Yaya48/dolphin
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #9849 from Tilka/x87

Browse Source 
x64Emitter: remove unused x87 instructions
This commit is contained in:
Tilka authored and GitHub committed 2021-06-28 17:53:54 +01:00
2 parents 2ef2eee2e0 afb9398059
commit bb2d576ecc
4 files changed
+1 -105
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.cpp
+1 -3
View File
@@ -34,9 +34,7 @@ alignas(16) static const __m128i double_bottom_bits = _mm_set_epi64x(0, 0x07ffff
// Since the following float conversion functions are used in non-arithmetic PPC float
// instructions, they must convert floats bitexact and never flush denormals to zero or turn SNaNs
// into QNaNs. This means we can't use CVTSS2SD/CVTSD2SS. The x87 FPU doesn't even support
// flush-to-zero so we can use FLD+FSTP even on denormals.
// If the number is a NaN, make sure to set the QNaN bit back to its original value.
// into QNaNs. This means we can't use CVTSS2SD/CVTSD2SS.
// Another problem is that officially, converting doubles to single format results in undefined
// behavior. Relying on undefined behavior is a bug so no software should ever do this.