2017-09-01 10:41:18 +01:00
|
|
|
/*
|
|
|
|
* Copyright © 2016 Red Hat.
|
|
|
|
* Copyright © 2016 Bas Nieuwenhuizen
|
|
|
|
*
|
|
|
|
* based in part on anv driver which is:
|
|
|
|
* Copyright © 2015 Intel Corporation
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
* Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
|
|
* IN THE SOFTWARE.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "util/mesa-sha1.h"
|
|
|
|
#include "util/u_atomic.h"
|
|
|
|
#include "radv_debug.h"
|
|
|
|
#include "radv_private.h"
|
|
|
|
#include "radv_shader.h"
|
2018-06-27 02:34:25 +01:00
|
|
|
#include "radv_shader_helper.h"
|
2019-11-11 17:05:03 +00:00
|
|
|
#include "radv_shader_args.h"
|
2017-09-01 10:41:18 +01:00
|
|
|
#include "nir/nir.h"
|
|
|
|
#include "nir/nir_builder.h"
|
|
|
|
#include "spirv/nir_spirv.h"
|
|
|
|
|
|
|
|
#include "sid.h"
|
|
|
|
#include "ac_binary.h"
|
|
|
|
#include "ac_llvm_util.h"
|
|
|
|
#include "ac_nir_to_llvm.h"
|
2019-07-01 00:29:24 +01:00
|
|
|
#include "ac_rtld.h"
|
2017-09-01 10:41:18 +01:00
|
|
|
#include "vk_format.h"
|
|
|
|
#include "util/debug.h"
|
|
|
|
#include "ac_exp_param.h"
|
|
|
|
|
2019-09-17 13:35:22 +01:00
|
|
|
#include "aco_interface.h"
|
|
|
|
|
|
|
|
static const struct nir_shader_compiler_options nir_options_llvm = {
|
2017-09-01 10:41:18 +01:00
|
|
|
.vertex_id_zero_based = true,
|
|
|
|
.lower_scmp = true,
|
2018-12-06 14:01:15 +00:00
|
|
|
.lower_flrp16 = true,
|
2017-09-01 10:41:18 +01:00
|
|
|
.lower_flrp32 = true,
|
2018-01-12 00:12:09 +00:00
|
|
|
.lower_flrp64 = true,
|
2018-01-21 16:13:26 +00:00
|
|
|
.lower_device_index_to_zero = true,
|
2017-09-01 10:41:18 +01:00
|
|
|
.lower_fsat = true,
|
|
|
|
.lower_fdiv = true,
|
2019-10-03 15:20:40 +01:00
|
|
|
.lower_fmod = true,
|
2019-01-25 15:08:38 +00:00
|
|
|
.lower_bitfield_insert_to_bitfield_select = true,
|
2019-01-25 15:24:55 +00:00
|
|
|
.lower_bitfield_extract = true,
|
2017-09-01 10:41:18 +01:00
|
|
|
.lower_sub = true,
|
|
|
|
.lower_pack_snorm_2x16 = true,
|
|
|
|
.lower_pack_snorm_4x8 = true,
|
|
|
|
.lower_pack_unorm_2x16 = true,
|
|
|
|
.lower_pack_unorm_4x8 = true,
|
|
|
|
.lower_unpack_snorm_2x16 = true,
|
|
|
|
.lower_unpack_snorm_4x8 = true,
|
|
|
|
.lower_unpack_unorm_2x16 = true,
|
|
|
|
.lower_unpack_unorm_4x8 = true,
|
|
|
|
.lower_extract_byte = true,
|
|
|
|
.lower_extract_word = true,
|
2020-09-24 16:46:31 +01:00
|
|
|
.lower_ffma16 = true,
|
|
|
|
.lower_ffma32 = true,
|
|
|
|
.lower_ffma64 = true,
|
2018-02-02 18:04:57 +00:00
|
|
|
.lower_fpow = true,
|
2019-03-06 21:35:31 +00:00
|
|
|
.lower_mul_2x32_64 = true,
|
2019-06-04 01:11:57 +01:00
|
|
|
.lower_rotate = true,
|
2020-07-17 19:17:10 +01:00
|
|
|
.use_scoped_barrier = true,
|
2019-05-10 09:44:20 +01:00
|
|
|
.max_unroll_iterations = 32,
|
|
|
|
.use_interpolated_input_intrinsics = true,
|
2019-10-15 20:43:39 +01:00
|
|
|
/* nir_lower_int64() isn't actually called for the LLVM backend, but
|
|
|
|
* this helps the loop unrolling heuristics. */
|
|
|
|
.lower_int64_options = nir_lower_imul64 |
|
|
|
|
nir_lower_imul_high64 |
|
|
|
|
nir_lower_imul_2x32_64 |
|
|
|
|
nir_lower_divmod64 |
|
|
|
|
nir_lower_minmax64 |
|
|
|
|
nir_lower_iabs64,
|
2020-06-23 17:29:22 +01:00
|
|
|
.lower_doubles_options = nir_lower_drcp |
|
|
|
|
nir_lower_dsqrt |
|
|
|
|
nir_lower_drsq |
|
|
|
|
nir_lower_ddiv,
|
2017-09-01 10:41:18 +01:00
|
|
|
};
|
|
|
|
|
2019-09-17 13:35:22 +01:00
|
|
|
static const struct nir_shader_compiler_options nir_options_aco = {
|
|
|
|
.vertex_id_zero_based = true,
|
|
|
|
.lower_scmp = true,
|
|
|
|
.lower_flrp16 = true,
|
|
|
|
.lower_flrp32 = true,
|
|
|
|
.lower_flrp64 = true,
|
|
|
|
.lower_device_index_to_zero = true,
|
|
|
|
.lower_fdiv = true,
|
2019-10-03 15:32:19 +01:00
|
|
|
.lower_fmod = true,
|
2019-09-17 13:35:22 +01:00
|
|
|
.lower_bitfield_insert_to_bitfield_select = true,
|
|
|
|
.lower_bitfield_extract = true,
|
|
|
|
.lower_pack_snorm_2x16 = true,
|
|
|
|
.lower_pack_snorm_4x8 = true,
|
|
|
|
.lower_pack_unorm_2x16 = true,
|
|
|
|
.lower_pack_unorm_4x8 = true,
|
|
|
|
.lower_unpack_snorm_2x16 = true,
|
|
|
|
.lower_unpack_snorm_4x8 = true,
|
|
|
|
.lower_unpack_unorm_2x16 = true,
|
|
|
|
.lower_unpack_unorm_4x8 = true,
|
|
|
|
.lower_unpack_half_2x16 = true,
|
|
|
|
.lower_extract_byte = true,
|
|
|
|
.lower_extract_word = true,
|
2020-09-24 16:46:31 +01:00
|
|
|
.lower_ffma16 = true,
|
|
|
|
.lower_ffma32 = true,
|
|
|
|
.lower_ffma64 = true,
|
2019-09-17 13:35:22 +01:00
|
|
|
.lower_fpow = true,
|
|
|
|
.lower_mul_2x32_64 = true,
|
|
|
|
.lower_rotate = true,
|
2020-07-17 19:17:10 +01:00
|
|
|
.use_scoped_barrier = true,
|
2019-09-17 13:35:22 +01:00
|
|
|
.max_unroll_iterations = 32,
|
|
|
|
.use_interpolated_input_intrinsics = true,
|
2019-10-15 20:43:39 +01:00
|
|
|
.lower_int64_options = nir_lower_imul64 |
|
|
|
|
nir_lower_imul_high64 |
|
|
|
|
nir_lower_imul_2x32_64 |
|
|
|
|
nir_lower_divmod64 |
|
|
|
|
nir_lower_minmax64 |
|
|
|
|
nir_lower_iabs64,
|
2020-06-23 17:29:22 +01:00
|
|
|
.lower_doubles_options = nir_lower_drcp |
|
|
|
|
nir_lower_dsqrt |
|
|
|
|
nir_lower_drsq |
|
|
|
|
nir_lower_ddiv,
|
2019-09-17 13:35:22 +01:00
|
|
|
};
|
|
|
|
|
2019-07-29 16:51:01 +01:00
|
|
|
bool
|
|
|
|
radv_can_dump_shader(struct radv_device *device,
|
|
|
|
struct radv_shader_module *module,
|
|
|
|
bool is_gs_copy_shader)
|
|
|
|
{
|
|
|
|
if (!(device->instance->debug_flags & RADV_DEBUG_DUMP_SHADERS))
|
|
|
|
return false;
|
2019-09-18 13:39:10 +01:00
|
|
|
if (module)
|
|
|
|
return !module->nir ||
|
|
|
|
(device->instance->debug_flags & RADV_DEBUG_DUMP_META_SHADERS);
|
2019-07-29 16:51:01 +01:00
|
|
|
|
2019-09-18 13:39:10 +01:00
|
|
|
return is_gs_copy_shader;
|
2019-07-29 16:51:01 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
bool
|
|
|
|
radv_can_dump_shader_stats(struct radv_device *device,
|
|
|
|
struct radv_shader_module *module)
|
|
|
|
{
|
|
|
|
/* Only dump non-meta shader stats. */
|
|
|
|
return device->instance->debug_flags & RADV_DEBUG_DUMP_SHADER_STATS &&
|
|
|
|
module && !module->nir;
|
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
VkResult radv_CreateShaderModule(
|
|
|
|
VkDevice _device,
|
|
|
|
const VkShaderModuleCreateInfo* pCreateInfo,
|
|
|
|
const VkAllocationCallbacks* pAllocator,
|
|
|
|
VkShaderModule* pShaderModule)
|
|
|
|
{
|
|
|
|
RADV_FROM_HANDLE(radv_device, device, _device);
|
|
|
|
struct radv_shader_module *module;
|
|
|
|
|
|
|
|
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO);
|
|
|
|
assert(pCreateInfo->flags == 0);
|
|
|
|
|
2020-04-29 09:16:32 +01:00
|
|
|
module = vk_alloc2(&device->vk.alloc, pAllocator,
|
2017-09-01 10:41:18 +01:00
|
|
|
sizeof(*module) + pCreateInfo->codeSize, 8,
|
|
|
|
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
|
|
|
|
if (module == NULL)
|
2018-05-31 00:06:41 +01:00
|
|
|
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-04-29 13:57:20 +01:00
|
|
|
vk_object_base_init(&device->vk, &module->base,
|
|
|
|
VK_OBJECT_TYPE_SHADER_MODULE);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
module->nir = NULL;
|
|
|
|
module->size = pCreateInfo->codeSize;
|
|
|
|
memcpy(module->data, pCreateInfo->pCode, module->size);
|
|
|
|
|
|
|
|
_mesa_sha1_compute(module->data, module->size, module->sha1);
|
|
|
|
|
|
|
|
*pShaderModule = radv_shader_module_to_handle(module);
|
|
|
|
|
|
|
|
return VK_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
void radv_DestroyShaderModule(
|
|
|
|
VkDevice _device,
|
|
|
|
VkShaderModule _module,
|
|
|
|
const VkAllocationCallbacks* pAllocator)
|
|
|
|
{
|
|
|
|
RADV_FROM_HANDLE(radv_device, device, _device);
|
|
|
|
RADV_FROM_HANDLE(radv_shader_module, module, _module);
|
|
|
|
|
|
|
|
if (!module)
|
|
|
|
return;
|
|
|
|
|
2020-04-29 13:57:20 +01:00
|
|
|
vk_object_base_finish(&module->base);
|
2020-04-29 09:16:32 +01:00
|
|
|
vk_free2(&device->vk.alloc, pAllocator, module);
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
|
2017-02-08 23:12:10 +00:00
|
|
|
void
|
2018-10-17 23:42:17 +01:00
|
|
|
radv_optimize_nir(struct nir_shader *shader, bool optimize_conservatively,
|
|
|
|
bool allow_copies)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
|
|
|
bool progress;
|
2018-08-19 00:42:04 +01:00
|
|
|
unsigned lower_flrp =
|
|
|
|
(shader->options->lower_flrp16 ? 16 : 0) |
|
|
|
|
(shader->options->lower_flrp32 ? 32 : 0) |
|
|
|
|
(shader->options->lower_flrp64 ? 64 : 0);
|
2017-09-01 10:41:18 +01:00
|
|
|
|
|
|
|
do {
|
|
|
|
progress = false;
|
|
|
|
|
2019-01-15 23:05:04 +00:00
|
|
|
NIR_PASS(progress, shader, nir_split_array_vars, nir_var_function_temp);
|
|
|
|
NIR_PASS(progress, shader, nir_shrink_vec_array_vars, nir_var_function_temp);
|
2018-10-18 00:19:16 +01:00
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS_V(shader, nir_lower_vars_to_ssa);
|
2018-04-27 08:28:48 +01:00
|
|
|
NIR_PASS_V(shader, nir_lower_pack);
|
2018-10-17 22:55:46 +01:00
|
|
|
|
2018-10-17 23:42:17 +01:00
|
|
|
if (allow_copies) {
|
|
|
|
/* Only run this pass in the first call to
|
|
|
|
* radv_optimize_nir. Later calls assume that we've
|
|
|
|
* lowered away any copy_deref instructions and we
|
|
|
|
* don't want to introduce any more.
