llvm-6502/test
Evan Cheng 48575f6ea7 Making use of VFP / NEON floating point multiply-accumulate / subtraction is
difficult on current ARM implementations for a few reasons.
1. Even though a single vmla has latency that is one cycle shorter than a pair
   of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause
   additional pipeline stall. So it's frequently better to single codegen
   vmul + vadd.
2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to
   stall for 4 cycles. We need to schedule them apart.
3. A vmla followed vmla is a special case. Obvious issuing back to back RAW
   vmla + vmla is very bad. But this isn't ideal either:
     vmul
     vadd
     vmla
   Instead, we want to expand the second vmla:
     vmla
     vmul
     vadd
   Even with the 4 cycle vmul stall, the second sequence is still 2 cycles
   faster.

Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough
but it isn't the optimial solution. This patch attempts to make it possible to
use vmla / vmls in cases where it is profitable.

A. Add missing isel predicates which cause vmla to be codegen'ed.
B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to
   compute a fmul and a fmla.
C. Add additional isel checks for vmla, avoid cases where vmla is feeding into
   fp instructions (except for the #3 exceptional case).
D. Add ARM hazard recognizer to model the vmla / vmls hazards.
E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the
   vmla / vmls will trigger one of the special hazards.

Work in progress, only A+B are enabled.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
..
Analysis remove fixme comment too. 2010-11-30 23:25:01 +00:00
Archive test/Archive/extract.ll: Use cmp instead of diff. Thanks to Danil Malyshev! 2010-12-02 09:16:14 +00:00
Assembler Fix PR 4170 by having ExtractValueInst::getIndexedType() reject out-of-bounds indexing. 2010-12-05 20:50:26 +00:00
Bindings/Ocaml
Bitcode
BugPoint
CodeGen Making use of VFP / NEON floating point multiply-accumulate / subtraction is 2010-12-05 22:04:16 +00:00
DebugInfo Hide tests, that check .loc, .file in output assembly, from darwin9 buildbot. 2010-12-02 23:29:58 +00:00
ExecutionEngine XFAIL for now. If someone with access to an ARM/Linux host wants to have a look 2010-12-02 20:20:32 +00:00
Feature
FrontendAda Adjust this test for the fact that the stores are no longer 2010-12-02 20:56:51 +00:00
FrontendC Test case for r120740. Radar 8712503. 2010-12-02 21:25:55 +00:00
FrontendC++ This test dates from the time when llvm-gcc had problems if two types were 2010-12-02 18:19:23 +00:00
FrontendFortran
FrontendObjC
FrontendObjC++
Integer
lib
Linker
LLVMC
MC There are two reasons why we might want to use 2010-12-04 03:21:47 +00:00
Other
Scripts
TableGen
Transforms Teach SimplifyCFG to turn 2010-12-05 18:29:03 +00:00
Unit
Verifier
CMakeLists.txt
lit.cfg
lit.site.cfg.in
Makefile
Makefile.tests
site.exp.in
TestRunner.sh