 J. Gareth "Curious Kit" Moreton
|
63a2395e2c
* x86: Made a check for BMI2 more explicit in the integer division code generator
|
2 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
645fe8a0c3
* x86: EDX is now deallocated before MULX instructions in the "magic div" algorithm so it can be used as the destination
|
2 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
3f24bd3886
* x86: Improved unsigned integer division for when BMI2 is available.
|
3 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
079905654d
* x86: 64-bit integer division reciprocal algorithm uses IMUL
|
2 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
6f24c8b4ef
* x86: Code generation fixes where FLAGS
|
3 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
3ce6c478cc
Bug fixes to magic division and vectorcall code that trigger -CriotR checks
|
3 gadi atpakaļ |
|
J. Gareth "Curious Kit" Moreton
|
671c674d65
Expanding division to 64-bit and tests
|
3 gadi atpakaļ |
 florian
|
fc41306c45
* fix alignment issues with -<single/double> when using SSE
|
3 gadi atpakaļ |
|
florian
|
5bcc530707
* handle -<mm reg.> more cleverly if SSE is used
|
3 gadi atpakaļ |
|
florian
|
9bd785c06b
* fix handling of -0.0 in sse/avx code, resolves #39357
|
3 gadi atpakaļ |
 pierre
|
85fa313e59
Use uvalue field to get unsigned and avoid range check or overflow errors
|
4 gadi atpakaļ |
 yury
|
64c586b86d
* Removed/ifdefed lots of unused variables.
|
4 gadi atpakaļ |
|
florian
|
ecc3ce64ea
* x86: some fixes to enable 8 and 16 bit operations
|
4 gadi atpakaļ |
|
florian
|
80f87edffd
* x86-64: correctly translate and with large constants in mod optimizations, resolves #38164
|
4 gadi atpakaļ |
|
florian
|
9af9f140a5
* do an unsigned division if one operand is unsigned an the other
|
4 gadi atpakaļ |
|
florian
|
637976e83f
* patch by Marģers to unify internal error numbers, resolves #37888
|
4 gadi atpakaļ |
|
florian
|
99d3651da9
* generate simplified code for unary minus when sse/avx is used
|
4 gadi atpakaļ |
|
florian
|
cf153b0854
* use (V)XORPS/D for unary minus
|
4 gadi atpakaļ |
|
florian
|
28f25b2df0
* reworked usage of tcgnotnode.handle_locjump
|
5 gadi atpakaļ |
|
florian
|
f6e6b807d5
+ optimize <signed> mod <power of 2> for x86-64 as well
|
5 gadi atpakaļ |
 Jonas Maebe
|
281b3ad276
* fix case completeness and unreachable code warnings in compiler that would
|
6 gadi atpakaļ |
|
yury
|
4357caaad8
* Removed unused local vars.
|
6 gadi atpakaļ |
|
florian
|
2385c47c28
* compilation on x86-64 fixed
|
7 gadi atpakaļ |
|
florian
|
8c5606b41d
+ support mmx shifting
|
7 gadi atpakaļ |
|
florian
|
810acd82b2
* patch by J. Gareth Moreton that makes some improvements to the Peephole Optimizer for x86 and x86-64 code, as well as some cleanup with formatting, code syntax consistency, and debug messages.
|
7 gadi atpakaļ |
|
florian
|
5c4b1737c4
* in the generate_code normally imaginary registers are used, so just resize the register to 8 bit, the register allocator takes care of the rest
|
7 gadi atpakaļ |
|
florian
|
4a98fcb9d3
* patch by J. Gareth Moreton: reorganises the produced machine code for large unsigned divisions, resolves #32984
|
7 gadi atpakaļ |
|
florian
|
52aa40c3b0
* remove explicit cast to int64 to avoid an internalerror 200706094, resolves #33004
|
7 gadi atpakaļ |
|
florian
|
724b822b54
* patch based on a proposal by J. Gareth Moreton to reduce register usage
|
7 gadi atpakaļ |
|
florian
|
11a3d8762a
* patch by J. Gareth Moreton:
|
7 gadi atpakaļ |
