[PATCH] D158695: [clang] Fix missing contract flag in sqrt intrinsic

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:501 if (CGF.Builder.getIsFPConstrained()) { CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E); Function *F = CGF.CGM.getIntrinsic(ConstrainedIntrinsicID, Src0->getType()); r

[PATCH] D158695: [clang] Fix missing contract flag in sqrt intrinsic

arsenm added inline comments. Comment at: clang/test/CodeGen/fp-contract-fast-pragma.cpp:77 +// CHECK: _Z13fp_contract_7f +// CHECK: tail call contract float @llvm.sqrt.f32(float %a) + return __builtin_sqrtf(a); This isn't demonstrating the strict support, proba

[PATCH] D159206: [Clang] Propagate target-features if compatible when using mlink-builtin-bitcode

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:2034 + } + + FuncAttr.addAttribute("target-features", llvm::join(MergedFeatures, ",")); Really it would be less bad if the incompatible functions were not imported rather than the backend p

[PATCH] D159206: [Clang] Propagate target-features if compatible when using mlink-builtin-bitcode

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:2035 + + FuncAttr.addAttribute("target-features", llvm::join(MergedFeatures, ",")); +} do you need to guard against adding the empty attribute? I don't want to see "target-features"="" Rep

[PATCH] D156743: clang/OpenCL: Add inline implementations of sqrt in builtin header

arsenm added a comment. ping The alternative is to directly put the !fpmath on the sqrt call sites but I have no idea how to do that CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156743/new/ https://reviews.llvm.org/D156743 ___ cfe-commits

[PATCH] D159206: [Clang] Propagate target-features if compatible when using mlink-builtin-bitcode

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.h:398-401 +/// If \p F "target-features" are incompatible with the \p TargetOpts features, +/// it is correct to drop the function. \return true if \p F is dropped +bool dropFunctionWithIncompatibleAttributes(llvm

[PATCH] D159206: [Clang] Propagate target-features if compatible when using mlink-builtin-bitcode

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:2017 +for (StringRef Feature : llvm::split(FFeatures.getValueAsString(), ',')) { + bool EnabledForFunc = Feature[0] == '+'; + StringRef Name = Feature.substr(1); Do you need to g

[PATCH] D156743: clang/OpenCL: Add inline implementations of sqrt in builtin header

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D156743/new/ https://reviews.llvm.org/D156743 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D159206: [Clang] Propagate target-features if compatible when using mlink-builtin-bitcode

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:2030-2031 + bool EnabledForTarget = TEntry->second; + if (EnabledForTarget != EnabledForFunc) +return; +} jmmartinez wrote: > arsenm wrote: > > Early return breaks the ot

[PATCH] D138274: Add version to all LLVM cmake package

arsenm accepted this revision. arsenm added a comment. This revision is now accepted and ready to land. I thought we had this already. The amount of boiler plate required repeated in each component is depressing, but I wouldn't be surprised if it's really needed Repository: rG LLVM Github Mon

[PATCH] D138392: clang/HIP: Fix broken implementations of __make_mantissa* functions

arsenm created this revision. arsenm added reviewers: yaxunl, JonChesterfield, saiislam, scchan, AlexVlx, b-sumner. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. The optimizer was folding the entire function to return 0. This meant to be c

[PATCH] D138394: HIP: Directly call fma builtins

arsenm created this revision. arsenm added reviewers: yaxunl, JonChesterfield, b-sumner. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. https://reviews.llvm.org/D138394 Files: clang/lib/Headers/__clang_hip_math.h clang/test/Headers/__cl

[PATCH] D138395: HIP: Directly call fmin/fmax builtins

arsenm created this revision. arsenm added reviewers: yaxunl, b-sumner, JonChesterfield. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. https://reviews.llvm.org/D138395 Files: clang/lib/Headers/__clang_hip_math.h clang/test/Headers/__cl

[PATCH] D138396: HIP: Directly call signbit builtins

arsenm created this revision. arsenm added reviewers: JonChesterfield, b-sumner, yaxunl. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. https://reviews.llvm.org/D138396 Files: clang/lib/Headers/__clang_hip_math.h clang/test/Headers/__cl

