zdoom/qzdoom
0
0
Fork 0
mirror of https://github.com/ZDoom/qzdoom.git synced 2024-11-28 23:12:24 +00:00
Code Issues Projects Releases Packages Wiki Activity

The precision of the SSE2 rcp opcode was a little bit too low

Browse source
This commit is contained in:
Magnus Norddahl 2016-11-30 09:08:49 +01:00
parent 34bf581957
commit 837d8f44d1
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
tools/drawergen/fixedfunction/drawtrianglecodegen.cpp
View file

@ -204,7 +204,7 @@ void DrawTriangleCodegen::LoopBlockX()
stack_x.store(minx); stack_x.store(minx);
stack_posx_w.store(posy_w); stack_posx_w.store(posy_w);
for (int i = 0; i < TriVertex::NumVarying; i++) for (int i = 0; i < TriVertex::NumVarying; i++)
stack_posx_varying[i].store(stack_posy_varying[i].load()); stack_posx_varying[i].store(posy_varying[i]);
SSAForBlock loop; SSAForBlock loop;
x = stack_x.load(); x = stack_x.load();
@ -277,7 +277,7 @@ void DrawTriangleCodegen::LoopBlockX()
void DrawTriangleCodegen::SetupAffineBlock() void DrawTriangleCodegen::SetupAffineBlock()
{ {
// Calculate varying variables for affine block // Calculate varying variables for affine block
SSAVec4f rcpW = SSAVec4f::rcp(SSAVec4f(posx_w, posx_w + gradWX, posx_w + gradWY, posx_w + gradWX + gradWY)); SSAVec4f rcpW = SSAVec4f(1.0f) / SSAVec4f(posx_w, posx_w + gradWX, posx_w + gradWY, posx_w + gradWX + gradWY);
for (int i = 0; i < TriVertex::NumVarying; i++) for (int i = 0; i < TriVertex::NumVarying; i++)
{ {
// Top left, top right, bottom left, bottom right: // Top left, top right, bottom left, bottom right: