Recommit "Use ValueOffsetPair to enhance value reuse during SCEV
expansion".
The fix for PR28705 will be committed consecutively.
In D12090, the ExprValueMap was added to reuse existing value during
SCEV expansion. However, const folding and sext/zext distribution can
make the reuse still difficult.
A simplified case is: suppose we know S1 expands to V1 in
ExprValueMap, and
S1 = S2 + C_a
S3 = S2 + C_b
where C_a and C_b are different SCEVConstants. Then we'd like to
expand S3 as V1 - C_a + C_b instead of expanding S2 literally. It is
helpful when S2 is a complex SCEV expr and S2 has no entry in
ExprValueMap, which is usually caused by the fact that S3 is
generated from S1 after const folding.
In order to do that, we represent ExprValueMap as a mapping from SCEV
to ValueOffsetPair. We will save both S1->{V1, 0} and S2->{V1, C_a}
into the ExprValueMap when we create SCEV for V1. When S3 is
expanded, it will first expand S2 to V1 - C_a because of S2->{V1,
C_a} in the map, then expand S3 to V1 - C_a + C_b.
Differential Revision: https://reviews.llvm.org/D21313
Pull in r278161 from upstream llvm trunk (by Wei Mi):
Fix the runtime error caused by "Use ValueOffsetPair to enhance value
reuse during SCEV expansion".
The patch is to fix the bug in PR28705. It was caused by setting
wrong return value for SCEVExpander::findExistingExpansion. The
return values of findExistingExpansion have different meanings when
the function is used in different ways so it is easy to make mistake.
The fix creates two new interfaces to replace
SCEVExpander::findExistingExpansion, and specifies where each
interface is expected to be used.
Differential Revision: https://reviews.llvm.org/D22942
Pull in r281439 from upstream llvm trunk (by Wei Mi):
Create a getelementptr instead of sub expr for ValueOffsetPair if the
value is a pointer.
This patch is to fix PR30213. When expanding an expr based on
ValueOffsetPair, if the value is of pointer type, we can only create
a getelementptr instead of sub expr.
Differential Revision: https://reviews.llvm.org/D24088
This should fix assertion failures when building OpenCV >= 3.1, and also
allow building lang/spidermonkey24 without any further assertions.
PR: 215649
MFC after: 1 week
under a shared read lock. This patch attempts to upgrade the lock to
an exclusive write lock. If the exclusive write lock fails to be
obtained, the current fragment is not placed at the head of the list.
This portion of the patch was inspired by NetBSD ip_frag.c r1.4 (which
effectively removed the section of code that performed the reordering).
The patch to sys/contrib/ipfilter/netinet/ip_compat.h adds the
MUTEX_TRY_UPGRADE macro to support the patch to ip_frag.c.
The patch to contrib/ipfilter/lib/rwlock_emul.c supports this patch
by emulating the mutex in userspace when exercised by ipftest(1).
Inspired by: NetBSD ip_frag.c r1.4
MFC after: 1 month
Use ValueOffsetPair to enhance value reuse during SCEV expansion.
In D12090, the ExprValueMap was added to reuse existing value during
SCEV expansion. However, const folding and sext/zext distribution can
make the reuse still difficult.
A simplified case is: suppose we know S1 expands to V1 in
ExprValueMap, and
S1 = S2 + C_a
S3 = S2 + C_b
where C_a and C_b are different SCEVConstants. Then we'd like to
expand S3 as V1 - C_a + C_b instead of expanding S2 literally. It is
helpful when S2 is a complex SCEV expr and S2 has no entry in
ExprValueMap, which is usually caused by the fact that S3 is
generated from S1 after const folding.
In order to do that, we represent ExprValueMap as a mapping from SCEV
to ValueOffsetPair. We will save both S1->{V1, 0} and S2->{V1, C_a}
into the ExprValueMap when we create SCEV for V1. When S3 is
expanded, it will first expand S2 to V1 - C_a because of S2->{V1,
C_a} in the map, then expand S3 to V1 - C_a + C_b.
Differential Revision: https://reviews.llvm.org/D21313
This should fix assertion failures when building OpenCV >= 3.1.
PR: 215649
MFC after: 3 days
Manually force the use of __decltype in C++03 with Clang 3.4.
<string> uses `decltype` in a way incompatible with `__typeof__`.
This is problematic when compiling <string> with Clang 3.4 because
even though it provides `__decltype` libc++ still used `__typeof__`
because clang 3.4 doesn't provide __is_identifier which libc++
uses to detect __decltype.
This patch manually detects Clang 3.4 and properly configures
for it.
This allows the graphics/openshadinglanguage port to build with
lang/clang34.
PR: 216054
Note that mandoc does not use anymore sqlite3 but a home made database format
An important improvement has been made as well in makewhatis performance:
Tests on my laptop shows makewhatis on the entire system goes from 26s to 12s
to a non-wildcard address. As documented in ip(4), doing sendmsg(2) with
IP_SENDSRCADDR on a socket that is bound to non-wildcard address is
completely different to using this control message on a wildcard one.
A fix is to add a bool to mark whether we did setsockopt(IP_RECVDSTADDR)
on the socket, and use IP_SENDSRCADDR control message only if we did.
While here, garbage collect absolutely useless udp_recv() function that
establishes some structures on stack to never use them later.
Fix use-after-free bug in AffectedValueCallbackVH::allUsesReplacedWith
When transferring affected values in the cache from an old value,
identified by the value of the current callback, to the specified new
value we might need to insert a new entry into the DenseMap which
constitutes the cache. Doing so might delete the current callback
object. Move the copying logic into a new function, a member of the
assumption cache itself, so that we don't run into UB should the
callback handle itself be removed mid-copy.
Differential Revision: https://reviews.llvm.org/D28749
This should fix crashes when building lld (as part of the llvmXY ports).
Reported by: jbeich
PR: 216117
It was used for nitems, but I converted it to __arraycount later, and
is already pulled in by header pollution on FreeBSD as well as NetBSD
ala sys/cdefs.h
Fix PR31644 introduced by r287138 and add a regression test.
Thanks Dimitry Andric for the report and fix!
This should restore -MP output to what it was before.
Reported by: jbeich
PR: 216043
unlinking them at the beginning of the testcase
This is a lot more intelligent in cleaning up the semaphores if the testcase
fails before sem_unlink is called.
Contributed back as bin/51872 upstream.
