};
static void allocate_user_sgprs(struct nir_to_llvm_context *ctx,
+ gl_shader_stage stage,
struct user_sgpr_info *user_sgpr_info)
{
memset(user_sgpr_info, 0, sizeof(struct user_sgpr_info));
/* until we sort out scratch/global buffers always assign ring offsets for gs/vs/es */
- if (ctx->stage == MESA_SHADER_GEOMETRY ||
- ctx->stage == MESA_SHADER_VERTEX ||
- ctx->stage == MESA_SHADER_TESS_CTRL ||
- ctx->stage == MESA_SHADER_TESS_EVAL ||
+ if (stage == MESA_SHADER_GEOMETRY ||
+ stage == MESA_SHADER_VERTEX ||
+ stage == MESA_SHADER_TESS_CTRL ||
+ stage == MESA_SHADER_TESS_EVAL ||
ctx->is_gs_copy_shader)
user_sgpr_info->need_ring_offsets = true;
- if (ctx->stage == MESA_SHADER_FRAGMENT &&
+ if (stage == MESA_SHADER_FRAGMENT &&
ctx->shader_info->info.ps.needs_sample_positions)
user_sgpr_info->need_ring_offsets = true;
user_sgpr_info->sgpr_count += 2;
}
- switch (ctx->stage) {
+ /* FIXME: fix the number of user sgprs for merged shaders on GFX9 */
+ switch (stage) {
case MESA_SHADER_COMPUTE:
if (ctx->shader_info->info.cs.uses_grid_size)
user_sgpr_info->sgpr_count += 3;
break;
}
- if (ctx->shader_info->info.needs_push_constants)
+ if (ctx->shader_info->info.loads_push_constants)
user_sgpr_info->sgpr_count += 2;
- uint32_t remaining_sgprs = 16 - user_sgpr_info->sgpr_count;
+ uint32_t available_sgprs = ctx->options->chip_class >= GFX9 ? 32 : 16;
+ uint32_t remaining_sgprs = available_sgprs - user_sgpr_info->sgpr_count;
+
if (remaining_sgprs / 2 < util_bitcount(ctx->shader_info->info.desc_set_used_mask)) {
user_sgpr_info->sgpr_count += 2;
user_sgpr_info->indirect_all_descriptor_sets = true;
add_array_arg(args, const_array(type, 32), desc_sets);
}
- if (ctx->shader_info->info.needs_push_constants) {
+ if (ctx->shader_info->info.loads_push_constants) {
/* 1 for push constants and dynamic descriptors */
add_array_arg(args, type, &ctx->push_constants);
}
{
add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->abi.vertex_id);
if (!ctx->is_gs_copy_shader) {
- add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->rel_auto_id);
- add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->vs_prim_id);
- add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->abi.instance_id);
+ if (ctx->options->key.vs.as_ls) {
+ add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->rel_auto_id);
+ add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->abi.instance_id);
+ } else {
+ add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->abi.instance_id);
+ add_arg(args, ARG_VGPR, ctx->ac.i32, &ctx->vs_prim_id);
+ }
+ add_arg(args, ARG_VGPR, ctx->ac.i32, NULL); /* unused */
}
}
ctx->shader_info->need_indirect_descriptor_sets = true;
}
- if (ctx->shader_info->info.needs_push_constants) {
