The goal of this patch is to maximize CPU utilization on multi-socket or high core count systems, so that parallel computations such as LLD/ThinLTO can use all hardware threads in the system. Before this patch, on Windows, a maximum of 64 hardware threads could be used at most, in some cases dispatched only on one CPU socket.
== Background ==
Windows doesn't have a flat cpu_set_t like Linux. Instead, it projects hardware CPUs (or NUMA nodes) to applications through a concept of "processor groups". A "processor" is the smallest unit of execution on a CPU, that is, an hyper-thread if SMT is active; a core otherwise. There's a limit of 32-bit processors on older 32-bit versions of Windows, which later was raised to 64-processors with 64-bit versions of Windows. This limit comes from the affinity mask, which historically is represented by the sizeof(void*). Consequently, the concept of "processor groups" was introduced for dealing with systems with more than 64 hyper-threads.
By default, the Windows OS assigns only one "processor group" to each starting application, in a round-robin manner. If the application wants to use more processors, it needs to programmatically enable it, by assigning threads to other "processor groups". This also means that affinity cannot cross "processor group" boundaries; one can only specify a "preferred" group on start-up, but the application is free to allocate more groups if it wants to.
This creates a peculiar situation, where newer CPUs like the AMD EPYC 7702P (64-cores, 128-hyperthreads) are projected by the OS as two (2) "processor groups". This means that by default, an application can only use half of the cores. This situation could only get worse in the years to come, as dies with more cores will appear on the market.
== The problem ==
The heavyweight_hardware_concurrency() API was introduced so that only *one hardware thread per core* was used. Once that API returns, that original intention is lost, only the number of threads is retained. Consider a situation, on Windows, where the system has 2 CPU sockets, 18 cores each, each core having 2 hyper-threads, for a total of 72 hyper-threads. Both heavyweight_hardware_concurrency() and hardware_concurrency() currently return 36, because on Windows they are simply wrappers over std:🧵:hardware_concurrency() -- which can only return processors from the current "processor group".
== The changes in this patch ==
To solve this situation, we capture (and retain) the initial intention until the point of usage, through a new ThreadPoolStrategy class. The number of threads to use is deferred as late as possible, until the moment where the std::threads are created (ThreadPool in the case of ThinLTO).
When using hardware_concurrency(), setting ThreadCount to 0 now means to use all the possible hardware CPU (SMT) threads. Providing a ThreadCount above to the maximum number of threads will have no effect, the maximum will be used instead.
The heavyweight_hardware_concurrency() is similar to hardware_concurrency(), except that only one thread per hardware *core* will be used.
When LLVM_ENABLE_THREADS is OFF, the threading APIs will always return 1, to ensure any caller loops will be exercised at least once.
Differential Revision: https://reviews.llvm.org/D71775
This reverts commit 80a34ae311 with fixes.
Previously, since bots turning on EXPENSIVE_CHECKS are essentially turning on
MachineVerifierPass by default on X86 and the fact that
inline-asm-avx-v-constraint-32bit.ll and inline-asm-avx512vl-v-constraint-32bit.ll
are not expected to generate functioning machine code, this would go
down to `report_fatal_error` in MachineVerifierPass. Here passing
`-verify-machineinstrs=0` to make the intent explicit.
This reverts commit 80a34ae311 with fixes.
On bots llvm-clang-x86_64-expensive-checks-ubuntu and
llvm-clang-x86_64-expensive-checks-debian only,
llc returns 0 for these two tests unexpectedly. I tweaked the RUN line a little
bit in the hope that LIT is the culprit since this change is not in the
codepath these tests are testing.
llvm\test\CodeGen\X86\inline-asm-avx-v-constraint-32bit.ll
llvm\test\CodeGen\X86\inline-asm-avx512vl-v-constraint-32bit.ll
Fixes https://bugs.llvm.org//show_bug.cgi?id=44878
When --strip-debug is specified, .debug* are removed from inputSections
while .rel[a].debug* (incorrectly) remain.
LinkerScript::addOrphanSections() requires the output section of a relocated
InputSectionBase to be created first.
.debug* are not in inputSections ->
output sections .debug* are not created ->
getOutputSectionName(.rel[a].debug*) dereferences a null pointer.
Fix the null pointer dereference by deleting .rel[a].debug* from inputSections as well.
Reviewed By: grimar, nickdesaulniers
Differential Revision: https://reviews.llvm.org/D74510
Recommit of 0b4a047bfb
(reverted in c29003813a) to incorporate
subsequent fix and add a warning when LLD's interworking behavior has
changed.
