aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/glsl/nir
diff options
context:
space:
mode:
Diffstat (limited to 'mesalib/src/glsl/nir')
-rw-r--r--mesalib/src/glsl/nir/glsl_to_nir.cpp15
-rw-r--r--mesalib/src/glsl/nir/nir_intrinsics.h20
-rw-r--r--mesalib/src/glsl/nir/nir_lower_atomics.c2
-rw-r--r--mesalib/src/glsl/nir/nir_lower_io.c2
-rw-r--r--mesalib/src/glsl/nir/nir_lower_phis_to_scalar.c5
-rw-r--r--mesalib/src/glsl/nir/nir_opt_algebraic.py2
-rw-r--r--mesalib/src/glsl/nir/nir_opt_peephole_ffma.c19
7 files changed, 37 insertions, 28 deletions
diff --git a/mesalib/src/glsl/nir/glsl_to_nir.cpp b/mesalib/src/glsl/nir/glsl_to_nir.cpp
index af758ceb0..95531bbcd 100644
--- a/mesalib/src/glsl/nir/glsl_to_nir.cpp
+++ b/mesalib/src/glsl/nir/glsl_to_nir.cpp
@@ -65,6 +65,7 @@ public:
virtual void visit(ir_dereference_variable *);
virtual void visit(ir_dereference_record *);
virtual void visit(ir_dereference_array *);
+ virtual void visit(ir_barrier *);
void create_function(ir_function *ir);
@@ -930,13 +931,9 @@ nir_visitor::evaluate_rvalue(ir_rvalue* ir)
}
nir_dest *dest = get_instr_dest(this->result);
-
assert(dest->is_ssa);
- nir_src src = NIR_SRC_INIT;
- src.is_ssa = true;
- src.ssa = &dest->ssa;
- return src;
+ return nir_src_for_ssa(&dest->ssa);
}
nir_alu_instr *
@@ -1893,3 +1890,11 @@ nir_visitor::visit(ir_dereference_array *ir)
ralloc_steal(this->deref_tail, deref);
this->deref_tail = &deref->deref;
}
+
+void
+nir_visitor::visit(ir_barrier *ir)
+{
+ nir_intrinsic_instr *instr =
+ nir_intrinsic_instr_create(this->shader, nir_intrinsic_barrier);
+ nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+}
diff --git a/mesalib/src/glsl/nir/nir_intrinsics.h b/mesalib/src/glsl/nir/nir_intrinsics.h
index 10192c531..bc6e6b8f4 100644
--- a/mesalib/src/glsl/nir/nir_intrinsics.h
+++ b/mesalib/src/glsl/nir/nir_intrinsics.h
@@ -67,6 +67,7 @@ INTRINSIC(interp_var_at_offset, 1, ARR(2), true, 0, 1, 0,
*/
#define BARRIER(name) INTRINSIC(name, 0, ARR(), false, 0, 0, 0, 0)
+BARRIER(barrier)
BARRIER(discard)
/*
@@ -138,12 +139,11 @@ SYSTEM_VALUE(sample_mask_in, 1)
SYSTEM_VALUE(invocation_id, 1)
/*
- * The first index is the address to load from, and the second index is the
- * number of array elements to load. Indirect loads have an additional
- * register input, which is added to the constant address to compute the
- * final address to load from. For UBO's (and SSBO's), the first source is
- * the (possibly constant) UBO buffer index and the indirect (if it exists)
- * is the second source.
+ * The first and only index is the base address to load from. Indirect
+ * loads have an additional register input, which is added to the constant
+ * address to compute the final address to load from. For UBO's (and
+ * SSBO's), the first source is the (possibly constant) UBO buffer index
+ * and the indirect (if it exists) is the second source.
*
* For vector backends, the address is in terms of one vec4, and so each array
* element is +4 scalar components from the previous array element. For scalar
@@ -152,9 +152,9 @@ SYSTEM_VALUE(invocation_id, 1)
*/
#define LOAD(name, extra_srcs, flags) \
- INTRINSIC(load_##name, extra_srcs, ARR(1), true, 0, 0, 2, flags) \
+ INTRINSIC(load_##name, extra_srcs, ARR(1), true, 0, 0, 1, flags) \
INTRINSIC(load_##name##_indirect, extra_srcs + 1, ARR(1, 1), \
- true, 0, 0, 2, flags)
+ true, 0, 0, 1, flags)
LOAD(uniform, 0, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
LOAD(ubo, 1, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
@@ -172,7 +172,7 @@ LOAD(input, 0, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
INTRINSIC(store_##name##_indirect, 2, ARR(0, 1), false, 0, 0, \
num_indices, flags) \
-STORE(output, 2, 0)
-/* STORE(ssbo, 3, 0) */
+STORE(output, 1, 0)
+/* STORE(ssbo, 2, 0) */
LAST_INTRINSIC(store_output_indirect)
diff --git a/mesalib/src/glsl/nir/nir_lower_atomics.c b/mesalib/src/glsl/nir/nir_lower_atomics.c
index f6f89020f..0457de60d 100644
--- a/mesalib/src/glsl/nir/nir_lower_atomics.c
+++ b/mesalib/src/glsl/nir/nir_lower_atomics.c
@@ -109,7 +109,7 @@ lower_instr(nir_intrinsic_instr *instr, nir_function_impl *impl)
}
