nv50/ir: disable mul+add to mad for precise instructions

fixes
    missrendering in TombRaider
    KHR-GL44.gpu_shader5.precise_qualifier
    KHR-GL45.gpu_shader5.precise_qualifier

v4: disable opt only for MAD, it's fine for SAD

Signed-off-by: Karol Herbst <karolherbst@gmail.com>
Reviewed-by: Pierre Moreau <pierre.morrow@free.fr>

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
index df83f18ae8338ca401894a1d08ad88d62f88e8ff..dac3e6f814c13b1b3965834a6b2c7ec31bbacda8 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
@@ -1677,7 +1677,8 @@ AlgebraicOpt::handleADD(Instruction *add)
       return false;
 
    bool changed = false;
-   if (!changed && prog->getTarget()->isOpSupported(OP_MAD, add->dType))
+   // we can't optimize to MAD if the add is precise
+   if (!add->precise && prog->getTarget()->isOpSupported(OP_MAD, add->dType))
       changed = tryADDToMADOrSAD(add, OP_MAD);
    if (!changed && prog->getTarget()->isOpSupported(OP_SAD, add->dType))
       changed = tryADDToMADOrSAD(add, OP_SAD);
@@ -1713,7 +1714,7 @@ AlgebraicOpt::tryADDToMADOrSAD(Instruction *add, operation toOp)
       return false;
 
    if (src->getInsn()->saturate || src->getInsn()->postFactor ||
-       src->getInsn()->dnz)
+       src->getInsn()->dnz || src->getInsn()->precise)
       return false;
 
    if (toOp == OP_SAD) {