From 0daf51e130f89d1e83fa897ef8df9229829f5b18 Mon Sep 17 00:00:00 2001 From: Kenneth Graunke Date: Sun, 29 Nov 2015 23:23:44 -0800 Subject: nir: Use writemasked store_vars in glsl_to_nir. Instead of performing the read-modify-write cycle in glsl->nir, we can simply emit a partial writemask. For locals, nir_lower_vars_to_ssa will do the equivalent read-modify-write cycle for us, so we continue to get the same SSA values we had before. Because glsl_to_nir calls nir_lower_outputs_to_temporaries, all outputs are shadowed with temporary values, and written out as whole vectors at the end of the shader. So, most consumers will still not see partial writemasks. However, nir_lower_outputs_to_temporaries bails for tessellation control shader outputs. So those remain actual variables, and stores to those variables now get a writemask. nir_lower_io passes that through. This means that TCS outputs should actually work now. This is a functional change for tessellation control shaders. v2: Relax the nir_validate assert to allow partial writemasks. Signed-off-by: Kenneth Graunke Reviewed-by: Jason Ekstrand --- src/glsl/nir/glsl_to_nir.cpp | 39 +++++++++------------------------------ 1 file changed, 9 insertions(+), 30 deletions(-) (limited to 'src/glsl/nir/glsl_to_nir.cpp') diff --git a/src/glsl/nir/glsl_to_nir.cpp b/src/glsl/nir/glsl_to_nir.cpp index c0de7c4af98..0f0f4a133e0 100644 --- a/src/glsl/nir/glsl_to_nir.cpp +++ b/src/glsl/nir/glsl_to_nir.cpp @@ -1129,44 +1129,23 @@ nir_visitor::visit(ir_assignment *ir) nir_ssa_def *src = evaluate_rvalue(ir->rhs); if (ir->write_mask != (1 << num_components) - 1 && ir->write_mask != 0) { - /* - * We have no good way to update only part of a variable, so just load - * the LHS and do a vec operation to combine the old with the new, and - * then store it - * back into the LHS. Copy propagation should get rid of the mess. + /* GLSL IR will give us the input to the write-masked assignment in a + * single packed vector. So, for example, if the writemask is xzw, then + * we have to swizzle x -> x, y -> z, and z -> w and get the y component + * from the load. */ - - nir_intrinsic_instr *load = - nir_intrinsic_instr_create(this->shader, nir_intrinsic_load_var); - load->num_components = ir->lhs->type->vector_elements; - nir_ssa_dest_init(&load->instr, &load->dest, num_components, NULL); - load->variables[0] = lhs_deref; - ralloc_steal(load, load->variables[0]); - nir_builder_instr_insert(&b, &load->instr); - - nir_ssa_def *srcs[4]; - + unsigned swiz[4]; unsigned component = 0; - for (unsigned i = 0; i < ir->lhs->type->vector_elements; i++) { - if (ir->write_mask & (1 << i)) { - /* GLSL IR will give us the input to the write-masked assignment - * in a single packed vector. So, for example, if the - * writemask is xzw, then we have to swizzle x -> x, y -> z, - * and z -> w and get the y component from the load. - */ - srcs[i] = nir_channel(&b, src, component++); - } else { - srcs[i] = nir_channel(&b, &load->dest.ssa, i); - } + for (unsigned i = 0; i < 4; i++) { + swiz[i] = ir->write_mask & (1 << i) ? component++ : 0; } - - src = nir_vec(&b, srcs, ir->lhs->type->vector_elements); + src = nir_swizzle(&b, src, swiz, num_components, !supports_ints); } nir_intrinsic_instr *store = nir_intrinsic_instr_create(this->shader, nir_intrinsic_store_var); store->num_components = ir->lhs->type->vector_elements; - store->const_index[0] = (1 << store->num_components) - 1; + store->const_index[0] = ir->write_mask; nir_deref *store_deref = nir_copy_deref(store, &lhs_deref->deref); store->variables[0] = nir_deref_as_var(store_deref); store->src[0] = nir_src_for_ssa(src); -- cgit v1.2.3