gamzinav/cling
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Use TLS for pointer check cache:

Browse Source 
With 16 threads, the cache is simply too small.
These are thread specific values anyway.
This commit is contained in:
Axel Naumann 2018-09-24 17:59:31 +02:00 committed by sftnight
parent 4ed58df12d
commit 2847412de4
1 changed files with 6 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
lib/Utils/PlatformPosix.cpp
View File

@ -41,8 +41,8 @@ namespace {
private:
// A simple round-robin cache: what enters first, leaves first.
// MRU cache wasn't worth the extra CPU cycles.
std::array<const void*, 8> lines = {};
std::atomic<unsigned> mostRecent = {0};
static thread_local std::array<const void*, 8> lines;
static thread_local unsigned mostRecent;
int FD;
// Concurrent writes to the same cache element can result in invalid cache
@ -51,12 +51,8 @@ namespace {
// slow-down, the cost for keeping the cache thread-local or atomic is
// much higher (yes, this was measured).
void push(const void* P) {
unsigned acquiredVal = mostRecent;
while(!mostRecent.compare_exchange_weak(acquiredVal,
(acquiredVal+1)%lines.size())) {
acquiredVal = mostRecent;
}
lines[acquiredVal] = P;
mostRecent = (mostRecent + 1) % lines.size();
lines[mostRecent] = P;
}
public:
@ -85,6 +81,8 @@ namespace {
return true;
}
};
thread_local std::array<const void*, 8> PointerCheck::lines = {};
thread_local unsigned PointerCheck::mostRecent = 0;
}
bool IsMemoryValid(const void *P) {