new_instr->src[0].is_ssa = true;
- new_instr->src[0].ssa = offset_def;;
+ new_instr->src[0].ssa = offset_def;
if (instr->dest.is_ssa) {
nir_ssa_dest_init(&new_instr->instr, &new_instr->dest,
diff --git a/mesalib/src/glsl/nir/nir_lower_io.c b/mesalib/src/glsl/nir/nir_lower_io.c
index 03eed04e1..6761d5bad 100644
--- a/mesalib/src/glsl/nir/nir_lower_io.c
+++ b/mesalib/src/glsl/nir/nir_lower_io.c
@@ -288,7 +288,6 @@ nir_lower_io_block(nir_block *block, void *void_state)
offset += intrin->variables[0]->var->data.driver_location;
load->const_index[0] = offset;
- load->const_index[1] = 1;
if (has_indirect)
load->src[0] = indirect;
@@ -331,7 +330,6 @@ nir_lower_io_block(nir_block *block, void *void_state)
offset += intrin->variables[0]->var->data.driver_location;
store->const_index[0] = offset;
- store->const_index[1] = 1;
nir_src_copy(&store->src[0], &intrin->src[0], state->mem_ctx);
diff --git a/mesalib/src/glsl/nir/nir_lower_phis_to_scalar.c b/mesalib/src/glsl/nir/nir_lower_phis_to_scalar.c
index 4bdb80072..a57d25397 100644
--- a/mesalib/src/glsl/nir/nir_lower_phis_to_scalar.c
+++ b/mesalib/src/glsl/nir/nir_lower_phis_to_scalar.c
@@ -153,6 +153,11 @@ should_lower_phi(nir_phi_instr *phi, struct lower_phis_to_scalar_state *state)
break;
}
+ /* The hash table entry for 'phi' may have changed while recursing the
+ * dependence graph, so we need to reset it */
+ entry = _mesa_hash_table_search(state->phi_table, phi);
+ assert(entry);
+
entry->data = (void *)(intptr_t)scalarizable;
return scalarizable;
diff --git a/mesalib/src/glsl/nir/nir_opt_algebraic.py b/mesalib/src/glsl/nir/nir_opt_algebraic.py
index fa039222f..eace791f5 100644
--- a/mesalib/src/glsl/nir/nir_opt_algebraic.py
+++ b/mesalib/src/glsl/nir/nir_opt_algebraic.py
@@ -156,6 +156,8 @@ optimizations = [
(('fpow', a, 2.0), ('fmul', a, a)),
(('fpow', a, 4.0), ('fmul', ('fmul', a, a), ('fmul', a, a))),
(('fpow', 2.0, a), ('fexp2', a)),
+ (('fpow', ('fpow', a, 2.2), 0.454545), a),
+ (('fpow', ('fabs', ('fpow', a, 2.2)), 0.454545), ('fabs', a)),
(('fsqrt', ('fexp2', a)), ('fexp2', ('fmul', 0.5, a))),
(('frcp', ('fexp2', a)), ('fexp2', ('fneg', a))),
(('frsq', ('fexp2', a)), ('fexp2', ('fmul', -0.5, a))),
diff --git a/mesalib/src/glsl/nir/nir_opt_peephole_ffma.c b/mesalib/src/glsl/nir/nir_opt_peephole_ffma.c
index b430eac8e..798506b75 100644
--- a/mesalib/src/glsl/nir/nir_opt_peephole_ffma.c
+++ b/mesalib/src/glsl/nir/nir_opt_peephole_ffma.c
@@ -73,7 +73,8 @@ are_all_uses_fadd(nir_ssa_def *def)
}
static nir_alu_instr *
-get_mul_for_src(nir_alu_src *src, uint8_t swizzle[4], bool *negate, bool *abs)
+get_mul_for_src(nir_alu_src *src, int num_components,
+ uint8_t swizzle[4], bool *negate, bool *abs)
{
assert(src->src.is_ssa && !src->abs && !src->negate);
@@ -85,16 +86,16 @@ get_mul_for_src(nir_alu_src *src, uint8_t swizzle[4], bool *negate, bool *abs)
switch (alu->op) {
case nir_op_imov:
case nir_op_fmov:
- alu = get_mul_for_src(&alu->src[0], swizzle, negate, abs);
+ alu = get_mul_for_src(&alu->src[0], num_components, swizzle, negate, abs);
break;
case nir_op_fneg:
- alu = get_mul_for_src(&alu->src[0], swizzle, negate, abs);
+ alu = get_mul_for_src(&alu->src[0], num_components, swizzle, negate, abs);
*negate = !*negate;
break;
case nir_op_fabs:
- alu = get_mul_for_src(&alu->src[0], swizzle, negate, abs);
+ alu = get_mul_for_src(&alu->src[0], num_components, swizzle, negate, abs);
*negate = false;
*abs = true;
break;
@@ -115,12 +116,8 @@ get_mul_for_src(nir_alu_src *src, uint8_t swizzle[4], bool *negate, bool *abs)
if (!alu)
return NULL;
- for (unsigned i = 0; i < 4; i++) {
- if (!(alu->dest.write_mask & (1 << i)))
- break;
-
+ for (unsigned i = 0; i < num_components; i++)
swizzle[i] = swizzle[src->swizzle[i]];
- }
return alu;
}
@@ -160,7 +157,9 @@ nir_opt_peephole_ffma_block(nir_block *block, void *void_state)
negate = false;
abs = false;
- mul = get_mul_for_src(&add->src[add_mul_src], swizzle, &negate, &abs);
+ mul = get_mul_for_src(&add->src[add_mul_src],
+ add->dest.dest.ssa.num_components,
+ swizzle, &negate, &abs);
if (mul != NULL)
break;
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-22 21:43:13 +0000