Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Revelator darkmod AVX* changes #631

Open
wants to merge 26 commits into
base: master
Choose a base branch
from
Open

Revelator darkmod AVX* changes #631

Changes from 1 commit
Commits
Show all changes
26 commits
Select commit Hold shift + click to select a range
5411625
AVX* changes to idlib
revelator Nov 17, 2024
eee2aaf
cmake changes for avx* support
revelator Nov 17, 2024
ca93e90
Changes to use the New AVX* based culling
revelator Nov 17, 2024
29de621
new detection routine for AVX changes.
revelator Nov 17, 2024
0a1f2a3
enums for new cpu types
revelator Nov 18, 2024
f50471b
Hopefully fixed some mistakes
revelator Nov 18, 2024
8dcbdc2
more forgotten changes...
revelator Nov 18, 2024
d0fc68c
trying to fix avx build failure on nix compiler
revelator Nov 18, 2024
59afa20
revert addition of ALIGNTYPE16
revelator Nov 18, 2024
f23f59c
changed ALIGNTYPE16 to ALIGN16 in DAZ detection
revelator Nov 18, 2024
5071ad5
Hybrid GLSL ARB2 renderer
revelator Nov 18, 2024
4d876e0
externs and shader code for GLSL
revelator Nov 18, 2024
a2d68e3
VBO changes
revelator Nov 18, 2024
ff4ab55
Another batch of misses
revelator Nov 18, 2024
bdae6b7
myGlMultiMatrix name change and SSE intrinsics
revelator Nov 18, 2024
453e122
myGlMultiMatrix name change
revelator Nov 18, 2024
92bf94a
missing type for DepthBoundsTest
revelator Nov 18, 2024
246602a
Try to make gcc happy
revelator Nov 18, 2024
ddb3af0
cast from const char to char gcc error
revelator Nov 18, 2024
3f3d2b7
turn of some warnings that make no sense
revelator Nov 18, 2024
50b1ed4
bring draft in line with current
revelator Nov 20, 2024
f7c55ee
further draft changes
revelator Nov 20, 2024
158c17b
and hpåefully the last batch needed for the draft
revelator Nov 20, 2024
6edbc44
mac build error with sse
revelator Nov 20, 2024
2240b7d
minor whoops in disabling SSE intrinsics for ppc
revelator Nov 21, 2024
0edfbdc
DWORD is not recognized
revelator Nov 21, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
53 changes: 38 additions & 15 deletions neo/renderer/tr_main.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@ If you have questions concerning this license or the applicable additional terms
===========================================================================
*/

// atleast ppc seems to have totally different intrinsics
#if defined(__ppc__) && defined(__APPLE__)
#include <vecLib/vecLib.h>
#endif
Expand All @@ -35,7 +36,7 @@ If you have questions concerning this license or the applicable additional terms
#endif

#if defined(_MSC_VER)
#include <intrin.h>
#include <xmmintrin.h>
#endif

#include "sys/platform.h"
Expand Down Expand Up @@ -473,9 +474,9 @@ void R_AxisToModelMatrix( const idMat3 &axis, const idVec3 &origin, float modelM
modelMatrix[15] = 1;
}

// FIXME: these assume no skewing or scaling transforms
// FIXME: these assume no skewing or scaling transforms (disabled SSE intrinsics on MAC ppc)
void R_LocalPointToGlobal( const float modelMatrix[16], const idVec3 &in, idVec3 &out ) {
#if !( defined(MACOS_X) || defined(__APPLE__) )
#if !( defined(__ppc__) || defined(MACOS_X) || defined(__APPLE__) )
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is a very good check for this purpose. It doesn't do what you want on

  • x86 Macs (there it should be enabled)
  • non-x86 CPUs that aren't PPC (unless they happen to run OSX)

Can't you just use #ifdef __SSE__?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yeah i newer coded for MAC before so was a bit unsure how to proceed :) ill see if just using SSE helps.

int cpuid = idLib::sys->GetProcessorId();

