freedreno/ir3: ir3_cmdline updates
[mesa.git] / src / gallium / drivers / freedreno / ir3 / ir3_cmdline.c
index b41c32d3756b09036cce81be3d43c28bfd64b263..5dad4da11d21cc6817cd05e165e83d06cb3e5540 100644 (file)
 #include <stdlib.h>
 #include <stdio.h>
 #include <err.h>
+#include <getopt.h>
 
+#include "nir/tgsi_to_nir.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_text.h"
 #include "tgsi/tgsi_dump.h"
 
-#include "freedreno_util.h"
+#include "ir3/ir3_compiler.h"
+#include "ir3/ir3_gallium.h"
+#include "ir3/ir3_nir.h"
+#include "ir3/instr-a3xx.h"
+#include "ir3/ir3.h"
 
-#include "ir3_compiler.h"
-#include "ir3_nir.h"
-#include "instr-a3xx.h"
-#include "ir3.h"
+#include "main/mtypes.h"
 
 #include "compiler/glsl/standalone.h"
 #include "compiler/glsl/glsl_to_nir.h"
 #include "compiler/nir_types.h"
 #include "compiler/spirv/nir_spirv.h"
 
-static void dump_info(struct ir3_shader_variant *so, const char *str)
+#include "pipe/p_context.h"
+
+static void
+dump_info(struct ir3_shader_variant *so, const char *str)
 {
        uint32_t *bin;
-       const char *type = ir3_shader_stage(so->shader);
-       bin = ir3_shader_assemble(so, so->shader->compiler->gpu_id);
-       debug_printf("; %s: %s\n", type, str);
+       const char *type = ir3_shader_stage(so);
+       bin = ir3_shader_assemble(so);
+       printf("; %s: %s\n", type, str);
        ir3_shader_disasm(so, bin, stdout);
-       free(bin);
 }
 
 static void
 insert_sorted(struct exec_list *var_list, nir_variable *new_var)
 {
-       nir_foreach_variable(var, var_list) {
+       nir_foreach_variable_in_list(var, var_list) {
                if (var->data.location > new_var->data.location) {
                        exec_node_insert_node_before(&var->node, &new_var->node);
                        return;
@@ -73,21 +78,21 @@ insert_sorted(struct exec_list *var_list, nir_variable *new_var)
 }
 
 static void
-sort_varyings(struct exec_list *var_list)
+sort_varyings(nir_shader *nir, nir_variable_mode mode)
 {
        struct exec_list new_list;
        exec_list_make_empty(&new_list);
-       nir_foreach_variable_safe(var, var_list) {
+       nir_foreach_variable_with_modes_safe(var, nir, mode) {
                exec_node_remove(&var->node);
                insert_sorted(&new_list, var);
        }
-       exec_list_move_nodes_to(&new_list, var_list);
+       exec_list_append(&nir->variables, &new_list);
 }
 
 static void
-fixup_varying_slots(struct exec_list *var_list)
+fixup_varying_slots(nir_shader *nir, nir_variable_mode mode)
 {
-       nir_foreach_variable(var, var_list) {
+       nir_foreach_variable_with_modes(var, nir, mode) {
                if (var->data.location >= VARYING_SLOT_VAR0) {
                        var->data.location += 9;
                } else if ((var->data.location >= VARYING_SLOT_TEX0) &&
@@ -103,65 +108,87 @@ static nir_shader *
 load_glsl(unsigned num_files, char* const* files, gl_shader_stage stage)
 {
        static const struct standalone_options options = {
-                       .glsl_version = 140,
+                       .glsl_version = 310,
                        .do_link = true,
+                       .lower_precision = true,
        };
        struct gl_shader_program *prog;
+       const nir_shader_compiler_options *nir_options =
+                       ir3_get_compiler_options(compiler);
+       static struct gl_context local_ctx;
 
-       prog = standalone_compile_shader(&options, num_files, files);
+       prog = standalone_compile_shader(&options, num_files, files, &local_ctx);
        if (!prog)
                errx(1, "couldn't parse `%s'", files[0]);
 
-       nir_shader *nir = glsl_to_nir(prog, stage, ir3_get_compiler_options(compiler));
+       nir_shader *nir = glsl_to_nir(&local_ctx, prog, stage, nir_options);
 
        /* required NIR passes: */
-       /* TODO cmdline args for some of the conditional lowering passes? */
+       if (nir_options->lower_all_io_to_temps ||
+                       nir->info.stage == MESA_SHADER_VERTEX ||
+                       nir->info.stage == MESA_SHADER_GEOMETRY) {
+               NIR_PASS_V(nir, nir_lower_io_to_temporaries,
+                               nir_shader_get_entrypoint(nir),
+                               true, true);
+       } else if (nir->info.stage == MESA_SHADER_FRAGMENT) {
+               NIR_PASS_V(nir, nir_lower_io_to_temporaries,
+                               nir_shader_get_entrypoint(nir),
+                               true, false);
+       }
 
-       NIR_PASS_V(nir, nir_lower_io_to_temporaries,
-                       nir_shader_get_entrypoint(nir),
-                       true, true);
        NIR_PASS_V(nir, nir_lower_global_vars_to_local);
        NIR_PASS_V(nir, nir_split_var_copies);
        NIR_PASS_V(nir, nir_lower_var_copies);
 
        NIR_PASS_V(nir, nir_split_var_copies);
        NIR_PASS_V(nir, nir_lower_var_copies);
+       nir_print_shader(nir, stdout);
+       NIR_PASS_V(nir, gl_nir_lower_atomics, prog, true);
+       NIR_PASS_V(nir, gl_nir_lower_buffers, prog);
+       NIR_PASS_V(nir, nir_lower_atomics_to_ssbo);
+       nir_print_shader(nir, stdout);
 
        switch (stage) {
        case MESA_SHADER_VERTEX:
-               nir_assign_var_locations(&nir->inputs,
+               nir_assign_var_locations(nir, nir_var_shader_in,
                                &nir->num_inputs,
                                ir3_glsl_type_size);
 
                /* Re-lower global vars, to deal with any dead VS inputs. */
                NIR_PASS_V(nir, nir_lower_global_vars_to_local);
 
-               sort_varyings(&nir->outputs);
-               nir_assign_var_locations(&nir->outputs,
+               sort_varyings(nir, nir_var_shader_out);
+               nir_assign_var_locations(nir, nir_var_shader_out,
                                &nir->num_outputs,
                                ir3_glsl_type_size);
-               fixup_varying_slots(&nir->outputs);
+               fixup_varying_slots(nir, nir_var_shader_out);
                break;
        case MESA_SHADER_FRAGMENT:
-               sort_varyings(&nir->inputs);
-               nir_assign_var_locations(&nir->inputs,
+               sort_varyings(nir, nir_var_shader_in);
+               nir_assign_var_locations(nir, nir_var_shader_in,
                                &nir->num_inputs,
                                ir3_glsl_type_size);
-               fixup_varying_slots(&nir->inputs);
-               nir_assign_var_locations(&nir->outputs,
+               fixup_varying_slots(nir, nir_var_shader_in);
+               nir_assign_var_locations(nir, nir_var_shader_out,
                                &nir->num_outputs,
                                ir3_glsl_type_size);
                break;
+       case MESA_SHADER_COMPUTE:
+       case MESA_SHADER_KERNEL:
+               break;
        default:
                errx(1, "unhandled shader stage: %d", stage);
        }
 
-       nir_assign_var_locations(&nir->uniforms,
+       nir_assign_var_locations(nir, nir_var_uniform,
                        &nir->num_uniforms,
                        ir3_glsl_type_size);
 
        NIR_PASS_V(nir, nir_lower_system_values);
-       NIR_PASS_V(nir, nir_lower_io, nir_var_all, ir3_glsl_type_size, 0);
+       NIR_PASS_V(nir, nir_lower_frexp);
+       NIR_PASS_V(nir, nir_lower_io,
+                  nir_var_shader_in | nir_var_shader_out | nir_var_uniform,
+                  ir3_glsl_type_size, (nir_lower_io_options)0);
        NIR_PASS_V(nir, gl_nir_lower_samplers, prog);
 
        return nir;
@@ -214,169 +241,84 @@ load_spirv(const char *filename, const char *entry, gl_shader_stage stage)
                        .int64 = true,
                        .variable_pointers = true,
                },
-               .lower_workgroup_access_to_offsets = true,
+               .lower_ubo_ssbo_access_to_offsets = true,
                .debug = {
                        .func = debug_func,
                }
        };
-       nir_function *entry_point;
+       nir_shader *nir;
        void *buf;
        size_t size;
 
        read_file(filename, &buf, &size);
 
-       entry_point = spirv_to_nir(buf, size / 4,
+       nir = spirv_to_nir(buf, size / 4,
                        NULL, 0, /* spec_entries */
                        stage, entry,
                        &spirv_options,
                        ir3_get_compiler_options(compiler));
 
-       nir_print_shader(entry_point->shader, stdout);
+       nir_print_shader(nir, stdout);
 
-       return entry_point->shader;
+       return nir;
 }
 
-static void print_usage(void)
+static const char *shortopts = "g:hv";
+
+static const struct option longopts[] = {
+       { "gpu",            required_argument, 0, 'g' },
+       { "help",           no_argument,       0, 'h' },
+       { "verbose",        no_argument,       0, 'v' },
+};
+
+static void
+print_usage(void)
 {
        printf("Usage: ir3_compiler [OPTIONS]... <file.tgsi | file.spv entry_point | (file.vert | file.frag)*>\n");
-       printf("    --verbose         - verbose compiler/debug messages\n");
-       printf("    --binning-pass    - generate binning pass shader (VERT)\n");
-       printf("    --color-two-side  - emulate two-sided color (FRAG)\n");
-       printf("    --half-precision  - use half-precision\n");
-       printf("    --saturate-s MASK - bitmask of samplers to saturate S coord\n");
-       printf("    --saturate-t MASK - bitmask of samplers to saturate T coord\n");
-       printf("    --saturate-r MASK - bitmask of samplers to saturate R coord\n");
-       printf("    --astc-srgb MASK  - bitmask of samplers to enable astc-srgb workaround\n");
-       printf("    --stream-out      - enable stream-out (aka transform feedback)\n");
-       printf("    --ucp MASK        - bitmask of enabled user-clip-planes\n");
-       printf("    --gpu GPU_ID      - specify gpu-id (default 320)\n");
-       printf("    --help            - show this message\n");
+       printf("    -g, --gpu GPU_ID - specify gpu-id (default 320)\n");
+       printf("    -h, --help       - show this message\n");
+       printf("    -v, --verbose    - verbose compiler/debug messages\n");
 }
 
-int main(int argc, char **argv)
+int
+main(int argc, char **argv)
 {
-       int ret = 0, n = 1;
+       int ret = 0, opt;
        char *filenames[2];
        int num_files = 0;
        unsigned stage = 0;
-       struct ir3_shader_variant v;
-       struct ir3_shader s;
        struct ir3_shader_key key = {};
-       /* TODO cmdline option to target different gpus: */
        unsigned gpu_id = 320;
        const char *info;
-       const char *entry;
+       const char *spirv_entry = NULL;
        void *ptr;
-       bool from_spirv = false;
+       bool from_tgsi = false;
        size_t size;
 
