AMDGPU: Delete probably wrong constant folding of expm1

It's not really correct to implement this as exp(x) - 1, it was maybe
OK for the restricted float-as-double case handled here. There's not a
s

AMDGPU: Delete probably wrong constant folding of expm1

It's not really correct to implement this as exp(x) - 1, it was maybe
OK for the restricted float-as-double case handled here. There's not a
strong reason to special case it with the host function, as the
implementation naturally constant folds anyway. InstSimplify can fully
handle everything in it, so just running the inliner alone with a
constant argument produces the fully constant folded result. This also
had no tests.

https://reviews.llvm.org/D156892

show more ...

AMDGPU: Simplify and improve sincos matching

The first trivial example I tried failed to merge due to the user scan
logic. Remove the complicated scan of users handling with distance
thresholds, wit

AMDGPU: Simplify and improve sincos matching

The first trivial example I tried failed to merge due to the user scan
logic. Remove the complicated scan of users handling with distance
thresholds, with a same block restriction. The actual expansion of
sincos is basically the same size as sin or cos individually. Copy the
technique the generic optimization uses, which is to just use the
input instruction as the insert point or just insert at the start of
the entry block.

https://reviews.llvm.org/D156706

show more ...

AMDGPU: Delete old PM support for libcall passes

This has no reason to run in the codegen pipeline.

AMDGPU: Don't try to fold wavefrontsize intrinsic in libcall simplify

It's not a libcall so doesn't really belong here to begin
with. Relying on checking the target name and explicit features isn't

AMDGPU: Don't try to fold wavefrontsize intrinsic in libcall simplify

It's not a libcall so doesn't really belong here to begin
with. Relying on checking the target name and explicit features isn't
particularly sound either. The library doesn't use the intrinsic
anymore, so it doesn't matter anyway.

show more ...

AMDGPU: Don't try memory optimizations in libcall optimizer

This was trying to find a loaded value for some reason when looking
for sincos arguments. This is untested and shouldn't be necessary.

ht

AMDGPU: Don't try memory optimizations in libcall optimizer

This was trying to find a loaded value for some reason when looking
for sincos arguments. This is untested and shouldn't be necessary.

https://reviews.llvm.org/D156746

show more ...

AMDGPU: Directly emit fabs intrinsic instead of new libcall

AMDGPU: Move check of compatible libcall

https://reviews.llvm.org/D156681

AMDGPU: Don't store current instruction in AMDGPULibCalls member

This was adding confusing global state which was shadowed most of the
time.

https://reviews.llvm.org/D156680

AMDGPU: Refactor libcall simplify to help with future refined fast math flag usage

https://reviews.llvm.org/D156678

AMDGPU: Don't parse name of sin/cos twice in libcall simplify

AMDGPU: Partially respect nobuiltin in libcall simplifier

There are more contexts where it's not handled correctly but this is
the simplest one.

https://reviews.llvm.org/D156682

AMDGPU: Some AMDGPULibCalls cleanups

dyn_cast instead of isa+cast, and initialize on declaration.

AMDGPU: Simplify early exit handling for libcall simplify

Early exit on intrinsics and don't duplicate indirect call
checks. Also let the IRBuilder constructor figure out the insert point
rather tha

AMDGPU: Simplify early exit handling for libcall simplify

Early exit on intrinsics and don't duplicate indirect call
checks. Also let the IRBuilder constructor figure out the insert point
rather than doing it manually. Also avoid debug print about trying to
simplify calls in more unhandled scenarios.

show more ...

AMDGPU: Remove some typed pointer handling

try to fix build yet more after 16544cbe64b8

[AMDGPU] Remove a redundant variable (NFC)

ArrayRef has operator[], so we don't need to access the contents via
data().

[AMDGPU] Declare TableRef in terms of ArrayRef (NFC)

[AMDGPU] Pull out repeated getVecSize() calls. NFC.

This is guaranteed to be evaluated so we can avoid repeated calls.

Helps the static analyzer as it couldn't recognise that each getVecSize() woul

[AMDGPU] Pull out repeated getVecSize() calls. NFC.

This is guaranteed to be evaluated so we can avoid repeated calls.

Helps the static analyzer as it couldn't recognise that each getVecSize() would return the same value.

show more ...

Cleanup header dependencies in LLVMCore

Based on the output of include-what-you-use.

This is a big chunk of changes. It is very likely to break downstream code
unless they took a lot of care in avo

Cleanup header dependencies in LLVMCore

Based on the output of include-what-you-use.

This is a big chunk of changes. It is very likely to break downstream code
unless they took a lot of care in avoiding hidden ehader dependencies, something
the LLVM codebase doesn't do that well :-/

I've tried to summarize the biggest change below:

- llvm/include/llvm-c/Core.h: no longer includes llvm-c/ErrorHandling.h
- llvm/IR/DIBuilder.h no longer includes llvm/IR/DebugInfo.h
- llvm/IR/IRBuilder.h no longer includes llvm/IR/IntrinsicInst.h
- llvm/IR/LLVMRemarkStreamer.h no longer includes llvm/Support/ToolOutputFile.h
- llvm/IR/LegacyPassManager.h no longer include llvm/Pass.h
- llvm/IR/Type.h no longer includes llvm/ADT/SmallPtrSet.h
- llvm/IR/PassManager.h no longer includes llvm/Pass.h nor llvm/Support/Debug.h

And the usual count of preprocessed lines:
$ clang++ -E -Iinclude -I../llvm/include ../llvm/lib/IR/*.cpp -std=c++14 -fno-rtti -fno-exceptions | wc -l
before: 6400831
after: 6189948

200k lines less to process is no that bad ;-)

Discourse thread on the topic: https://llvm.discourse.group/t/include-what-you-use-include-cleanup

Differential Revision: https://reviews.llvm.org/D118652

show more ...

[AMDGPU] Remove unused declarations fold_exp* and fold_log* (NFC)

[AMDGPU] Remove replaceWithNative (NFC)

The function was introduced without any use on Aug 11, 2017 in commit
7f37794ebd2c6c36224597800e4d1e5a99ad80e9.

[llvm] Use nullptr instead of 0 (NFC)

Identified with modernize-use-nullptr.

[AMDGPU][NFC] Fix typos

Differential Revision: https://reviews.llvm.org/D113672

[Target, Transforms] Use StringRef::contains (NFC)

AMDGPULibCalls - constify some FuncInfo& arguments. NFCI.