mirrors/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-04-21 12:05:23 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

ppu Fast/Precise: Fix SIMD instructions VSUM2SWS, VPKSWSS, VPKSHUS, VPKSHSS

Browse source 
Also rewrite VPKSHUS for speed.
This commit is contained in:
eladash 2019-03-19 22:05:10 +02:00 • committed by Ivan
parent e21504d52d
commit a9014a8cac
1 changed files with 16 additions and 38 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

54
rpcs3/Emu/Cell/PPUInterpreter.cpp
View file

@ -1426,7 +1426,7 @@ bool ppu_interpreter_precise::VPKSHSS(ppu_thread& ppu, ppu_opcode_t op)
{
const auto& a = ppu.vr[op.va];
const auto& b = ppu.vr[op.vb];
auto& d = ppu.vr[op.vd];
v128 d;
for (u8 i = 0; i < 8; i++)
{
@ -1465,6 +1465,7 @@ bool ppu_interpreter_precise::VPKSHSS(ppu_thread& ppu, ppu_opcode_t op)
}
}
ppu.vr[op.vd] = d;
return true;
}
@ -1476,47 +1477,20 @@ bool ppu_interpreter_fast::VPKSHUS(ppu_thread& ppu, ppu_opcode_t op)
bool ppu_interpreter_precise::VPKSHUS(ppu_thread& ppu, ppu_opcode_t op)
{
const auto& a = ppu.vr[op.va];
const auto& b = ppu.vr[op.vb];
auto& d = ppu.vr[op.vd];
const auto a = ppu.vr[op.va];
const auto b = ppu.vr[op.vb];
for (u8 i = 0; i < 8; i++)
// Detect saturation
{
s16 result = a._s16[i];
if (result < 0)
const u64 mask = 0xFF00FF00FF00FF00ULL;
const auto all_bits = v128::fromV(_mm_or_si128(a.vi, b.vi));
if ((all_bits._u64[0] | all_bits._u64[1]) & mask)
{
d._u8[i + 8] = 0;
ppu.sat = true;
}
else if (result > UINT8_MAX)
{
d._u8[i + 8] = UINT8_MAX;
ppu.sat = true;
}
else
{
d._u8[i + 8] = (u8)result;
}
result = b._s16[i];
if (result < 0)
{
d._u8[i] = 0;
ppu.sat = true;
}
else if (result > UINT8_MAX)
{
d._u8[i] = UINT8_MAX;
ppu.sat = true;
}
else
{
d._u8[i] = (u8)result;
}
}
ppu.vr[op.vd].vi = _mm_packus_epi16(b.vi, a.vi);
return true;
}
@ -1530,7 +1504,7 @@ bool ppu_interpreter_precise::VPKSWSS(ppu_thread& ppu, ppu_opcode_t op)
{
const auto& a = ppu.vr[op.va];
const auto& b = ppu.vr[op.vb];
auto& d = ppu.vr[op.vd];
v128 d;
for (u8 i = 0; i < 4; i++)
{
@ -1569,6 +1543,7 @@ bool ppu_interpreter_precise::VPKSWSS(ppu_thread& ppu, ppu_opcode_t op)
}
}
ppu.vr[op.vd] = d;
return true;
}
@ -2495,7 +2470,7 @@ bool ppu_interpreter_precise::VSUMSWS(ppu_thread& ppu, ppu_opcode_t op)
bool ppu_interpreter_fast::VSUM2SWS(ppu_thread& ppu, ppu_opcode_t op)
{
auto& d = ppu.vr[op.vd];
v128 d;
const auto& a = ppu.vr[op.va];
const auto& b = ppu.vr[op.vb];
@ -2516,12 +2491,13 @@ bool ppu_interpreter_fast::VSUM2SWS(ppu_thread& ppu, ppu_opcode_t op)
}
d._s32[1] = 0;
d._s32[3] = 0;
ppu.vr[op.vd] = d;
return true;
}
bool ppu_interpreter_precise::VSUM2SWS(ppu_thread& ppu, ppu_opcode_t op)
{
auto& d = ppu.vr[op.vd];
v128 d;
const auto& a = ppu.vr[op.va];
const auto& b = ppu.vr[op.vb];
@ -2542,8 +2518,10 @@ bool ppu_interpreter_precise::VSUM2SWS(ppu_thread& ppu, ppu_opcode_t op)
else
d._s32[n * 2] = (s32)sum;
}
d._s32[1] = 0;
d._s32[3] = 0;
ppu.vr[op.vd] = d;
return true;
}