#include <stdarg.h>
#include <stdio.h>
-#include "main/core.h" /* for struct gl_shader */
+#include "main/mtypes.h"
#include "main/shaderobj.h"
#include "ir_builder.h"
#include "glsl_parser_extras.h"
compatibility_vs_only(const _mesa_glsl_parse_state *state)
{
return state->stage == MESA_SHADER_VERTEX &&
- state->language_version <= 130 &&
+ (state->compat_shader || state->ARB_compatibility_enable) &&
!state->es_shader;
}
return state->is_version(130, 0);
}
+static bool
+v460_desktop(const _mesa_glsl_parse_state *state)
+{
+ return state->is_version(460, 0);
+}
+
static bool
v130_fs_only(const _mesa_glsl_parse_state *state)
{
return state->OES_EGL_image_external_enable;
}
+static bool
+texture_external_es3(const _mesa_glsl_parse_state *state)
+{
+ return state->OES_EGL_image_external_essl3_enable &&
+ state->es_shader &&
+ state->is_version(0, 300);
+}
+
/** True if texturing functions with explicit LOD are allowed. */
static bool
lod_exists_in_stage(const _mesa_glsl_parse_state *state)
{
return state->stage == MESA_SHADER_FRAGMENT &&
(state->is_version(110, 300) ||
- state->OES_standard_derivatives_enable);
+ state->OES_standard_derivatives_enable ||
+ state->ctx->Const.AllowGLSLRelaxedES);
}
static bool
return state->ARB_shader_atomic_counter_ops_enable;
}
+static bool
+shader_atomic_counter_ops_or_v460_desktop(const _mesa_glsl_parse_state *state)
+{
+ return state->ARB_shader_atomic_counter_ops_enable || v460_desktop(state);
+}
+
static bool
shader_ballot(const _mesa_glsl_parse_state *state)
{
return state->ARB_shader_ballot_enable;
}
+static bool
+supports_arb_fragment_shader_interlock(const _mesa_glsl_parse_state *state)
+{
+ return state->ARB_fragment_shader_interlock_enable;
+}
+
+static bool
+supports_nv_fragment_shader_interlock(const _mesa_glsl_parse_state *state)
+{
+ return state->NV_fragment_shader_interlock_enable;
+}
+
+static bool
+supports_intel_fragment_shader_ordering(const _mesa_glsl_parse_state *state)
+{
+ return state->INTEL_fragment_shader_ordering_enable;
+}
+
static bool
shader_clock(const _mesa_glsl_parse_state *state)
{
shader_clock_int64(const _mesa_glsl_parse_state *state)
{
return state->ARB_shader_clock_enable &&
- state->ARB_gpu_shader_int64_enable;
+ (state->ARB_gpu_shader_int64_enable ||
+ state->AMD_gpu_shader_int64_enable);
}
static bool
{
return (state->is_version(450, 320) ||
state->ARB_ES3_1_compatibility_enable ||
- state->OES_shader_image_atomic_enable);
+ state->OES_shader_image_atomic_enable ||
+ state->NV_shader_atomic_float_enable);
+}
+
+static bool
+shader_image_atomic_add_float(const _mesa_glsl_parse_state *state)
+{
+ return state->NV_shader_atomic_float_enable;
}
static bool
return state->ARB_shader_group_vote_enable;
}
+static bool
+vote_or_v460_desktop(const _mesa_glsl_parse_state *state)
+{
+ return state->ARB_shader_group_vote_enable || v460_desktop(state);
+}
+
static bool
integer_functions_supported(const _mesa_glsl_parse_state *state)
{
return state->extensions->MESA_shader_integer_functions;
}
+
+static bool
+NV_shader_atomic_float_supported(const _mesa_glsl_parse_state *state)
+{
+ return state->extensions->NV_shader_atomic_float;
+}
+
+static bool
+shader_atomic_float_add(const _mesa_glsl_parse_state *state)
+{
+ return state->NV_shader_atomic_float_enable;
+}
+
+static bool
+shader_atomic_float_exchange(const _mesa_glsl_parse_state *state)
+{
+ return state->NV_shader_atomic_float_enable ||
+ state->INTEL_shader_atomic_float_minmax_enable;
+}
+
+static bool
+INTEL_shader_atomic_float_minmax_supported(const _mesa_glsl_parse_state *state)
+{
+ return state->extensions->INTEL_shader_atomic_float_minmax;
+}
+
+static bool
+shader_atomic_float_minmax(const _mesa_glsl_parse_state *state)
