Andre Werthmann wrote:
>
> I have committed the remaining transform-functions written in SSE-assembly.
> That means now all the transform_points* functions are done in SSE-assembly :)
>
> These functions are faster than the fpu-ones, but to get the most speed out of the
>SSE the data, the SSE-instructions
> are working on (transform-matrix, src_vertices and dst_vertices) must be 16byte
>aligned.
>
> I have to find a way how to do it best (while keeping the code as clean as possible
>!).
>
I updated my copy of the Makefile.fx (for Windows 32 compilation) to
include the katmai files. I only have a PPRO so I don't know if it
really
works, but it compiles and doesn't seem to cause problems on my
computer.
So if there are any windows users who can test this please do so...
I will attach the patch into this message.
While doing this I noticed that the previous version
of Makefile.fx did not contain the -DUSE_3DNOW_ASM flag,
which I included here.
This addition should propably be done also to the
3.2 sources.
Eero
Index: Makefile.fx
===================================================================
RCS file: /cvs/mesa3d/Mesa/src/Makefile.fx,v
retrieving revision 1.7
diff -c -r1.7 Makefile.fx
*** Makefile.fx 1999/11/17 21:13:03 1.7
--- Makefile.fx 1999/12/08 13:29:55
***************
*** 41,47 ****
!if "$(NASM)" != ""
! X86OPT = -DUSE_X86_ASM
!else
X86OPT =
!endif
--- 41,47 ----
!if "$(NASM)" != ""
! X86OPT = -DUSE_X86_ASM -DUSE_3DNOW_ASM -DUSE_KATMAI_ASM
!else
X86OPT =
!endif
***************
*** 107,113 ****
X86\3dnow_xform_raw2.obj X86\3dnow_xform_raw3.obj\
X86\3dnow_xform_raw4.obj\
X86\mmx_blend.obj X86\vertex.obj X86\vertex_3dnow.obj \
! X86\x86a.obj X86\common_x86.obj X86\common_x86asm.obj X86\x86.obj
!else
X86OBJS =
!endif
--- 107,120 ----
X86\3dnow_xform_raw2.obj X86\3dnow_xform_raw3.obj\
X86\3dnow_xform_raw4.obj\
X86\mmx_blend.obj X86\vertex.obj X86\vertex_3dnow.obj \
! X86\x86a.obj X86\common_x86.obj X86\common_x86asm.obj X86\x86.obj\
! X86\katmai.obj X86\katmai_norm_raw.obj \
! X86\katmai_xform_masked1.obj X86\katmai_xform_masked2.obj\
! X86\katmai_xform_masked3.obj X86\katmai_xform_masked4.obj\
! X86\katmai_xform_raw1.obj X86\katmai_xform_raw2.obj\
! X86\katmai_xform_raw3.obj X86\katmai_xform_raw4.obj\
! X86\vertex_katmai.obj
!
!else
X86OBJS =
!endif
