nir/lower_io: Apply alignments from derefs when available
If the deref has no explicit alignment in the chain, we assume component
alignment which is what we currently assume for all derefs today. This
should be correct for all APIs in the sense that we can usually assume
at least component alignment. However, for some APIs such as OpenCL, we
could potentially make larger alignment assumptions. The intention is
that those will be handled via alignment-increasing casts.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>
diff --git a/src/compiler/nir/nir_lower_io.c b/src/compiler/nir/nir_lower_io.c
index 7cfbe28..bac6b8e 100644
--- a/src/compiler/nir/nir_lower_io.c
+++ b/src/compiler/nir/nir_lower_io.c
@@ -952,6 +952,7 @@
static nir_ssa_def *
build_explicit_io_load(nir_builder *b, nir_intrinsic_instr *intrin,
nir_ssa_def *addr, nir_address_format addr_format,
+ uint32_t align_mul, uint32_t align_offset,
unsigned num_components)
{
nir_variable_mode mode = nir_src_as_deref(intrin->src[0])->mode;
@@ -1027,10 +1028,7 @@
bit_size = 32;
}
- /* TODO: We should try and provide a better alignment. For OpenCL, we need
- * to plumb the alignment through from SPIR-V when we have one.
- */
- nir_intrinsic_set_align(load, bit_size / 8, 0);
+ nir_intrinsic_set_align(load, align_mul, align_offset);
assert(intrin->dest.is_ssa);
load->num_components = num_components;
@@ -1079,6 +1077,7 @@
static void
build_explicit_io_store(nir_builder *b, nir_intrinsic_instr *intrin,
nir_ssa_def *addr, nir_address_format addr_format,
+ uint32_t align_mul, uint32_t align_offset,
nir_ssa_def *value, nir_component_mask_t write_mask)
{
nir_variable_mode mode = nir_src_as_deref(intrin->src[0])->mode;
@@ -1145,10 +1144,7 @@
if (nir_intrinsic_has_access(store))
nir_intrinsic_set_access(store, nir_intrinsic_access(intrin));
- /* TODO: We should try and provide a better alignment. For OpenCL, we need
- * to plumb the alignment through from SPIR-V when we have one.
- */
- nir_intrinsic_set_align(store, value->bit_size / 8, 0);
+ nir_intrinsic_set_align(store, align_mul, align_offset);
assert(value->num_components == 1 ||
value->num_components == intrin->num_components);
@@ -1301,19 +1297,31 @@
assert(vec_stride == 0 || glsl_type_is_vector(deref->type));
assert(vec_stride == 0 || vec_stride >= scalar_size);
+ uint32_t align_mul, align_offset;
+ if (!nir_get_explicit_deref_align(deref, true, &align_mul, &align_offset)) {
+ /* If we don't have an alignment from the deref, assume scalar */
+ align_mul = scalar_size;
+ align_offset = 0;
+ }
+
if (intrin->intrinsic == nir_intrinsic_load_deref) {
nir_ssa_def *value;
if (vec_stride > scalar_size) {
nir_ssa_def *comps[4] = { NULL, };
for (unsigned i = 0; i < intrin->num_components; i++) {
+ unsigned comp_offset = i * vec_stride;
nir_ssa_def *comp_addr = build_addr_iadd_imm(b, addr, addr_format,
- vec_stride * i);
+ comp_offset);
comps[i] = build_explicit_io_load(b, intrin, comp_addr,
- addr_format, 1);
+ addr_format, align_mul,
+ (align_offset + comp_offset) %
+ align_mul,
+ 1);
}
value = nir_vec(b, comps, intrin->num_components);
} else {
value = build_explicit_io_load(b, intrin, addr, addr_format,
+ align_mul, align_offset,
intrin->num_components);
}
nir_ssa_def_rewrite_uses(&intrin->dest.ssa, nir_src_for_ssa(value));
@@ -1326,13 +1334,17 @@
if (!(write_mask & (1 << i)))
continue;
+ unsigned comp_offset = i * vec_stride;
nir_ssa_def *comp_addr = build_addr_iadd_imm(b, addr, addr_format,
- vec_stride * i);
+ comp_offset);
build_explicit_io_store(b, intrin, comp_addr, addr_format,
+ align_mul,
+ (align_offset + comp_offset) % align_mul,
nir_channel(b, value, i), 1);
}
} else {
build_explicit_io_store(b, intrin, addr, addr_format,
+ align_mul, align_offset,
value, write_mask);
}
} else {