+{
+ return state->INTEL_shader_atomic_float_minmax_enable;
+}
/** @} */
/******************************************************************************/
ir_expression_operation opcode,
const glsl_type *return_type,
const glsl_type *param0_type,
- const glsl_type *param1_type);
+ const glsl_type *param1_type,
+ bool swap_operands = false);
#define B0(X) ir_function_signature *_##X();
#define B1(X) ir_function_signature *_##X(const glsl_type *);
ir_function_signature *_read_invocation_intrinsic(const glsl_type *type);
ir_function_signature *_read_invocation(const glsl_type *type);
+
+ ir_function_signature *_invocation_interlock_intrinsic(
+ builtin_available_predicate avail,
+ enum ir_intrinsic_id id);
+ ir_function_signature *_invocation_interlock(
+ const char *intrinsic_name,
+ builtin_available_predicate avail);
+
ir_function_signature *_shader_clock_intrinsic(builtin_available_predicate avail,
const glsl_type *type);
ir_function_signature *_shader_clock(builtin_available_predicate avail,
ir_function_signature *_vote_intrinsic(builtin_available_predicate avail,
enum ir_intrinsic_id id);
- ir_function_signature *_vote(const char *intrinsic_name);
+ ir_function_signature *_vote(const char *intrinsic_name,
+ builtin_available_predicate avail);
#undef B0
#undef B1
IMAGE_FUNCTION_WRITE_ONLY = (1 << 5),
IMAGE_FUNCTION_AVAIL_ATOMIC = (1 << 6),
IMAGE_FUNCTION_MS_ONLY = (1 << 7),
- IMAGE_FUNCTION_AVAIL_ATOMIC_EXCHANGE = (1 << 8)
+ IMAGE_FUNCTION_AVAIL_ATOMIC_EXCHANGE = (1 << 8),
+ IMAGE_FUNCTION_AVAIL_ATOMIC_ADD = (1 << 9),
};
} /* anonymous namespace */
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_add),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_intrinsic2(NV_shader_atomic_float_supported,
+ glsl_type::float_type,
+ ir_intrinsic_generic_atomic_add),
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_add),
NULL);
add_function("__intrinsic_atomic_min",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_min),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_intrinsic2(INTEL_shader_atomic_float_minmax_supported,
+ glsl_type::float_type,
+ ir_intrinsic_generic_atomic_min),
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_min),
NULL);
add_function("__intrinsic_atomic_max",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_max),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_intrinsic2(INTEL_shader_atomic_float_minmax_supported,
+ glsl_type::float_type,
+ ir_intrinsic_generic_atomic_max),
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_max),
NULL);
add_function("__intrinsic_atomic_and",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_and),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_and),
NULL);
add_function("__intrinsic_atomic_or",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_or),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_or),
NULL);
add_function("__intrinsic_atomic_xor",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_xor),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_xor),
NULL);
add_function("__intrinsic_atomic_exchange",
_atomic_intrinsic2(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_exchange),
- _atomic_counter_intrinsic1(shader_atomic_counter_ops,
+ _atomic_intrinsic2(NV_shader_atomic_float_supported,
+ glsl_type::float_type,
+ ir_intrinsic_generic_atomic_exchange),
+ _atomic_counter_intrinsic1(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_exchange),
NULL);
add_function("__intrinsic_atomic_comp_swap",
_atomic_intrinsic3(buffer_atomics_supported,
glsl_type::int_type,
ir_intrinsic_generic_atomic_comp_swap),
- _atomic_counter_intrinsic2(shader_atomic_counter_ops,
