diff options
author | Samuel Pitoiset <[email protected]> | 2016-03-31 00:50:39 +0200 |
---|---|---|
committer | Samuel Pitoiset <[email protected]> | 2016-04-01 22:26:24 +0200 |
commit | 4f58b78c309db372d408912ca87e88d319b895da (patch) | |
tree | e4f0fffbfb8fa8764d4f02cf671f72fecacf4810 /src/gallium/drivers/nouveau/codegen | |
parent | 3b246a71d7fe12c4b0670a9dadf566ea3eca1128 (diff) |
nvc0/ir: add support for compute UBOs on Kepler
Make sure to avoid out of bounds access in presence of indirect
array indexing by loading the size from the driver constant buffer.
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
Diffstat (limited to 'src/gallium/drivers/nouveau/codegen')
-rw-r--r-- | src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp | 55 | ||||
-rw-r--r-- | src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h | 3 |
2 files changed, 57 insertions, 1 deletions
diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp index 850147b62e9..da58ced4d7c 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp @@ -1322,6 +1322,24 @@ NVC0LoweringPass::loadBufLength32(Value *ptr, uint32_t off) } inline Value * +NVC0LoweringPass::loadUboInfo32(Value *ptr, uint32_t off) +{ + return loadResInfo32(ptr, off, prog->driver->io.uboInfoBase); +} + +inline Value * +NVC0LoweringPass::loadUboInfo64(Value *ptr, uint32_t off) +{ + return loadResInfo64(ptr, off, prog->driver->io.uboInfoBase); +} + +inline Value * +NVC0LoweringPass::loadUboLength32(Value *ptr, uint32_t off) +{ + return loadResLength32(ptr, off, prog->driver->io.uboInfoBase); +} + +inline Value * NVC0LoweringPass::loadMsInfo32(Value *ptr, uint32_t off) { uint8_t b = prog->driver->io.msInfoCBSlot; @@ -1711,7 +1729,42 @@ NVC0LoweringPass::handleLDST(Instruction *i) assert(prog->getType() != Program::TYPE_FRAGMENT); // INTERP } } else if (i->src(0).getFile() == FILE_MEMORY_CONST) { - if (i->src(0).isIndirect(1)) { + if (targ->getChipset() >= NVISA_GK104_CHIPSET && + prog->getType() == Program::TYPE_COMPUTE) { + // The launch descriptor only allows to set up 8 CBs, but OpenGL + // requires at least 12 UBOs. To bypass this limitation, we store the + // addrs into the driver constbuf and we directly load from the global + // memory. + int8_t fileIndex = i->getSrc(0)->reg.fileIndex - 1; + Value *ind = i->getIndirect(0, 1); + Value *ptr = loadUboInfo64(ind, fileIndex * 16); + + // TODO: clamp the offset to the maximum number of const buf. + if (i->src(0).isIndirect(1)) { + Value *offset = bld.loadImm(NULL, i->getSrc(0)->reg.data.offset + typeSizeof(i->sType)); + Value *length = loadUboLength32(ind, fileIndex * 16); + Value *pred = new_LValue(func, FILE_PREDICATE); + if (i->src(0).isIndirect(0)) { + bld.mkOp2(OP_ADD, TYPE_U64, ptr, ptr, i->getIndirect(0, 0)); + bld.mkOp2(OP_ADD, TYPE_U32, offset, offset, i->getIndirect(0, 0)); + } + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; + i->setIndirect(0, 1, NULL); + i->setIndirect(0, 0, ptr); + bld.mkCmp(OP_SET, CC_GT, TYPE_U32, pred, TYPE_U32, offset, length); + i->setPredicate(CC_NOT_P, pred); + if (i->defExists(0)) { + bld.mkMov(i->getDef(0), bld.mkImm(0)); + } + } else if (fileIndex >= 0) { + if (i->src(0).isIndirect(0)) { + bld.mkOp2(OP_ADD, TYPE_U64, ptr, ptr, i->getIndirect(0, 0)); + } + i->getSrc(0)->reg.file = FILE_MEMORY_GLOBAL; + i->setIndirect(0, 1, NULL); + i->setIndirect(0, 0, ptr); + } + } else if (i->src(0).isIndirect(1)) { Value *ptr; if (i->src(0).isIndirect(0)) ptr = bld.mkOp3v(OP_INSBF, TYPE_U32, bld.getSSA(), diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h index be81d29eb0a..aa192494d9f 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.h @@ -127,6 +127,9 @@ private: Value *loadBufInfo32(Value *ptr, uint32_t off); Value *loadBufInfo64(Value *ptr, uint32_t off); Value *loadBufLength32(Value *ptr, uint32_t off); + Value *loadUboInfo32(Value *ptr, uint32_t off); + Value *loadUboInfo64(Value *ptr, uint32_t off); + Value *loadUboLength32(Value *ptr, uint32_t off); Value *loadMsInfo32(Value *ptr, uint32_t off); Value *loadTexHandle(Value *ptr, unsigned int slot); |