[PATCH] D138399: HIP: Directly call isinf builtins

arsenm created this revision. arsenm added reviewers: b-sumner, JonChesterfield, yaxunl. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. https://reviews.llvm.org/D138399 Files: clang/lib/Headers/__clang_hip_math.h clang/test/Headers/__cl

[PATCH] D138439: clang: Fix cast failure when using -fsanitize=undefined for HIP

arsenm created this revision. arsenm added reviewers: yaxunl, samsonov, tra, bkramer. Herald added a subscriber: arichardson. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. This was assuming a direct reference to the global variable. The con

[PATCH] D138391: clang/HIP: Add new header test for math IR gen

arsenm closed this revision. arsenm added a comment. 2edafe8393fa7c9ad2e95d691191469adb7bf4b0 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138391/new/ https://reviews.llvm.org/D138391 __

[PATCH] D138504: clang/HIP: Remove __llvm_amdgcn_* wrapper hacks

arsenm created this revision. arsenm added reviewers: yaxunl, JonChesterfield, b-sumner. Herald added a subscriber: kosarev. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. These are leftover hacks from using asm declaratios to access intrins

[PATCH] D138394: HIP: Directly call fma builtins

arsenm updated this revision to Diff 477213. arsenm added a comment. Missed one CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138394/new/ https://reviews.llvm.org/D138394 Files: clang/lib/Headers/__clang_hip_math.h clang/test/Headers/__clang_hip_math.hip Index: clang/test/Headers/

[PATCH] D138504: clang/HIP: Remove __llvm_amdgcn_* wrapper hacks

arsenm updated this revision to Diff 477219. arsenm added a comment. Remove some of the externally defined ones CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138504/new/ https://reviews.llvm.org/D138504 Files: clang/lib/Headers/__clang_hip_libdevice_declares.h clang/lib/Headers/__c

[PATCH] D138507: HIP: Directly use sqrt builtins instead of calling ocml (f32 case)

arsenm added a comment. In D138507#3944358 , @b-sumner wrote: > __builtin_sqrtf does not produce a correctly rounded result. I don't > recommend this change. It's supposed to. I'm working towards correctly handling these in the backend CHANGES SINCE

[PATCH] D138509: clang/HIP: Add another math header test

arsenm created this revision. arsenm added reviewers: yaxunl, JonChesterfield, b-sumner. Herald added a project: All. arsenm requested review of this revision. Herald added a subscriber: wdng. This needs more exhaustive checks for the other things here; for now just test the ones directly calling

[PATCH] D138394: HIP: Directly call fma builtins

arsenm updated this revision to Diff 477238. arsenm added a comment. Catch another hidden in another header CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138394/new/ https://reviews.llvm.org/D138394 Files: clang/lib/Headers/__clang_hip_cmath.h clang/lib/Headers/__clang_hip_math.h

[PATCH] D138473: clang/HIP: Inline frexp/frexpf implementations

arsenm added a comment. In D138473#3944586 , @yaxunl wrote: > Are you sure these functions are equivalent? we do not have a comprehensive > test for these functions regarding accuracy. These are ultimately simple bit extracts. The library implementation

[PATCH] D91195: Add Annotation2MD pass to add !annotate metadata from llvm.global.annotations

arsenm added inline comments. Herald added subscribers: kosarev, ormris. Herald added a project: All. Comment at: llvm/lib/Transforms/IPO/Annotation2Metadata.cpp:51 +auto *StrGEP = dyn_cast(OpC->getOperand(1)); +if (!StrGEP || StrGEP->getNumOperands() < 2) + continue

[PATCH] D138702: support for HIP non hostcall printf

arsenm requested changes to this revision. arsenm added a comment. Herald added a subscriber: wdng. I have a few questions. First, why surface this to users? If we really need to, I don't think this is the right flag name/design. A named argument to some kind of printf lowering flag would be bet

[PATCH] D138509: clang/HIP: Add another math header test

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138509/new/ https://reviews.llvm.org/D138509 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D138532: [SPARC][clang] Enable frame pointer optimization by default

arsenm accepted this revision. arsenm added a comment. This revision is now accepted and ready to land. LGTM. I still find it strange that by default clang explicitly disables an optimization that's the default for unattributed functions Repository: rG LLVM Github Monorepo CHANGES SINCE LAST

