glsl: do vs attribute validation in NIR linker

This allows us to tidy up the code and call the attribute location
code a single time rather than doing a "dry run".

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24035>
This commit is contained in:
Timothy Arceri 2023-07-07 15:45:04 +10:00
parent 9cbdf5c9e5
commit 80c001013c
2 changed files with 22 additions and 520 deletions

View file

@ -1082,6 +1082,12 @@ prelink_lowering(const struct gl_constants *consts,
consts->ShaderCompilerOptions[shader->Stage].NirOptions; consts->ShaderCompilerOptions[shader->Stage].NirOptions;
struct gl_program *prog = shader->Program; struct gl_program *prog = shader->Program;
/* ES vertex shaders still have dead varyings but its now safe to remove
* them as validation is now done according to the spec.
*/
if (shader_program->IsES && i == MESA_SHADER_VERTEX)
remove_dead_varyings_pre_linking(prog->nir);
preprocess_shader(consts, exts, prog, shader_program, shader->Stage); preprocess_shader(consts, exts, prog, shader_program, shader->Stage);
if (prog->nir->info.shared_size > consts->MaxComputeSharedMemorySize) { if (prog->nir->info.shared_size > consts->MaxComputeSharedMemorySize) {
@ -1328,7 +1334,22 @@ gl_nir_link_glsl(const struct gl_constants *consts,
if (prog->_LinkedShaders[i]) { if (prog->_LinkedShaders[i]) {
linked_shader[num_shaders++] = prog->_LinkedShaders[i]; linked_shader[num_shaders++] = prog->_LinkedShaders[i];
remove_dead_varyings_pre_linking(prog->_LinkedShaders[i]->Program->nir); /* Section 13.46 (Vertex Attribute Aliasing) of the OpenGL ES 3.2
* specification says:
*
* "In general, the behavior of GLSL ES should not depend on
* compiler optimizations which might be implementation-dependent.
* Name matching rules in most languages, including C++ from which
* GLSL ES is derived, are based on declarations rather than use.
*
* RESOLUTION: The existence of aliasing is determined by
* declarations present after preprocessing."
*
* Because of this rule, we don't remove dead attributes before
* attribute assignment for vertex shader inputs here.
*/
if (!(prog->IsES && i == MESA_SHADER_VERTEX))
remove_dead_varyings_pre_linking(prog->_LinkedShaders[i]->Program->nir);
} }
} }

View file

@ -2590,504 +2590,6 @@ resize_tes_inputs(const struct gl_constants *consts,
} }
} }
/**
* Find a contiguous set of available bits in a bitmask.
*
* \param used_mask Bits representing used (1) and unused (0) locations
* \param needed_count Number of contiguous bits needed.
*
* \return
* Base location of the available bits on success or -1 on failure.
*/
static int
find_available_slots(unsigned used_mask, unsigned needed_count)
{
unsigned needed_mask = (1 << needed_count) - 1;
const int max_bit_to_test = (8 * sizeof(used_mask)) - needed_count;
/* The comparison to 32 is redundant, but without it GCC emits "warning:
* cannot optimize possibly infinite loops" for the loop below.
*/
if ((needed_count == 0) || (max_bit_to_test < 0) || (max_bit_to_test > 32))
return -1;
for (int i = 0; i <= max_bit_to_test; i++) {
if ((needed_mask & ~used_mask) == needed_mask)
return i;
needed_mask <<= 1;
}
return -1;
}
#define SAFE_MASK_FROM_INDEX(i) (((i) >= 32) ? ~0 : ((1 << (i)) - 1))
/**
* Assign locations for either VS inputs or FS outputs.
*
* \param mem_ctx Temporary ralloc context used for linking.
* \param prog Shader program whose variables need locations
* assigned.
* \param constants Driver specific constant values for the program.
* \param target_index Selector for the program target to receive location
* assignmnets. Must be either \c MESA_SHADER_VERTEX or
* \c MESA_SHADER_FRAGMENT.
* \param do_assignment Whether we are actually marking the assignment or we
* are just doing a dry-run checking.