+ _atomic_intrinsic3(INTEL_shader_atomic_float_minmax_supported,
+ glsl_type::float_type,
+ ir_intrinsic_generic_atomic_comp_swap),
+ _atomic_counter_intrinsic2(shader_atomic_counter_ops_or_v460_desktop,
ir_intrinsic_atomic_counter_comp_swap),
NULL);
ir_intrinsic_memory_barrier_shared),
NULL);
+ add_function("__intrinsic_begin_invocation_interlock",
+ _invocation_interlock_intrinsic(
+ supports_arb_fragment_shader_interlock,
+ ir_intrinsic_begin_invocation_interlock), NULL);
+
+ add_function("__intrinsic_end_invocation_interlock",
+ _invocation_interlock_intrinsic(
+ supports_arb_fragment_shader_interlock,
+ ir_intrinsic_end_invocation_interlock), NULL);
+
+ add_function("__intrinsic_begin_fragment_shader_ordering",
+ _invocation_interlock_intrinsic(
+ supports_intel_fragment_shader_ordering,
+ ir_intrinsic_begin_fragment_shader_ordering), NULL);
+
add_function("__intrinsic_shader_clock",
_shader_clock_intrinsic(shader_clock,
glsl_type::uvec2_type),
NULL);
add_function("__intrinsic_vote_all",
- _vote_intrinsic(vote, ir_intrinsic_vote_all),
+ _vote_intrinsic(vote_or_v460_desktop, ir_intrinsic_vote_all),
NULL);
add_function("__intrinsic_vote_any",
- _vote_intrinsic(vote, ir_intrinsic_vote_any),
+ _vote_intrinsic(vote_or_v460_desktop, ir_intrinsic_vote_any),
NULL);
add_function("__intrinsic_vote_eq",
- _vote_intrinsic(vote, ir_intrinsic_vote_eq),
+ _vote_intrinsic(vote_or_v460_desktop, ir_intrinsic_vote_eq),
NULL);
add_function("__intrinsic_ballot", _ballot_intrinsic(), NULL);
_texture(ir_tex, v130, glsl_type::float_type, glsl_type::sampler2DRectShadow_type, glsl_type::vec3_type),
+ _texture(ir_tex, texture_external_es3, glsl_type::vec4_type, glsl_type::samplerExternalOES_type, glsl_type::vec2_type),
+
_texture(ir_txb, v130_fs_only, glsl_type::vec4_type, glsl_type::sampler1D_type, glsl_type::float_type),
_texture(ir_txb, v130_fs_only, glsl_type::ivec4_type, glsl_type::isampler1D_type, glsl_type::float_type),
_texture(ir_txb, v130_fs_only, glsl_type::uvec4_type, glsl_type::usampler1D_type, glsl_type::float_type),
_texture(ir_tex, v130, glsl_type::vec4_type, glsl_type::sampler2DRect_type, glsl_type::vec3_type, TEX_PROJECT),
_texture(ir_tex, v130, glsl_type::ivec4_type, glsl_type::isampler2DRect_type, glsl_type::vec3_type, TEX_PROJECT),
+ _texture(ir_tex, texture_external_es3, glsl_type::vec4_type, glsl_type::samplerExternalOES_type, glsl_type::vec3_type, TEX_PROJECT),
+ _texture(ir_tex, texture_external_es3, glsl_type::vec4_type, glsl_type::samplerExternalOES_type, glsl_type::vec4_type, TEX_PROJECT),
+
_texture(ir_tex, v130, glsl_type::uvec4_type, glsl_type::usampler2DRect_type, glsl_type::vec3_type, TEX_PROJECT),
_texture(ir_tex, v130, glsl_type::vec4_type, glsl_type::sampler2DRect_type, glsl_type::vec4_type, TEX_PROJECT),
_texture(ir_tex, v130, glsl_type::ivec4_type, glsl_type::isampler2DRect_type, glsl_type::vec4_type, TEX_PROJECT),
_texelFetch(texture_multisample_array, glsl_type::vec4_type, glsl_type::sampler2DMSArray_type, glsl_type::ivec3_type),
_texelFetch(texture_multisample_array, glsl_type::ivec4_type, glsl_type::isampler2DMSArray_type, glsl_type::ivec3_type),
_texelFetch(texture_multisample_array, glsl_type::uvec4_type, glsl_type::usampler2DMSArray_type, glsl_type::ivec3_type),
+
+ _texelFetch(texture_external_es3, glsl_type::vec4_type, glsl_type::samplerExternalOES_type, glsl_type::ivec2_type),
+
NULL);
+
add_function("texelFetchOffset",
_texelFetch(v130, glsl_type::vec4_type, glsl_type::sampler1D_type, glsl_type::int_type, glsl_type::int_type),
_texelFetch(v130, glsl_type::ivec4_type, glsl_type::isampler1D_type, glsl_type::int_type, glsl_type::int_type),
shader_atomic_counter_ops),
NULL);
+ add_function("atomicCounterAdd",