D73474 disabled the generation of interworking thunks for branch
relocations to non STT_FUNC symbols. This patch handles the case of BL and
BLX instructions to non STT_FUNC symbols. LLD would normally look at the
state of the caller and the callee and write a BL if the states are the
same and a BLX if the states are different.
This patch disables BL/BLX substitution when the destination symbol does
not have type STT_FUNC. This brings our behavior in line with GNU ld which
may prevent difficult to diagnose runtime errors when switching to lld.
This change does change how LLD handles interworking of symbols that do not
have type STT_FUNC from previous versions including the 10.0 release. This
brings LLD in line with ld.bfd but there may be programs that have not been
linked with ld.bfd that depend on LLD's previous behavior. We emit a warning
when the behavior changes.
A summary of the difference between 10.0 and 11.0 is that for symbols
that do not have a type of STT_FUNC LLD will not change a BL to a BLX or
vice versa. The table below enumerates the changes
| relocation | STT_FUNC | bit(0) | in | 10.0- out | 11.0+ out |
| R_ARM_CALL | no | 1 | BL | BLX | BL |
| R_ARM_CALL | no | 0 | BLX | BL | BLX |
| R_ARM_THM_CALL | no | 1 | BLX | BL | BLX |
| R_ARM_THM_CALL | no | 0 | BL | BLX | BL |
Differential Revision: https://reviews.llvm.org/D73542
We do not keep the actual value of the CIE ID field, because it is
predefined, and use a constant when dumping a CIE record. The issue
was that the predefined value is different for .debug_frame and
.eh_frame sections, but we always printed the one which corresponds
to .debug_frame. The patch fixes that by choosing an appropriate
constant to print.
See the following for more information about .eh_frame sections:
https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/ehframechpt.html
Differential Revision: https://reviews.llvm.org/D73627
Summary:
Even though this test is a check for failure, lld still attempts
to open the final output file, which fails when the default "a.out"
file is used and the current directory is read-only. Specifying an
output file works around this problem.
Reviewers: espindola
Subscribers: emaste, MaskRay, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74523
D43468+D44380 added INSERT [AFTER|BEFORE] for non-orphan sections. This patch
makes INSERT work for orphan sections as well.
`SECTIONS {...} INSERT [AFTER|BEFORE] .foo` does not set `hasSectionCommands`, so the result
will be similar to a regular link without a linker script. The differences when `hasSectionCommands` is set include:
* image base is different
* -z noseparate-code/-z noseparate-loadable-segments are unavailable
* some special symbols such as `_end _etext _edata` are not defined
The behavior is similar to GNU ld:
INSERT is not considered an external linker script.
This feature makes the section layout more flexible. It can be used to:
* Place .nv_fatbin before other readonly SHT_PROGBITS sections to mitigate relocation overflows.
* Disturb the layout to expose address sensitive application bugs.
Reviewed By: grimar
Differential Revision: https://reviews.llvm.org/D74375
GNU ld has a counterintuitive lang_propagate_lma_regions rule.
```
// .foo's LMA region is propagated to .bar because their VMA region is the same,
// and .bar does not have an explicit output section address (addr_tree).
.foo : { *(.foo) } >RAM AT> FLASH
.bar : { *(.bar) } >RAM
// An explicit output section address disables propagation.
.foo : { *(.foo) } >RAM AT> FLASH
.bar . : { *(.bar) } >RAM
```
In both cases, lld thinks .foo's LMA region is propagated and
places .bar in the same PT_LOAD, so lld diverges from GNU ld w.r.t. the
second case (lma-align.test).
This patch changes Writer<ELFT>::createPhdrs to disable propagation
(start a new PT_LOAD). A user of the first case can make linker scripts
portable by explicitly specifying `AT>`. By contrast, there was no
workaround for the old behavior.
This change uncovers another LMA related bug in assignOffsets() where
`ctx->lmaOffset = 0;` was omitted. It caused a spurious "load address
range overlaps" error for at2.test
The new PT_LOAD rule is complex. For convenience, I listed the origins of some subexpressions:
* rL323449: `sec->memRegion == load->firstSec->memRegion`; linkerscript/at3.test
* D43284: `load->lastSec == Out::programHeaders` (don't start a new PT_LOAD after program headers); linkerscript/at4.test
* D58892: `sec != relroEnd` (start a new PT_LOAD after PT_GNU_RELRO)
Reviewed By: psmith
Differential Revision: https://reviews.llvm.org/D74297
When lmaRegion is non-null, respect `sec->alignment`
This rule is analogous to `switchTo(sec)` which advances sh_addr (VMA).