*
* \return
* If locations are (or can be, in case of dry-running) successfully assigned,
* true is returned. Otherwise an error is emitted to the shader link log and
* false is returned.
*/
static bool
assign_attribute_or_color_locations(void *mem_ctx,
gl_shader_program *prog,
const struct gl_constants *constants,
unsigned target_index,
bool do_assignment)
{
/* Maximum number of generic locations. This corresponds to either the
* maximum number of draw buffers or the maximum number of generic
* attributes.
*/
unsigned max_index = (target_index == MESA_SHADER_VERTEX) ?
constants->Program[target_index].MaxAttribs :
MAX2(constants->MaxDrawBuffers, constants->MaxDualSourceDrawBuffers);
/* Mark invalid locations as being used.
*/
unsigned used_locations = ~SAFE_MASK_FROM_INDEX(max_index);
unsigned double_storage_locations = 0;
assert((target_index == MESA_SHADER_VERTEX)
|| (target_index == MESA_SHADER_FRAGMENT));
gl_linked_shader *const sh = prog->_LinkedShaders[target_index];
if (sh == NULL)
return true;
/* Operate in a total of four passes.
*
* 1. Invalidate the location assignments for all vertex shader inputs.
*
* 2. Assign locations for inputs that have user-defined (via
* glBindVertexAttribLocation) locations and outputs that have
* user-defined locations (via glBindFragDataLocation).
*
* 3. Sort the attributes without assigned locations by number of slots
* required in decreasing order. Fragmentation caused by attribute
* locations assigned by the application may prevent large attributes
* from having enough contiguous space.
*
* 4. Assign locations to any inputs without assigned locations.
*/
const int generic_base = (target_index == MESA_SHADER_VERTEX)
? (int) VERT_ATTRIB_GENERIC0 : (int) FRAG_RESULT_DATA0;
const enum ir_variable_mode direction =
(target_index == MESA_SHADER_VERTEX)
? ir_var_shader_in : ir_var_shader_out;
/* Temporary storage for the set of attributes that need locations assigned.
*/
struct temp_attr {
unsigned slots;
ir_variable *var;
/* Used below in the call to qsort. */
static int compare(const void *a, const void *b)
{
const temp_attr *const l = (const temp_attr *) a;
const temp_attr *const r = (const temp_attr *) b;
/* Reversed because we want a descending order sort below. */
return r->slots - l->slots;
}
} to_assign[32];
assert(max_index <= 32);
/* Temporary array for the set of attributes that have locations assigned,
* for the purpose of checking overlapping slots/components of (non-ES)
* fragment shader outputs.
*/
ir_variable *assigned[12 * 4]; /* (max # of FS outputs) * # components */
unsigned assigned_attr = 0;
unsigned num_attr = 0;
foreach_in_list(ir_instruction, node, sh->ir) {
ir_variable *const var = node->as_variable();
if ((var == NULL) || (var->data.mode != (unsigned) direction))
continue;
if (var->data.explicit_location) {
if ((var->data.location >= (int)(max_index + generic_base))
|| (var->data.location < 0)) {
linker_error(prog,
"invalid explicit location %d specified for `%s'\n",
(var->data.location < 0)
? var->data.location
: var->data.location - generic_base,
var->name);
return false;
}
} else if (target_index == MESA_SHADER_VERTEX) {
unsigned binding;
if (prog->AttributeBindings->get(binding, var->name)) {
assert(binding >= VERT_ATTRIB_GENERIC0);
var->data.location = binding;
}
} else if (target_index == MESA_SHADER_FRAGMENT) {
unsigned binding;
unsigned index;
const char *name = var->name;
const glsl_type *type = var->type;
while (type) {
/* Check if there's a binding for the variable name */
if (prog->FragDataBindings->get(binding, name)) {
assert(binding >= FRAG_RESULT_DATA0);
var->data.location = binding;
if (prog->FragDataIndexBindings->get(index, name)) {
var->data.index = index;
}
break;
}
/* If not, but it's an array type, look for name[0] */
if (type->is_array()) {
name = ralloc_asprintf(mem_ctx, "%s[0]", name);
type = type->fields.array;
continue;
}
break;
}
}
if (strcmp(var->name, "gl_LastFragData") == 0)
continue;
/* From GL4.5 core spec, section 15.2 (Shader Execution):
*
* "Output binding assignments will cause LinkProgram to fail:
* ...