|
|
|
|
*/
|
|
|
|
NIR_PASS(progress, shader, nir_opt_find_array_copies);
|
|
|
|
}
|
|
|
|
|
2018-10-17 22:55:46 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_copy_prop_vars);
|
|
|
|
NIR_PASS(progress, shader, nir_opt_dead_write_vars);
|
2019-06-26 13:03:31 +01:00
|
|
|
NIR_PASS(progress, shader, nir_remove_dead_variables,
|
2020-05-28 01:59:28 +01:00
|
|
|
nir_var_function_temp | nir_var_shader_in | nir_var_shader_out,
|
|
|
|
NULL);
|
2018-10-17 22:55:46 +01:00
|
|
|
|
2019-08-30 05:14:54 +01:00
|
|
|
NIR_PASS_V(shader, nir_lower_alu_to_scalar, NULL, NULL);
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS_V(shader, nir_lower_phis_to_scalar);
|
|
|
|
|
|
|
|
NIR_PASS(progress, shader, nir_copy_prop);
|
|
|
|
NIR_PASS(progress, shader, nir_opt_remove_phis);
|
|
|
|
NIR_PASS(progress, shader, nir_opt_dce);
|
|
|
|
if (nir_opt_trivial_continues(shader)) {
|
|
|
|
progress = true;
|
|
|
|
NIR_PASS(progress, shader, nir_copy_prop);
|
2017-09-13 03:49:31 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_remove_phis);
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_dce);
|
|
|
|
}
|
2019-04-08 11:13:49 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_if, true);
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_dead_cf);
|
|
|
|
NIR_PASS(progress, shader, nir_opt_cse);
|
2018-06-19 00:11:55 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_peephole_select, 8, true, true);
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_constant_folding);
|
2019-05-02 04:38:52 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_algebraic);
|
2018-08-19 00:42:04 +01:00
|
|
|
|
|
|
|
if (lower_flrp != 0) {
|
2019-05-08 15:32:43 +01:00
|
|
|
bool lower_flrp_progress = false;
|
2018-08-19 00:42:04 +01:00
|
|
|
NIR_PASS(lower_flrp_progress,
|
|
|
|
shader,
|
|
|
|
nir_lower_flrp,
|
|
|
|
lower_flrp,
|
2020-07-23 03:13:16 +01:00
|
|
|
false /* always_precise */);
|
2018-08-19 00:42:04 +01:00
|
|
|
if (lower_flrp_progress) {
|
|
|
|
NIR_PASS(progress, shader,
|
|
|
|
nir_opt_constant_folding);
|
|
|
|
progress = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Nothing should rematerialize any flrps, so we only
|
|
|
|
* need to do this lowering once.
|
|
|
|
*/
|
|
|
|
lower_flrp = 0;
|
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_undef);
|
2020-08-19 17:16:15 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_shrink_vectors);
|
2017-09-01 10:41:18 +01:00
|
|
|
if (shader->options->max_unroll_iterations) {
|
|
|
|
NIR_PASS(progress, shader, nir_opt_loop_unroll, 0);
|
|
|
|
}
|
2018-05-08 05:57:55 +01:00
|
|
|
} while (progress && !optimize_conservatively);
|
2018-01-29 16:19:18 +00:00
|
|
|
|
2019-07-20 18:21:14 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_conditional_discard);
|
2019-07-24 19:23:21 +01:00
|
|
|
NIR_PASS(progress, shader, nir_opt_move, nir_move_load_ubo);
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
|
2019-11-07 14:56:35 +00:00
|
|
|
static void
|
|
|
|
shared_var_info(const struct glsl_type *type, unsigned *size, unsigned *align)
|
|
|
|
{
|
|
|
|
assert(glsl_type_is_vector_or_scalar(type));
|
|
|
|
|
|
|
|
uint32_t comp_size = glsl_type_is_boolean(type) ? 4 : glsl_get_bit_size(type) / 8;
|
|
|
|
unsigned length = glsl_get_vector_elements(type);
|
|
|
|
*size = comp_size * length,
|
|
|
|
*align = comp_size;
|
|
|
|
}
|
|
|
|
|
2020-08-14 12:59:16 +01:00
|
|
|
struct radv_shader_debug_data {
|
2020-08-07 14:23:03 +01:00
|
|
|
struct radv_device *device;
|
|
|
|
const struct radv_shader_module *module;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void radv_spirv_nir_debug(void *private_data,
|
|
|
|
enum nir_spirv_debug_level level,
|
|
|
|
size_t spirv_offset,
|
|
|
|
const char *message)
|
|
|
|
{
|
2020-08-14 12:59:16 +01:00
|
|
|
struct radv_shader_debug_data *debug_data = private_data;
|
2020-08-07 14:23:03 +01:00
|
|
|
struct radv_instance *instance = debug_data->device->instance;
|
|
|
|
|
|
|
|
static const VkDebugReportFlagsEXT vk_flags[] = {
|
|
|
|
[NIR_SPIRV_DEBUG_LEVEL_INFO] = VK_DEBUG_REPORT_INFORMATION_BIT_EXT,
|
|
|
|
[NIR_SPIRV_DEBUG_LEVEL_WARNING] = VK_DEBUG_REPORT_WARNING_BIT_EXT,
|
|
|
|
[NIR_SPIRV_DEBUG_LEVEL_ERROR] = VK_DEBUG_REPORT_ERROR_BIT_EXT,
|
|
|
|
};
|
|
|
|
char buffer[256];
|
|
|
|
|
|
|
|
snprintf(buffer, sizeof(buffer), "SPIR-V offset %lu: %s",
|
|
|
|
(unsigned long)spirv_offset, message);
|
|
|
|
|
|
|
|
vk_debug_report(&instance->debug_report_callbacks,
|
|
|
|
vk_flags[level],
|
|
|
|
VK_DEBUG_REPORT_OBJECT_TYPE_SHADER_MODULE_EXT,
|
|
|
|
(uint64_t)(uintptr_t)debug_data->module,
|
|
|
|
0, 0, "radv", buffer);
|
|
|
|
}
|
|
|
|
|
2020-08-14 12:59:16 +01:00
|
|
|
static void radv_compiler_debug(void *private_data,
|
|
|
|
enum radv_compiler_debug_level level,
|
|
|
|
const char *message)
|
|
|
|
{
|
|
|
|
struct radv_shader_debug_data *debug_data = private_data;
|
|
|
|
struct radv_instance *instance = debug_data->device->instance;
|
|
|
|
|
|
|
|
static const VkDebugReportFlagsEXT vk_flags[] = {
|
|
|
|
[RADV_COMPILER_DEBUG_LEVEL_PERFWARN] = VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
|
|
|
|
[RADV_COMPILER_DEBUG_LEVEL_ERROR] = VK_DEBUG_REPORT_ERROR_BIT_EXT,
|
|
|
|
};
|
|
|
|
|
|
|
|
/* VK_DEBUG_REPORT_DEBUG_BIT_EXT specifies diagnostic information
|
|
|
|
* from the implementation and layers.
|
|
|
|
*/
|
|
|
|
vk_debug_report(&instance->debug_report_callbacks,
|
|
|
|
vk_flags[level] | VK_DEBUG_REPORT_DEBUG_BIT_EXT,
|
|
|
|
VK_DEBUG_REPORT_OBJECT_TYPE_SHADER_MODULE_EXT,
|
|
|
|
(uint64_t)(uintptr_t)debug_data->module,
|
|
|
|
0, 0, "radv", message);
|
|
|
|
}
|
|
|
|
|
radv: do not lower UBO/SSBO access to offsets
Use nir_lower_explicit_io instead of lowering to offsets. Extra
(useless) additions are removed by lowering load_vulkan_descriptor
to vec2(src.x, 0).
fossils-db (Navi):
Totals from 18236 (13.21% of 138013) affected shaders:
SGPRs: 1172766 -> 1168278 (-0.38%); split: -0.89%, +0.50%
VGPRs: 940156 -> 952232 (+1.28%); split: -0.08%, +1.37%
SpillSGPRs: 30286 -> 31109 (+2.72%); split: -0.78%, +3.50%
SpillVGPRs: 1893 -> 1909 (+0.85%)
CodeSize: 87910396 -> 88113592 (+0.23%); split: -0.35%, +0.58%
Scratch: 819200 -> 823296 (+0.50%)
MaxWaves: 205535 -> 202102 (-1.67%); split: +0.05%, -1.72%
Instrs: 17052527 -> 17113484 (+0.36%); split: -0.32%, +0.67%
Cycles: 670794876 -> 669084540 (-0.25%); split: -0.38%, +0.13%
VMEM: 5274728 -> 5388556 (+2.16%); split: +3.10%, -0.94%
SMEM: 1196146 -> 1165850 (-2.53%); split: +2.06%, -4.60%
VClause: 381463 -> 399217 (+4.65%); split: -1.08%, +5.73%
SClause: 666216 -> 631135 (-5.27%); split: -5.44%, +0.18%
Copies: 1292720 -> 1289318 (-0.26%); split: -1.28%, +1.01%
Branches: 467336 -> 473028 (+1.22%); split: -0.67%, +1.89%
PreSGPRs: 766459 -> 772175 (+0.75%); split: -0.53%, +1.28%
PreVGPRs: 819746 -> 825327 (+0.68%); split: -0.05%, +0.73%
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6202>
2020-06-02 12:20:34 +01:00
|
|
|
static bool
|
|
|
|
lower_load_vulkan_descriptor(nir_shader *nir)
|
|
|
|
{
|
|
|
|
nir_function_impl *entry = nir_shader_get_entrypoint(nir);
|
|
|
|
bool progress = false;
|
|
|
|
nir_builder b;
|
|
|
|
|
|
|
|
nir_builder_init(&b, entry);
|
|
|
|
|
|
|
|
nir_foreach_block(block, entry) {
|
|
|
|
nir_foreach_instr_safe(instr, block) {
|
|
|
|
if (instr->type != nir_instr_type_intrinsic)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
|
|
|
|
if (intrin->intrinsic != nir_intrinsic_load_vulkan_descriptor)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
b.cursor = nir_before_instr(&intrin->instr);
|
|
|
|
|
|
|
|
nir_ssa_def *def = nir_vec2(&b,
|
|
|
|
nir_channel(&b, intrin->src[0].ssa, 0),
|
|
|
|
nir_imm_int(&b, 0));
|
|
|
|
nir_ssa_def_rewrite_uses(&intrin->dest.ssa,
|
|
|
|
nir_src_for_ssa(def));
|
|
|
|
|
|
|
|
nir_instr_remove(instr);
|
|
|
|
progress = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return progress;
|
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
nir_shader *
|
|
|
|
radv_shader_compile_to_nir(struct radv_device *device,
|
|
|
|
struct radv_shader_module *module,
|
|
|
|
const char *entrypoint_name,
|
|
|
|
gl_shader_stage stage,
|
2018-05-08 05:57:55 +01:00
|
|
|
const VkSpecializationInfo *spec_info,
|
2019-03-30 13:28:06 +00:00
|
|
|
const VkPipelineCreateFlags flags,
|
2020-03-16 16:29:33 +00:00
|
|
|
const struct radv_pipeline_layout *layout,
|
2020-03-16 17:44:18 +00:00
|
|
|
unsigned subgroup_size, unsigned ballot_bit_size)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
|
|
|
nir_shader *nir;
|
2020-03-12 13:20:05 +00:00
|
|
|
const nir_shader_compiler_options *nir_options =
|
2020-08-06 16:50:33 +01:00
|
|
|
radv_use_llvm_for_stage(device, stage) ? &nir_options_llvm : &nir_options_aco;
|
2020-03-12 13:20:05 +00:00
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
if (module->nir) {
|
|
|
|
/* Some things such as our meta clear/blit code will give us a NIR
|
|
|
|
* shader directly. In that case, we just ignore the SPIR-V entirely
|
|
|
|
* and just use the NIR shader */
|
|
|
|
nir = module->nir;
|
2019-09-17 13:35:22 +01:00
|
|
|
nir->options = nir_options;
|
2018-10-18 21:18:30 +01:00
|
|
|
nir_validate_shader(nir, "in internal shader");
|
2017-09-01 10:41:18 +01:00
|
|
|
|
|
|
|
assert(exec_list_length(&nir->functions) == 1);
|
|
|
|
} else {
|
|
|
|
uint32_t *spirv = (uint32_t *) module->data;
|
|
|
|
assert(module->size % 4 == 0);
|
|
|
|
|
2017-10-11 01:59:20 +01:00
|
|
|
if (device->instance->debug_flags & RADV_DEBUG_DUMP_SPIRV)
|
2019-10-28 15:56:15 +00:00
|
|
|
radv_print_spirv(module->data, module->size, stderr);