This fixes the p_paddr misalignment issue as reported by
https://android-review.googlesource.com/c/trusty/external/trusted-firmware-a/+/1230058
Note, `sec->alignment` is the maximum of ALIGN and input section alignments. We may overalign LMA than GNU ld.
linkerscript/align-lma.s has a FIXME that demonstrates another bug:
`.bss ... >RAM` should be placed in a different PT_LOAD (GNU ld
behavior) because its lmaRegion (nullptr) is different from the previous
section's lmaRegion (ROM).
Reviewed By: psmith
Differential Revision: https://reviews.llvm.org/D74286
Summary:
GNU objdump prints the file format in lowercase, e.g. `elf64-x86-64`. llvm-objdump prints `ELF64-x86-64` right now, even though piping that into llvm-objcopy refuses that as a valid arch to use.
As an example of a problem this causes, see: https://github.com/ClangBuiltLinux/linux/issues/779
Reviewers: MaskRay, jhenderson, alexshap
Reviewed By: MaskRay
Subscribers: tpimh, sbc100, grimar, jvesely, nhaehnle, kerbowa, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D74433
Also remove some test duplication and add a test case that shows the
maximum version is rejected (this also shows that the value in the error
message is actually in decimal, and not just missing an 0x prefix).
Reviewed by: dblaikie
Differential Revision: https://reviews.llvm.org/D74403
This reverts commit rGcd5b308b828e, rGcd5b308b828e, rG8cedf0e2994c.
There are issues to be investigated for polly bots and bots turning on
EXPENSIVE_CHECKS.
This can happen if lto::LTO::getRuntimeLibcallSymbols doesn't return
an complete/accurate list of libcalls. In this case new bitcode
object can be linked in after LTO.
For example the WebAssembly backend currently calls:
setLibcallName(RTLIB::FPROUND_F32_F16, "__truncsfhf2");
But `__truncsfhf2` is not part of `getRuntimeLibcallSymbols` so if
this symbol is generated during LTO the link will currently fail.
Without this change the linker crashes because the bitcode symbol
makes it all the way to the output phase.
See: https://bugs.llvm.org/show_bug.cgi?id=44353
Differential Revision: https://reviews.llvm.org/D71632
There are still problems after the fix in
"[ELF][ARM] Fix regression of BL->BLX substitution after D73542"
so let's revert to get trunk back to green while we investigate.
See https://reviews.llvm.org/D73542
This reverts commit 5461fa2b1f.
This reverts commit 0b4a047bfb.
This adds some of LLD specific scopes and picks up optimisation scopes
via LTO/ThinLTO. Makes use of TimeProfiler multi-thread support added in
77e6bb3c.
Differential Revision: https://reviews.llvm.org/D71060
D73542 made a typo (`rel.type == R_PLT_PC`; should be `rel.expr`) and introduced a regression:
BL->BLX substitution was disabled when the target symbol is preemptible
(expr is R_PLT_PC).
The two added bl instructions in arm-thumb-interwork-shared.s check that
we patch BL to BLX.
Fixes https://bugs.chromium.org/p/chromium/issues/detail?id=1047531
For an out-of-range relocation referencing a non-local symbol, report the symbol name and the object file that defines the symbol. As an example:
```
t.o:(function func: .text.func+0x3): relocation R_X86_64_32S out of range: -281474974609120 is not in [-2147483648, 2147483647]
```
=>
```
t.o:(function func: .text.func+0x3): relocation R_X86_64_32S out of range: -281474974609120 is not in [-2147483648, 2147483647]; references func
>>> defined in t1.o
```
Reviewed By: grimar
Differential Revision: https://reviews.llvm.org/D73518
D73474 disabled the generation of interworking thunks for branch
relocations to non STT_FUNC symbols. This patch handles the case of BL and
BLX instructions to non STT_FUNC symbols. LLD would normally look at the
state of the caller and the callee and write a BL if the states are the
same and a BLX if the states are different.
This patch disables BL/BLX substitution when the destination symbol does
not have type STT_FUNC. This brings our behavior in line with GNU ld which
may prevent difficult to diagnose runtime errors when switching to lld.
