extended vector types. Remove the special SDNode opcodes used for pre-legalize
vector operations, and the special MVT::Vector type used with them. Adjust
lowering and legalize to work with the normal SDNode kinds instead, and to
use the normal MVT functions to work with vector types instead of using the
two special operands that the pre-legalize nodes held.
This allows pre-legalize and post-legalize DAGs, and the code that operates
on them, to be more consistent. Pre-legalize vector operators can be handled
more consistently with scalar operators. And, -view-dag-combine1-dags and
-view-legalize-dags now look prettier for vector code.
llvm-svn: 37719
TargetLowering to SelectionDAG so that they have more convenient
access to the current DAG, in preparation for the ValueType routines
being changed from standalone functions to members of SelectionDAG for
the pre-legalize vector type changes.
llvm-svn: 37704
with a general target hook to identify rematerializable instructions. Some
instructions are only rematerializable with specific operands, such as loads
from constant pools, while others are always rematerializable. This hook
allows both to be identified as being rematerializable with the same
mechanism.
llvm-svn: 37644
1. Consider all possible ifcvt cases at once. No longer restricted to bottom
up iterative approach.
2. Sort all possible cases based on a cost function. Perform the most profitable
ones first invalidate others that target the same blocks.
3. Fixed a number of bugs related to block duplication.
llvm-svn: 37613
VCONCAT_VECTORS. Use these for CopyToReg and CopyFromReg legalizing in
the case that the full register is to be split into subvectors instead
of scalars. This replaces uses of VBIT_CONVERT to present values as
vector-of-vector types in order to make whole subvectors accessible via
BUILD_VECTOR and EXTRACT_VECTOR_ELT.
This is in preparation for adding extended ValueType values, where
having vector-of-vector types is undesirable.
llvm-svn: 37569
crashing but breaks exception handling. The problem
described in PR1224 is that invoke is a terminator that
can produce a value. The value may be needed in other
blocks. The code that writes to registers values needed
in other blocks runs before terminators are lowered (in
this case invoke) so asserted because the value was not
yet available. The fix that was applied was to do invoke
lowering earlier, before writing values to registers.
The problem this causes is that the code to copy values
to registers can be output after the invoke call. If
an exception is raised and control is passed to the
landing pad then this copy-code will never execute. If
the value is needed in some code path reached via the
landing pad then that code will get something bogus.
So revert the original fix and simply skip invoke values
in the general copying to registers code. Instead copy
the invoke value to a register in the invoke lowering code.
llvm-svn: 37567
simultaneously. Move that pass to SimpleRegisterCoalescing.
This makes it easier to implement alternative register allocation and
coalescing strategies while maintaining reuse of the existing live
interval analysis.
llvm-svn: 37520
that the CSE map always contains explicit alignment information. This allows
more loads to be CSE'd when there is a mix of explicit-alignment loads and
implicit-alignment loads.
Also, in SelectionDAG::FindModifiedNodeSlot, add the operands to the
FoldingSetNodeID before the load/store information instead of after, so
that it matches what is done elsewhere.
llvm-svn: 37411
(landing pad) when an exception unwinds through the call. This doesn't
quite match the way the dwarf unwinder works: by default it only jumps to
the landing pad if the catch or filter specification matches, and otherwise
it keeps on unwinding. There are two ways of specifying to the unwinder
that it should "always" (more on why there are quotes here later) jump to
the landing pad: follow the specification by a 0 typeid, or follow it by
the typeid for the NULL typeinfo. GCC does the first, and this patch makes
LLVM do the same as gcc. However there is a problem: the unwinder performs
optimizations based on C++ semantics (it only expects destructors to be
run if the 0 typeid fires - known as "cleanups"), meaning it assumes that no
exceptions will be raised and that the raised exception will be reraised
at the end of the cleanup code. So if someone writes their own LLVM code
using the exception intrinsics they will get a nasty surprise if they don't
follow these rules. The other possibility of using the typeid corresponding
to NULL (catch-all) causes the unwinder to make no assumptions, so this is
probably what we should use in the long-run. However since we are still
having trouble getting exception handling working properly, for the moment
it seems best to closely imitate GCC.
llvm-svn: 37399
simplifies the code in DwarfWriter, allows for multiple filters and
makes it trivial to specify filters accompanied by cleanups or catch-all
specifications (see next patch). What a deal! Patch blessed by Anton.
llvm-svn: 37398
See test/CodeGen/X86/test-pic-jtbl.ll for a case where it works well;
shaves another 10K off our favorite benchmark. I was hesitant about
this because of compile speed, but seems to do OK on a bootstrap.
llvm-svn: 37392
smaller than the preferred alignment, but so that the target can actually
specify a minimum alignment if needed. This fixes some objc protocol
failures Devang tracked down.
llvm-svn: 37373
Duraid Madina