|
2017-09-01 10:41:18 +01:00
|
|
|
|
|
|
|
uint32_t num_spec_entries = 0;
|
|
|
|
struct nir_spirv_specialization *spec_entries = NULL;
|
|
|
|
if (spec_info && spec_info->mapEntryCount > 0) {
|
|
|
|
num_spec_entries = spec_info->mapEntryCount;
|
2020-04-22 20:05:13 +01:00
|
|
|
spec_entries = calloc(num_spec_entries, sizeof(*spec_entries));
|
2017-09-01 10:41:18 +01:00
|
|
|
for (uint32_t i = 0; i < num_spec_entries; i++) {
|
|
|
|
VkSpecializationMapEntry entry = spec_info->pMapEntries[i];
|
|
|
|
const void *data = spec_info->pData + entry.offset;
|
|
|
|
assert(data + entry.size <= spec_info->pData + spec_info->dataSize);
|
|
|
|
|
|
|
|
spec_entries[i].id = spec_info->pMapEntries[i].constantID;
|
2020-04-22 19:42:12 +01:00
|
|
|
switch (entry.size) {
|
|
|
|
case 8:
|
2020-04-22 20:05:13 +01:00
|
|
|
spec_entries[i].value.u64 = *(const uint64_t *)data;
|
2020-04-22 19:42:12 +01:00
|
|
|
break;
|
|
|
|
case 4:
|
2020-04-22 20:05:13 +01:00
|
|
|
spec_entries[i].value.u32 = *(const uint32_t *)data;
|
2020-04-22 19:42:12 +01:00
|
|
|
break;
|
|
|
|
case 2:
|
2020-04-22 20:05:13 +01:00
|
|
|
spec_entries[i].value.u16 = *(const uint16_t *)data;
|
2020-04-22 19:42:12 +01:00
|
|
|
break;
|
|
|
|
case 1:
|
2020-04-22 20:05:13 +01:00
|
|
|
spec_entries[i].value.u8 = *(const uint8_t *)data;
|
2020-04-22 19:42:12 +01:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(!"Invalid spec constant size");
|
|
|
|
break;
|
|
|
|
}
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
}
|
2020-08-07 14:23:03 +01:00
|
|
|
|
2020-08-14 12:59:16 +01:00
|
|
|
struct radv_shader_debug_data spirv_debug_data = {
|
2020-08-07 14:23:03 +01:00
|
|
|
.device = device,
|
|
|
|
.module = module,
|
|
|
|
};
|
2017-10-19 01:28:19 +01:00
|
|
|
const struct spirv_to_nir_options spirv_options = {
|
|
|
|
.caps = {
|
2019-05-16 11:26:16 +01:00
|
|
|
.amd_fragment_mask = true,
|
2018-05-09 19:41:23 +01:00
|
|
|
.amd_gcn_shader = true,
|
2020-05-19 13:49:10 +01:00
|
|
|
.amd_image_gather_bias_lod = true,
|
2020-01-03 10:51:14 +00:00
|
|
|
.amd_image_read_write_lod = true,
|
2020-06-26 08:36:46 +01:00
|
|
|
.amd_shader_ballot = true,
|
2020-01-27 10:08:26 +00:00
|
|
|
.amd_shader_explicit_vertex_parameter = true,
|
2018-05-09 19:41:23 +01:00
|
|
|
.amd_trinary_minmax = true,
|
2020-06-05 14:58:36 +01:00
|
|
|
.demote_to_helper_invocation = true,
|
2019-04-19 11:40:37 +01:00
|
|
|
.derivative_group = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.descriptor_array_dynamic_indexing = true,
|
2019-04-29 16:05:13 +01:00
|
|
|
.descriptor_array_non_uniform_indexing = true,
|
|
|
|
.descriptor_indexing = true,
|
2018-01-21 16:13:26 +00:00
|
|
|
.device_group = true,
|
2017-10-19 01:28:19 +01:00
|
|
|
.draw_parameters = true,
|
2019-10-14 10:27:32 +01:00
|
|
|
.float_controls = true,
|
2020-06-05 08:27:01 +01:00
|
|
|
.float16 = device->physical_device->rad_info.has_packed_math_16bit,
|
2020-07-20 17:38:17 +01:00
|
|
|
.float32_atomic_add = true,
|
2017-10-19 01:28:19 +01:00
|
|
|
.float64 = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.geometry_streams = true,
|
2019-12-12 17:22:34 +00:00
|
|
|
.image_ms_array = true,
|
2017-10-19 01:28:19 +01:00
|
|
|
.image_read_without_format = true,
|
|
|
|
.image_write_without_format = true,
|
2020-06-04 10:33:28 +01:00
|
|
|
.int8 = true,
|
2020-05-07 09:54:12 +01:00
|
|
|
.int16 = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.int64 = true,
|
2019-04-16 09:38:24 +01:00
|
|
|
.int64_atomics = true,
|
2020-05-08 18:26:19 +01:00
|
|
|
.min_lod = true,
|
2017-10-19 01:28:19 +01:00
|
|
|
.multiview = true,
|
2019-01-24 01:06:27 +00:00
|
|
|
.physical_storage_buffer_address = true,
|
2019-07-16 16:11:50 +01:00
|
|
|
.post_depth_coverage = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.runtime_descriptor_array = true,
|
2019-10-07 09:26:22 +01:00
|
|
|
.shader_clock = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.shader_viewport_index_layer = true,
|
|
|
|
.stencil_export = true,
|
2020-05-05 08:07:33 +01:00
|
|
|
.storage_8bit = true,
|
|
|
|
.storage_16bit = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.storage_image_ms = true,
|
2018-09-18 14:27:52 +01:00
|
|
|
.subgroup_arithmetic = true,
|
2018-03-06 14:05:13 +00:00
|
|
|
.subgroup_ballot = true,
|
2018-01-21 14:06:10 +00:00
|
|
|
.subgroup_basic = true,
|
2018-03-06 14:05:13 +00:00
|
|
|
.subgroup_quad = true,
|
|
|
|
.subgroup_shuffle = true,
|
|
|
|
.subgroup_vote = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.tessellation = true,
|
2018-10-05 17:04:56 +01:00
|
|
|
.transform_feedback = true,
|
2019-01-07 16:28:23 +00:00
|
|
|
.variable_pointers = true,
|
2020-07-30 11:07:22 +01:00
|
|
|
.vk_memory_model = true,
|
|
|
|
.vk_memory_model_device_scope = true,
|
2018-02-23 12:55:01 +00:00
|
|
|
},
|
2019-05-01 22:15:32 +01:00
|
|
|
.ubo_addr_format = nir_address_format_32bit_index_offset,
|
|
|
|
.ssbo_addr_format = nir_address_format_32bit_index_offset,
|
|
|
|
.phys_ssbo_addr_format = nir_address_format_64bit_global,
|
|
|
|
.push_const_addr_format = nir_address_format_logical,
|
|
|
|
.shared_addr_format = nir_address_format_32bit_offset,
|
2019-05-13 14:39:54 +01:00
|
|
|
.frag_coord_is_sysval = true,
|
2020-08-07 14:23:03 +01:00
|
|
|
.debug = {
|
|
|
|
.func = radv_spirv_nir_debug,
|
|
|
|
.private_data = &spirv_debug_data,
|
|
|
|
},
|
2017-09-01 10:41:18 +01:00
|
|
|
};
|
2019-05-19 08:22:17 +01:00
|
|
|
nir = spirv_to_nir(spirv, module->size / 4,
|
|
|
|
spec_entries, num_spec_entries,
|
|
|
|
stage, entrypoint_name,
|
2019-09-17 13:35:22 +01:00
|
|
|
&spirv_options, nir_options);
|
2017-09-15 03:52:38 +01:00
|
|
|
assert(nir->info.stage == stage);
|
2018-10-18 21:18:30 +01:00
|
|
|
nir_validate_shader(nir, "after spirv_to_nir");
|
2017-09-01 10:41:18 +01:00
|
|
|
|
|
|
|
free(spec_entries);
|
|
|
|
|
|
|
|
/* We have to lower away local constant initializers right before we
|
|
|
|
* inline functions. That way they get properly initialized at the top
|
|
|
|
* of the function and not at the top of its caller.
|
|
|
|
*/
|
2020-02-07 20:18:49 +00:00
|
|
|
NIR_PASS_V(nir, nir_lower_variable_initializers, nir_var_function_temp);
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_returns);
|
|
|
|
NIR_PASS_V(nir, nir_inline_functions);
|
2020-06-11 19:29:02 +01:00
|
|
|
NIR_PASS_V(nir, nir_copy_prop);
|
2018-12-13 17:08:13 +00:00
|
|
|
NIR_PASS_V(nir, nir_opt_deref);
|
2017-09-01 10:41:18 +01:00
|
|
|
|
|
|
|
/* Pick off the single entrypoint that we want */
|
|
|
|
foreach_list_typed_safe(nir_function, func, node, &nir->functions) {
|
2019-05-19 08:11:37 +01:00
|
|
|
if (func->is_entrypoint)
|
|
|
|
func->name = ralloc_strdup(func, "main");
|
|
|
|
else
|
2017-09-01 10:41:18 +01:00
|
|
|
exec_node_remove(&func->node);
|
|
|
|
}
|
|
|
|
assert(exec_list_length(&nir->functions) == 1);
|
|
|
|
|
2018-03-19 04:27:49 +00:00
|
|
|
/* Make sure we lower constant initializers on output variables so that
|
|
|
|
* nir_remove_dead_variables below sees the corresponding stores
|
|
|
|
*/
|
2020-02-07 20:18:49 +00:00
|
|
|
NIR_PASS_V(nir, nir_lower_variable_initializers, nir_var_shader_out);
|
2018-03-19 04:27:49 +00:00
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
/* Now that we've deleted all but the main function, we can go ahead and
|
|
|
|
* lower the rest of the constant initializers.
|
|
|
|
*/
|
2020-02-07 20:18:49 +00:00
|
|
|
NIR_PASS_V(nir, nir_lower_variable_initializers, ~0);
|
2018-03-22 00:30:22 +00:00
|
|
|
|
|
|
|
/* Split member structs. We do this before lower_io_to_temporaries so that
|
|
|
|
* it doesn't lower system values to temporaries by accident.
|
|
|
|
*/
|
|
|
|
NIR_PASS_V(nir, nir_split_var_copies);
|
|
|
|
NIR_PASS_V(nir, nir_split_per_member_structs);
|
|
|
|
|
2020-09-28 09:13:44 +01:00
|
|
|
if (nir->info.stage == MESA_SHADER_FRAGMENT)
|
2019-09-17 13:35:22 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_io_to_vector, nir_var_shader_out);
|
2019-04-05 10:01:39 +01:00
|
|
|
if (nir->info.stage == MESA_SHADER_FRAGMENT)
|
2020-07-01 15:55:46 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_input_attachments,
|
|
|
|
&(nir_input_attachment_options) {
|
|
|
|
.use_fragcoord_sysval = true,
|
2020-07-01 16:19:11 +01:00
|
|
|
.use_layer_id_sysval = false,
|
2020-07-01 15:55:46 +01:00
|
|
|
});
|
2019-04-05 10:01:39 +01:00
|
|
|
|
2018-08-22 11:34:13 +01:00
|
|
|
NIR_PASS_V(nir, nir_remove_dead_variables,
|
2020-05-28 01:59:28 +01:00
|
|
|
nir_var_shader_in | nir_var_shader_out | nir_var_system_value | nir_var_mem_shared,
|
|
|
|
NULL);
|
2018-08-22 11:34:13 +01:00
|
|
|
|
2019-09-05 12:57:11 +01:00
|
|
|
NIR_PASS_V(nir, nir_propagate_invariant);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_system_values);
|
2020-08-21 18:40:45 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_compute_system_values, NULL);
|
2020-08-21 18:18:14 +01:00
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_clip_cull_distance_arrays);
|
2020-07-26 12:22:12 +01:00
|
|
|
|
2020-03-04 16:35:26 +00:00
|
|
|
if (device->instance->debug_flags & RADV_DEBUG_DISCARD_TO_DEMOTE)
|
|
|
|
NIR_PASS_V(nir, nir_lower_discard_to_demote);
|
2020-06-23 17:29:22 +01:00
|
|
|
|
|
|
|
nir_lower_doubles_options lower_doubles =
|
|
|
|
nir->options->lower_doubles_options;
|
|
|
|
|
2020-06-24 08:15:47 +01:00
|
|
|
if (device->physical_device->rad_info.chip_class == GFX6) {
|
|
|
|
/* GFX6 doesn't support v_floor_f64 and the precision
|
|
|
|
* of v_fract_f64 which is used to implement 64-bit
|
|
|
|
* floor is less than what Vulkan requires.