* If the program has an active output assigned to a location greater
* than or equal to the value of MAX_DUAL_SOURCE_DRAW_BUFFERS and has
* an active output assigned an index greater than or equal to one;"
*/
if (target_index == MESA_SHADER_FRAGMENT && var->data.index >= 1 &&
var->data.location - generic_base >=
(int) constants->MaxDualSourceDrawBuffers) {
linker_error(prog,
"output location %d >= GL_MAX_DUAL_SOURCE_DRAW_BUFFERS "
"with index %u for %s\n",
var->data.location - generic_base, var->data.index,
var->name);
return false;
}
const unsigned slots = var->type->count_attribute_slots(target_index == MESA_SHADER_VERTEX);
/* If the variable is not a built-in and has a location statically
* assigned in the shader (presumably via a layout qualifier), make sure
* that it doesn't collide with other assigned locations. Otherwise,
* add it to the list of variables that need linker-assigned locations.
*/
if (var->data.location != -1) {
if (var->data.location >= generic_base && var->data.index < 1) {
/* From page 61 of the OpenGL 4.0 spec:
*
* "LinkProgram will fail if the attribute bindings assigned
* by BindAttribLocation do not leave not enough space to
* assign a location for an active matrix attribute or an
* active attribute array, both of which require multiple
* contiguous generic attributes."
*
* I think above text prohibits the aliasing of explicit and
* automatic assignments. But, aliasing is allowed in manual
* assignments of attribute locations. See below comments for
* the details.
*
* From OpenGL 4.0 spec, page 61:
*
* "It is possible for an application to bind more than one
* attribute name to the same location. This is referred to as
* aliasing. This will only work if only one of the aliased
* attributes is active in the executable program, or if no
* path through the shader consumes more than one attribute of
* a set of attributes aliased to the same location. A link
* error can occur if the linker determines that every path
* through the shader consumes multiple aliased attributes,
* but implementations are not required to generate an error
* in this case."
*
* From GLSL 4.30 spec, page 54:
*
* "A program will fail to link if any two non-vertex shader
* input variables are assigned to the same location. For
* vertex shaders, multiple input variables may be assigned
* to the same location using either layout qualifiers or via
* the OpenGL API. However, such aliasing is intended only to
* support vertex shaders where each execution path accesses
* at most one input per each location. Implementations are
* permitted, but not required, to generate link-time errors
* if they detect that every path through the vertex shader
* executable accesses multiple inputs assigned to any single
* location. For all shader types, a program will fail to link
* if explicit location assignments leave the linker unable
* to find space for other variables without explicit
* assignments."
*
* From OpenGL ES 3.0 spec, page 56:
*
* "Binding more than one attribute name to the same location
* is referred to as aliasing, and is not permitted in OpenGL
* ES Shading Language 3.00 vertex shaders. LinkProgram will
* fail when this condition exists. However, aliasing is
* possible in OpenGL ES Shading Language 1.00 vertex shaders.
* This will only work if only one of the aliased attributes
* is active in the executable program, or if no path through
* the shader consumes more than one attribute of a set of
* attributes aliased to the same location. A link error can
* occur if the linker determines that every path through the
* shader consumes multiple aliased attributes, but implemen-
* tations are not required to generate an error in this case."
*
* After looking at above references from OpenGL, OpenGL ES and
* GLSL specifications, we allow aliasing of vertex input variables
* in: OpenGL 2.0 (and above) and OpenGL ES 2.0.
*
* NOTE: This is not required by the spec but its worth mentioning
* here that we're not doing anything to make sure that no path
* through the vertex shader executable accesses multiple inputs
* assigned to any single location.
*/
/* Mask representing the contiguous slots that will be used by
* this attribute.
*/
const unsigned attr = var->data.location - generic_base;
const unsigned use_mask = (1 << slots) - 1;
const char *const string = (target_index == MESA_SHADER_VERTEX)
? "vertex shader input" : "fragment shader output";
/* Generate a link error if the requested locations for this
* attribute exceed the maximum allowed attribute location.
