Summary:
Truncating the result of a merge means that most likely we could have done without merge in the first place and just used the input merge inputs directly. This can be done in three cases:
1. If the truncation result is smaller than the merge source, we can use the source in the trunc directly
2. If the sizes are the same, we can replace the register or use a copy
3. If the truncation size is a multiple of the merge source size, we can build a smaller merge
This gets rid of most of the larger, hard-to-legalize merges.
Reviewers: qcolombet, aditya_nandakumar, aemerson, paquette, arsenm, Petar.Avramovic
Reviewed By: arsenm
Subscribers: sdardis, jvesely, wdng, nhaehnle, rovka, jrtc27, atanasyan, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D75915
Summary:
This is a part of the series of efforts for correcting alignment of memory operations.
(Another related bugs: https://bugs.llvm.org/show_bug.cgi?id=44388 , https://bugs.llvm.org/show_bug.cgi?id=44543 )
This fixes https://bugs.llvm.org/show_bug.cgi?id=43880 by giving default alignment of loads to 1.
The test CodeGen/AArch64/bcmp-inline-small.ll should have been changed; it was introduced by https://reviews.llvm.org/D64805 . I talked with @evandro, and confirmed that the test is okay to be changed.
Other two tests from PowerPC needed changes as well, but fixes were straightforward.
Reviewers: courbet
Reviewed By: courbet
Subscribers: nlopes, gchatelet, wuzish, nemanjai, kristof.beyls, hiraditya, steven.zhang, danielkiss, llvm-commits, evandro
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D76113
Commit [1] added a declaration of function-member
StackFrame::BehavesLikeZerothFrame but hasn't added an implementation
for the function. This commit removes this declation, because the
function is not used anywhere.
[1] 31e6dbe1c6 Fix PC adjustment in StackFrame::GetSymbolContext
Differential Revision: https://reviews.llvm.org/D75979
Patch by Anton Kolesov <Anton.Kolesov@synopsys.com>
Fix to code from https://reviews.llvm.org/D64993.
Field StackFrame::m_behaves_like_zeroth_frame was introduced in commit
[1], however that commit hasn't added a copying of the field to
UpdatePreviousFrameFromCurrentFrame, therefore the value wouldn't change
when updating frames to reflect the current situation.
The particular scenario, where this matters is following. Assume we have
function main that invokes function func1. We set breakpoint at
func1 entry and in main after the func1 call, and do not stop at
the main entry. Therefore, when debugger stops for the first time,
func1 is frame#0, while main is frame#1, thus
m_behaves_like_zeroth_frame is set to 0 for main frame. Execution is
resumed, and stops now in main, where it is now frame#0. However while
updating the frame object, m_behaves_like_zeroth_frame remains false.
This field plays an important role when calculating line information for
backtrace: for frame#0, PC is the current line, therefore line
information is retrieved for PC, however for all other frames this is
not the case - calculated PC is a return-PC, i.e. instruction after the
function call line, therefore for those frames LLDB needs to step back
by one instruction. Initial implementation did this strictly for frames
that have index != 0 (and index is updated properly in
UpdatePreviousFrameFromCurrentFrame), but m_behaves_like_zeroth_frame
added a capability for middle-of-stack frames to behave in a similar
manner. But because current code now doesn't check frame idx,
m_behaves_like_zeroth_frame must be set to true for frames with 0 index,
not only for frame that behave like one. In the described test case,
after stopping in main, LLDB would still consider frame#0 as
non-zeroth, and would subtract instruction from the PC, and would report
previous like as current line.
The error doesn't manifest itself in LLDB interpreter though - it can be
reproduced through LLDB-MI and when using SB API, but not when we
interpreter command "continue" is executed. Honestly, I didn't fully
understand why it works in interpreter, I did found that bug "fixes"
itself if I enable DEBUG_STACK_FRAMES in StackFrameList.cpp, because
that calls StackFrame::Dump and that calls
GetSymbolContext(eSymbolContextEverything), which fills the context of
frame on the first breakpoint, therefore it doesn't have to be
recalculated (improperly) on a second frame. However, on first
breakpoint symbol context is calculated for the "call" line, not the
next one, therefore it should be recalculated anyway on a second
breakpoint, and it is done correctly, even though
m_behaves_like_zeroth_frame is still incorrect, as long as
GetSymbolContext(eSymbolContextEverything) has been called.
[1] 31e6dbe1c6 Fix PC adjustment in StackFrame::GetSymbolContext
Differential Revision: https://reviews.llvm.org/D75975
Patch by Anton Kolesov <Anton.Kolesov@synopsys.com>
We currently only ever use this for lowering constant uniform values (shift/rotate by immediate) so we can safely enable it by default (it treats the undef bits as zero when extracting constants).
