llvm-project/llvm/test/CodeGen
Eli Friedman cbed30c501 [ARM] Split 128-bit vectors in BUILD_VECTOR lowering
Given that INSERT_VECTOR_ELT operates on D registers anyway, combining
64-bit vectors into a 128-bit vector is basically free. Therefore, try
to split BUILD_VECTOR nodes before giving up and lowering them to a series
of INSERT_VECTOR_ELT instructions. Sometimes this allows dramatically
better lowerings; see testcases for examples. Inspired by similar code
in the x86 backend for AVX.

Differential Revision: https://reviews.llvm.org/D27624

llvm-svn: 289706
2016-12-14 20:44:38 +00:00
..
AArch64 Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
AMDGPU AMDGPU: Emit runtime metadata version 2 as YAML 2016-12-14 17:16:52 +00:00
ARM [ARM] Split 128-bit vectors in BUILD_VECTOR lowering 2016-12-14 20:44:38 +00:00
AVR [AVR] Add a function instrumentation pass 2016-12-14 10:15:00 +00:00
BPF Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
Generic Add -O0 support for @llvm.invariant.group.barrier by discarding it if it gets to ISel. 2016-11-07 16:47:20 +00:00
Hexagon Move .mir tests to appropriate directories 2016-12-09 19:08:15 +00:00
Inputs
Lanai Move .mir tests to appropriate directories 2016-12-09 19:08:15 +00:00
MIR AMDGPU: Fix handling of 16-bit immediates 2016-12-10 00:39:12 +00:00
MSP430 Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
Mips Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
NVPTX [NVPTX] Remove NVPTXFavorNonGenericAddrSpaces pass. 2016-10-31 21:51:42 +00:00
PowerPC Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
SPARC ScheduleDAGInstrs: Add condjump deps to addSchedBarrierDeps() 2016-11-11 01:34:21 +00:00
SystemZ Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
Thumb Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
Thumb2 Revert "[Thumb] Teach ISel how to lower compares of AND bitmasks efficiently" 2016-11-03 14:08:01 +00:00
WebAssembly [WebAssembly] Emit .import_global assembler directives 2016-12-01 00:11:15 +00:00
WinEH Avoid infinite loops in branch folding 2016-12-12 23:05:38 +00:00
X86 Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00
XCore Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." 2016-12-14 16:43:44 +00:00