|
|
|
|
*/
|
|
|
|
lower_doubles |= nir_lower_dfloor;
|
|
|
|
}
|
|
|
|
|
2020-06-23 17:29:22 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_doubles, NULL, lower_doubles);
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Vulkan uses the separate-shader linking model */
|
|
|
|
nir->info.separate_shader = true;
|
|
|
|
|
2019-05-19 08:11:37 +01:00
|
|
|
nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir));
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-03-13 10:23:07 +00:00
|
|
|
if (nir->info.stage == MESA_SHADER_GEOMETRY)
|
2019-10-14 17:46:02 +01:00
|
|
|
nir_lower_gs_intrinsics(nir, true);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
static const nir_lower_tex_options tex_options = {
|
|
|
|
.lower_txp = ~0,
|
2019-03-19 18:55:21 +00:00
|
|
|
.lower_tg4_offsets = true,
|
2017-09-01 10:41:18 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
nir_lower_tex(nir, &tex_options);
|
|
|
|
|
|
|
|
nir_lower_vars_to_ssa(nir);
|
2018-05-23 13:31:55 +01:00
|
|
|
|
2018-05-23 13:31:56 +01:00
|
|
|
if (nir->info.stage == MESA_SHADER_VERTEX ||
|
2019-05-10 09:44:20 +01:00
|
|
|
nir->info.stage == MESA_SHADER_GEOMETRY ||
|
|
|
|
nir->info.stage == MESA_SHADER_FRAGMENT) {
|
2018-05-23 13:31:56 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_io_to_temporaries,
|
|
|
|
nir_shader_get_entrypoint(nir), true, true);
|
2019-05-10 09:44:20 +01:00
|
|
|
} else if (nir->info.stage == MESA_SHADER_TESS_EVAL) {
|
2018-05-23 13:31:56 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_io_to_temporaries,
|
|
|
|
nir_shader_get_entrypoint(nir), true, false);
|
|
|
|
}
|
|
|
|
|
2018-05-23 13:31:55 +01:00
|
|
|
nir_split_var_copies(nir);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
nir_lower_global_vars_to_local(nir);
|
2020-05-28 01:59:28 +01:00
|
|
|
nir_remove_dead_variables(nir, nir_var_function_temp, NULL);
|
2020-03-11 14:05:47 +00:00
|
|
|
bool gfx7minus = device->physical_device->rad_info.chip_class <= GFX7;
|
2018-01-21 14:06:10 +00:00
|
|
|
nir_lower_subgroups(nir, &(struct nir_lower_subgroups_options) {
|
2020-03-16 16:29:33 +00:00
|
|
|
.subgroup_size = subgroup_size,
|
2020-03-16 17:44:18 +00:00
|
|
|
.ballot_bit_size = ballot_bit_size,
|
2018-01-21 14:06:10 +00:00
|
|
|
.lower_to_scalar = 1,
|
|
|
|
.lower_subgroup_masks = 1,
|
|
|
|
.lower_shuffle = 1,
|
2018-03-06 14:05:13 +00:00
|
|
|
.lower_shuffle_to_32bit = 1,
|
|
|
|
.lower_vote_eq_to_ballot = 1,
|
2020-03-11 14:05:47 +00:00
|
|
|
.lower_quad_broadcast_dynamic = 1,
|
|
|
|
.lower_quad_broadcast_dynamic_to_const = gfx7minus,
|
2020-06-23 17:38:37 +01:00
|
|
|
.lower_shuffle_to_swizzle_amd = 1,
|
2018-01-21 14:06:10 +00:00
|
|
|
});
|
|
|
|
|
2018-09-24 09:18:48 +01:00
|
|
|
nir_lower_load_const_to_scalar(nir);
|
|
|
|
|
2018-05-08 05:57:55 +01:00
|
|
|
if (!(flags & VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT))
|
2018-10-17 23:42:17 +01:00
|
|
|
radv_optimize_nir(nir, false, true);
|
|
|
|
|
2020-07-26 12:22:12 +01:00
|
|
|
/* call radv_nir_lower_ycbcr_textures() late as there might still be
|
|
|
|
* tex with undef texture/sampler before first optimization */
|
|
|
|
NIR_PASS_V(nir, radv_nir_lower_ycbcr_textures, layout);
|
|
|
|
|
2018-10-17 23:42:17 +01:00
|
|
|
/* We call nir_lower_var_copies() after the first radv_optimize_nir()
|
|
|
|
* to remove any copies introduced by nir_opt_find_array_copies().
|
|
|
|
*/
|
|
|
|
nir_lower_var_copies(nir);
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-05-27 23:09:33 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_explicit_io, nir_var_mem_push_const,
|
|
|
|
nir_address_format_32bit_offset);
|
|
|
|
|
radv: do not lower UBO/SSBO access to offsets
Use nir_lower_explicit_io instead of lowering to offsets. Extra
(useless) additions are removed by lowering load_vulkan_descriptor
to vec2(src.x, 0).
fossils-db (Navi):
Totals from 18236 (13.21% of 138013) affected shaders:
SGPRs: 1172766 -> 1168278 (-0.38%); split: -0.89%, +0.50%
VGPRs: 940156 -> 952232 (+1.28%); split: -0.08%, +1.37%
SpillSGPRs: 30286 -> 31109 (+2.72%); split: -0.78%, +3.50%
SpillVGPRs: 1893 -> 1909 (+0.85%)
CodeSize: 87910396 -> 88113592 (+0.23%); split: -0.35%, +0.58%
Scratch: 819200 -> 823296 (+0.50%)
MaxWaves: 205535 -> 202102 (-1.67%); split: +0.05%, -1.72%
Instrs: 17052527 -> 17113484 (+0.36%); split: -0.32%, +0.67%
Cycles: 670794876 -> 669084540 (-0.25%); split: -0.38%, +0.13%
VMEM: 5274728 -> 5388556 (+2.16%); split: +3.10%, -0.94%
SMEM: 1196146 -> 1165850 (-2.53%); split: +2.06%, -4.60%
VClause: 381463 -> 399217 (+4.65%); split: -1.08%, +5.73%
SClause: 666216 -> 631135 (-5.27%); split: -5.44%, +0.18%
Copies: 1292720 -> 1289318 (-0.26%); split: -1.28%, +1.01%
Branches: 467336 -> 473028 (+1.22%); split: -0.67%, +1.89%
PreSGPRs: 766459 -> 772175 (+0.75%); split: -0.53%, +1.28%
PreVGPRs: 819746 -> 825327 (+0.68%); split: -0.05%, +0.73%
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6202>
2020-06-02 12:20:34 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_explicit_io,
|
|
|
|
nir_var_mem_ubo | nir_var_mem_ssbo,
|
|
|
|
nir_address_format_32bit_index_offset);
|
|
|
|
|
|
|
|
NIR_PASS_V(nir, lower_load_vulkan_descriptor);
|
|
|
|
|
2019-11-07 14:56:35 +00:00
|
|
|
/* Lower deref operations for compute shared memory. */
|
|
|
|
if (nir->info.stage == MESA_SHADER_COMPUTE) {
|
|
|
|
NIR_PASS_V(nir, nir_lower_vars_to_explicit_types,
|
|
|
|
nir_var_mem_shared, shared_var_info);
|
|
|
|
NIR_PASS_V(nir, nir_lower_explicit_io,
|
|
|
|
nir_var_mem_shared, nir_address_format_32bit_offset);
|
|
|
|
}
|
|
|
|
|
2020-06-04 07:27:50 +01:00
|
|
|
nir_lower_explicit_io(nir, nir_var_mem_global,
|
|
|
|
nir_address_format_64bit_global);
|
|
|
|
|
2019-08-30 15:08:47 +01:00
|
|
|
/* Lower large variables that are always constant with load_constant
|
|
|
|
* intrinsics, which get turned into PC-relative loads from a data
|
|
|
|
* section next to the shader.
|
|
|
|
*/
|
|
|
|
NIR_PASS_V(nir, nir_opt_large_constants,
|
|
|
|
glsl_get_natural_size_align_bytes, 16);
|
|
|
|
|
2018-03-08 05:20:48 +00:00
|
|
|
/* Indirect lowering must be called after the radv_optimize_nir() loop
|
|
|
|
* has been called at least once. Otherwise indirect lowering can
|
|
|
|
* bloat the instruction count of the loop and cause it to be
|
|
|
|
* considered too large for unrolling.
|
|
|
|
*/
|
|
|
|
ac_lower_indirect_derefs(nir, device->physical_device->rad_info.chip_class);
|
2018-10-17 23:42:17 +01:00
|
|
|
radv_optimize_nir(nir, flags & VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT, false);
|
2018-03-08 05:20:48 +00:00
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
return nir;
|
|
|
|
}
|
|
|
|
|
2019-05-10 09:44:20 +01:00
|
|
|
static int
|
|
|
|
type_size_vec4(const struct glsl_type *type, bool bindless)
|
|
|
|
{
|
|
|
|
return glsl_count_attribute_slots(type, false);
|
|
|
|
}
|
|
|
|
|
|
|
|
static nir_variable *
|
|
|
|
find_layer_in_var(nir_shader *nir)
|
|
|
|
{
|
|
|
|
nir_variable *var =
|
2020-07-23 05:37:27 +01:00
|
|
|
nir_find_variable_with_location(nir, nir_var_shader_in, VARYING_SLOT_LAYER);
|
|
|
|
if (var != NULL)
|
|
|
|
return var;
|
|
|
|
|
|
|
|
var = nir_variable_create(nir, nir_var_shader_in, glsl_int_type(), "layer id");
|
2019-05-10 09:44:20 +01:00
|
|
|
var->data.location = VARYING_SLOT_LAYER;
|
|
|
|
var->data.interpolation = INTERP_MODE_FLAT;
|
|
|
|
return var;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We use layered rendering to implement multiview, which means we need to map
|
2020-07-01 16:19:11 +01:00
|
|
|
* view_index to gl_Layer. The code generates a load from the layer_id sysval,
|
|
|
|
* but since we don't have a way to get at this information from the fragment
|
|
|
|
* shader, we also need to lower this to the gl_Layer varying. This pass
|
|
|
|
* lowers both to a varying load from the LAYER slot, before lowering io, so
|
|
|
|
* that nir_assign_var_locations() will give the LAYER varying the correct
|
|
|
|
* driver_location.