+ if (ctx->shader_info->info.loads_push_constants) {
set_loc_shader(ctx, AC_UD_PUSH_CONSTANTS, user_sgpr_idx, 2);
}
}
struct arg_info args = {};
LLVMValueRef desc_sets;
- allocate_user_sgprs(ctx, &user_sgpr_info);
+ allocate_user_sgprs(ctx, stage, &user_sgpr_info);
if (user_sgpr_info.need_ring_offsets && !ctx->options->supports_spill) {
add_arg(&args, ARG_SGPR, const_array(ctx->ac.v4i32, 16),
break;
case nir_op_frcp:
src[0] = ac_to_float(&ctx->ac, src[0]);
- result = ac_build_fdiv(&ctx->ac, ctx->ac.f32_1, src[0]);
+ result = ac_build_fdiv(&ctx->ac, instr->dest.dest.ssa.bit_size == 32 ? ctx->ac.f32_1 : ctx->ac.f64_1,
+ src[0]);
break;
case nir_op_iand:
result = LLVMBuildAnd(ctx->ac.builder, src[0], src[1], "");
case nir_op_frsq:
result = emit_intrin_1f_param(&ctx->ac, "llvm.sqrt",
ac_to_float_type(&ctx->ac, def_type), src[0]);
- result = ac_build_fdiv(&ctx->ac, ctx->ac.f32_1, result);
+ result = ac_build_fdiv(&ctx->ac, instr->dest.dest.ssa.bit_size == 32 ? ctx->ac.f32_1 : ctx->ac.f64_1,
+ result);
break;
case nir_op_fpow:
result = emit_intrin_2f_param(&ctx->ac, "llvm.pow",
case nir_texop_txf:
case nir_texop_txf_ms:
case nir_texop_samples_identical:
- args->opcode = instr->sampler_dim == GLSL_SAMPLER_DIM_MS ? ac_image_load : ac_image_load_mip;
+ args->opcode = lod_is_zero ||
+ instr->sampler_dim == GLSL_SAMPLER_DIM_MS ?
+ ac_image_load : ac_image_load_mip;
args->compare = false;
args->offset = false;
break;
static LLVMValueRef visit_load_ubo_buffer(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
- LLVMValueRef results[8], ret;
+ LLVMValueRef ret;
LLVMValueRef rsrc = get_src(ctx, instr->src[0]);
LLVMValueRef offset = get_src(ctx, instr->src[1]);
int num_components = instr->num_components;
if (instr->dest.ssa.bit_size == 64)
num_components *= 2;
- for (unsigned i = 0; i < num_components; ++i) {
- LLVMValueRef params[] = {
- rsrc,
- LLVMBuildAdd(ctx->ac.builder, LLVMConstInt(ctx->ac.i32, 4 * i, 0),
- offset, "")
- };
- results[i] = ac_build_intrinsic(&ctx->ac, "llvm.SI.load.const.v4i32", ctx->ac.f32,
- params, 2,
- AC_FUNC_ATTR_READNONE |
- AC_FUNC_ATTR_LEGACY);
- }
+ ret = ac_build_buffer_load(&ctx->ac, rsrc, num_components, NULL, offset,
+ NULL, 0, false, false, true, true);
-
- ret = ac_build_gather_values(&ctx->ac, results, num_components);
return LLVMBuildBitCast(ctx->ac.builder, ret,
get_def_type(ctx, &instr->dest.ssa), "");
}
ac_build_waitcnt(&ctx->ac, waitcnt);
}
-static void emit_barrier(struct nir_to_llvm_context *ctx)
+static void emit_barrier(struct ac_llvm_context *ac, gl_shader_stage stage)
{
/* SI only (thanks to a hw bug workaround):
* The real barrier instruction isn’t needed, because an entire patch
* always fits into a single wave.