+ _atomic_counter_op1("__intrinsic_atomic_add",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterSubtract",
+ _atomic_counter_op1("__intrinsic_atomic_sub",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterMin",
+ _atomic_counter_op1("__intrinsic_atomic_min",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterMax",
+ _atomic_counter_op1("__intrinsic_atomic_max",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterAnd",
+ _atomic_counter_op1("__intrinsic_atomic_and",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterOr",
+ _atomic_counter_op1("__intrinsic_atomic_or",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterXor",
+ _atomic_counter_op1("__intrinsic_atomic_xor",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterExchange",
+ _atomic_counter_op1("__intrinsic_atomic_exchange",
+ v460_desktop),
+ NULL);
+ add_function("atomicCounterCompSwap",
+ _atomic_counter_op2("__intrinsic_atomic_comp_swap",
+ v460_desktop),
+ NULL);
+
add_function("atomicAdd",
_atomic_op2("__intrinsic_atomic_add",
buffer_atomics_supported,
_atomic_op2("__intrinsic_atomic_add",
buffer_atomics_supported,
glsl_type::int_type),
+ _atomic_op2("__intrinsic_atomic_add",
+ shader_atomic_float_add,
+ glsl_type::float_type),
NULL);
add_function("atomicMin",
_atomic_op2("__intrinsic_atomic_min",
_atomic_op2("__intrinsic_atomic_min",
buffer_atomics_supported,
glsl_type::int_type),
+ _atomic_op2("__intrinsic_atomic_min",
+ shader_atomic_float_minmax,
+ glsl_type::float_type),
NULL);
add_function("atomicMax",
_atomic_op2("__intrinsic_atomic_max",
_atomic_op2("__intrinsic_atomic_max",
buffer_atomics_supported,
glsl_type::int_type),
+ _atomic_op2("__intrinsic_atomic_max",
+ shader_atomic_float_minmax,
+ glsl_type::float_type),
NULL);
add_function("atomicAnd",
_atomic_op2("__intrinsic_atomic_and",
_atomic_op2("__intrinsic_atomic_exchange",
buffer_atomics_supported,
glsl_type::int_type),
+ _atomic_op2("__intrinsic_atomic_exchange",
+ shader_atomic_float_exchange,
+ glsl_type::float_type),
NULL);
add_function("atomicCompSwap",
_atomic_op3("__intrinsic_atomic_comp_swap",
_atomic_op3("__intrinsic_atomic_comp_swap",
buffer_atomics_supported,
glsl_type::int_type),
+ _atomic_op3("__intrinsic_atomic_comp_swap",
+ shader_atomic_float_minmax,
+ glsl_type::float_type),
NULL);
add_function("min3",
glsl_type::uint64_t_type),
NULL);
- add_function("anyInvocationARB", _vote("__intrinsic_vote_any"), NULL);
- add_function("allInvocationsARB", _vote("__intrinsic_vote_all"), NULL);
- add_function("allInvocationsEqualARB", _vote("__intrinsic_vote_eq"), NULL);
+ add_function("beginInvocationInterlockARB",
+ _invocation_interlock(
+ "__intrinsic_begin_invocation_interlock",
+ supports_arb_fragment_shader_interlock),
+ NULL);
+
+ add_function("endInvocationInterlockARB",
+ _invocation_interlock(
+ "__intrinsic_end_invocation_interlock",
+ supports_arb_fragment_shader_interlock),
+ NULL);
+
+ add_function("beginInvocationInterlockNV",
+ _invocation_interlock(
+ "__intrinsic_begin_invocation_interlock",
+ supports_nv_fragment_shader_interlock),
+ NULL);
+
+ add_function("endInvocationInterlockNV",
+ _invocation_interlock(
+ "__intrinsic_end_invocation_interlock",
+ supports_nv_fragment_shader_interlock),
+ NULL);
+
+ add_function("beginFragmentShaderOrderingINTEL",
+ _invocation_interlock(
+ "__intrinsic_begin_fragment_shader_ordering",
+ supports_intel_fragment_shader_ordering),
+ NULL);
+
+ add_function("anyInvocationARB",
+ _vote("__intrinsic_vote_any", vote),
+ NULL);
+
+ add_function("allInvocationsARB",
+ _vote("__intrinsic_vote_all", vote),
+ NULL);
+
+ add_function("allInvocationsEqualARB",
+ _vote("__intrinsic_vote_eq", vote),
+ NULL);
+
+ add_function("anyInvocation",