-       memset(&s, 0, sizeof(s));
-       memset(&v, 0, sizeof(v));
-
-       /* cmdline args which impact shader variant get spit out in a
-        * comment on the first line..  a quick/dirty way to preserve
-        * that info so when ir3test recompiles the shader with a new
-        * compiler version, we use the same shader-key settings:
-        */
-       debug_printf("; options:");
-
-       while (n < argc) {
-               if (!strcmp(argv[n], "--verbose")) {
-                       fd_mesa_debug |= FD_DBG_MSGS | FD_DBG_OPTMSGS | FD_DBG_DISASM;
-                       n++;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--binning-pass")) {
-                       debug_printf(" %s", argv[n]);
-                       key.binning_pass = true;
-                       n++;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--color-two-side")) {
-                       debug_printf(" %s", argv[n]);
-                       key.color_two_side = true;
-                       n++;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--half-precision")) {
-                       debug_printf(" %s", argv[n]);
-                       key.half_precision = true;
-                       n++;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--saturate-s")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       key.vsaturate_s = key.fsaturate_s = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--saturate-t")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       key.vsaturate_t = key.fsaturate_t = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--saturate-r")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       key.vsaturate_r = key.fsaturate_r = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--astc-srgb")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       key.vastc_srgb = key.fastc_srgb = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--stream-out")) {
-                       struct pipe_stream_output_info *so = &s.stream_output;
-                       debug_printf(" %s", argv[n]);
-                       /* TODO more dynamic config based on number of outputs, etc
-                        * rather than just hard-code for first output:
-                        */
-                       so->num_outputs = 1;
-                       so->stride[0] = 4;
-                       so->output[0].register_index = 0;
-                       so->output[0].start_component = 0;
-                       so->output[0].num_components = 4;
-                       so->output[0].output_buffer = 0;
-                       so->output[0].dst_offset = 2;
-                       so->output[0].stream = 0;
-                       n++;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--ucp")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       key.ucp_enables = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--gpu")) {
-                       debug_printf(" %s %s", argv[n], argv[n+1]);
-                       gpu_id = strtol(argv[n+1], NULL, 0);
-                       n += 2;
-                       continue;
-               }
-
-               if (!strcmp(argv[n], "--help")) {
+       while ((opt = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
+               switch (opt) {
+               case 'g':
+                       gpu_id = strtol(optarg, NULL, 0);
+                       break;
+               case 'v':
+                       ir3_shader_debug |= IR3_DBG_OPTMSGS | IR3_DBG_DISASM;
+                       break;
+               default:
+                       printf("unrecognized arg: %c\n", opt);
+                       /* fallthrough */
+               case 'h':
                        print_usage();
                        return 0;
                }
+       }
 
-               break;
+       if (optind >= argc) {
+               fprintf(stderr, "no file specified!\n");
+               print_usage();
+               return 0;
        }
-       debug_printf("\n");
 
+       unsigned n = optind;
        while (n < argc) {
                char *filename = argv[n];
                char *ext = strrchr(filename, '.');
@@ -384,26 +326,31 @@ int main(int argc, char **argv)
                if (strcmp(ext, ".tgsi") == 0) {
                        if (num_files != 0)
                                errx(1, "in TGSI mode, only a single file may be specified");
-                       s.from_tgsi = true;
+                       from_tgsi = true;
                } else if (strcmp(ext, ".spv") == 0) {
                        if (num_files != 0)
                                errx(1, "in SPIR-V mode, only a single file may be specified");
                        stage = MESA_SHADER_COMPUTE;
-                       from_spirv = true;
                        filenames[num_files++] = filename;
                        n++;
                        if (n == argc)
                                errx(1, "in SPIR-V mode, an entry point must be specified");
-                       entry = argv[n];
+                       spirv_entry = argv[n];
                        n++;
+               } else if (strcmp(ext, ".comp") == 0) {
+                       if (from_tgsi || spirv_entry)
+                               errx(1, "cannot mix GLSL/TGSI/SPIRV");
+                       if (num_files >= ARRAY_SIZE(filenames))
+                               errx(1, "too many GLSL files");
+                       stage = MESA_SHADER_COMPUTE;
                } else if (strcmp(ext, ".frag") == 0) {
-                       if (s.from_tgsi || from_spirv)
+                       if (from_tgsi || spirv_entry)
                                errx(1, "cannot mix GLSL/TGSI/SPIRV");
                        if (num_files >= ARRAY_SIZE(filenames))
                                errx(1, "too many GLSL files");
                        stage = MESA_SHADER_FRAGMENT;
                } else if (strcmp(ext, ".vert") == 0) {
-                       if (s.from_tgsi)
+                       if (from_tgsi)
                                errx(1, "cannot mix GLSL and TGSI");
                        if (num_files >= ARRAY_SIZE(filenames))
                                errx(1, "too many GLSL files");
@@ -422,8 +369,10 @@ int main(int argc, char **argv)
 
        compiler = ir3_compiler_create(NULL, gpu_id);
 