|
2019-05-10 09:44:20 +01:00
|
|
|
*/
|
|
|
|
|
|
|
|
static bool
|
|
|
|
lower_view_index(nir_shader *nir)
|
|
|
|
{
|
|
|
|
bool progress = false;
|
|
|
|
nir_function_impl *entry = nir_shader_get_entrypoint(nir);
|
|
|
|
nir_builder b;
|
|
|
|
nir_builder_init(&b, entry);
|
|
|
|
|
|
|
|
nir_variable *layer = NULL;
|
|
|
|
nir_foreach_block(block, entry) {
|
|
|
|
nir_foreach_instr_safe(instr, block) {
|
|
|
|
if (instr->type != nir_instr_type_intrinsic)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
nir_intrinsic_instr *load = nir_instr_as_intrinsic(instr);
|
2020-07-01 16:19:11 +01:00
|
|
|
if (load->intrinsic != nir_intrinsic_load_view_index)
|
2019-05-10 09:44:20 +01:00
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!layer)
|
|
|
|
layer = find_layer_in_var(nir);
|
|
|
|
|
|
|
|
b.cursor = nir_before_instr(instr);
|
|
|
|
nir_ssa_def *def = nir_load_var(&b, layer);
|
|
|
|
nir_ssa_def_rewrite_uses(&load->dest.ssa,
|
|
|
|
nir_src_for_ssa(def));
|
|
|
|
|
|
|
|
nir_instr_remove(instr);
|
|
|
|
progress = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return progress;
|
|
|
|
}
|
|
|
|
|
2019-08-29 10:16:44 +01:00
|
|
|
void
|
2020-09-28 16:16:40 +01:00
|
|
|
radv_lower_io(struct radv_device *device, nir_shader *nir)
|
2019-05-10 09:44:20 +01:00
|
|
|
{
|
2020-09-28 16:16:40 +01:00
|
|
|
if (nir->info.stage == MESA_SHADER_COMPUTE)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* TODO: Lower IO for all stages with LLVM. */
|
|
|
|
if (nir->info.stage != MESA_SHADER_FRAGMENT &&
|
|
|
|
radv_use_llvm_for_stage(device, nir->info.stage))
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (nir->info.stage == MESA_SHADER_FRAGMENT) {
|
|
|
|
NIR_PASS_V(nir, lower_view_index);
|
|
|
|
nir_assign_io_var_locations(nir, nir_var_shader_in, &nir->num_inputs,
|
|
|
|
MESA_SHADER_FRAGMENT);
|
|
|
|
}
|
2019-05-10 09:44:20 +01:00
|
|
|
|
2020-09-25 15:11:27 +01:00
|
|
|
NIR_PASS_V(nir, nir_lower_io, nir_var_shader_in | nir_var_shader_out, type_size_vec4, 0);
|
2019-05-10 09:44:20 +01:00
|
|
|
|
|
|
|
/* This pass needs actual constants */
|
|
|
|
nir_opt_constant_folding(nir);
|
|
|
|
|
2020-09-25 15:09:13 +01:00
|
|
|
NIR_PASS_V(nir, nir_io_add_const_offset_to_base,
|
|
|
|
nir_var_shader_in | nir_var_shader_out);
|
2019-05-10 09:44:20 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2020-06-24 00:28:54 +01:00
|
|
|
static void *
|
2017-09-01 10:41:18 +01:00
|
|
|
radv_alloc_shader_memory(struct radv_device *device,
|
|
|
|
struct radv_shader_variant *shader)
|
|
|
|
{
|
|
|
|
mtx_lock(&device->shader_slab_mutex);
|
|
|
|
list_for_each_entry(struct radv_shader_slab, slab, &device->shader_slabs, slabs) {
|
|
|
|
uint64_t offset = 0;
|
|
|
|
list_for_each_entry(struct radv_shader_variant, s, &slab->shaders, slab_list) {
|
|
|
|
if (s->bo_offset - offset >= shader->code_size) {
|
|
|
|
shader->bo = slab->bo;
|
|
|
|
shader->bo_offset = offset;
|
|
|
|
list_addtail(&shader->slab_list, &s->slab_list);
|
|
|
|
mtx_unlock(&device->shader_slab_mutex);
|
|
|
|
return slab->ptr + offset;
|
|
|
|
}
|
|
|
|
offset = align_u64(s->bo_offset + s->code_size, 256);
|
|
|
|
}
|
2020-04-14 20:15:46 +01:00
|
|
|
if (offset <= slab->size && slab->size - offset >= shader->code_size) {
|
2017-09-01 10:41:18 +01:00
|
|
|
shader->bo = slab->bo;
|
|
|
|
shader->bo_offset = offset;
|
|
|
|
list_addtail(&shader->slab_list, &slab->shaders);
|
|
|
|
mtx_unlock(&device->shader_slab_mutex);
|
|
|
|
return slab->ptr + offset;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
mtx_unlock(&device->shader_slab_mutex);
|
|
|
|
struct radv_shader_slab *slab = calloc(1, sizeof(struct radv_shader_slab));
|
|
|
|
|
2020-04-14 20:15:46 +01:00
|
|
|
slab->size = MAX2(256 * 1024, shader->code_size);
|
2017-09-01 10:41:18 +01:00
|
|
|
slab->bo = device->ws->buffer_create(device->ws, slab->size, 256,
|
2018-01-04 14:19:47 +00:00
|
|
|
RADEON_DOMAIN_VRAM,
|
|
|
|
RADEON_FLAG_NO_INTERPROCESS_SHARING |
|
2019-08-20 16:20:42 +01:00
|
|
|
(device->physical_device->rad_info.cpdma_prefetch_writes_memory ?
|
2019-01-27 23:28:05 +00:00
|
|
|
0 : RADEON_FLAG_READ_ONLY),
|
|
|
|
RADV_BO_PRIORITY_SHADER);
|
2020-06-20 20:12:01 +01:00
|
|
|
if (!slab->bo) {
|
|
|
|
free(slab);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
slab->ptr = (char*)device->ws->buffer_map(slab->bo);
|
2020-06-20 20:12:01 +01:00
|
|
|
if (!slab->ptr) {
|
|
|
|
device->ws->buffer_destroy(slab->bo);
|
|
|
|
free(slab);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
list_inithead(&slab->shaders);
|
|
|
|
|
|
|
|
mtx_lock(&device->shader_slab_mutex);
|
|
|
|
list_add(&slab->slabs, &device->shader_slabs);
|
|
|
|
|
|
|
|
shader->bo = slab->bo;
|
|
|
|
shader->bo_offset = 0;
|
|
|
|
list_add(&shader->slab_list, &slab->shaders);
|
|
|
|
mtx_unlock(&device->shader_slab_mutex);
|
|
|
|
return slab->ptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
radv_destroy_shader_slabs(struct radv_device *device)
|
|
|
|
{
|
|
|
|
list_for_each_entry_safe(struct radv_shader_slab, slab, &device->shader_slabs, slabs) {
|
|
|
|
device->ws->buffer_destroy(slab->bo);
|
|
|
|
free(slab);
|
|
|
|
}
|
|
|
|
mtx_destroy(&device->shader_slab_mutex);
|
|
|
|
}
|
|
|
|
|
2018-06-27 09:39:51 +01:00
|
|
|
/* For the UMR disassembler. */
|
|
|
|
#define DEBUGGER_END_OF_CODE_MARKER 0xbf9f0000 /* invalid instruction */
|
|
|
|
#define DEBUGGER_NUM_MARKERS 5
|
|
|
|
|
|
|
|
static unsigned
|
2019-07-01 00:29:24 +01:00
|
|
|
radv_get_shader_binary_size(size_t code_size)
|
2018-06-27 09:39:51 +01:00
|
|
|
{
|
2019-07-01 00:29:24 +01:00
|
|
|
return code_size + DEBUGGER_NUM_MARKERS * 4;
|
2018-06-27 09:39:51 +01:00
|
|
|
}
|
|
|
|
|
2020-08-18 07:49:11 +01:00
|
|
|
static void radv_postprocess_config(const struct radv_device *device,
|
2019-07-01 00:29:24 +01:00
|
|
|
const struct ac_shader_config *config_in,
|
2019-09-03 16:39:23 +01:00
|
|
|
const struct radv_shader_info *info,
|
2019-07-01 00:29:24 +01:00
|
|
|
gl_shader_stage stage,
|
|
|
|
struct ac_shader_config *config_out)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
2020-08-18 07:49:11 +01:00
|
|
|
const struct radv_physical_device *pdevice = device->physical_device;
|
2019-07-01 00:29:24 +01:00
|
|
|
bool scratch_enabled = config_in->scratch_bytes_per_wave > 0;
|
2020-08-25 07:43:56 +01:00
|
|
|
bool trap_enabled = !!device->trap_handler_shader;
|
2017-09-01 10:41:18 +01:00
|
|
|
unsigned vgpr_comp_cnt = 0;
|
2019-07-01 01:19:13 +01:00
|
|
|
unsigned num_input_vgprs = info->num_input_vgprs;
|
|
|
|
|
|
|
|
if (stage == MESA_SHADER_FRAGMENT) {
|
2019-09-25 15:40:07 +01:00
|
|
|
num_input_vgprs = ac_get_fs_input_vgpr_cnt(config_in, NULL, NULL);
|
2019-07-01 01:19:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
unsigned num_vgprs = MAX2(config_in->num_vgprs, num_input_vgprs);
|
|
|
|
/* +3 for scratch wave offset and VCC */
|
|
|
|
unsigned num_sgprs = MAX2(config_in->num_sgprs, info->num_input_sgprs + 3);
|
2019-09-13 14:53:09 +01:00
|
|
|
unsigned num_shared_vgprs = config_in->num_shared_vgprs;
|
|
|
|
/* shared VGPRs are introduced in Navi and are allocated in blocks of 8 (RDNA ref 3.6.5) */
|
|
|
|
assert((pdevice->rad_info.chip_class >= GFX10 && num_shared_vgprs % 8 == 0)
|
|
|
|
|| (pdevice->rad_info.chip_class < GFX10 && num_shared_vgprs == 0));
|
|
|
|
unsigned num_shared_vgpr_blocks = num_shared_vgprs / 8;
|
2020-08-25 07:43:56 +01:00
|
|
|
unsigned excp_en = 0;
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
*config_out = *config_in;
|
2019-07-01 01:19:13 +01:00
|
|
|
config_out->num_vgprs = num_vgprs;
|
|
|
|
config_out->num_sgprs = num_sgprs;
|
2019-09-13 14:53:09 +01:00
|
|
|
config_out->num_shared_vgprs = num_shared_vgprs;
|
2019-07-01 01:19:13 +01:00
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
config_out->rsrc2 = S_00B12C_USER_SGPR(info->num_user_sgprs) |
|
2020-08-25 07:43:56 +01:00
|
|
|
S_00B12C_SCRATCH_EN(scratch_enabled) |
|
|
|
|
S_00B12C_TRAP_PRESENT(trap_enabled);
|
2019-09-09 09:23:30 +01:00
|
|
|
|
2020-08-25 07:43:56 +01:00
|
|
|
if (trap_enabled) {
|
|
|
|
/* Configure the shader exceptions like memory violation, etc.
|
2020-08-18 07:49:11 +01:00
|
|
|
* TODO: Enable (and validate) more exceptions.
|
|
|
|
*/
|
2020-08-25 07:43:56 +01:00
|
|
|
excp_en = 1 << 8; /* mem_viol */
|
2020-08-18 07:49:11 +01:00
|
|
|
}
|
|
|
|
|
2019-09-09 09:23:30 +01:00
|
|
|
if (!pdevice->use_ngg_streamout) {
|
|
|
|
config_out->rsrc2 |= S_00B12C_SO_BASE0_EN(!!info->so.strides[0]) |
|
|
|
|
S_00B12C_SO_BASE1_EN(!!info->so.strides[1]) |
|
|
|
|
S_00B12C_SO_BASE2_EN(!!info->so.strides[2]) |
|
|
|
|
S_00B12C_SO_BASE3_EN(!!info->so.strides[3]) |
|
|
|
|
S_00B12C_SO_EN(!!info->so.num_outputs);
|
|
|
|
}
|
2019-07-01 00:29:24 +01:00
|
|
|
|
2019-07-30 17:32:42 +01:00
|
|
|
config_out->rsrc1 = S_00B848_VGPRS((num_vgprs - 1) /
|
2019-09-03 16:39:23 +01:00
|
|
|
(info->wave_size == 32 ? 8 : 4)) |
|
2019-07-01 00:29:24 +01:00
|
|
|
S_00B848_DX10_CLAMP(1) |
|
2019-07-01 01:19:13 +01:00
|
|
|
S_00B848_FLOAT_MODE(config_out->float_mode);
|
2017-10-17 23:59:16 +01:00
|
|
|
|
2019-06-25 12:33:03 +01:00
|
|
|
if (pdevice->rad_info.chip_class >= GFX10) {
|
|
|
|
config_out->rsrc2 |= S_00B22C_USER_SGPR_MSB_GFX10(info->num_user_sgprs >> 5);
|
|
|
|
} else {
|
|
|
|
config_out->rsrc1 |= S_00B228_SGPRS((num_sgprs - 1) / 8);
|
2019-07-23 13:55:16 +01:00
|
|
|
config_out->rsrc2 |= S_00B22C_USER_SGPR_MSB_GFX9(info->num_user_sgprs >> 5);
|
2019-06-25 12:33:03 +01:00
|
|
|
}
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
switch (stage) {
|
|
|
|
case MESA_SHADER_TESS_EVAL:
|
2019-07-08 22:44:32 +01:00
|
|
|
if (info->is_ngg) {
|
|
|
|
config_out->rsrc1 |= S_00B228_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10);
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B22C_OC_LDS_EN(1) |
|
|
|
|
S_00B22C_EXCP_EN(excp_en);
|
2019-07-08 22:44:32 +01:00
|
|
|
} else if (info->tes.as_es) {
|
2019-07-01 00:29:24 +01:00
|
|
|
assert(pdevice->rad_info.chip_class <= GFX8);
|
2019-09-03 16:39:23 +01:00
|
|
|
vgpr_comp_cnt = info->uses_prim_id ? 3 : 2;
|
2019-07-08 22:44:32 +01:00
|
|
|
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B12C_OC_LDS_EN(1) |
|
|
|
|
S_00B12C_EXCP_EN(excp_en);
|
2019-06-26 14:11:00 +01:00
|
|
|
} else {
|
2019-09-03 16:39:23 +01:00
|
|
|
bool enable_prim_id = info->tes.export_prim_id || info->uses_prim_id;
|
2019-06-26 14:11:00 +01:00
|
|
|
vgpr_comp_cnt = enable_prim_id ? 3 : 2;
|
2019-07-06 11:31:25 +01:00
|
|
|
|
|
|
|
config_out->rsrc1 |= S_00B128_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10);
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B12C_OC_LDS_EN(1) |
|
|
|
|
S_00B12C_EXCP_EN(excp_en);
|
2019-06-26 14:11:00 +01:00
|
|
|
}
|
2019-09-13 14:53:09 +01:00
|
|
|
config_out->rsrc2 |= S_00B22C_SHARED_VGPR_CNT(num_shared_vgpr_blocks);
|
2017-10-17 23:59:16 +01:00
|
|
|
break;
|
2017-09-01 10:41:18 +01:00
|
|
|
case MESA_SHADER_TESS_CTRL:
|
2019-07-01 00:29:24 +01:00
|
|
|
if (pdevice->rad_info.chip_class >= GFX9) {
|
2019-06-26 14:11:03 +01:00
|
|
|
/* We need at least 2 components for LS.
