 florian
|
9a223100d1
+ x86: proper flag (de)alloc in a_bit_scan
|
5 months ago |
 Rika Ichinose
|
d1ba97d462
Make use of not_zero in x86 bitscans.
|
6 months ago |
|
florian
|
b2f6214b33
+ a_bit_scan_reg_reg gets a flag if src cannot be zero: this simplifies the generated code
|
6 months ago |
|
florian
|
210cb8cb74
* load 64 bit constants through a register into memory to avoid store-forwarding stalls, resolves #40945
|
9 months ago |
|
florian
|
2f5fe67e56
* FSRM works only with movsb
|
10 months ago |
|
Rika Ichinose
|
fc1050a834
Make use of CPUX86_HINT_BSX_DEST_UNCHANGED_ON_ZF_1 in Bsf*/Bsr*.
|
1 year ago |
|
florian
|
cab3b8c06f
* factor out tcgx86.getcopymode
|
1 year ago |
|
florian
|
e29d1ecda9
* reorganize Tcgx86.g_concatcopy
|
1 year ago |
|
florian
|
b826ad8b7e
+ CPUX86_HINT_FAST_SHORT_REP_MOVS
|
1 year ago |
|
florian
|
346962beaa
* compiler part of #35433
|
1 year ago |
 J. Gareth "Curious Kit" Moreton
|
6cbb3d982d
* x86: Added SARX to BMI2 code generator
|
2 years ago |
|
florian
|
e8e53d13cd
* fix #39730 as suggested by Jonas
|
3 years ago |
|
J. Gareth "Curious Kit" Moreton
|
6f24c8b4ef
* x86: Code generation fixes where FLAGS
|
3 years ago |
|
florian
|
ee3b5f5a62
* x86_64: do not use copy_mm in g_concatcopy for 8 byte sized copies
|
3 years ago |
|
J. Gareth "Curious Kit" Moreton
|
fe39f9ca21
Fixed bug on x86 platforms where the stack pointer was not allocated if it was different to the frame pointer
|
3 years ago |
|
florian
|
4966e93764
* small fix for ie
|
3 years ago |
|
J. Gareth "Curious Kit" Moreton
|
2dc0995067
- Bug fix to new ADD/SUB optimisation where conditions are concerned
|
3 years ago |
|
florian
|
8f68988aef
* another location to use source register as second register in VCVTSD2SS and VCVTSS2SD
|
3 years ago |
|
florian
|
4752230c8f
* use source register as second register in VCVTSD2SS and VCVTSS2SD, this should break
|
3 years ago |
 yury
|
75491ae21c
* Removed/ifdefed the assigned and unused variables.
|
4 years ago |
|
yury
|
64c586b86d
* Removed/ifdefed lots of unused variables.
|
4 years ago |
|
florian
|
ecc3ce64ea
* x86: some fixes to enable 8 and 16 bit operations
|
4 years ago |
|
florian
|
792b629aa6
+ -Xa option: generate code which allows to use more than 2 GB of static data on 64 Bit targets
|
4 years ago |
|
florian
|
9c7cb3dbd7
* fix spilling of avx(512) registers
|
4 years ago |
|
florian
|
c3a0e0c6d2
* compilation on i386 fixed
|
4 years ago |
|
florian
|
78244d89dc
* use VPXORD in avx-512 mode as VPXOR does not work for the upper 16 registers
|
4 years ago |
|
florian
|
47066f0ce7
* moved UseAVX from cgx86 to cpubase
|
4 years ago |
|
florian
|
3ae370b190
* also use avx copying for mod 4=0 memory blocks
|
4 years ago |
|
florian
|
f8d33dc188
* fix tcgx86.a_loadmm_reg_ref and tcgx86.a_loadmm_ref_reg for 64 OS_M512
|
4 years ago |
|
florian
|
60bb8600a1
+ copy mode avx512
|
4 years ago |