[PATCH] D138439: clang: Fix cast failure when using -fsanitize=undefined for HIP

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138439/new/ https://reviews.llvm.org/D138439 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D138392: clang/HIP: Fix broken implementations of __make_mantissa* functions

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138392/new/ https://reviews.llvm.org/D138392 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D138868: AMDGPU/clang: Remove target features from address space test builtins

arsenm created this revision. arsenm added reviewers: yaxunl, b-sumner, JonChesterfield. Herald added subscribers: kosarev, kerbowa, tpr, dstuttard, jvesely, kzhuravl. Herald added a project: All. arsenm requested review of this revision. Herald added subscribers: llvm-commits, wdng. Herald added a

[PATCH] D138870: clang/AMDGPU: Remove flat-address-space from feature map

arsenm created this revision. arsenm added reviewers: yaxunl, JonChesterfield, b-sumner, jdoerfert. Herald added subscribers: kosarev, kerbowa, tpr, dstuttard, jvesely, kzhuravl. Herald added a project: All. arsenm requested review of this revision. Herald added subscribers: sstefan1, wdng. This w

[PATCH] D138509: clang/HIP: Add another math header test

arsenm closed this revision. arsenm added a comment. aa4acea8cd8d30b6c8b5d539217be0d5c76db625 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138509/new/ https://reviews.llvm.org/D138509 __

[PATCH] D138439: clang: Fix cast failure when using -fsanitize=undefined for HIP

arsenm closed this revision. arsenm added a comment. 91ba8b2b8dc44355851f2953ea3eeea21c230c53 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138439/new/ https://reviews.llvm.org/D138439 __

[PATCH] D138393: HIP: Directly call fabs builtins

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138393/new/ https://reviews.llvm.org/D138393 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D138722: Overload all llvm.annotation intrinsics for globals argument

arsenm added inline comments. Comment at: llvm/test/Bitcode/upgrade-annotation.ll:14 + +declare i32 @llvm.annotation.i32(i32, i8*, i8*, i32) +; CHECK: declare i32 @llvm.annotation.i32.p0i8(i32, i8*, i8*, i32) New tests should use opaque pointers. For autoupgrade

[PATCH] D76472: AMDGPU: Emit llvm.fshr for __builtin_amdgcn_alignbit

arsenm closed this revision. arsenm added a comment. 3f533006ba8c8ae6f3596f49f480aa794ed4e347 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76472/new/ https://reviews.llvm.org/D76472 __

[PATCH] D76772: [AMDGPU] Add __builtin_amdgcn_workgroup_size_x/y/z

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGBuiltin.cpp:13428 +return Call; + return CGF.Builder.CreateAddrSpaceCast(Call, RetTy); +} Why is this necessary? The builtin always has the same return type? Comment at: clang/l

[PATCH] D59321: WIP: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 252630. arsenm added a comment. Herald added a subscriber: kerbowa. Rebase CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/clang/Dri

[PATCH] D59321: WIP: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 252692. arsenm added a comment. Use -nogpulib instead of -nodefaultlibs CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/clang/Driver

[PATCH] D59321: WIP: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 252708. arsenm added a comment. Cleanup a bit. This should eventually merge with the HIP library handling when it switches to using clang cc1 instead of llc CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Fi

[PATCH] D59321: WIP: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 252715. arsenm added a comment. Herald added a subscriber: hiraditya. Handle wave64 library CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/

[PATCH] D76772: [AMDGPU] Add __builtin_amdgcn_workgroup_size_x/y/z

arsenm accepted this revision. arsenm added inline comments. This revision is now accepted and ready to land. Comment at: clang/test/CodeGenCUDA/amdgpu-workgroup-size.cu:2 +// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa \ +// RUN: -fcuda-is-device -emit-llvm -o - -x hip %s \ +/

[PATCH] D76862: HIP: Ensure new denormal mode attributes are set

arsenm created this revision. arsenm added reviewers: yaxunl, gregrodgers. Herald added subscribers: kerbowa, tpr, nhaehnle, wdng, jvesely. Apparently HIPToolChain does not subclass from AMDGPUToolChain, so this was not applying the new denormal attributes. I'm not sure why this doesn't subclass.

