mirror/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-01-17 01:15:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

SPU LLVM: Optimize branching

Browse Source 
- Detect a pattern where vpmovmskb and a check against the sign bit can be used instead of checking against zero
This commit is contained in:
Malcolm Jestadt 2021-05-15 06:30:38 -04:00 committed by Ivan
parent 57cd4ff3c6
commit 52780e65e7
1 changed files with 46 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
rpcs3/Emu/Cell/SPURecompiler.cpp
View File

@ -8742,6 +8742,29 @@ public:
const u32 target = spu_branch_target(m_pos, op.i16); const u32 target = spu_branch_target(m_pos, op.i16);
// Check sign bit instead (optimization)
if (match_vr<s32[4], s64[2]>(op.rt, [&](auto c, auto MP)
{
using VT = typename decltype(MP)::type;
if (auto [ok, x] = match_expr(c, sext<VT>(match<bool[std::extent_v<VT>]>())); ok)
{
if (target != m_pos + 4)
{
m_block->block_end = m_ir->GetInsertBlock();
const auto a = get_vr<s8[16]>(op.rt);
const auto cond = eval(bitcast<s16>(trunc<bool[16]>(a)) >= 0);
m_ir->CreateCondBr(cond.value, add_block(target), add_block(m_pos + 4));
return true;
}
}
return false;
}))
{
return;
}
if (target != m_pos + 4) if (target != m_pos + 4)
{ {
m_block->block_end = m_ir->GetInsertBlock(); m_block->block_end = m_ir->GetInsertBlock();
@ -8763,6 +8786,29 @@ public:
const u32 target = spu_branch_target(m_pos, op.i16); const u32 target = spu_branch_target(m_pos, op.i16);
// Check sign bit instead (optimization)
if (match_vr<s32[4], s64[2]>(op.rt, [&](auto c, auto MP)
{
using VT = typename decltype(MP)::type;
if (auto [ok, x] = match_expr(c, sext<VT>(match<bool[std::extent_v<VT>]>())); ok)
{
if (target != m_pos + 4)
{
m_block->block_end = m_ir->GetInsertBlock();
const auto a = get_vr<s8[16]>(op.rt);
const auto cond = eval(bitcast<s16>(trunc<bool[16]>(a)) < 0);
m_ir->CreateCondBr(cond.value, add_block(target), add_block(m_pos + 4));
return true;
}
}
return false;
}))
{
return;
}
if (target != m_pos + 4) if (target != m_pos + 4)
{ {
m_block->block_end = m_ir->GetInsertBlock(); m_block->block_end = m_ir->GetInsertBlock();