// a more general approach here, bonus also supports SSE instead of minimum SSE2
Expand All @@ -487,19 +488,23 @@ void R_LocalPointToGlobal( const float modelMatrix[16], const idVec3 &in, idVec3
__m128 xxxx = _mm_set1_ps( in.x );
__m128 yyyy = _mm_set1_ps( in.y );
__m128 zzzz = _mm_set1_ps( in.z );
__m128 res = _mm_add_ps( _mm_add_ps( _mm_mul_ps( row0, xxxx ), _mm_mul_ps( row1, yyyy ) ),
__m128 res = _mm_add_ps( _mm_add_ps( _mm_mul_ps( row0, xxxx ),
_mm_mul_ps( row1, yyyy ) ),
_mm_add_ps( _mm_mul_ps( row2, zzzz ), row3 ) );
// unaligned float x 3 store
_mm_storel_pi( ( __m64 * )&out[0], res );
_mm_store_ss( &out[2], _mm_movehl_ps( res, res ) );
} else
#else
{
} else {
// fall back to non SSE
out[0] = in[0] * modelMatrix[0] + in[1] * modelMatrix[4] + in[2] * modelMatrix[8] + modelMatrix[12];
out[1] = in[0] * modelMatrix[1] + in[1] * modelMatrix[5] + in[2] * modelMatrix[9] + modelMatrix[13];
out[2] = in[0] * modelMatrix[2] + in[1] * modelMatrix[6] + in[2] * modelMatrix[10] + modelMatrix[14];
}
#else
// fall back to non SSE on mac
out[0] = in[0] * modelMatrix[0] + in[1] * modelMatrix[4] + in[2] * modelMatrix[8] + modelMatrix[12];
out[1] = in[0] * modelMatrix[1] + in[1] * modelMatrix[5] + in[2] * modelMatrix[9] + modelMatrix[13];
out[2] = in[0] * modelMatrix[2] + in[1] * modelMatrix[6] + in[2] * modelMatrix[10] + modelMatrix[14];
#endif
}

Expand Down Expand Up @@ -763,11 +768,11 @@ void R_TransformClipToDevice( const idPlane &clip, const viewDef_t *view, idVec3
==========================
R_MatrixMultiply

Changed name to be more in line with RBDoom
Changed name to be more in line with RBDoom (disabled SSE intrinsics on MAC ppc)
==========================
*/
void R_MatrixMultiply( const float a[16], const float b[16], float out[16] ) {
#if !( defined(MACOS_X) || defined(__APPLE__) )
#if !( defined(__ppc__) || defined(MACOS_X) || defined(__APPLE__) )
int cpuid = idLib::sys->GetProcessorId();