Differential Revision: https://reviews.llvm.org/D73542
Many of the debug line prologue errors are not inherently fatal. In most
cases, we can make reasonable assumptions and carry on. This patch does
exactly that. In the case of length problems, the approach of "assume
stated length is correct" is taken which means the offset might need
adjusting.
This is a relanding of b94191fe, fixing an LLD test and the LLDB build.
Reviewed by: dblaikie, labath
Differential Revision: https://reviews.llvm.org/D72158
This is how it should've been and brings it more in line with
std::string_view. There should be no functional change here.
This is mostly mechanical from a custom clang-tidy check, with a lot of
manual fixups. It uncovers a lot of minor inefficiencies.
This doesn't actually modify StringRef yet, I'll do that in a follow-up.
Similar to R_MIPS_GPREL16 and R_MIPS_GPREL32 (D45972).
If the addend of an R_PPC_PLTREL24 is >= 0x8000, it indicates that r30
is relative to the input section .got2.
```
addis 30, 30, .got2+0x8000-.L1$pb@ha
addi 30, 30, .got2+0x8000-.L1$pb@l
...
bl foo+0x8000@PLT
```
After linking, the relocation will be relative to the output section .got2.
To compensate for the shift `address(input section .got2) - address(output section .got2) = ppc32Got2OutSecOff`, adjust by `ppc32Got2OutSecOff`:
```
addis 30, 30, .got2+0x8000-.L1+ppc32Got2OutSecOff$pb@ha
addi 30, 30, .got2+0x8000-.L1+ppc32Got2OutSecOff$pb@ha$pb@l
...
bl foo+0x8000+ppc32Got2OutSecOff@PLT
```
This rule applys to a relocatable link or a non-relocatable link with --emit-relocs.
Reviewed By: Bdragon28
Differential Revision: https://reviews.llvm.org/D73532
In ErrorHandler::error(), rearrange code to avoid calling exitLld with
the mutex locked. Acquire mutex lock when flushing the output streams in
exitLld.
Differential Revision: https://reviews.llvm.org/D73281
ELF for the ARM architecture requires linkers to provide
interworking for symbols that are of type STT_FUNC. Interworking for
other symbols must be encoded directly in the object file. LLD was always
providing interworking, regardless of the symbol type, this breaks some
programs that have branches from Thumb state targeting STT_NOTYPE symbols
that have bit 0 clear, but they are in fact internal labels in a Thumb
function. LLD treats these symbols as ARM and inserts a transition to Arm.
This fixes the problem for in range branches, R_ARM_JUMP24,
R_ARM_THM_JUMP24 and R_ARM_THM_JUMP19. This is expected to be the vast
majority of problem cases as branching to an internal label close to the
function.
There is at least one follow up patch required.
- R_ARM_CALL and R_ARM_THM_CALL may do interworking via BL/BLX
substitution.
In theory range-extension thunks can be altered to not change state when
the symbol type is not STT_FUNC. I will need to check with ld.bfd to see if
this is the case in practice.
Fixes (part of) https://github.com/ClangBuiltLinux/linux/issues/773
Differential Revision: https://reviews.llvm.org/D73474
This restores 59733525d3 (D71913), along
with bot fix 19c76989bb.
The bot failure should be fixed by D73418, committed as
af954e441a.
I also added a fix for non-x86 bot failures by requiring x86 in new test
lld/test/ELF/lto/devirt_vcall_vis_public.ll.
* Generalize the code added in D70637 and D70937. We should eventually remove the EM_MIPS special case.
* Handle R_PPC_LOCAL24PC the same way as R_PPC_REL24.
Reviewed By: Bdragon28
Differential Revision: https://reviews.llvm.org/D73424
-fno-pie produces a pair of non-GOT-non-PLT relocations R_PPC_ADDR16_{HA,LO} (R_ABS) referencing external
functions.
```
lis 3, func@ha
la 3, func@l(3)
```
In a -no-pie/-pie link, if func is not defined in the executable, a canonical PLT entry (st_value>0, st_shndx=0) will be needed.
References to func in shared objects will be resolved to this address.
-fno-pie -pie should fail with "can't create dynamic relocation ... against ...", so we just need to think about -no-pie.
On x86, the PLT entry passes the JMP_SLOT offset to the rtld PLT resolver.
On x86-64: the PLT entry passes the JUMP_SLOT index to the rtld PLT resolver.
On ARM/AArch64: the PLT entry passes &.got.plt[n]. The PLT header passes &.got.plt[fixed-index]. The rtld PLT resolver can compute the JUMP_SLOT index from the two addresses.