[PATCH] D76887: AMDGPU: Make HIPToolChain a subclass of ROCMToolChain

arsenm created this revision. arsenm added reviewers: yaxunl, gregrodgers. Herald added subscribers: t-tye, tpr, dstuttard, wdng, kzhuravl. arsenm added parent revisions: D59321: WIP: AMDGPU: Teach toolchain to link rocm device libs, D76862: HIP: Ensure new denormal mode attributes are set. This

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 252978. arsenm retitled this revision from "WIP: AMDGPU: Teach toolchain to link rocm device libs" to "AMDGPU: Teach toolchain to link rocm device libs". arsenm edited the summary of this revision. Herald added a subscriber: Anastasia. CHANGES SINCE LAST ACTIO

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 253156. arsenm added a comment. Cleanup wave64 check CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/clang/Driver/Options.td clang

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 253194. arsenm added a comment. Fix negating backwards logic for default FTZ mode CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/cl

[PATCH] D76950: HIP: Link correct denormal mode library

arsenm created this revision. arsenm added a reviewer: yaxunl. Herald added a subscriber: wdng. arsenm added parent revisions: D59321: AMDGPU: Teach toolchain to link rocm device libs, D76862: HIP: Ensure new denormal mode attributes are set. This wasn't respecting the flush mode based on the def

[PATCH] D76957: HIP: Merge builtin library handling

arsenm created this revision. arsenm added reviewers: yaxunl, gregrodgers. Herald added subscribers: kerbowa, Anastasia, nhaehnle, wdng, jvesely. arsenm added parent revisions: D76862: HIP: Ensure new denormal mode attributes are set, D76950: HIP: Link correct denormal mode library, D59321: AMDGPU

[PATCH] D77013: [AMDGPU] Add options -mamdgpu-ieee -mno-amdgpu-ieee

arsenm added inline comments. Comment at: clang/include/clang/Basic/CodeGenOptions.def:399 +/// Whether to emit IEEE754-2008 NaN compliant instructions if available (AMDGPU Only) +CODEGENOPT(EmitIEEENaNCompliantInsts, 1, 1) Description is misleading. Better de

[PATCH] D77013: [AMDGPU] Add options -mamdgpu-ieee -mno-amdgpu-ieee

arsenm added inline comments. Comment at: clang/test/CodeGenOpenCL/amdgpu-ieee.cl:20 +} + +// ON-NOT: attributes [[ATTRS]] = {{.*}} "amdgpu-ieee" arsenm wrote: > Should also test a non-kernel function I think we should also have some ISA check run lines that show

[PATCH] D76389: [NewPM] Run the Speculative Execution Pass if the target has divergent branches

arsenm added a comment. Commit message should say only if? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76389/new/ https://reviews.llvm.org/D76389 ___ cfe-commits mailing list cfe-commits@lists.llvm.o

[PATCH] D76862: HIP: Ensure new denormal mode attributes are set

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76862/new/ https://reviews.llvm.org/D76862 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D76862: HIP: Ensure new denormal mode attributes are set

arsenm added a comment. In D76862#1953013 , @yaxunl wrote: > Are there any other clang options affecting flushing denormals? If so, are > they working properly after this change? Do we need to have tests for them? > Thanks. I already ensured these are

[PATCH] D76887: AMDGPU: Make HIPToolChain a subclass of ROCMToolChain

arsenm closed this revision. arsenm added a comment. 175e42303bb2a4253c6512b1ae05b32b0004 , rebased to avoid dependence on ROCM toolchain patch CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76887/new/ https://rev

[PATCH] D76862: HIP: Ensure new denormal mode attributes are set

arsenm closed this revision. arsenm added a comment. c9d65a48af1d7bbfed7e785613cc9d9acf71821b CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76862/new/ https://reviews.llvm.org/D76862 __

[PATCH] D76950: HIP: Link correct denormal mode library

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76950/new/ https://reviews.llvm.org/D76950 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D76950: HIP: Link correct denormal mode library

arsenm closed this revision. arsenm marked an inline comment as done. arsenm added a comment. 4ea3650c212ae471657d3a253cd424ce9d1316ac CHANGES SINCE LAST ACTION https://reviews.llvm.org/D76950/new/ https://reviews.llvm.org