/* fast path use sse */
Expand All @@ -782,14 +787,14 @@ void R_MatrixMultiply( const float a[16], const float b[16], float out[16] ) {
__m128 Ai1 = _mm_set1_ps( a[4 * i + 1] );
__m128 Ai2 = _mm_set1_ps( a[4 * i + 2] );
__m128 Ai3 = _mm_set1_ps( a[4 * i + 3] );
__m128 Rix = _mm_add_ps( _mm_add_ps( _mm_mul_ps( Ai0, B0x ), _mm_mul_ps( Ai1, B1x ) ),
_mm_add_ps( _mm_mul_ps( Ai2, B2x ), _mm_mul_ps( Ai3, B3x ) ) );
__m128 Rix = _mm_add_ps( _mm_add_ps( _mm_mul_ps( Ai0, B0x ),
_mm_mul_ps( Ai1, B1x ) ),
_mm_add_ps( _mm_mul_ps( Ai2, B2x ),
_mm_mul_ps( Ai3, B3x ) ) );
_mm_storeu_ps( &out[4 * i], Rix );
}
} else
#else
{
/* fallback to matrix */
} else {
/* fallback to non SSE matrix */
out[0 * 4 + 0] = a[0 * 4 + 0] * b[0 * 4 + 0] + a[0 * 4 + 1] * b[1 * 4 + 0] + a[0 * 4 + 2] * b[2 * 4 + 0] + a[0 * 4 + 3] * b[3 * 4 + 0];
out[0 * 4 + 1] = a[0 * 4 + 0] * b[0 * 4 + 1] + a[0 * 4 + 1] * b[1 * 4 + 1] + a[0 * 4 + 2] * b[2 * 4 + 1] + a[0 * 4 + 3] * b[3 * 4 + 1];
out[0 * 4 + 2] = a[0 * 4 + 0] * b[0 * 4 + 2] + a[0 * 4 + 1] * b[1 * 4 + 2] + a[0 * 4 + 2] * b[2 * 4 + 2] + a[0 * 4 + 3] * b[3 * 4 + 2];
Expand All @@ -807,6 +812,24 @@ void R_MatrixMultiply( const float a[16], const float b[16], float out[16] ) {
out[3 * 4 + 2] = a[3 * 4 + 0] * b[0 * 4 + 2] + a[3 * 4 + 1] * b[1 * 4 + 2] + a[3 * 4 + 2] * b[2 * 4 + 2] + a[3 * 4 + 3] * b[3 * 4 + 2];
out[3 * 4 + 3] = a[3 * 4 + 0] * b[0 * 4 + 3] + a[3 * 4 + 1] * b[1 * 4 + 3] + a[3 * 4 + 2] * b[2 * 4 + 3] + a[3 * 4 + 3] * b[3 * 4 + 3];
}
#else
/* fallback to non SSE matrix on mac */
out[0 * 4 + 0] = a[0 * 4 + 0] * b[0 * 4 + 0] + a[0 * 4 + 1] * b[1 * 4 + 0] + a[0 * 4 + 2] * b[2 * 4 + 0] + a[0 * 4 + 3] * b[3 * 4 + 0];
out[0 * 4 + 1] = a[0 * 4 + 0] * b[0 * 4 + 1] + a[0 * 4 + 1] * b[1 * 4 + 1] + a[0 * 4 + 2] * b[2 * 4 + 1] + a[0 * 4 + 3] * b[3 * 4 + 1];
out[0 * 4 + 2] = a[0 * 4 + 0] * b[0 * 4 + 2] + a[0 * 4 + 1] * b[1 * 4 + 2] + a[0 * 4 + 2] * b[2 * 4 + 2] + a[0 * 4 + 3] * b[3 * 4 + 2];
out[0 * 4 + 3] = a[0 * 4 + 0] * b[0 * 4 + 3] + a[0 * 4 + 1] * b[1 * 4 + 3] + a[0 * 4 + 2] * b[2 * 4 + 3] + a[0 * 4 + 3] * b[3 * 4 + 3];
out[1 * 4 + 0] = a[1 * 4 + 0] * b[0 * 4 + 0] + a[1 * 4 + 1] * b[1 * 4 + 0] + a[1 * 4 + 2] * b[2 * 4 + 0] + a[1 * 4 + 3] * b[3 * 4 + 0];
out[1 * 4 + 1] = a[1 * 4 + 0] * b[0 * 4 + 1] + a[1 * 4 + 1] * b[1 * 4 + 1] + a[1 * 4 + 2] * b[2 * 4 + 1] + a[1 * 4 + 3] * b[3 * 4 + 1];
out[1 * 4 + 2] = a[1 * 4 + 0] * b[0 * 4 + 2] + a[1 * 4 + 1] * b[1 * 4 + 2] + a[1 * 4 + 2] * b[2 * 4 + 2] + a[1 * 4 + 3] * b[3 * 4 + 2];
out[1 * 4 + 3] = a[1 * 4 + 0] * b[0 * 4 + 3] + a[1 * 4 + 1] * b[1 * 4 + 3] + a[1 * 4 + 2] * b[2 * 4 + 3] + a[1 * 4 + 3] * b[3 * 4 + 3];
out[2 * 4 + 0] = a[2 * 4 + 0] * b[0 * 4 + 0] + a[2 * 4 + 1] * b[1 * 4 + 0] + a[2 * 4 + 2] * b[2 * 4 + 0] + a[2 * 4 + 3] * b[3 * 4 + 0];
out[2 * 4 + 1] = a[2 * 4 + 0] * b[0 * 4 + 1] + a[2 * 4 + 1] * b[1 * 4 + 1] + a[2 * 4 + 2] * b[2 * 4 + 1] + a[2 * 4 + 3] * b[3 * 4 + 1];
out[2 * 4 + 2] = a[2 * 4 + 0] * b[0 * 4 + 2] + a[2 * 4 + 1] * b[1 * 4 + 2] + a[2 * 4 + 2] * b[2 * 4 + 2] + a[2 * 4 + 3] * b[3 * 4 + 2];
out[2 * 4 + 3] = a[2 * 4 + 0] * b[0 * 4 + 3] + a[2 * 4 + 1] * b[1 * 4 + 3] + a[2 * 4 + 2] * b[2 * 4 + 3] + a[2 * 4 + 3] * b[3 * 4 + 3];
out[3 * 4 + 0] = a[3 * 4 + 0] * b[0 * 4 + 0] + a[3 * 4 + 1] * b[1 * 4 + 0] + a[3 * 4 + 2] * b[2 * 4 + 0] + a[3 * 4 + 3] * b[3 * 4 + 0];
out[3 * 4 + 1] = a[3 * 4 + 0] * b[0 * 4 + 1] + a[3 * 4 + 1] * b[1 * 4 + 1] + a[3 * 4 + 2] * b[2 * 4 + 1] + a[3 * 4 + 3] * b[3 * 4 + 1];
out[3 * 4 + 2] = a[3 * 4 + 0] * b[0 * 4 + 2] + a[3 * 4 + 1] * b[1 * 4 + 2] + a[3 * 4 + 2] * b[2 * 4 + 2] + a[3 * 4 + 3] * b[3 * 4 + 2];
out[3 * 4 + 3] = a[3 * 4 + 0] * b[0 * 4 + 3] + a[3 * 4 + 1] * b[1 * 4 + 3] + a[3 * 4 + 2] * b[2 * 4 + 3] + a[3 * 4 + 3] * b[3 * 4 + 3];
#endif
}

Expand Down
Loading