llvm-svn: 127155

llvm-svn: 127154

llvm-svn: 127153

Patch by Jack Howarth.

llvm-svn: 127152

right thing, instead of merely being unable to analyze and the transform
doesn't occur.

llvm-svn: 127149

then the result could go either way. If it's provably positive then so is the
srem. Fixes PR9343 #7!

llvm-svn: 127146

llvm-svn: 127145

llvm-svn: 127140

thing for variables.

llvm-svn: 127138

llvm-svn: 127135

llvm-svn: 127131

the value splatted into every element. Extend this to getTrue and getFalse which
by providing new overloads that take Types that are either i1 or <N x i1>. Use
it in InstCombine to add vector support to some code, fixing PR8469!

llvm-svn: 127116

llvm-svn: 127114

Disable a couple of experimental heuristics to get the best results from the current implementation of -pre-RA-sched=list-ilp.

llvm-svn: 127113

llvm-svn: 127110

Revert r127073: "Introduce $(ECHOPATH) to print DOSish path string on MSYS bash for alternative of $(ECHO)."

It broke the llvm-gcc-native-mingw32 buildbot, and we need all of them to be green for the 2.9 branch.

Takumi, please reapply after we branch, preferably with a fix ;-)

llvm-svn: 127107

ARM assembler stuff is crazy: for .setfp positive values of offset corresponds to "add" instruction, not to "sub" as in .pad case

llvm-svn: 127106

In Thumb1 mode the constant might be materialized via the load from constpool. Emit unwinding information in case when this load from constpool is used to change the stack pointer in the prologue.

llvm-svn: 127105

Implement frame unwinding information emission for Thumb1. Not finished yet because there is no way given the constpool index to examine the actual entry: the reason is clones inserted by constant island pass, which are not tracked at all! The only connection is done during asmprinting time via magic label names which is really gross and needs to be eventually fixed.

llvm-svn: 127104

llvm-svn: 127103

llvm-svn: 127102

This is just very first approximation how the stuff should be done
(e.g. ARM-only for now). More to follow.

llvm-svn: 127101

llvm-svn: 127100

llvm-svn: 127099

llvm-svn: 127098

llvm-svn: 127097

The coalescer can in very rare cases leave too large live intervals around after
rematerializing cheap-as-a-move instructions.

Linear scan doesn't really care, but live range splitting gets very confused
when a live range is killed by a ghost instruction.

I will fix this properly in the coalescer after 2.9 branches.

llvm-svn: 127096

llvm-svn: 127090

InstCombine: We know the number of items initially added to the worklist map, reserve space early to avoid rehashing.

llvm-svn: 127089

This makes lookup slightly more expensive but it's worth it, unused
DenseMaps are common in LLVM code apparently.

1% speedup on clang -O3 bzip2.c
4% speedup on clang -O3 oggenc.c (Release build of clang on i386/linux)

llvm-svn: 127088

llvm-svn: 127084

llvm-svn: 127075

utils/lit/lit/TestRunner.py: bash is available with MSYS on Python/W32. Then we can execute "bash tests".

llvm-svn: 127074

On mingw and python/w32, lit would not be expected to understand MSYS-style path.

llvm-svn: 127073

On Windows hosts, Python scripts in test/Scripts did not accept binary files from stdin. The environment variable "PYTHONUNBUFFERED" makes stdin as binary. Thanks to Danil Malyshev!

llvm-svn: 127072

llvm-svn: 127071

llvm-svn: 127070

the percentage of time spent in CodeGenPrepare when llcing 403.gcc from 12.6% to
1.8% of total llc time.

llvm-svn: 127069

llvm-svn: 127068

regs. This is the only change in this checkin that may affects the
default scheduler. With better register tracking and heuristics, it
doesn't make sense to artificially lower the register limit so much.

Added -sched-high-latency-cycles and X86InstrInfo::isHighLatencyDef to
give the scheduler a way to account for div and sqrt on targets that
don't have an itinerary. It is currently defaults to 10 (the actual
number doesn't matter much), but only takes effect on non-default
schedulers: list-hybrid and list-ilp.

Added several heuristics that can be individually disabled for the
non-default sched=list-ilp mode. This helps us determine how much
better we can do on a given benchmark than the default
scheduler. Certain compute intensive loops run much faster in this
mode with the right set of heuristics, and it doesn't seem to have
much negative impact elsewhere. Not all of the heuristics are needed,
but we still need to experiment to decide which should be disabled by
default for sched=list-ilp.

llvm-svn: 127067