*/
- if (ctx->options->chip_class == SI &&
- ctx->stage == MESA_SHADER_TESS_CTRL) {
- ac_build_waitcnt(&ctx->ac, LGKM_CNT & VM_CNT);
+ if (ac->chip_class == SI && stage == MESA_SHADER_TESS_CTRL) {
+ ac_build_waitcnt(ac, LGKM_CNT & VM_CNT);
return;
}
- ac_build_intrinsic(&ctx->ac, "llvm.amdgcn.s.barrier",
- ctx->ac.voidt, NULL, 0, AC_FUNC_ATTR_CONVERGENT);
+ ac_build_intrinsic(ac, "llvm.amdgcn.s.barrier",
+ ac->voidt, NULL, 0, AC_FUNC_ATTR_CONVERGENT);
}
static void emit_discard_if(struct ac_nir_context *ctx,
}
static LLVMValueRef
-visit_load_tess_coord(struct nir_to_llvm_context *ctx,
- const nir_intrinsic_instr *instr)
+load_tess_coord(struct ac_shader_abi *abi, LLVMTypeRef type,
+ unsigned num_components)
{
+ struct nir_to_llvm_context *ctx = nir_to_llvm_context_from_abi(abi);
+
LLVMValueRef coord[4] = {
ctx->tes_u,
ctx->tes_v,
coord[2] = LLVMBuildFSub(ctx->builder, ctx->ac.f32_1,
LLVMBuildFAdd(ctx->builder, coord[0], coord[1], ""), "");
- LLVMValueRef result = ac_build_gather_values(&ctx->ac, coord, instr->num_components);
- return LLVMBuildBitCast(ctx->builder, result,
- get_def_type(ctx->nir, &instr->dest.ssa), "");
+ LLVMValueRef result = ac_build_gather_values(&ctx->ac, coord, num_components);
+ return LLVMBuildBitCast(ctx->builder, result, type, "");
+}
+
+static LLVMValueRef
+load_patch_vertices_in(struct ac_shader_abi *abi)
+{
+ struct nir_to_llvm_context *ctx = nir_to_llvm_context_from_abi(abi);
+ return LLVMConstInt(ctx->ac.i32, ctx->options->key.tcs.input_vertices, false);
}
static void visit_intrinsic(struct ac_nir_context *ctx,
emit_membar(ctx->nctx, instr);
break;
case nir_intrinsic_barrier:
- emit_barrier(ctx->nctx);
+ emit_barrier(&ctx->ac, ctx->stage);
break;
case nir_intrinsic_var_atomic_add:
case nir_intrinsic_var_atomic_imin:
case nir_intrinsic_end_primitive:
visit_end_primitive(ctx->nctx, instr);
break;
- case nir_intrinsic_load_tess_coord:
- result = visit_load_tess_coord(ctx->nctx, instr);
+ case nir_intrinsic_load_tess_coord: {
+ LLVMTypeRef type = ctx->nctx ?
+ get_def_type(ctx->nctx->nir, &instr->dest.ssa) :
+ NULL;
+ result = ctx->abi->load_tess_coord(ctx->abi, type, instr->num_components);
+ break;
+ }
+ case nir_intrinsic_load_tess_level_outer:
+ result = ctx->abi->load_tess_level(ctx->abi, VARYING_SLOT_TESS_LEVEL_OUTER);
+ break;
+ case nir_intrinsic_load_tess_level_inner:
+ result = ctx->abi->load_tess_level(ctx->abi, VARYING_SLOT_TESS_LEVEL_INNER);
break;
case nir_intrinsic_load_patch_vertices_in:
- result = LLVMConstInt(ctx->ac.i32, ctx->nctx->options->key.tcs.input_vertices, false);
+ result = ctx->abi->load_patch_vertices_in(ctx->abi);
break;
default:
fprintf(stderr, "Unknown intrinsic: ");
if (ctx->options->key.vs.instance_rate_inputs & (1u << index)) {
buffer_index = LLVMBuildAdd(ctx->builder, ctx->abi.instance_id,
ctx->abi.start_instance, "");
- ctx->shader_info->vs.vgpr_comp_cnt = MAX2(3,
- ctx->shader_info->vs.vgpr_comp_cnt);
+ if (ctx->options->key.vs.as_ls) {
+ ctx->shader_info->vs.vgpr_comp_cnt =
+ MAX2(2, ctx->shader_info->vs.vgpr_comp_cnt);
+ } else {
+ ctx->shader_info->vs.vgpr_comp_cnt =
+ MAX2(1, ctx->shader_info->vs.vgpr_comp_cnt);
+ }
} else
buffer_index = LLVMBuildAdd(ctx->builder, ctx->abi.vertex_id,
ctx->abi.base_vertex, "");
nir_foreach_variable(variable, &func->impl->locals) {