|
|
|
|
* VGPR0-3: (VertexID, RelAutoindex, InstanceID / StepRate0, InstanceID).
|
|
|
|
* StepRate0 is set to 1. so that VGPR3 doesn't have to be loaded.
|
|
|
|
*/
|
2019-07-08 22:44:32 +01:00
|
|
|
if (pdevice->rad_info.chip_class >= GFX10) {
|
2019-09-03 16:39:23 +01:00
|
|
|
vgpr_comp_cnt = info->vs.needs_instance_id ? 3 : 1;
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B42C_LDS_SIZE_GFX10(info->tcs.num_lds_blocks) |
|
2020-08-21 13:09:58 +01:00
|
|
|
S_00B42C_EXCP_EN_GFX6(excp_en);
|
2019-07-08 22:44:32 +01:00
|
|
|
} else {
|
2019-09-03 16:39:23 +01:00
|
|
|
vgpr_comp_cnt = info->vs.needs_instance_id ? 2 : 1;
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B42C_LDS_SIZE_GFX9(info->tcs.num_lds_blocks) |
|
|
|
|
S_00B42C_EXCP_EN_GFX9(excp_en);
|
2019-07-08 22:44:32 +01:00
|
|
|
}
|
2018-05-11 08:46:46 +01:00
|
|
|
} else {
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B12C_OC_LDS_EN(1) |
|
|
|
|
S_00B12C_EXCP_EN(excp_en);
|
2018-05-11 08:46:46 +01:00
|
|
|
}
|
2019-07-12 11:17:16 +01:00
|
|
|
config_out->rsrc1 |= S_00B428_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10) |
|
|
|
|
S_00B848_WGP_MODE(pdevice->rad_info.chip_class >= GFX10);
|
2019-09-13 14:53:09 +01:00
|
|
|
config_out->rsrc2 |= S_00B42C_SHARED_VGPR_CNT(num_shared_vgpr_blocks);
|
2017-09-01 10:41:18 +01:00
|
|
|
break;
|
|
|
|
case MESA_SHADER_VERTEX:
|
2019-07-05 07:33:06 +01:00
|
|
|
if (info->is_ngg) {
|
|
|
|
config_out->rsrc1 |= S_00B228_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10);
|
|
|
|
} else if (info->vs.as_ls) {
|
2019-07-01 00:29:24 +01:00
|
|
|
assert(pdevice->rad_info.chip_class <= GFX8);
|
2019-06-26 14:11:03 +01:00
|
|
|
/* We need at least 2 components for LS.
|
|
|
|
* VGPR0-3: (VertexID, RelAutoindex, InstanceID / StepRate0, InstanceID).
|
|
|
|
* StepRate0 is set to 1. so that VGPR3 doesn't have to be loaded.
|
|
|
|
*/
|
2019-09-03 16:39:23 +01:00
|
|
|
vgpr_comp_cnt = info->vs.needs_instance_id ? 2 : 1;
|
2019-07-01 00:29:24 +01:00
|
|
|
} else if (info->vs.as_es) {
|
|
|
|
assert(pdevice->rad_info.chip_class <= GFX8);
|
2019-06-26 14:11:03 +01:00
|
|
|
/* VGPR0-3: (VertexID, InstanceID / StepRate0, ...) */
|
2019-09-03 16:39:23 +01:00
|
|
|
vgpr_comp_cnt = info->vs.needs_instance_id ? 1 : 0;
|
2019-06-26 14:11:03 +01:00
|
|
|
} else {
|
|
|
|
/* VGPR0-3: (VertexID, InstanceID / StepRate0, PrimID, InstanceID)
|
|
|
|
* If PrimID is disabled. InstanceID / StepRate1 is loaded instead.
|
|
|
|
* StepRate0 is set to 1. so that VGPR3 doesn't have to be loaded.
|
|
|
|
*/
|
2019-09-03 16:39:23 +01:00
|
|
|
if (info->vs.needs_instance_id && pdevice->rad_info.chip_class >= GFX10) {
|
2019-08-21 00:50:53 +01:00
|
|
|
vgpr_comp_cnt = 3;
|
|
|
|
} else if (info->vs.export_prim_id) {
|
2019-06-26 14:11:03 +01:00
|
|
|
vgpr_comp_cnt = 2;
|
2019-09-03 16:39:23 +01:00
|
|
|
} else if (info->vs.needs_instance_id) {
|
2019-08-21 00:50:53 +01:00
|
|
|
vgpr_comp_cnt = 1;
|
2019-06-26 14:11:03 +01:00
|
|
|
} else {
|
|
|
|
vgpr_comp_cnt = 0;
|
|
|
|
}
|
2019-07-06 11:31:25 +01:00
|
|
|
|
|
|
|
config_out->rsrc1 |= S_00B128_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10);
|
2019-06-26 14:11:03 +01:00
|
|
|
}
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B12C_SHARED_VGPR_CNT(num_shared_vgpr_blocks) |
|
|
|
|
S_00B12C_EXCP_EN(excp_en);
|
2017-09-01 10:41:18 +01:00
|
|
|
break;
|
|
|
|
case MESA_SHADER_FRAGMENT:
|
2019-07-06 11:31:25 +01:00
|
|
|
config_out->rsrc1 |= S_00B028_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10);
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B02C_SHARED_VGPR_CNT(num_shared_vgpr_blocks) |
|
|
|
|
S_00B02C_TRAP_PRESENT(1) |
|
|
|
|
S_00B02C_EXCP_EN(excp_en);
|
2019-07-06 11:31:25 +01:00
|
|
|
break;
|
2019-06-26 14:11:01 +01:00
|
|
|
case MESA_SHADER_GEOMETRY:
|
2019-07-12 11:17:16 +01:00
|
|
|
config_out->rsrc1 |= S_00B228_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10) |
|
|
|
|
S_00B848_WGP_MODE(pdevice->rad_info.chip_class >= GFX10);
|
2020-08-25 07:43:56 +01:00
|
|
|
config_out->rsrc2 |= S_00B22C_SHARED_VGPR_CNT(num_shared_vgpr_blocks) |
|
|
|
|
S_00B22C_EXCP_EN(excp_en);
|
2017-09-01 10:41:18 +01:00
|
|
|
break;
|
2018-05-11 08:46:46 +01:00
|
|
|
case MESA_SHADER_COMPUTE:
|
2019-07-12 11:17:16 +01:00
|
|
|
config_out->rsrc1 |= S_00B848_MEM_ORDERED(pdevice->rad_info.chip_class >= GFX10) |
|
|
|
|
S_00B848_WGP_MODE(pdevice->rad_info.chip_class >= GFX10);
|
2019-07-01 00:29:24 +01:00
|
|
|
config_out->rsrc2 |=
|
2019-09-03 16:39:23 +01:00
|
|
|
S_00B84C_TGID_X_EN(info->cs.uses_block_id[0]) |
|
|
|
|
S_00B84C_TGID_Y_EN(info->cs.uses_block_id[1]) |
|
|
|
|
S_00B84C_TGID_Z_EN(info->cs.uses_block_id[2]) |
|
|
|
|
S_00B84C_TIDIG_COMP_CNT(info->cs.uses_thread_id[2] ? 2 :
|
|
|
|
info->cs.uses_thread_id[1] ? 1 : 0) |
|
|
|
|
S_00B84C_TG_SIZE_EN(info->cs.uses_local_invocation_idx) |
|
2020-08-25 07:43:56 +01:00
|
|
|
S_00B84C_LDS_SIZE(config_in->lds_size) |
|
|
|
|
S_00B84C_EXCP_EN(excp_en);
|
2019-09-13 14:53:09 +01:00
|
|
|
config_out->rsrc3 |= S_00B8A0_SHARED_VGPR_CNT(num_shared_vgpr_blocks);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
unreachable("unsupported shader type");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2019-07-16 15:39:16 +01:00
|
|
|
if (pdevice->rad_info.chip_class >= GFX10 && info->is_ngg &&
|
2019-07-09 07:44:01 +01:00
|
|
|
(stage == MESA_SHADER_VERTEX || stage == MESA_SHADER_TESS_EVAL || stage == MESA_SHADER_GEOMETRY)) {
|
2019-07-05 07:33:06 +01:00
|
|
|
unsigned gs_vgpr_comp_cnt, es_vgpr_comp_cnt;
|
2019-07-11 07:44:15 +01:00
|
|
|
gl_shader_stage es_stage = stage;
|
|
|
|
if (stage == MESA_SHADER_GEOMETRY)
|
|
|
|
es_stage = info->gs.es_type;
|
2019-07-05 07:33:06 +01:00
|
|
|
|
|
|
|
/* VGPR5-8: (VertexID, UserVGPR0, UserVGPR1, UserVGPR2 / InstanceID) */
|
2019-07-11 07:44:15 +01:00
|
|
|
if (es_stage == MESA_SHADER_VERTEX) {
|
2019-09-03 16:39:23 +01:00
|
|
|
es_vgpr_comp_cnt = info->vs.needs_instance_id ? 3 : 0;
|
2019-07-11 07:44:15 +01:00
|
|
|
} else if (es_stage == MESA_SHADER_TESS_EVAL) {
|
2019-09-03 16:39:23 +01:00
|
|
|
bool enable_prim_id = info->tes.export_prim_id || info->uses_prim_id;
|
2019-07-09 07:27:30 +01:00
|
|
|
es_vgpr_comp_cnt = enable_prim_id ? 3 : 2;
|
2019-07-18 23:00:03 +01:00
|
|
|
} else
|
|
|
|
unreachable("Unexpected ES shader stage");
|
2019-07-08 22:44:32 +01:00
|
|
|
|
|
|
|
bool tes_triangles = stage == MESA_SHADER_TESS_EVAL &&
|
|
|
|
info->tes.primitive_mode >= 4; /* GL_TRIANGLES */
|
2019-09-03 16:39:23 +01:00
|
|
|
if (info->uses_invocation_id || stage == MESA_SHADER_VERTEX) {
|
2019-07-08 22:44:32 +01:00
|
|
|
gs_vgpr_comp_cnt = 3; /* VGPR3 contains InvocationID. */
|
2019-09-03 16:39:23 +01:00
|
|
|
} else if (info->uses_prim_id) {
|
2019-07-08 22:44:32 +01:00
|
|
|
gs_vgpr_comp_cnt = 2; /* VGPR2 contains PrimitiveID. */
|
|
|
|
} else if (info->gs.vertices_in >= 3 || tes_triangles) {
|
|
|
|
gs_vgpr_comp_cnt = 1; /* VGPR1 contains offsets 2, 3 */
|
|
|
|
} else {
|
|
|
|
gs_vgpr_comp_cnt = 0; /* VGPR0 contains offsets 0, 1 */
|
|
|
|
}
|
2019-07-05 07:33:06 +01:00
|
|
|
|
2019-07-12 11:17:16 +01:00
|
|
|
config_out->rsrc1 |= S_00B228_GS_VGPR_COMP_CNT(gs_vgpr_comp_cnt) |
|
|
|
|
S_00B228_WGP_MODE(1);
|
2019-07-05 07:33:06 +01:00
|
|
|
config_out->rsrc2 |= S_00B22C_ES_VGPR_COMP_CNT(es_vgpr_comp_cnt) |
|
2019-07-15 17:46:48 +01:00
|
|
|
S_00B22C_LDS_SIZE(config_in->lds_size) |
|
|
|
|
S_00B22C_OC_LDS_EN(es_stage == MESA_SHADER_TESS_EVAL);
|
2019-07-05 07:33:06 +01:00
|
|
|
} else if (pdevice->rad_info.chip_class >= GFX9 &&
|
|
|
|
stage == MESA_SHADER_GEOMETRY) {
|
2019-07-01 00:29:24 +01:00
|
|
|
unsigned es_type = info->gs.es_type;
|
2018-01-09 15:01:10 +00:00
|
|
|
unsigned gs_vgpr_comp_cnt, es_vgpr_comp_cnt;
|
|
|
|
|
|
|
|
if (es_type == MESA_SHADER_VERTEX) {
|
2019-06-26 14:11:03 +01:00
|
|
|
/* VGPR0-3: (VertexID, InstanceID / StepRate0, ...) */
|
2019-09-03 16:39:23 +01:00
|
|
|
if (info->vs.needs_instance_id) {
|
2019-07-23 10:52:36 +01:00
|
|
|
es_vgpr_comp_cnt = pdevice->rad_info.chip_class >= GFX10 ? 3 : 1;
|
|
|
|
} else {
|
|
|
|
es_vgpr_comp_cnt = 0;
|
|
|
|
}
|
2018-01-09 15:01:10 +00:00
|
|
|
} else if (es_type == MESA_SHADER_TESS_EVAL) {
|
2019-09-03 16:39:23 +01:00
|
|
|
es_vgpr_comp_cnt = info->uses_prim_id ? 3 : 2;
|
2018-01-09 15:01:10 +00:00
|
|
|
} else {
|
2018-01-17 22:23:02 +00:00
|
|
|
unreachable("invalid shader ES type");
|
2018-01-09 15:01:10 +00:00
|
|
|
}
|
2017-12-20 19:56:57 +00:00
|
|
|
|
|
|
|
/* If offsets 4, 5 are used, GS_VGPR_COMP_CNT is ignored and
|
|
|
|
* VGPR[0:4] are always loaded.
