archive/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-07-05 22:41:25 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

SPU LLVM: Microfixes

Browse source 
- Avoid vpermb path in shufb when op.ra == op.rb
- Reverse indices with (c ^ 0xf) rather than (~c) in vpermb path, vpternlogd is a 3 input operation and requires needless mov instructions to avoid destroying inputs
This commit is contained in:
Malcolm Jestadt 2022-06-07 22:32:41 -04:00 • committed by Ivan
parent 004d9b09b8
commit 64616f1408
1 changed files with 6 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

12
rpcs3/Emu/Cell/SPURecompiler.cpp
View file

@ -7695,7 +7695,7 @@ public:
if (auto [ok, bs] = match_expr(b, byteswap(match<u8[16]>())); ok) if (auto [ok, bs] = match_expr(b, byteswap(match<u8[16]>())); ok)
{ {
// Undo endian swapping, and rely on pshufb/vperm2b to re-reverse endianness // Undo endian swapping, and rely on pshufb/vperm2b to re-reverse endianness
if (m_use_avx512_icl) if (m_use_avx512_icl && (op.ra != op.rb))
{ {
if (perm_only) if (perm_only)
{ {
@ -7757,19 +7757,19 @@ public:
} }
} }
if (m_use_avx512_icl) if (m_use_avx512_icl && (op.ra != op.rb))
{ {
if (perm_only) if (perm_only)
{ {
set_vr(op.rt4, vperm2b256to128(b, a, eval(~c))); set_vr(op.rt4, vperm2b256to128(a, b, eval(c ^ 0xf)));
return; return;
} }
const auto m = gf2p8affineqb(c, build<u8[16]>(0x40, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x40, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20), 0x7f); const auto m = gf2p8affineqb(c, build<u8[16]>(0x40, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x40, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20), 0x7f);
const auto mm = select(noncast<s8[16]>(m) >= 0, splat<u8[16]>(0), m); const auto mm = select(noncast<s8[16]>(m) >= 0, splat<u8[16]>(0), m);
const auto cr = eval(~c); const auto cr = eval(c ^ 0xf);
const auto ab = vperm2b256to128(b, a, cr); const auto ab = vperm2b256to128(a, b, cr);
set_vr(op.rt4, select(noncast<s8[16]>(cr) >= 0, mm, ab)); set_vr(op.rt4, select(noncast<s8[16]>(c) >= 0, ab, mm));
return; return;
} }