Displaying 7 results from an estimated 7 matches for "cloneforward".
2017 Apr 03
5
[PATCH v2 0/3] nv50/ir: Preapre for running Opts inside a loop
Slowly we are getting to the point, that we miss enough optimization
opportunities as the result of our own passes.
For this we need to fix AlgebraicOpt to be able to handle mods on sources
without creating new issues.
The last patch enables looping opts.
v2: update commit author
Karol Herbst (3):
nv50/ir: fix AlgebraicOpt for slcts with mods
nv50/ir: handle logops with NOT in AlgebraicOpt
2017 Dec 20
2
[PATCH] gm107/ir: use lane 0 for manual textureGrad handling
...p2(OP_QUADOP, TYPE_F32, crd[c], tmp, crd[c]);
- add->subOp = qOps[l][1];
+ add->subOp = qOps[1];
add->lanes = 1; /* abused for .ndv */
}
@@ -164,8 +166,20 @@ GM107LoweringPass::handleManualTXD(TexInstruction *i)
// texture
bld.insert(tex = cloneForward(func, i));
+ if (l != 0) {
+ if (array)
+ tex->setSrc(0, arr);
+ if (i->tex.target.isShadow())
+ tex->setSrc(array + dim, shadow);
+ }
for (c = 0; c < dim; ++c)
tex->setSrc(c + array, src[c]);
+ // broadcast result...
2017 Apr 03
0
[PATCH v2 2/3] nv50/ir: handle logops with NOT in AlgebraicOpt
...gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
index bd60a84998..0de84fe9fc 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
@@ -1856,6 +1856,12 @@ AlgebraicOpt::handleLOGOP(Instruction *logop)
set0 = cloneForward(func, set0);
set1 = cloneShallow(func, set1);
+
+ if (logop->src(0).mod == Modifier(NV50_IR_MOD_NOT))
+ set0->asCmp()->setCond = inverseCondCode(set0->asCmp()->setCond);
+ if (logop->src(1).mod == Modifier(NV50_IR_MOD_NOT))
+ set1->asCmp()->s...
2014 Jul 05
0
[PATCH] nvc0: do quadops on the right texture coordinates for TXD
...>getSrc(c + array), zero);
// add dPdx from lane l to lanes dx
for (c = 0; c < dim; ++c)
bld.mkQuadop(qOps[l][0], crd[c], l, i->dPdx[c].get(), crd[c]);
@@ -732,7 +733,7 @@ NVC0LoweringPass::handleManualTXD(TexInstruction *i)
// texture
bld.insert(tex = cloneForward(func, i));
for (c = 0; c < dim; ++c)
- tex->setSrc(c, crd[c]);
+ tex->setSrc(c + array, crd[c]);
// save results
for (c = 0; i->defExists(c); ++c) {
Instruction *mov;
--
1.8.5.5
2017 Apr 03
3
[PATCH 0/3] nv50/ir: Preapre for running Opts inside a loop
Slowly we are getting to the point, that we miss enough optimization
opportunities as the result of our own passes.
For this we need to fix AlgebraicOpt to be able to handle mods on sources
without creating new issues.
The last patch enables looping opts.
Karol Herbst (3):
nv50/ir: fix AlgebraicOpt for slcts with mods
nv50/ir: handle logops with NOT in AlgebraicOpt
nv50/ir: run some
2017 Dec 20
0
[PATCH] gm107/ir: use lane 0 for manual textureGrad handling
...c]);
> - add->subOp = qOps[l][1];
> + add->subOp = qOps[1];
> add->lanes = 1; /* abused for .ndv */
> }
>
> @@ -164,8 +166,20 @@ GM107LoweringPass::handleManualTXD(TexInstruction *i)
>
> // texture
> bld.insert(tex = cloneForward(func, i));
> + if (l != 0) {
> + if (array)
> + tex->setSrc(0, arr);
> + if (i->tex.target.isShadow())
> + tex->setSrc(array + dim, shadow);
> + }
> for (c = 0; c < dim; ++c)
> tex->setSrc(c + ar...
2019 Oct 14
1
[PATCH] gm107/ir: fix loading z offset for layered 3d image bindings
...ot;pred" which (optionally) contains whether to do the surface
+ // op at all, and a "pred2d" which indicates that, in case of doing the
+ // surface op, we have to create a 2d and 3d version, conditioned on pred2d.
+ TexInstruction *su2d = NULL;
+ if (pred2d) {
+ su2d = cloneForward(func, su)->asTex();
+ for (unsigned i = 0; su->defExists(i); ++i)
+ su2d->setDef(i, bld.getSSA());
+ su2d->moveSources(dim + 1, -1);
+ su2d->tex.target = nv50_ir::TEX_TARGET_2D;
+ }
+ if (pred2d && pred) {
+ Instruction *pred3d = bld.mkOp2(OP_A...