dahjah/Cemu
1
0
Fork 0
mirror of https://github.com/cemu-project/Cemu.git synced 2026-07-10 01:24:41 -06:00
Code Issues Packages Projects Releases Wiki Activity

Espresso: use std::countr_zero for register-set bit scans

Browse Source 
Replace the byte-then-bit linear scan in IMLPhysRegisterSet::GetFirstAvailableReg / GetNextAvailableReg with std::countr_zero (single tzcnt / rbit+clz). Behaviour-identical (both return the index of the lowest set bit), branchless, and ~16x faster on the isolated scan in a micro-benchmark - it removes the mispredicted inner-loop branches.
This commit is contained in:
1e1 2026-07-02 22:40:59 +02:00
parent 88b07d4798
commit 3f835f563f
1 changed files with 3 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
src/Cafe/HW/Espresso/Recompiler/IML/IMLRegisterAllocator.h
View File

@ -63,19 +63,7 @@ public:
IMLPhysReg GetFirstAvailableReg()
{
cemu_assert_debug(m_regBitmask != 0);
sint32 regIndex = 0;
auto tmp = m_regBitmask;
while ((tmp & 0xFF) == 0)
{
regIndex += 8;
tmp >>= 8;
}
while ((tmp & 0x1) == 0)
{
regIndex++;
tmp >>= 1;
}
return regIndex;
return (IMLPhysReg)std::countr_zero(m_regBitmask);
}
// returns index of next available register (search includes any register index >= startIndex)
@ -84,22 +72,10 @@ public:
{
if (startIndex >= 64)
return -1;
uint32 regIndex = startIndex;
auto tmp = m_regBitmask;
tmp >>= regIndex;
uint64 tmp = m_regBitmask >> startIndex;
if (!tmp)
return -1;
while ((tmp & 0xFF) == 0)
{
regIndex += 8;
tmp >>= 8;
}
while ((tmp & 0x1) == 0)
{
regIndex++;
tmp >>= 1;
}
return regIndex;
return startIndex + (IMLPhysReg)std::countr_zero(tmp);
}
sint32 CountAvailableRegs() const