+ _vote("__intrinsic_vote_any", v460_desktop),
+ NULL);
+
+ add_function("allInvocations",
+ _vote("__intrinsic_vote_all", v460_desktop),
+ NULL);
+
+ add_function("allInvocationsEqual",
+ _vote("__intrinsic_vote_eq", v460_desktop),
+ NULL);
add_function("__builtin_idiv64",
generate_ir::idiv64(mem_ctx, integer_functions_supported),
add_image_function(glsl ? "imageAtomicAdd" : "__intrinsic_image_atomic_add",
"__intrinsic_image_atomic_add",
- &builtin_builder::_image_prototype, 1, atom_flags,
+ &builtin_builder::_image_prototype, 1,
+ (flags | IMAGE_FUNCTION_AVAIL_ATOMIC_ADD |
+ IMAGE_FUNCTION_SUPPORTS_FLOAT_DATA_TYPE),
ir_intrinsic_image_atomic_add);
add_image_function(glsl ? "imageAtomicMin" : "__intrinsic_image_atomic_min",
ir_expression_operation opcode,
const glsl_type *return_type,
const glsl_type *param0_type,
- const glsl_type *param1_type)
+ const glsl_type *param1_type,
+ bool swap_operands)
{
ir_variable *x = in_var(param0_type, "x");
ir_variable *y = in_var(param1_type, "y");
MAKE_SIG(return_type, avail, 2, x, y);
- body.emit(ret(expr(opcode, x, y)));
+
+ if (swap_operands)
+ body.emit(ret(expr(opcode, y, x)));
+ else
+ body.emit(ret(expr(opcode, x, y)));
+
return sig;
}
builtin_builder::_lessThanEqual(builtin_available_predicate avail,
const glsl_type *type)
{
- return binop(avail, ir_binop_lequal,
- glsl_type::bvec(type->vector_elements), type, type);
+ return binop(avail, ir_binop_gequal,
+ glsl_type::bvec(type->vector_elements), type, type,
+ true);
}
ir_function_signature *
builtin_builder::_greaterThan(builtin_available_predicate avail,
const glsl_type *type)
{
- return binop(avail, ir_binop_greater,
- glsl_type::bvec(type->vector_elements), type, type);
+ return binop(avail, ir_binop_less,
+ glsl_type::bvec(type->vector_elements), type, type,
+ true);
}
ir_function_signature *
type->sampled_type == GLSL_TYPE_FLOAT)
return shader_image_atomic_exchange_float;
+ if ((flags & IMAGE_FUNCTION_AVAIL_ATOMIC_ADD) &&
+ type->sampled_type == GLSL_TYPE_FLOAT)
+ return shader_image_atomic_add_float;
+
else if (flags & (IMAGE_FUNCTION_AVAIL_ATOMIC_EXCHANGE |
+ IMAGE_FUNCTION_AVAIL_ATOMIC_ADD |
IMAGE_FUNCTION_AVAIL_ATOMIC))
return shader_image_atomic;
return sig;
}
+ir_function_signature *
+builtin_builder::_invocation_interlock_intrinsic(builtin_available_predicate avail,
+ enum ir_intrinsic_id id)
+{
+ MAKE_INTRINSIC(glsl_type::void_type, id, avail, 0);
+ return sig;
+}
+
+ir_function_signature *
+builtin_builder::_invocation_interlock(const char *intrinsic_name,
+ builtin_available_predicate avail)
+{
+ MAKE_SIG(glsl_type::void_type, avail, 0);
+ body.emit(call(shader->symbols->get_function(intrinsic_name),
+ NULL, sig->parameters));
+ return sig;
+}
+
ir_function_signature *
builtin_builder::_shader_clock_intrinsic(builtin_available_predicate avail,
const glsl_type *type)
}
ir_function_signature *
-builtin_builder::_vote(const char *intrinsic_name)
+builtin_builder::_vote(const char *intrinsic_name,
+ builtin_available_predicate avail)
{
ir_variable *value = in_var(glsl_type::bool_type, "value");
- MAKE_SIG(glsl_type::bool_type, vote, 1, value);
+ MAKE_SIG(glsl_type::bool_type, avail, 1, value);
ir_variable *retval = body.make_temp(glsl_type::bool_type, "retval");
s = builtins.find(state, name, actual_parameters);
mtx_unlock(&builtins_lock);
- if (s == NULL)
- return NULL;
-
- struct hash_table *ht =
- _mesa_hash_table_create(NULL, _mesa_hash_pointer, _mesa_key_pointer_equal);
- void *mem_ctx = state;
- ir_function *f = s->function()->clone(mem_ctx, ht);
- _mesa_hash_table_destroy(ht, NULL);
-
- return f->matching_signature(state, actual_parameters, true);
+ return s;
}
bool
projects / mesa.git / blobdiff