For these targets, the canonical PLT entry can just reuse the regular PLT entry (in PltSection).
On PPC32: PltSection (.glink) consists of `b PLTresolve` instructions and `PLTresolve`. The rtld PLT resolver depends on r11 having been set up to the .plt (GotPltSection) entry.
On PPC64 ELFv2: PltSection (.glink) consists of `__glink_PLTresolve` and `bl __glink_PLTresolve`. The rtld PLT resolver depends on r12 having been set up to the .plt (GotPltSection) entry.
We cannot reuse a `b PLTresolve`/`bl __glink_PLTresolve` in PltSection as a canonical PLT entry. PPC64 ELFv2 avoids the problem by using TOC for any external reference, even in non-pic code, so the canonical PLT entry scenario should not happen in the first place.
For PPC32, we have to create a PLT call stub as the canonical PLT entry. The code sequence sets up r11.
Reviewed By: Bdragon28
Differential Revision: https://reviews.llvm.org/D73399
Symbol information can be used to improve out-of-range/misalignment diagnostics.
It also helps R_ARM_CALL/R_ARM_THM_CALL which has different behaviors with different symbol types.
There are many (67) relocateOne() call sites used in thunks, {Arm,AArch64}errata, PLT, etc.
Rename them to `relocateNoSym()` to be clearer that there is no symbol information.
Reviewed By: grimar, peter.smith
Differential Revision: https://reviews.llvm.org/D73254
Summary:
Third part in series to support Safe Whole Program Devirtualization
Enablement, see RFC here:
http://lists.llvm.org/pipermail/llvm-dev/2019-December/137543.html
This patch adds type test metadata under -fwhole-program-vtables,
even for classes without hidden visibility. It then changes WPD to skip
devirtualization for a virtual function call when any of the compatible
vtables has public vcall visibility.
Additionally, internal LLVM options as well as lld and gold-plugin
options are added which enable upgrading all public vcall visibility
to linkage unit (hidden) visibility during LTO. This enables the more
aggressive WPD to kick in based on LTO time knowledge of the visibility
guarantees.
Support was added to all flavors of LTO WPD (regular, hybrid and
index-only), and to both the new and old LTO APIs.
Unfortunately it was not simple to split the first and second parts of
this part of the change (the unconditional emission of type tests and
the upgrading of the vcall visiblity) as I needed a way to upgrade the
public visibility on legacy WPD llvm assembly tests that don't include
linkage unit vcall visibility specifiers, to avoid a lot of test churn.
I also added a mechanism to LowerTypeTests that allows dropping type
test assume sequences we now aggressively insert when we invoke
distributed ThinLTO backends with null indexes, which is used in testing
mode, and which doesn't invoke the normal ThinLTO backend pipeline.
Depends on D71907 and D71911.
Reviewers: pcc, evgeny777, steven_wu, espindola
Subscribers: emaste, Prazek, inglorion, arichardson, hiraditya, MaskRay, dexonsmith, dang, davidxl, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D71913
I felt really sad to push this commit for my selfish purpose to make
glibc -static-pie build with lld. Some code constructs in glibc require
R_X86_64_GOTPCREL/R_X86_64_REX_GOTPCRELX referencing undefined weak to
be resolved to a GOT entry not relocated by R_X86_64_GLOB_DAT (GNU ld
behavior), e.g.
csu/libc-start.c
if (__pthread_initialize_minimal != NULL)
__pthread_initialize_minimal ();
elf/dl-object.c
void
_dl_add_to_namespace_list (struct link_map *new, Lmid_t nsid)
{
/* We modify the list of loaded objects. */
__rtld_lock_lock_recursive (GL(dl_load_write_lock));
Emitting a GLOB_DAT will make the address equal &__ehdr_start (true
value) and cause elf/ldconfig to segfault. glibc really should move away
from weak references, which do not have defined semantics.
Temporarily special case --no-dynamic-linker.
These functions call relocateOne(). This patch is a prerequisite for
making relocateOne() aware of `Symbol` (D73254).
Reviewed By: grimar, nickdesaulniers
Differential Revision: https://reviews.llvm.org/D73250
When annotating a symbol with __declspec(selectany), Clang assigns it
comdat 2 while GCC assigns it comdat 3. This patch enables two object
files that contain a __declspec(selectany) symbol, one created by gcc
and the other by clang, to be linked together instead of issuing a
duplicate symbol error.
Differential Revision: https://reviews.llvm.org/D73139
Alexandre Ganea