|
|
|
|
*/
|
2019-09-03 16:39:23 +01:00
|
|
|
if (info->uses_invocation_id) {
|
2017-12-20 19:56:57 +00:00
|
|
|
gs_vgpr_comp_cnt = 3; /* VGPR3 contains InvocationID. */
|
2019-09-03 16:39:23 +01:00
|
|
|
} else if (info->uses_prim_id) {
|
2017-12-20 19:56:57 +00:00
|
|
|
gs_vgpr_comp_cnt = 2; /* VGPR2 contains PrimitiveID. */
|
2019-07-01 00:29:24 +01:00
|
|
|
} else if (info->gs.vertices_in >= 3) {
|
2018-01-05 16:18:52 +00:00
|
|
|
gs_vgpr_comp_cnt = 1; /* VGPR1 contains offsets 2, 3 */
|
2018-05-11 08:46:46 +01:00
|
|
|
} else {
|
2018-01-05 16:18:52 +00:00
|
|
|
gs_vgpr_comp_cnt = 0; /* VGPR0 contains offsets 0, 1 */
|
2018-05-11 08:46:46 +01:00
|
|
|
}
|
2017-12-20 19:56:57 +00:00
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
config_out->rsrc1 |= S_00B228_GS_VGPR_COMP_CNT(gs_vgpr_comp_cnt);
|
|
|
|
config_out->rsrc2 |= S_00B22C_ES_VGPR_COMP_CNT(es_vgpr_comp_cnt) |
|
2019-06-30 00:47:30 +01:00
|
|
|
S_00B22C_OC_LDS_EN(es_type == MESA_SHADER_TESS_EVAL);
|
2019-07-01 00:29:24 +01:00
|
|
|
} else if (pdevice->rad_info.chip_class >= GFX9 &&
|
2018-05-11 08:46:46 +01:00
|
|
|
stage == MESA_SHADER_TESS_CTRL) {
|
2019-07-01 00:29:24 +01:00
|
|
|
config_out->rsrc1 |= S_00B428_LS_VGPR_COMP_CNT(vgpr_comp_cnt);
|
2018-05-11 08:46:46 +01:00
|
|
|
} else {
|
2019-07-01 00:29:24 +01:00
|
|
|
config_out->rsrc1 |= S_00B128_VGPR_COMP_CNT(vgpr_comp_cnt);
|
2018-05-11 08:46:46 +01:00
|
|
|
}
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_shader_variant *
|
|
|
|
radv_shader_variant_create(struct radv_device *device,
|
2019-05-31 00:06:27 +01:00
|
|
|
const struct radv_shader_binary *binary,
|
|
|
|
bool keep_shader_info)
|
2019-07-01 00:29:24 +01:00
|
|
|
{
|
|
|
|
struct ac_shader_config config = {0};
|
|
|
|
struct ac_rtld_binary rtld_binary = {0};
|
|
|
|
struct radv_shader_variant *variant = calloc(1, sizeof(struct radv_shader_variant));
|
|
|
|
if (!variant)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
variant->ref_count = 1;
|
|
|
|
|
|
|
|
if (binary->type == RADV_BINARY_TYPE_RTLD) {
|
2019-09-03 12:01:54 +01:00
|
|
|
struct ac_rtld_symbol lds_symbols[2];
|
2019-07-01 00:29:24 +01:00
|
|
|
unsigned num_lds_symbols = 0;
|
|
|
|
const char *elf_data = (const char *)((struct radv_shader_binary_rtld *)binary)->data;
|
|
|
|
size_t elf_size = ((struct radv_shader_binary_rtld *)binary)->elf_size;
|
|
|
|
|
|
|
|
if (device->physical_device->rad_info.chip_class >= GFX9 &&
|
2019-09-03 12:01:54 +01:00
|
|
|
(binary->stage == MESA_SHADER_GEOMETRY || binary->info.is_ngg) &&
|
|
|
|
!binary->is_gs_copy_shader) {
|
2019-07-01 00:29:24 +01:00
|
|
|
/* We add this symbol even on LLVM <= 8 to ensure that
|
|
|
|
* shader->config.lds_size is set correctly below.
|
|
|
|
*/
|
|
|
|
struct ac_rtld_symbol *sym = &lds_symbols[num_lds_symbols++];
|
|
|
|
sym->name = "esgs_ring";
|
2019-09-18 08:58:54 +01:00
|
|
|
sym->size = binary->info.ngg_info.esgs_ring_size;
|
2019-07-01 00:29:24 +01:00
|
|
|
sym->align = 64 * 1024;
|
2019-09-03 12:01:54 +01:00
|
|
|
}
|
2019-07-11 07:44:16 +01:00
|
|
|
|
2019-09-03 12:01:54 +01:00
|
|
|
if (binary->info.is_ngg &&
|
|
|
|
binary->stage == MESA_SHADER_GEOMETRY) {
|
|
|
|
struct ac_rtld_symbol *sym = &lds_symbols[num_lds_symbols++];
|
|
|
|
sym->name = "ngg_emit";
|
|
|
|
sym->size = binary->info.ngg_info.ngg_emit_size * 4;
|
|
|
|
sym->align = 4;
|
2019-07-01 00:29:24 +01:00
|
|
|
}
|
2019-07-30 17:32:42 +01:00
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
struct ac_rtld_open_info open_info = {
|
|
|
|
.info = &device->physical_device->rad_info,
|
|
|
|
.shader_type = binary->stage,
|
2019-09-03 16:39:23 +01:00
|
|
|
.wave_size = binary->info.wave_size,
|
2019-07-01 00:29:24 +01:00
|
|
|
.num_parts = 1,
|
|
|
|
.elf_ptrs = &elf_data,
|
|
|
|
.elf_sizes = &elf_size,
|
|
|
|
.num_shared_lds_symbols = num_lds_symbols,
|
|
|
|
.shared_lds_symbols = lds_symbols,
|
|
|
|
};
|
|
|
|
|
|
|
|
if (!ac_rtld_open(&rtld_binary, open_info)) {
|
|
|
|
free(variant);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2020-01-29 04:35:49 +00:00
|
|
|
if (!ac_rtld_read_config(&device->physical_device->rad_info,
|
|
|
|
&rtld_binary, &config)) {
|
2019-07-01 00:29:24 +01:00
|
|
|
ac_rtld_close(&rtld_binary);
|
|
|
|
free(variant);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (rtld_binary.lds_size > 0) {
|
|
|
|
unsigned alloc_granularity = device->physical_device->rad_info.chip_class >= GFX7 ? 512 : 256;
|
|
|
|
config.lds_size = align(rtld_binary.lds_size, alloc_granularity) / alloc_granularity;
|
|
|
|
}
|
|
|
|
|
|
|
|
variant->code_size = rtld_binary.rx_size;
|
2019-08-29 16:15:46 +01:00
|
|
|
variant->exec_size = rtld_binary.exec_size;
|
2019-07-01 00:29:24 +01:00
|
|
|
} else {
|
|
|
|
assert(binary->type == RADV_BINARY_TYPE_LEGACY);
|
|
|
|
config = ((struct radv_shader_binary_legacy *)binary)->config;
|
2019-08-29 16:15:46 +01:00
|
|
|
variant->code_size = radv_get_shader_binary_size(((struct radv_shader_binary_legacy *)binary)->code_size);
|
2019-09-17 13:35:22 +01:00
|
|
|
variant->exec_size = ((struct radv_shader_binary_legacy *)binary)->exec_size;
|
2019-07-01 00:29:24 +01:00
|
|
|
}
|
|
|
|
|
2019-09-03 16:39:23 +01:00
|
|
|
variant->info = binary->info;
|
2020-08-18 07:49:11 +01:00
|
|
|
radv_postprocess_config(device, &config, &binary->info,
|
2019-07-01 00:29:24 +01:00
|
|
|
binary->stage, &variant->config);
|
2019-07-31 04:57:16 +01:00
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
void *dest_ptr = radv_alloc_shader_memory(device, variant);
|
2020-06-20 20:12:01 +01:00
|
|
|
if (!dest_ptr) {
|
|
|
|
if (binary->type == RADV_BINARY_TYPE_RTLD)
|
|
|
|
ac_rtld_close(&rtld_binary);
|
|
|
|
free(variant);
|
|
|
|
return NULL;
|
|
|
|
}
|
2019-07-01 00:29:24 +01:00
|
|
|
|
|
|
|
if (binary->type == RADV_BINARY_TYPE_RTLD) {
|
|
|
|
struct radv_shader_binary_rtld* bin = (struct radv_shader_binary_rtld *)binary;
|
|
|
|
struct ac_rtld_upload_info info = {
|
|
|
|
.binary = &rtld_binary,
|
|
|
|
.rx_va = radv_buffer_get_va(variant->bo) + variant->bo_offset,
|
|
|
|
.rx_ptr = dest_ptr,
|
|
|
|
};
|
|
|
|
|
|
|
|
if (!ac_rtld_upload(&info)) {
|
|
|
|
radv_shader_variant_destroy(device, variant);
|
|
|
|
ac_rtld_close(&rtld_binary);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2019-05-31 00:06:27 +01:00
|
|
|
if (keep_shader_info ||
|
2019-07-23 08:55:24 +01:00
|
|
|
(device->instance->debug_flags & RADV_DEBUG_DUMP_SHADERS)) {
|
2019-07-17 05:20:55 +01:00
|
|
|
const char *disasm_data;
|
|
|
|
size_t disasm_size;
|
|
|
|
if (!ac_rtld_get_section_by_name(&rtld_binary, ".AMDGPU.disasm", &disasm_data, &disasm_size)) {
|
|
|
|
radv_shader_variant_destroy(device, variant);
|
|
|
|
ac_rtld_close(&rtld_binary);
|
|
|
|
return NULL;
|
|
|
|
}
|
2019-07-01 00:29:24 +01:00
|
|
|
|
2019-09-25 11:48:04 +01:00
|
|
|
variant->ir_string = bin->llvm_ir_size ? strdup((const char*)(bin->data + bin->elf_size)) : NULL;
|
2019-07-17 05:20:55 +01:00
|
|
|
variant->disasm_string = malloc(disasm_size + 1);
|
|
|
|
memcpy(variant->disasm_string, disasm_data, disasm_size);
|
|
|
|
variant->disasm_string[disasm_size] = 0;
|
|
|
|
}
|
2019-07-01 00:29:24 +01:00
|
|
|
|
|
|
|
ac_rtld_close(&rtld_binary);
|
|
|
|
} else {
|
|
|
|
struct radv_shader_binary_legacy* bin = (struct radv_shader_binary_legacy *)binary;
|
2019-12-04 14:46:31 +00:00
|
|
|
memcpy(dest_ptr, bin->data + bin->stats_size, bin->code_size);
|
2019-07-01 00:29:24 +01:00
|
|
|
|
|
|
|
/* Add end-of-code markers for the UMR disassembler. */
|
|
|
|
uint32_t *ptr32 = (uint32_t *)dest_ptr + bin->code_size / 4;
|
|
|
|
for (unsigned i = 0; i < DEBUGGER_NUM_MARKERS; i++)
|
|
|
|
ptr32[i] = DEBUGGER_END_OF_CODE_MARKER;
|
|
|
|
|
2019-12-04 14:46:31 +00:00
|
|
|
variant->ir_string = bin->ir_size ? strdup((const char*)(bin->data + bin->stats_size + bin->code_size)) : NULL;
|
|
|
|
variant->disasm_string = bin->disasm_size ? strdup((const char*)(bin->data + bin->stats_size + bin->code_size + bin->ir_size)) : NULL;
|
|
|
|
|
|
|
|
if (bin->stats_size) {
|
|
|
|
variant->statistics = calloc(bin->stats_size, 1);
|
|
|
|
memcpy(variant->statistics, bin->data, bin->stats_size);
|
|
|
|
}
|
2019-07-01 00:29:24 +01:00
|
|
|
}
|
|
|
|
return variant;
|
|
|
|
}
|
|
|
|
|
2019-06-01 19:54:35 +01:00
|
|
|
static char *
|
|
|
|
radv_dump_nir_shaders(struct nir_shader * const *shaders,
|
|
|
|
int shader_count)
|
|
|
|
{
|
|
|
|
char *data = NULL;
|
|
|
|
char *ret = NULL;
|
|
|
|
size_t size = 0;
|
|
|
|
FILE *f = open_memstream(&data, &size);
|
|
|
|
if (f) {
|
|
|
|
for (int i = 0; i < shader_count; ++i)
|
|
|
|
nir_print_shader(shaders[i], f);
|
|
|
|
fclose(f);
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = malloc(size + 1);
|
|
|
|
if (ret) {
|
|
|
|
memcpy(ret, data, size);
|
|
|
|
ret[size] = 0;
|
|
|
|
}
|
|
|
|
free(data);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-09-01 15:51:12 +01:00
|
|
|
static struct radv_shader_variant *
|
2019-07-01 00:29:24 +01:00
|
|
|
