add png logo for Annapurna Labs

fix fmls fp16

[aarch64] add def-pats for dot product

[aarch64] move custom isel of extract_vector_elt to td file - NFC

[aarch64] Add combine patterns for fp16 fmla

[AArch64][asan] fix typo in AsanStats::Print

[AArch64] Speed-up leak and address sanitizers on AArch64 for 48-bit VMA

[hot-cold-split] fix failing testcases

[hot-cold-split] fix static analysis of cold regions

update call for papers for the LLVM Performance Workshop

HotColdSplit: fix invalid SSA due to outlining

HotColdSplit: fix isSingleEntrySingleExit

HotColdSplit: add back propagation to extend cold regions

HotColdSplitting: fix test failing because of last commit

HotColdSplitting: check that target supports cold calling convention

add flag instead of using a constant [NFC]

make flag name more specific to gvn [NFC]

add new upcoming events, move down old workshops

add preliminary call for talks for the third LLVM Performance Workshop

[InstCombine] reassociate loop invariant GEP chains to enable LICM

[x86][aarch64] ask the backend whether it has a vector blend instruction

[AArch64] add missing pattern for insert_subvector undef

DA: remove uses of GEP, only ask SCEV

[AArch64] define isExtractSubvectorCheap

fix PR36582

[AArch64] generate vuzp instead of mov

update deadlines for llvm performance workshop

Handle UnreachableInst in isGuaranteedToTransferExecutionToSuccessor

[www] add slides from the LLVM Perf meeting

[www] add slides from the LLVM Performance Workshop

[www] add slides from LLVM Performance Workshop

Summary: [www] LLVM Performace workshop: add missing commas

Summary: [www] LLVM Performace workshop: add abstracts

Summary: [www] LLVM Performace workshop: list accepted talks

improve performance of string::find

remove pretty-print test that requires debug

machine combiner: fix pretty printer

[test-suite] fix beamformer for CFLAGS="-ffp-contract=on"

[test-suite] fix telecomm-FFT for CFLAGS="-ffp-contract=on"

[SCEVExpand] do not hoist divisions by zero (PR30935)

instr-combiner: sum up all latencies of the transformed instructions

[polybench] remove all __attribute__((optnone))

[polybench] fix some places I have missed in the large patch

[www] add property executable to index.html

[www] add call for speakers for the LLVM Performance Workshop at CGO 2017

add FP_ABSTOLERANCE missing definition to mvt Makefile

[polybench] check against strict fp

remove warnings from google-benchmarks in libcxx

Memory-SSA: strengthen defClobbersUseOrDef interface

commit back "GVN-hoist: fix store past load dependence analysis (PR30216…

Memory-SSA cleanup of clobbers interface, NFC

GVN-hoist: fix store past load dependence analysis (PR30216, PR30499)

[AArch64] Avoid generating indexed vector instructions for Exynos

set the underlying value of “#pragma STDC FP_CONTRACT” on by default

GVN-hoist: do not dereference null pointers

GVN-hoist: fix store past load dependence analysis (PR30216)

GVN-hoist: fix typo

GVN-hoist: move hoist testcase to GVNHoist dir

GVN-hoist: only hoist relevant scalar instructions

GVN-hoist: invalidate MD cache (PR29144)

GVN-hoist: fix hoistingFromAllPaths for loops (PR29034)

Add 'inline' attribute to __init to inline the basic_string's constructor

GVN-hoist: enable by default

GVN-hoist: enable by default

GVN-hoist: fix early exit logic

Make GVN Hoisting obey optnone/bisect.

GVN-hoist: enable by default

GVN-hoist: limit the length of dependent instructions

GVN-hoist: compute DFS numbers once

GVN-hoist: compute MSSA once per function (PR28670)

Pass EphValues by const-ref as it is not modified in the callee