ShuriZma
/
rpcs3
mirror of https://github.com/RPCS3/rpcs3.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

SPU LLVM: Rewrite fma32x4 to match FM and older asmjit stuff

This commit is contained in:
Nekotekina 2020-03-02 23:28:13 +03:00
parent de1774d8f2
commit 01db83bc36
1 changed files with 7 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
rpcs3/Emu/Cell/SPURecompiler.cpp
View File

@ -7381,27 +7381,25 @@ public:
value_t<f32[4]> fma32x4(value_t<f32[4]> a, value_t<f32[4]> b, value_t<f32[4]> c)
{
// Compare absolute values with max positive float in normal range.
const auto aa = bitcast<s32[4]>(fabs(a));
const auto ab = bitcast<s32[4]>(fabs(b));
const auto sc = eval(max(aa, ab) > 0x7f7fffff);
const auto ma = eval(sext<s32[4]>(fcmp_uno(a != fsplat<f32[4]>(0.))));
const auto mb = eval(sext<s32[4]>(fcmp_uno(b != fsplat<f32[4]>(0.))));
const auto ca = eval(bitcast<f32[4]>(bitcast<s32[4]>(a) & mb));
const auto cb = eval(bitcast<f32[4]>(bitcast<s32[4]>(b) & ma));
if (m_use_fma)
{
value_t<f32[4]> r;
r.value = m_ir->CreateCall(get_intrinsic<f32[4]>(llvm::Intrinsic::fma), {a.value, b.value, c.value});
r.value = m_ir->CreateSelect(sc.value, c.value, r.value);
r.value = m_ir->CreateCall(get_intrinsic<f32[4]>(llvm::Intrinsic::fma), {ca.value, cb.value, c.value});
return r;
}
// Convert to doubles
const auto xa = m_ir->CreateFPExt(a.value, get_type<f64[4]>());
const auto xb = m_ir->CreateFPExt(b.value, get_type<f64[4]>());
const auto xa = m_ir->CreateFPExt(ca.value, get_type<f64[4]>());
const auto xb = m_ir->CreateFPExt(cb.value, get_type<f64[4]>());
const auto xc = m_ir->CreateFPExt(c.value, get_type<f64[4]>());
const auto xr = m_ir->CreateCall(get_intrinsic<f64[4]>(llvm::Intrinsic::fmuladd), {xa, xb, xc});
value_t<f32[4]> r;
r.value = m_ir->CreateFPTrunc(xr, get_type<f32[4]>());
r.value = m_ir->CreateSelect(sc.value, c.value, r.value);
return r;
}