This is necessary for an upcoming patch that will use SimplifyDemandedBits more aggressively on funnel shift amounts and causes regressions in vXi64 constant without it.
These tests passed in my Windows 10 VM, but are failing on Windows bots
with errors which look related to unicode encodings. Disable the tests
on Windows for now.
The type legalizer will scalarize vector conversions from integer to floating
point if the source element size is less than that of the result.
This is avoided now by inserting a zero/sign-extension of the source vector
before type legalization.
Review: Ulrich Weigand
Differential revision: https://reviews.llvm.org/D75978
Summary:
This is a preparatory change for allowing LLVM to emit DW_OP_convert
operations converting to the generic type.
If DW_OP_convert's operand is 0, it converts the top of the stack to the
generic type, as specified by DWARFv5 section 2.5.1.6:
"[...] takes one operand, which is an unsigned LEB128 integer that
represents the offset of a debugging information entry in the current
compilation unit, or value 0 which represents the generic type."
This adds support for such operations to dsymutil.
Reviewers: aprantl, markus, friss, JDevlieghere
Reviewed By: aprantl, JDevlieghere
Subscribers: hiraditya, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D76142
Add a unit test for the compact DWARF expression printer which will be
used by the llvm-objdump --debug-vars option.
Differential revision: https://reviews.llvm.org/D75250
This adds the --debug-vars option to llvm-objdump, which prints
locations (registers/memory) of source-level variables alongside the
disassembly based on DWARF info. A vertical line is printed for each
live-range, with a label at the top giving the variable name and
location, and the position and length of the line indicating the program
counter range in which it is valid.
Currently, this only works for object files, not executables or shared
libraries.
Differential revision: https://reviews.llvm.org/D70720
This is a first patch in a series for the SveEmitter to generate the arm_sve.h
header file and builtins.
I've tried my best to strip down this patch as best as I could, but there
are still a few changes that are not necessarily exercised by the load intrinsics
in this patch, mostly around the SVEType class which has some common logic to
represent types from a type and prototype string. I thought it didn't make
much sense to remove that from this patch and split it up.
Reviewers: efriedma, rovka, SjoerdMeijer, rsandifo-arm, rengolin
Reviewed By: SjoerdMeijer
Tags: #clang
Differential Revision: https://reviews.llvm.org/D75470
Summary:
This is a preparatory change for allowing LLVM to emit DW_OP_convert
operations converting to the generic type.
If DW_OP_convert's operand is 0, it converts the top of stack to the
generic type, as specified by DWARFv5 section 2.5.1.6:
"[...] takes one operand, which is an unsigned LEB128 integer that
represents the offset of a debugging information entry in the current
compilation unit, or value 0 which represents the generic type."
This adds support for such operations to llvm-dwarfdump.
Reviewers: aprantl, markus, jdoerfert, jhenderson
Reviewed By: aprantl
Subscribers: hiraditya, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D76141
alignBranches is X86 specific, change the name in a
more general one since other target can do some state
chang before and after emitting the instruction.
According to LangRef for unordered atomic memory transfer intrinsics
"The first three arguments are the same as they are in the @llvm.memcpy intrinsic, with the added constraint that
len is required to be a positive integer multiple of the element_size. If len is not a positive integer multiple
of element_size, then the behaviour of the intrinsic is undefined."
So the len is not multiple of element size is just an undefined behavior and verifier should not complain about that
as undefined behavior is allowed in LLVM IR.
This change removes the verifier check for this condition
Reviewers: reames
Reviewed By: reames
Subscribers: dantrushin, hiraditya, jfb, llvm-commits
Differential Revision: https://reviews.llvm.org/D76116
This makes life easier for downstream users who maintain exotic
target platforms.
Patch by Vince Bridgers!
Differential Revision: https://reviews.llvm.org/D75529
Summary:
This is a simple fix for CodeGenPrepare that freezes branch condition when transforming select to branch.
If it is not frozen, instsimplify or the later pipeline can potentially exploit undefined behavior.
The diff shows optimized form becase D75859 and D76048 already made a few changes to CodeGenPrepare for optimizing freeze(cmp).
Reviewers: jdoerfert, spatel, lebedev.ri, efriedma
Reviewed By: lebedev.ri
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D76179
Ideally we'd want all shebangs to be configurable, but that's not a
viable solution. Given that lldb-dotest is already configured, we might
as well make sure it uses the correct interpreter.
Differential revision: https://reviews.llvm.org/D76167
Now that D75203 has landed and baked for a few days, extend the basic approach to prefix padding as well. The patch itself is fairly straight forward.
For the moment, this patch adds the functional support and some basic testing there of, but defaults to not enabling prefix padding. I want to be able to phrase a separate patch which adds the target specific reasoning and test it cleanly. I haven't decided whether I want to common it with the nop logic or not.
Differential Revision: https://reviews.llvm.org/D75300
Nico Weber