-       if (s.from_tgsi) {
+       if (from_tgsi) {
                struct tgsi_token toks[65536];
+               const nir_shader_compiler_options *nir_options =
+                       ir3_get_compiler_options(compiler);
 
                ret = read_file(filenames[0], &ptr, &size);
                if (ret) {
@@ -431,25 +380,26 @@ int main(int argc, char **argv)
                        return ret;
                }
 
-               if (fd_mesa_debug & FD_DBG_OPTMSGS)
-                       debug_printf("%s\n", (char *)ptr);
+               if (ir3_shader_debug & IR3_DBG_OPTMSGS)
+                       printf("%s\n", (char *)ptr);
 
                if (!tgsi_text_translate(ptr, toks, ARRAY_SIZE(toks)))
                        errx(1, "could not parse `%s'", filenames[0]);
 
-               if (fd_mesa_debug & FD_DBG_OPTMSGS)
+               if (ir3_shader_debug & IR3_DBG_OPTMSGS)
                        tgsi_dump(toks, 0);
 
-               nir = ir3_tgsi_to_nir(toks);
+               nir = tgsi_to_nir_noscreen(toks, nir_options);
                NIR_PASS_V(nir, nir_lower_global_vars_to_local);
-       } else if (from_spirv) {
-               nir = load_spirv(filenames[0], entry, stage);
+       } else if (spirv_entry) {
+               nir = load_spirv(filenames[0], spirv_entry, stage);
 
-               NIR_PASS_V(nir, nir_lower_io, nir_var_all, ir3_glsl_type_size,
-                               (nir_lower_io_options)0);
+               NIR_PASS_V(nir, nir_lower_io,
+                          nir_var_shader_in | nir_var_shader_out,
+                          ir3_glsl_type_size, (nir_lower_io_options)0);
 
                /* TODO do this somewhere else */
-               nir_lower_int64(nir, ~0);
+               nir_lower_int64(nir);
                nir_lower_system_values(nir);
        } else if (num_files > 0) {
                nir = load_glsl(num_files, filenames, stage);
@@ -458,31 +408,32 @@ int main(int argc, char **argv)
                return -1;
        }
 
-       s.compiler = compiler;
-       s.nir = ir3_optimize_nir(&s, nir, NULL);
+       ir3_finalize_nir(compiler, nir);
+       ir3_nir_post_finalize(compiler, nir);
 
-       v.key = key;
-       v.shader = &s;
+       struct ir3_shader *shader = rzalloc_size(NULL, sizeof(*shader));
+       shader->compiler = compiler;
+       shader->type = stage;
+       shader->nir = nir;
 
-       switch (nir->info.stage) {
-       case MESA_SHADER_FRAGMENT:
-               s.type = v.type = SHADER_FRAGMENT;
-               break;
-       case MESA_SHADER_VERTEX:
-               s.type = v.type = SHADER_VERTEX;
-               break;
-       case MESA_SHADER_COMPUTE:
-               s.type = v.type = SHADER_COMPUTE;
-               break;
-       default:
-               errx(1, "unhandled shader stage: %d", nir->info.stage);
-       }
+       struct ir3_shader_variant *v = rzalloc_size(shader, sizeof(*v));
+       v->type = shader->type;
+       v->shader = shader;
+       v->key = key;
+       v->const_state = rzalloc_size(v, sizeof(*v->const_state));
+
+       shader->variants = v;
+       shader->variant_count = 1;
+
+       ir3_nir_lower_variant(v, nir);
 
        info = "NIR compiler";
-       ret = ir3_compile_shader_nir(s.compiler, &v);
+       ret = ir3_compile_shader_nir(compiler, v);
        if (ret) {
                fprintf(stderr, "compiler failed!\n");
                return ret;
        }
-       dump_info(&v, info);
+       dump_info(v, info);
+
+       return 0;
 }