Re: Device.c and set render states
2008/7/28 Chris Ahrendt [EMAIL PROTECTED]: Please comment on the attached patch and let me know what you think.. Almost all of my comments on the original patch still stand. In addition, you're adding trailing spaces, and you've got an unnecessary whitespace change in there.
RE: Device.c and set render states
My performance concern is about the switch statement itself, less about the
actual value checking. If you have something like
switch(renderstate) {
case WINED3DRS_A:
case WINED3DRS_B:
doSomething();
break;
default: break;
}
The binary code generated by the compiler still checks 239 states against
D3DRS_A and D3DRS_B(the only exception is D3DRS_A, which is never compared
against D3DRS_B). That is where my worries come from, not the doSomething()
itself.
But beyond that, I think checking the value ranges is just wrong, at least
we need a test to show Windows does that as well.
Re: Device.c and set render states
Stefan Dösinger wrote:
My performance concern is about the switch statement itself, less about the
actual value checking. If you have something like
switch(renderstate) {
case WINED3DRS_A:
case WINED3DRS_B:
doSomething();
break;
default: break;
}
The binary code generated by the compiler still checks 239 states against
D3DRS_A and D3DRS_B(the only exception is D3DRS_A, which is never compared
against D3DRS_B). That is where my worries come from, not the doSomething()
itself.
But beyond that, I think checking the value ranges is just wrong, at least
we need a test to show Windows does that as well.
Ok I will work on the test case when I get a chance... (I think I can do
this in a vmware session if not it is going to have to wait till I get
to the house where I have a windows machine).
Ahh ok... I also this morning moved the check to see if the call is
just resetting the same value over and over to before the switch
statement so that if we are getting the same value it just returns like
it did before however it does not set it over again.
Chris
RE: Device.c and set render states
Ok I will work on the test case when I get a chance... (I think I can do this in a vmware session if not it is going to have to wait till I get to the house where I have a windows machine). Actually, you have to be careful with D3D and vmware. D3D is highly driver dependent, up to the point where we diverge from the reference rasterizer because games depend on a driver bug. VMware has its own D3D driver, which fails miserably in Wine tests(last someone checked). So you can certainly use vmware to write the test, but before using it to implement the same behavior in Wine or submitting the test, you should really test with a Windows machine, preferably one that has a GeForce or Radeon card
Re: Device.c and set render states
Stefan Dösinger wrote: Ok I will work on the test case when I get a chance... (I think I can do this in a vmware session if not it is going to have to wait till I get to the house where I have a windows machine). Actually, you have to be careful with D3D and vmware. D3D is highly driver dependent, up to the point where we diverge from the reference rasterizer because games depend on a driver bug. VMware has its own D3D driver, which fails miserably in Wine tests(last someone checked). So you can certainly use vmware to write the test, but before using it to implement the same behavior in Wine or submitting the test, you should really test with a Windows machine, preferably one that has a GeForce or Radeon card Ok I have a machine at home that I think has a GeForce card in it or a Radeon (I cant remember) that I will run on. Would it be better to write the test code or find something like 3D mark or the like? Chris
RE: Device.c and set render states
Ok I have a machine at home that I think has a GeForce card in it or a Radeon (I cant remember) that I will run on. Would it be better to write the test code or find something like 3D mark or the like? Look at dlls/d3d9/tests/*.c for how the Wine tests work
Re: Device.c and set render states
Stefan Dösinger wrote: Ok I have a machine at home that I think has a GeForce card in it or a Radeon (I cant remember) that I will run on. Would it be better to write the test code or find something like 3D mark or the like? Look at dlls/d3d9/tests/*.c for how the Wine tests work Ok thanks... will do =) Chris
Re: Device.c and set render states
Stefan Dösinger wrote: Ok I have a machine at home that I think has a GeForce card in it or a Radeon (I cant remember) that I will run on. Would it be better to write the test code or find something like 3D mark or the like? Look at dlls/d3d9/tests/*.c for how the Wine tests work Ok looking at the d3d9 directory I found visual.c which has quite a few calls to the set render state.. So would I go through and make sure that all of the calls which I added in the type checking are in the test? What about failure tests to make sure the function is working right esp on windows? Sorry for the questions but I am not finding anything out there on writing the wine test cases =) chris
