PixelShaderGen: Cleanups and fixes for tev combiners. #193
Conversation
| "(((%s.a&255) == (%s.a&255)) ? (%s.a&255) : 0)" // TEVCMP_A8_EQ | ||
| "((tevin_a.r > tevin_b.r) ? tevin_c.a : 0)", // TEVCMP_R8_GT | ||
| "((tevin_a.r == tevin_b.r) ? tevin_c.a : 0)", // TEVCMP_R8_EQ | ||
| "((idot(tevin_a.rgb, comp16) > idot(tevin_b.rgb, comp16)) ? tevin_c.a : 0)", // TEVCMP_GR16_GT |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
|
Thanks for @degasus for pointing out the "other" functional change. I'll verify that this is indeed correct hw behavior before we merge this change. |
|
Additionally, this branch now implements the bit-perfect lerping algorithm now, which I found via hardware tests. I still have to verify the first change, however. |
| // combine the color channel | ||
| if (cc.bias != TevBias_COMPARE) // if not compare | ||
| { | ||
| out.Write("(((tevin_d.rgb%s)%s) %s ((((tevin_a.rgb*256 + (tevin_b.rgb-tevin_a.rgb)*(tevin_c.rgb+(tevin_c.rgb>>7)))%s)%s)>>8))%s", tevBiasTable[cc.bias], tevScaleTableLeft[cc.shift], tevOpTable[cc.op], tevScaleTableLeft[cc.shift], tevLerpBias[2*cc.op+(cc.shift==3)], tevScaleTableRight[cc.shift]); |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
|
Verified against hw that the new behavior indeed is correct. Will clean up the lerper, then it should be good for merging. |
|
Do you also want to fix videosw in the same PR? |
|
Probably, because it's the same fix after all. |
|
Should be good to merge. Tell me what you think :) |
|
LGTM |
|
LTGM |
|
|
||
| out.Write("(((tevin_d.%s%s)%s)", components, tevBiasTable[bias], tevScaleTableLeft[shift]); | ||
| out.Write(" %s ", tevOpTable[op]); | ||
| out.Write("((((tevin_a.%s*256 + (tevin_b.%s-tevin_a.%s)*(tevin_c.%s+(tevin_c.%s>>7)))%s)%s)>>8)", |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
Fixes issue 4674.
…also reproduces hardware behavior perfectly. The new behavior has been verified to be correct by hardware tests. This is an improvement over the old code, which was just a guess.
…r compare mode inputs. As pointed out by dolphin-emu/hwtests@f684f2498.
|
Added comments explaining wtf is going on. |
PixelShaderGen: Cleanups and fixes for tev combiners.
This was meant to be merely a cleanup, but it turned out to also fix the shadows issue in Luigi's Mansion.
It also fixes a small issue in the tev compare function, where
was incorrectly used when it should have actually been
However, this is not what fixed the shadows in Luigi's Mansion. There seems to be another functional chance with this patch, but I wasn't able to figure out where. If you find what else it could be, I'm curious to hear it :p
UPDATE: Turned out two very critical issues was fixed by these patches: The alpha combiner was using incorrect input values in compare mode. But when fixing that issue, the code also needed to be rearchitectured a bit since otherwise the color combiner would overwrite the inputs for the alpha combiner.