diff options
| author | Tony Wasserka <neobrainx@gmail.com> | 2014-03-14 22:57:40 +0100 |
|---|---|---|
| committer | Tony Wasserka <neobrainx@gmail.com> | 2014-03-14 22:57:40 +0100 |
| commit | a9a8c730748b8c8a2e83feb4c38626361cf9daa1 (patch) | |
| tree | b847e06006198ac3d5b5dbb701ff4b1ab4e2de00 /Source/Core/VideoBackends/Software/Tev.cpp | |
| parent | 450bde27bb2d7a47335c9800d3a29e012c9d67e9 (diff) | |
| parent | 0f81cbd6be009ac14f036a05db8f4e516ea626d0 (diff) | |
Merge pull request #68 from neobrain/tev_fixes_new
Change pixel processing to use integer arithmetic.
Diffstat (limited to 'Source/Core/VideoBackends/Software/Tev.cpp')
| -rw-r--r-- | Source/Core/VideoBackends/Software/Tev.cpp | 26 |
1 files changed, 14 insertions, 12 deletions
diff --git a/Source/Core/VideoBackends/Software/Tev.cpp b/Source/Core/VideoBackends/Software/Tev.cpp index ad21496e88..3b517175eb 100644 --- a/Source/Core/VideoBackends/Software/Tev.cpp +++ b/Source/Core/VideoBackends/Software/Tev.cpp @@ -478,7 +478,7 @@ void Tev::Indirect(unsigned int stageNum, s32 s, s32 t) case ITBA_OFF: AlphaBump = 0; break; - case ITBA_S: + case ITBA_S: AlphaBump = indmap[TextureSampler::ALP_SMP]; break; case ITBA_T: @@ -528,9 +528,9 @@ void Tev::Indirect(unsigned int stageNum, s32 s, s32 t) return; } - s64 indtevtrans[2] = { 0,0 }; + s32 indtevtrans[2] = { 0,0 }; - // matrix multiply + // matrix multiply - results might overflow, but we don't care since we only use the lower 24 bits of the result. int indmtxid = indirect.mid & 3; if (indmtxid) { @@ -544,19 +544,21 @@ void Tev::Indirect(unsigned int stageNum, s32 s, s32 t) switch (indirect.mid & 12) { case 0: - shift = 3 + (17 - scale); - indtevtrans[0] = indmtx.col0.ma * indcoord[0] + indmtx.col1.mc * indcoord[1] + indmtx.col2.me * indcoord[2]; - indtevtrans[1] = indmtx.col0.mb * indcoord[0] + indmtx.col1.md * indcoord[1] + indmtx.col2.mf * indcoord[2]; + // matrix values are S0.10, output format is S17.7, so divide by 8 + shift = (17 - scale); + indtevtrans[0] = (indmtx.col0.ma * indcoord[0] + indmtx.col1.mc * indcoord[1] + indmtx.col2.me * indcoord[2]) >> 3; + indtevtrans[1] = (indmtx.col0.mb * indcoord[0] + indmtx.col1.md * indcoord[1] + indmtx.col2.mf * indcoord[2]) >> 3; break; case 4: // s matrix - shift = 8 + (17 - scale); - indtevtrans[0] = s * indcoord[0]; - indtevtrans[1] = t * indcoord[0]; + // s is S17.7, matrix elements are divided by 256, output is S17.7, so divide by 256. - TODO: Maybe, since s is actually stored as S24, we should divide by 256*64? + shift = (17 - scale); + indtevtrans[0] = s * indcoord[0] / 256; + indtevtrans[1] = t * indcoord[0] / 256; break; case 8: // t matrix - shift = 8 + (17 - scale); - indtevtrans[0] = s * indcoord[1]; - indtevtrans[1] = t * indcoord[1]; + shift = (17 - scale); + indtevtrans[0] = s * indcoord[1] / 256; + indtevtrans[1] = t * indcoord[1] / 256; break; default: return; |