*/
if (attr + slots > max_index) {
linker_error(prog,
"insufficient contiguous locations "
"available for %s `%s' %d %d %d\n", string,
var->name, used_locations, use_mask, attr);
return false;
}
/* Generate a link error if the set of bits requested for this
* attribute overlaps any previously allocated bits.
*/
if ((~(use_mask << attr) & used_locations) != used_locations) {
if (target_index == MESA_SHADER_FRAGMENT && !prog->IsES) {
/* From section 4.4.2 (Output Layout Qualifiers) of the GLSL
* 4.40 spec:
*
* "Additionally, for fragment shader outputs, if two
* variables are placed within the same location, they
* must have the same underlying type (floating-point or
* integer). No component aliasing of output variables or
* members is allowed.
*/
for (unsigned i = 0; i < assigned_attr; i++) {
unsigned assigned_slots =
assigned[i]->type->count_attribute_slots(false);
unsigned assig_attr =
assigned[i]->data.location - generic_base;
unsigned assigned_use_mask = (1 << assigned_slots) - 1;
if ((assigned_use_mask << assig_attr) &
(use_mask << attr)) {
const glsl_type *assigned_type =
assigned[i]->type->without_array();
const glsl_type *type = var->type->without_array();
if (assigned_type->base_type != type->base_type) {
linker_error(prog, "types do not match for aliased"
" %ss %s and %s\n", string,
assigned[i]->name, var->name);
return false;
}
unsigned assigned_component_mask =
((1 << assigned_type->vector_elements) - 1) <<
assigned[i]->data.location_frac;
unsigned component_mask =
((1 << type->vector_elements) - 1) <<
var->data.location_frac;
if (assigned_component_mask & component_mask) {
linker_error(prog, "overlapping component is "
"assigned to %ss %s and %s "
"(component=%d)\n",
string, assigned[i]->name, var->name,
var->data.location_frac);
return false;
}
}
}
} else if (target_index == MESA_SHADER_FRAGMENT ||
(prog->IsES && prog->GLSL_Version >= 300)) {
linker_error(prog, "overlapping location is assigned "
"to %s `%s' %d %d %d\n", string, var->name,
used_locations, use_mask, attr);
return false;
} else {
linker_warning(prog, "overlapping location is assigned "
"to %s `%s' %d %d %d\n", string, var->name,
used_locations, use_mask, attr);
}
}
if (target_index == MESA_SHADER_FRAGMENT && !prog->IsES) {
/* Only track assigned variables for non-ES fragment shaders
* to avoid overflowing the array.
*
* At most one variable per fragment output component should
* reach this.
*/
assert(assigned_attr < ARRAY_SIZE(assigned));
assigned[assigned_attr] = var;
assigned_attr++;
}
used_locations |= (use_mask << attr);
/* From the GL 4.5 core spec, section 11.1.1 (Vertex Attributes):
*
* "A program with more than the value of MAX_VERTEX_ATTRIBS
* active attribute variables may fail to link, unless
* device-dependent optimizations are able to make the program
* fit within available hardware resources. For the purposes
* of this test, attribute variables of the type dvec3, dvec4,
* dmat2x3, dmat2x4, dmat3, dmat3x4, dmat4x3, and dmat4 may
* count as consuming twice as many attributes as equivalent
* single-precision types. While these types use the same number
* of generic attributes as their single-precision equivalents,
* implementations are permitted to consume two single-precision
* vectors of internal storage for each three- or four-component
* double-precision vector."
*
* Mark this attribute slot as taking up twice as much space
* so we can count it properly against limits. According to
* issue (3) of the GL_ARB_vertex_attrib_64bit behavior, this
* is optional behavior, but it seems preferable.
*/
if (var->type->without_array()->is_dual_slot())
double_storage_locations |= (use_mask << attr);
}
continue;
}
if (num_attr >= max_index) {
linker_error(prog, "too many %s (max %u)",
target_index == MESA_SHADER_VERTEX ?
