI__builtin_amdgcn_flat_atomic_fmax_f64: { +Intrinsic::ID IID; +llvm::Type *ArgTy = llvm::Type::getDoubleTy(getLLVMContext())

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. @rampitec Since remarks cannot be emitted in SIISelLowering because it isn't a pass, in what form can I emit the diagnostics in SIISelLowering? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. How can I construct an ORE to start off with? I don't think its appropriate to construct it in `shouldExpandAtomicRMWInsts(RMW)` Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D10689

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added inline comments. Comment at: llvm/lib/CodeGen/AtomicExpandPass.cpp:597 + OptimizationRemark Remark(DEBUG_TYPE, "Passed", RMW->getFunction()); + Remark << "A hardware instruction was generated"; + return Remark; arsenm wrote: > Thi

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. @rampitec That would mean changing the interfaces in TargetLowering(), AMDGPUISelLowering, and so on. Is that okay? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 __

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. Callback sounds way nicer, I will do that. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-commits mailing list cfe-commits

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363753. gandhi21299 added a comment. passing ORE into SIISelLowering::shouldExpandAtomicRMW() to be able to emit ORE for unsafe hardware atomics. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ h

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 marked 5 inline comments as done. gandhi21299 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:16270 +llvm::Function *F = CGM.getIntrinsic(IID, {ArgTy}); +return Builder.CreateCall(F, {Addr, Val, ZeroI32, ZeroI32, ZeroI1}); + } -

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added a comment. @rampitec how do I handle the following? builtins-fp-atomics.cl:38:10: error: '__builtin_amdgcn_global_atomic_fadd_f64' needs target feature atomic-fadd-insts *rtn = __builtin_amdgcn_global_atomic_fadd_f64(addr, x, memory_order_relaxed); ^ Reposit

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 363786. gandhi21299 added a comment. refreshing patch, passed all tests associated with the `check-all` CMake option Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D10689

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added a comment. @rampitec what should I be testing exactly in the IR test? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 ___ cfe-commits mailing

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:16270 +llvm::Function *F = CGM.getIntrinsic(IID, {ArgTy}); +return Builder.CreateCall(F, {Addr, Val, ZeroI32, ZeroI32, ZeroI1}); + } rampitec wrote: > gandhi21299 wrote: > > r

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added a comment. Passed PSDB, @rampitec Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 updated this revision to Diff 364132. gandhi21299 added a comment. - eliminated the scope argument as per discussion - added more tests Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files:

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks when an FP atomic instruction is converted into a CAS loop or unsafe hardware instruction for GFX90A

gandhi21299 added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:16270 +llvm::Function *F = CGM.getIntrinsic(IID, {ArgTy}); +return Builder.CreateCall(F, {Addr, Val, ZeroI32, ZeroI32, ZeroI1}); + } @rampitec @arsenm I don't quite remember i

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 364152. gandhi21299 added a comment. updating test Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 Files: clang/include/clang/Basic/BuiltinsAMDGPU.def clang/l

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added inline comments. Comment at: clang/include/clang/Basic/BuiltinsAMDGPU.def:201 +TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fadd_f32, "ff*1f", "t", "gfx90a-insts") +TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fadd_v2f16, "V2hV2h*1V2h", "t", "gfx90a-insts")

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics in GFX90A

gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12139 +OptimizationRemark Remark(DEBUG_TYPE, "Passed", RMW->getFunction()); +Remark << "A hardware instruction was generated"; +return Remark; rampite

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 364242. gandhi21299 added a comment. - added more negative tests - fixed some tests Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 Files: clang/include/clang/B

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364256. gandhi21299 added a comment. - fixed remarks in SIISelLowering for hw instruction generation - fixed test accordingly Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.o

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 updated this revision to Diff 364266. gandhi21299 added a comment. - combined tests into a single file - renamed tests for consistency Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 Files:

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added a comment. Thanks a lot for the review, I will merge this patch in :) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 ___ cfe-commits mailing

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

This revision was automatically updated to reflect the committed changes. Closed by commit rG39dac1f7f656: [clang] Add clang builtins support for gfx90a (authored by gandhi21299). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://rev

[PATCH] D106909: [clang] Add clang builtins support for gfx90a

gandhi21299 added a comment. I will merge this patch in as soon as the builds are successful. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106909/new/ https://reviews.llvm.org/D106909 ___ cfe-commits ma

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364525. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - addressed reviewer feedback and updated test accordingly Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https:

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364527. gandhi21299 added a comment. added missing newline Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: llvm/include/llvm/CodeGen/TargetLowering.h l

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 marked an inline comment as done. gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12155 + OptimizationRemark Remark(DEBUG_TYPE, "Passed", RMW->getFunction()); + Remark << "A floating-point atomic instruction w

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364574. gandhi21299 marked 4 inline comments as done. gandhi21299 added a comment. requested changes Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: llvm

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364608. gandhi21299 marked 2 inline comments as done. gandhi21299 added a comment. - code formatting - @arsenm Being specific can be useful to some users. I don't see generating optimization remarks as a popular thing that every user will trigger so only

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364609. gandhi21299 added a comment. - removed unnecessary header includes Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atom

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364619. gandhi21299 added a comment. - eliminated trailing whitespaces in tests Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12120 + OptimizationRemarkEmitter *ORE, + OptimizationRemark OptRemark) { + ORE->emit([&]() { return OptRemark; }); rampitec wrote: > Why

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364687. gandhi21299 added a comment. - changed remarks Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364825. gandhi21299 added a comment. - changed 'CAS instruction' to 'CAS loop' Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 marked 2 inline comments as done. gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12120 + OptimizationRemarkEmitter *ORE, + OptimizationRemark OptRemark) { + ORE->emit([&]() { return OptRema

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364853. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - improved code style in AtomicExpand and SIISelLowering Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364854. gandhi21299 added a comment. created a patch against llvm trunk Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12120 + OptimizationRemarkEmitter *ORE, + OptimizationRemark OptRemark) { + ORE->emit([&]() { return OptRemark; }); rampitec wrote: > gan

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 marked an inline comment as done. gandhi21299 added inline comments. Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12120 + OptimizationRemarkEmitter *ORE, + OptimizationRemark OptRemark) { + ORE->emit([&]() { return OptRema

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 364891. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. - turned reportAtomicExpand(...) into a lambda as requested Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365066. gandhi21299 added a comment. refreshing patch Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: llvm/include/llvm/CodeGen/TargetLowering.h llvm/l

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365199. gandhi21299 added a comment. rebase Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-atomics-optremarks.cu clang/test/

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365202. gandhi21299 added a comment. - emit CAS loop remark only if the instruction is of floating point type Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 File

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365228. gandhi21299 added a comment. - floating point check is not required for cas loop check, getting rid of it Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365522. gandhi21299 added a comment. - restricting remarks emission on AMDGPU targets only Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/Cod

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 added a comment. @rampitec comments? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https:/

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365564. gandhi21299 marked an inline comment as done. gandhi21299 added a comment. requested change Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 added a comment. @rampitec besides the remarks, am I missing anything else in the patch? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 ___ cfe-com

[PATCH] D106891: [AMDGPU] [Remarks] Emit optimization remarks for FP atomics

gandhi21299 updated this revision to Diff 365625. gandhi21299 added a comment. - reverted the patch to the previous one Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106891/new/ https://reviews.llvm.org/D106891 Files: clang/test/CodeGenCUDA/fp-a

  1   2   3   >