unsigned attrib_count = glsl_count_attribute_slots(variable->type, false);
variable->data.driver_location = ctx->num_locals * 4;
+ variable->data.location_frac = 0;
ctx->num_locals += attrib_count;
}
ctx->locals = malloc(4 * ctx->num_locals * sizeof(LLVMValueRef));
LLVMValueRef lds_base, lds_inner, lds_outer, byteoffset, buffer;
LLVMValueRef out[6], vec0, vec1, tf_base, inner[4], outer[4];
int i;
- emit_barrier(ctx);
+ emit_barrier(&ctx->ac, ctx->stage);
switch (ctx->options->key.tcs.primitive_mode) {
case GL_ISOLINES:
ctx.tcs_outputs_read = shaders[i]->info.outputs_read;
ctx.tcs_patch_outputs_read = shaders[i]->info.patch_outputs_read;
ctx.abi.load_tess_inputs = load_tcs_input;
+ ctx.abi.load_patch_vertices_in = load_patch_vertices_in;
ctx.abi.store_tcs_outputs = store_tcs_output;
} else if (shaders[i]->info.stage == MESA_SHADER_TESS_EVAL) {
ctx.tes_primitive_mode = shaders[i]->info.tess.primitive_mode;
ctx.abi.load_tess_inputs = load_tes_input;
+ ctx.abi.load_tess_coord = load_tess_coord;
+ ctx.abi.load_patch_vertices_in = load_patch_vertices_in;
} else if (shaders[i]->info.stage == MESA_SHADER_VERTEX) {
if (shader_info->info.vs.needs_instance_id) {
- ctx.shader_info->vs.vgpr_comp_cnt =
- MAX2(3, ctx.shader_info->vs.vgpr_comp_cnt);
+ if (ctx.ac.chip_class == GFX9 &&
+ shaders[shader_count - 1]->info.stage == MESA_SHADER_TESS_CTRL) {
+ ctx.shader_info->vs.vgpr_comp_cnt =
+ MAX2(2, ctx.shader_info->vs.vgpr_comp_cnt);
+ } else {
+ ctx.shader_info->vs.vgpr_comp_cnt =
+ MAX2(1, ctx.shader_info->vs.vgpr_comp_cnt);
+ }
}
} else if (shaders[i]->info.stage == MESA_SHADER_FRAGMENT) {
shader_info->fs.can_discard = shaders[i]->info.fs.uses_discard;
}
if (i)
- emit_barrier(&ctx);
+ emit_barrier(&ctx.ac, ctx.stage);
ac_setup_rings(&ctx);
/* +3 for scratch wave offset and VCC */
config->num_sgprs = MAX2(config->num_sgprs,
shader_info->num_input_sgprs + 3);
+
+ /* Enable 64-bit and 16-bit denormals, because there is no performance
+ * cost.
+ *
+ * If denormals are enabled, all floating-point output modifiers are
+ * ignored.
+ *
+ * Don't enable denormals for 32-bit floats, because:
+ * - Floating-point output modifiers would be ignored by the hw.
+ * - Some opcodes don't support denormals, such as v_mad_f32. We would
+ * have to stop using those.
+ * - SI & CI would be very slow.
+ */
+ config->float_mode |= V_00B028_FP_64_DENORMS;
}
static void
case MESA_SHADER_VERTEX:
shader_info->vs.as_es = options->key.vs.as_es;
shader_info->vs.as_ls = options->key.vs.as_ls;
- /* in LS mode we need at least 1, invocation id needs 3, handled elsewhere */
+ /* in LS mode we need at least 1, invocation id needs 2, handled elsewhere */
if (options->key.vs.as_ls)
shader_info->vs.vgpr_comp_cnt = MAX2(1, shader_info->vs.vgpr_comp_cnt);
break;
ac_compile_llvm_module(tm, llvm_module, binary, config, shader_info, nir[0]->info.stage, dump_shader, options->supports_spill);
for (int i = 0; i < nir_count; ++i)
ac_fill_shader_info(shader_info, nir[i], options);
+
+ /* Determine the ES type (VS or TES) for the GS on GFX9. */
+ if (options->chip_class == GFX9) {
+ if (nir_count == 2 &&
+ nir[1]->info.stage == MESA_SHADER_GEOMETRY) {
+ shader_info->gs.es_type = nir[0]->info.stage;
+ }
+ }
}
static void
projects / mesa.git / blobdiff