"vertex shader inputs" : "fragment shader outputs",
max_index);
return false;
}
to_assign[num_attr].slots = slots;
to_assign[num_attr].var = var;
num_attr++;
}
if (!do_assignment)
return true;
if (target_index == MESA_SHADER_VERTEX) {
unsigned total_attribs_size =
util_bitcount(used_locations & SAFE_MASK_FROM_INDEX(max_index)) +
util_bitcount(double_storage_locations);
if (total_attribs_size > max_index) {
linker_error(prog,
"attempt to use %d vertex attribute slots only %d available ",
total_attribs_size, max_index);
return false;
}
}
/* If all of the attributes were assigned locations by the application (or
* are built-in attributes with fixed locations), return early. This should
* be the common case.
*/
if (num_attr == 0)
return true;
qsort(to_assign, num_attr, sizeof(to_assign[0]), temp_attr::compare);
if (target_index == MESA_SHADER_VERTEX) {
/* VERT_ATTRIB_GENERIC0 is a pseudo-alias for VERT_ATTRIB_POS. It can
* only be explicitly assigned by via glBindAttribLocation. Mark it as
* reserved to prevent it from being automatically allocated below.
*/
find_deref_visitor find("gl_Vertex");
find.run(sh->ir);
if (find.variable_found())
used_locations |= (1 << 0);
}
for (unsigned i = 0; i < num_attr; i++) {
/* Mask representing the contiguous slots that will be used by this
* attribute.
*/
const unsigned use_mask = (1 << to_assign[i].slots) - 1;
int location = find_available_slots(used_locations, to_assign[i].slots);
if (location < 0) {
const char *const string = (target_index == MESA_SHADER_VERTEX)
? "vertex shader input" : "fragment shader output";
linker_error(prog,
"insufficient contiguous locations "
"available for %s `%s'\n",
string, to_assign[i].var->name);
return false;
}
to_assign[i].var->data.location = generic_base + location;
used_locations |= (use_mask << location);
if (to_assign[i].var->type->without_array()->is_dual_slot())
double_storage_locations |= (use_mask << location);
}
/* Now that we have all the locations, from the GL 4.5 core spec, section
* 11.1.1 (Vertex Attributes), dvec3, dvec4, dmat2x3, dmat2x4, dmat3,
* dmat3x4, dmat4x3, and dmat4 count as consuming twice as many attributes
* as equivalent single-precision types.
*/
if (target_index == MESA_SHADER_VERTEX) {
unsigned total_attribs_size =
util_bitcount(used_locations & SAFE_MASK_FROM_INDEX(max_index)) +
util_bitcount(double_storage_locations);
if (total_attribs_size > max_index) {
linker_error(prog,
"attempt to use %d vertex attribute slots only %d available ",
total_attribs_size, max_index);
return false;
}
}
return true;
}
/** /**
* Initializes explicit location slots to INACTIVE_UNIFORM_EXPLICIT_LOCATION * Initializes explicit location slots to INACTIVE_UNIFORM_EXPLICIT_LOCATION
@ -3678,27 +3180,6 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
if (consts->ShaderCompilerOptions[i].LowerCombinedClipCullDistance) { if (consts->ShaderCompilerOptions[i].LowerCombinedClipCullDistance) {
lower_clip_cull_distance(prog, prog->_LinkedShaders[i]); lower_clip_cull_distance(prog, prog->_LinkedShaders[i]);
} }
/* Section 13.46 (Vertex Attribute Aliasing) of the OpenGL ES 3.2
* specification says:
*
* "In general, the behavior of GLSL ES should not depend on compiler
* optimizations which might be implementation-dependent. Name matching
* rules in most languages, including C++ from which GLSL ES is derived,
* are based on declarations rather than use.
*
* RESOLUTION: The existence of aliasing is determined by declarations
* present after preprocessing."
*
* Because of this rule, we do a 'dry-run' of attribute assignment for
* vertex shader inputs here.
*/
if (prog->IsES && i == MESA_SHADER_VERTEX) {
if (!assign_attribute_or_color_locations(mem_ctx, prog, consts,
MESA_SHADER_VERTEX, false)) {
goto done;
}
}
} }
/* Check and validate stream emissions in geometry shaders */ /* Check and validate stream emissions in geometry shaders */