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm added inline comments. Comment at: clang/include/clang/Basic/DiagnosticDriverKinds.td:264 def err_drv_invalid_malign_branch_EQ : Error< "invalid argument '%0' to -malign-branch=; each element must be one of: %1">; yaxunl wrote: > could you please reb

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 254229. arsenm marked an inline comment as done. arsenm added a comment. Rebase again CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/includ

[PATCH] D77234: clang/AMDGPU: Stop setting old denormal subtarget features

arsenm created this revision. arsenm added a reviewer: yaxunl. Herald added subscribers: kerbowa, t-tye, tpr, dstuttard, nhaehnle, wdng, jvesely, kzhuravl. https://reviews.llvm.org/D77234 Files: clang/lib/Basic/Targets/AMDGPU.cpp clang/lib/Basic/Targets/AMDGPU.h clang/test/CodeGenCUDA/flus

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm marked 2 inline comments as done. arsenm added a comment. In D59321#1955405 , @hliao wrote: > Do we have a better way to avoid adding those empty bitcode files? No, we need the files to exist for tests. This is what existing bitcode link tests do

[PATCH] D76389: [NewPM] Run the Speculative Execution Pass only if the target has divergent branches

arsenm added inline comments. Comment at: clang/test/CodeGen/thinlto-distributed-newpm.ll:110 ; CHECK-O: Running analysis: OuterAnalysisManagerProxy -; CHECK-O: Running pass: SpeculativeExecutionPass on main +; CHECK-O: Running pass: SpeculativeExecutionIfHasBranchDivergencePass

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 254584. arsenm added a comment. Rebase CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/clang/Driver/Options.td clang/lib/Driver/Dr

[PATCH] D77329: [AMDGPU] Allow AGPR in inline asm

arsenm added inline comments. Comment at: clang/test/CodeGenOpenCL/inline-asm-amdgcn.cl:11-14 + float acc_c; + float reg_a; + float reg_b; + float reg_c; These mostly aren't the right types? CHANGES SINCE LAST ACTION https://reviews.llvm.org/D77329/new/

[PATCH] D59321: AMDGPU: Teach toolchain to link rocm device libs

arsenm updated this revision to Diff 254591. arsenm added a comment. Correct group CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59321/new/ https://reviews.llvm.org/D59321 Files: clang/include/clang/Basic/DiagnosticDriverKinds.td clang/include/clang/Driver/Options.td clang/lib/Dr

[PATCH] D77234: clang/AMDGPU: Stop setting old denormal subtarget features

arsenm closed this revision. arsenm added a comment. ce2258c1cd5dc9cf20040d1b1e540d80250c1435 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D77234/new/ https://reviews.llvm.org/D77234 __

[PATCH] D74500: clang: Treat ieee mode as the default for denormal-fp-math

arsenm updated this revision to Diff 245972. arsenm edited the summary of this revision. arsenm added a comment. Rebase CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74500/new/ https://reviews.llvm.org/D74500 Files: clang/include/clang/Basic/CodeGenOptions.h clang/include/clang/Dri

[PATCH] D74935: [LangRef][AliasAnalysis] Clarify `noalias` affects only modified objects

arsenm added a comment. Another related point I’ve never been clear on is if a readnone function is allowed to read constant memory Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74935/new/ https://reviews.llvm.org/D74935 ___

[PATCH] D75028: Make __builtin_amdgcn_dispatch_ptr dereferenceable and align at 4

arsenm accepted this revision. arsenm added a comment. This revision is now accepted and ready to land. LGTM CHANGES SINCE LAST ACTION https://reviews.llvm.org/D75028/new/ https://reviews.llvm.org/D75028 ___ cfe-commits mailing list cfe-commits@l

[PATCH] D75028: Make __builtin_amdgcn_dispatch_ptr dereferenceable and align at 4

arsenm added inline comments. Comment at: llvm/include/llvm/IR/IntrinsicsAMDGPU.td:144 def int_amdgcn_dispatch_ptr : - GCCBuiltin<"__builtin_amdgcn_dispatch_ptr">, Intrinsic<[LLVMQualPointerType], [], I don't think you actually need to delete this. It should

[PATCH] D74500: clang: Treat ieee mode as the default for denormal-fp-math

arsenm added a comment. ping CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74500/new/ https://reviews.llvm.org/D74500 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D74500: clang: Treat ieee mode as the default for denormal-fp-math

arsenm updated this revision to Diff 248349. arsenm added a comment. Split out constexpr change CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74500/new/ https://reviews.llvm.org/D74500 Files: clang/include/clang/Basic/CodeGenOptions.h clang/include/clang/Driver/ToolChain.h clang/

[PATCH] D74500: clang: Treat ieee mode as the default for denormal-fp-math

arsenm closed this revision. arsenm added a comment. c64ca93053af235bac0ca4dcdcd21c8882478310 CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74500/new/ https://reviews.llvm.org/D74500 __

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. In D69498#1725819 , @mehdi_amini wrote: > Maybe we can start by looking into the motivation for this patch: > > > There is a burden on frontends in environments that care about convergent > > operations to add the attribute just in

[PATCH] D69598: WIP: Work on cleaning up denormal mode handling

arsenm created this revision. arsenm added reviewers: scanon, andrew.w.kaylor, cameron.mcinally, spatel, RKSimon, olista01, SjoerdMeijer. Herald added subscribers: dexonsmith, hiraditya, tpr, mgorny, wdng. Herald added a project: LLVM. Cleanup handling of the denormal-fp-math attribute. Consolida

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. In D69498#1727419 , @mehdi_amini wrote: > In D69498#1727080 , @jdoerfert wrote: > > > Let me quote @arsenm here because this is so important: "Basically no > > frontend has gotten this right

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. In D69498#1727626 , @mehdi_amini wrote: > In D69498#1727546 , @jdoerfert wrote: > > > In D69498#1727419 , @mehdi_amini > > wrote: > > > > > In D69498

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. In D69498#1728039 , @tra wrote: > Perhaps we can deal with that by providing a way to specify per-module > default for the assumed convergence of the functions and then checking in the > back-end (only those that do care about con

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. As far as optimization inhibition is concerned, noconvergent will be inferred for all functions that don't call convergent intrinsics (i.e. the state of the world for all functions on all CPU targets). The frontend needing to do something for optimization comes up in rel

[PATCH] D69666: clang: Fix assert on void pointer arithmetic with address_space

arsenm created this revision. arsenm added reviewers: yaxunl, jdoerfert, Anastasia. Herald added a subscriber: wdng. This attempted to always use the default address space void pointer type instead of preserving the source address space. https://reviews.llvm.org/D69666 Files: clang/lib/CodeGe

[PATCH] D69666: clang: Fix assert on void pointer arithmetic with address_space

arsenm added a comment. In D69666#1728946 , @yaxunl wrote: > Is the description reversed? > > This attempts to preserve the source address space instead of always using > the default address space for void pointer type. I don't think so? This now preser

[PATCH] D69666: clang: Fix assert on void pointer arithmetic with address_space

arsenm closed this revision. arsenm added a comment. c6da9ec0e90ea8798ecae583bb8d26bdf6b9b79f CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69666/new/ https://reviews.llvm.org/D69666 ___ cfe-commits mailing list cfe-commits@lists.llvm.org h

[PATCH] D69756: [opaque pointer types] Add element type argument to IRBuilder CreatePreserveStructAccessIndex and CreatePreserveArrayAccessIndex

arsenm accepted this revision. arsenm added a comment. This revision is now accepted and ready to land. Herald added a subscriber: wdng. LGTM Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69756/new/ https://reviews.llvm.org/D69756 __

[PATCH] D69770: Add recoverable string parsing errors to APFloat

arsenm added inline comments. Comment at: llvm/lib/Support/APFloat.cpp:273 + if (p != end) +return createError("Invalid exponent in exponent"); Error message sounds like nonsense Comment at: llvm/unittests/ADT/APFloatTest.cpp:1322 + EX

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm updated this revision to Diff 227775. arsenm retitled this revision from "WIP: Work on cleaning up denormal mode handling" to "Work on cleaning up denormal mode handling". arsenm added a comment. Defer any behavior changes until a future patch, so all tests now pass CHANGES SINCE LAST AC

[PATCH] D69826: [hip] Enable pointer argument lowering through coercing type.

arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:1308-1310 + if (isa(SrcTy) && + isa(DstTy) && + SrcTy->getPointerAddressSpace() != DstTy->getPointerAddressSpace()) { I would somewhat prefer 2 dyn_cast and getAddressSpace, this is

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm marked an inline comment as done. arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:1736-1737 +if (CodeGenOpts.FPSubnormalMode != llvm::SubnormalMode::Invalid) + FuncAttrs.addAttribute("denormal-fp-math", + llvm::s

[PATCH] D69878: Consoldiate internal denormal flushing controls

arsenm created this revision. arsenm added reviewers: scanon, spatel, cameron.mcinally, andrew.w.kaylor, tra, jlebar, Anastasia, yaxunl. Herald added subscribers: hiraditya, kristof.beyls, tpr, nhaehnle, wdng, jvesely, jholewinski. Herald added a project: LLVM. arsenm added parent revisions: D695

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm marked an inline comment as done. arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:1736-1737 +if (CodeGenOpts.FPSubnormalMode != llvm::SubnormalMode::Invalid) + FuncAttrs.addAttribute("denormal-fp-math", + llvm::s

[PATCH] D69498: IR: Invert convergent attribute handling

arsenm added a comment. In D69498#1731265 , @mehdi_amini wrote: > In D69498#1727650 , @dexonsmith > wrote: > > > In D69498#1723606 , @rjmccall > > wrote: > > > > > Perhaps

[PATCH] D69878: Consoldiate internal denormal flushing controls

arsenm marked an inline comment as done. arsenm added inline comments. Comment at: llvm/docs/LangRef.rst:1828-1831 + be flushed to zero by standard floating point operations. It is not + mandated that flushing to zero occurs, but if a subnormal output is + flushed to zero,

[PATCH] D69878: Consoldiate internal denormal flushing controls

arsenm marked an inline comment as done. arsenm added inline comments. Comment at: llvm/docs/LangRef.rst:1828-1831 + be flushed to zero by standard floating point operations. It is not + mandated that flushing to zero occurs, but if a subnormal output is + flushed to zero,

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm updated this revision to Diff 228161. arsenm added a comment. Rename to denormal CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69598/new/ https://reviews.llvm.org/D69598 Files: clang/include/clang/Basic/CodeGenOptions.h clang/lib/CodeGen/CGCall.cpp clang/lib/Frontend/Compi

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm marked an inline comment as done. arsenm added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:1736-1737 +if (CodeGenOpts.FPSubnormalMode != llvm::SubnormalMode::Invalid) + FuncAttrs.addAttribute("denormal-fp-math", + llvm::s

[PATCH] D69878: Consoldiate internal denormal flushing controls

arsenm updated this revision to Diff 228167. arsenm added a comment. Rename subnormal to denormal. Will defer splitting input and output setting into a future patch before switching default behavior CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69878/new/ https://reviews.llvm.org/D6987

[PATCH] D69598: Work on cleaning up denormal mode handling

arsenm updated this revision to Diff 228168. arsenm added a comment. Missed a spot to rename CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69598/new/ https://reviews.llvm.org/D69598 Files: clang/include/clang/Basic/CodeGenOptions.h clang/lib/CodeGen/CGCall.cpp clang/lib/Frontend/

[PATCH] D69878: Consoldiate internal denormal flushing controls

arsenm marked an inline comment as done. arsenm added a comment. In D69878#1736865 , @Anastasia wrote: > > Stop emitting the denorms-are-zero attribute for the OpenCL flag. It > > has no in-tree users. The meaning would also be target dependent, such > >

[PATCH] D69978: Separately track input and output denormal mode

arsenm created this revision. arsenm added reviewers: scanon, cameron.mcinally, spatel, andrew.w.kaylor, mibintc, SjoerdMeijer. Herald added subscribers: dexonsmith, hiraditya, tpr, nhaehnle, wdng, jvesely, jholewinski. Herald added a project: LLVM. arsenm added parent revisions: D69878: Consoldi

<    4   5   6   7   8   9   10   11   12   13   >