shader_variant_compile(struct radv_device *device,
|
|
|
|
struct radv_shader_module *module,
|
|
|
|
struct nir_shader * const *shaders,
|
|
|
|
int shader_count,
|
|
|
|
gl_shader_stage stage,
|
2019-09-03 09:29:19 +01:00
|
|
|
struct radv_shader_info *info,
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_nir_compiler_options *options,
|
|
|
|
bool gs_copy_shader,
|
2020-08-18 17:44:07 +01:00
|
|
|
bool trap_handler_shader,
|
2019-05-31 00:06:27 +01:00
|
|
|
bool keep_shader_info,
|
2020-03-19 15:09:31 +00:00
|
|
|
bool keep_statistic_info,
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_shader_binary **binary_out)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
|
|
|
enum radeon_family chip_family = device->physical_device->rad_info.family;
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_shader_binary *binary = NULL;
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-08-14 12:59:16 +01:00
|
|
|
struct radv_shader_debug_data debug_data = {
|
|
|
|
.device = device,
|
|
|
|
.module = module,
|
|
|
|
};
|
|
|
|
|
2017-09-01 15:51:12 +01:00
|
|
|
options->family = chip_family;
|
|
|
|
options->chip_class = device->physical_device->rad_info.chip_class;
|
2018-05-11 15:36:02 +01:00
|
|
|
options->dump_shader = radv_can_dump_shader(device, module, gs_copy_shader);
|
2018-03-14 09:28:49 +00:00
|
|
|
options->dump_preoptir = options->dump_shader &&
|
2018-01-19 11:12:02 +00:00
|
|
|
device->instance->debug_flags & RADV_DEBUG_PREOPTIR;
|
2019-09-25 11:48:04 +01:00
|
|
|
options->record_ir = keep_shader_info;
|
2020-03-19 15:09:31 +00:00
|
|
|
options->record_stats = keep_statistic_info;
|
2018-06-14 13:28:58 +01:00
|
|
|
options->check_ir = device->instance->debug_flags & RADV_DEBUG_CHECKIR;
|
2018-02-19 07:14:04 +00:00
|
|
|
options->tess_offchip_block_dw_size = device->tess_offchip_block_dw_size;
|
2018-05-16 15:02:04 +01:00
|
|
|
options->address32_hi = device->physical_device->rad_info.address32_hi;
|
2019-08-23 07:55:53 +01:00
|
|
|
options->has_ls_vgpr_init_bug = device->physical_device->rad_info.has_ls_vgpr_init_bug;
|
2019-09-09 09:23:30 +01:00
|
|
|
options->use_ngg_streamout = device->physical_device->use_ngg_streamout;
|
2020-06-05 13:24:33 +01:00
|
|
|
options->enable_mrt_output_nan_fixup = device->instance->enable_mrt_output_nan_fixup;
|
2020-08-14 12:59:16 +01:00
|
|
|
options->debug.func = radv_compiler_debug;
|
|
|
|
options->debug.private_data = &debug_data;
|
2019-08-03 23:48:05 +01:00
|
|
|
|
2019-11-11 17:05:03 +00:00
|
|
|
struct radv_shader_args args = {};
|
|
|
|
args.options = options;
|
|
|
|
args.shader_info = info;
|
|
|
|
args.is_gs_copy_shader = gs_copy_shader;
|
2020-08-18 17:44:07 +01:00
|
|
|
args.is_trap_handler_shader = trap_handler_shader;
|
|
|
|
|
2019-11-11 17:05:03 +00:00
|
|
|
radv_declare_shader_args(&args,
|
|
|
|
gs_copy_shader ? MESA_SHADER_VERTEX
|
|
|
|
: shaders[shader_count - 1]->info.stage,
|
|
|
|
shader_count >= 2,
|
|
|
|
shader_count >= 2 ? shaders[shader_count - 2]->info.stage
|
|
|
|
: MESA_SHADER_VERTEX);
|
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
if (radv_use_llvm_for_stage(device, stage) ||
|
2020-03-12 13:20:05 +00:00
|
|
|
options->dump_shader || options->record_ir)
|
2019-09-17 13:35:22 +01:00
|
|
|
ac_init_llvm_once();
|
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
if (radv_use_llvm_for_stage(device, stage)) {
|
2020-03-12 13:49:55 +00:00
|
|
|
llvm_compile_shader(device, shader_count, shaders, &binary, &args);
|
2020-06-11 10:43:02 +01:00
|
|
|
} else {
|
|
|
|
aco_compile_shader(shader_count, shaders, &binary, &args);
|
2019-09-17 13:35:22 +01:00
|
|
|
}
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-03-12 13:49:55 +00:00
|
|
|
binary->info = *info;
|
|
|
|
|
2019-05-31 00:06:27 +01:00
|
|
|
struct radv_shader_variant *variant = radv_shader_variant_create(device, binary,
|
|
|
|
keep_shader_info);
|
2019-07-01 00:29:24 +01:00
|
|
|
if (!variant) {
|
|
|
|
free(binary);
|
|
|
|
return NULL;
|
|
|
|
}
|
2017-09-01 12:45:33 +01:00
|
|
|
|
2019-07-01 01:19:13 +01:00
|
|
|
if (options->dump_shader) {
|
2020-04-01 12:43:50 +01:00
|
|
|
fprintf(stderr, "%s", radv_get_shader_name(info, shaders[0]->info.stage));
|
|
|
|
for (int i = 1; i < shader_count; ++i)
|
|
|
|
fprintf(stderr, " + %s", radv_get_shader_name(info, shaders[i]->info.stage));
|
|
|
|
|
|
|
|
fprintf(stderr, "\ndisasm:\n%s\n", variant->disasm_string);
|
2019-07-01 01:19:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2019-05-31 00:06:27 +01:00
|
|
|
if (keep_shader_info) {
|
2019-06-01 19:54:35 +01:00
|
|
|
variant->nir_string = radv_dump_nir_shaders(shaders, shader_count);
|
2020-08-18 17:44:07 +01:00
|
|
|
if (!gs_copy_shader && !trap_handler_shader && !module->nir) {
|
2019-10-28 15:56:15 +00:00
|
|
|
variant->spirv = malloc(module->size);
|
|
|
|
if (!variant->spirv) {
|
|
|
|
free(variant);
|
|
|
|
free(binary);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(variant->spirv, module->data, module->size);
|
2017-09-22 15:56:40 +01:00
|
|
|
variant->spirv_size = module->size;
|
2017-09-22 15:44:08 +01:00
|
|
|
}
|
2017-09-01 12:45:33 +01:00
|
|
|
}
|
|
|
|
|
2019-07-01 00:29:24 +01:00
|
|
|
if (binary_out)
|
|
|
|
*binary_out = binary;
|
|
|
|
else
|
|
|
|
free(binary);
|
|
|
|
|
2017-09-01 10:41:18 +01:00
|
|
|
return variant;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct radv_shader_variant *
|
2019-07-01 00:29:24 +01:00
|
|
|
radv_shader_variant_compile(struct radv_device *device,
|
2017-09-22 15:44:08 +01:00
|
|
|
struct radv_shader_module *module,
|
2017-10-16 12:18:02 +01:00
|
|
|
struct nir_shader *const *shaders,
|
|
|
|
int shader_count,
|
2017-09-01 15:51:12 +01:00
|
|
|
struct radv_pipeline_layout *layout,
|
2018-03-13 13:54:04 +00:00
|
|
|
const struct radv_shader_variant_key *key,
|
2019-09-03 09:29:19 +01:00
|
|
|
struct radv_shader_info *info,
|
2020-03-19 15:09:31 +00:00
|
|
|
bool keep_shader_info, bool keep_statistic_info,
|
2020-08-28 07:41:20 +01:00
|
|
|
bool disable_optimizations,
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_shader_binary **binary_out)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
2020-08-06 16:50:33 +01:00
|
|
|
gl_shader_stage stage = shaders[shader_count - 1]->info.stage;
|
2018-03-13 13:54:04 +00:00
|
|
|
struct radv_nir_compiler_options options = {0};
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2017-09-01 15:51:12 +01:00
|
|
|
options.layout = layout;
|
|
|
|
if (key)
|
|
|
|
options.key = *key;
|
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
options.explicit_scratch_args = !radv_use_llvm_for_stage(device, stage);
|
2019-08-02 11:40:17 +01:00
|
|
|
options.robust_buffer_access = device->robust_buffer_access;
|
2020-08-28 07:41:20 +01:00
|
|
|
options.disable_optimizations = disable_optimizations;
|
2017-09-01 15:51:12 +01:00
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
return shader_variant_compile(device, module, shaders, shader_count, stage, info,
|
2020-08-18 17:44:07 +01:00
|
|
|
&options, false, false,
|
|
|
|
keep_shader_info, keep_statistic_info, binary_out);
|
2017-09-01 15:51:12 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
struct radv_shader_variant *
|
|
|
|
radv_create_gs_copy_shader(struct radv_device *device,
|
|
|
|
struct nir_shader *shader,
|
2019-09-03 09:29:19 +01:00
|
|
|
struct radv_shader_info *info,
|
2019-07-01 00:29:24 +01:00
|
|
|
struct radv_shader_binary **binary_out,
|
2020-03-19 15:09:31 +00:00
|
|
|
bool keep_shader_info, bool keep_statistic_info,
|
2020-08-28 07:41:20 +01:00
|
|
|
bool multiview, bool disable_optimizations)
|
2017-09-01 15:51:12 +01:00
|
|
|
{
|
2018-03-13 13:54:04 +00:00
|
|
|
struct radv_nir_compiler_options options = {0};
|
2020-08-06 16:50:33 +01:00
|
|
|
gl_shader_stage stage = MESA_SHADER_VERTEX;
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
options.explicit_scratch_args = !radv_use_llvm_for_stage(device, stage);
|
2017-09-01 15:51:12 +01:00
|
|
|
options.key.has_multiview_view_index = multiview;
|
2020-08-28 07:41:20 +01:00
|
|
|
options.disable_optimizations = disable_optimizations;
|
2017-09-01 10:41:18 +01:00
|
|
|
|
2020-08-06 16:50:33 +01:00
|
|
|
return shader_variant_compile(device, NULL, &shader, 1, stage,
|
2020-08-18 17:44:07 +01:00
|
|
|
info, &options, true, false,
|
|
|
|
keep_shader_info, keep_statistic_info, binary_out);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct radv_shader_variant *
|
|
|
|
radv_create_trap_handler_shader(struct radv_device *device)
|
|
|
|
{
|
|
|
|
struct radv_nir_compiler_options options = {0};
|
|
|
|
struct radv_shader_variant *shader = NULL;
|
|
|
|
struct radv_shader_binary *binary = NULL;
|
|
|
|
struct radv_shader_info info = {0};
|
|
|
|
|
|
|
|
nir_builder b;
|
|
|
|
nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_COMPUTE, NULL);
|
|
|
|
b.shader->info.name = ralloc_strdup(b.shader, "meta_trap_handler");
|
|
|
|
|
|
|
|
options.explicit_scratch_args = true;
|
|
|
|
info.wave_size = 64;
|
|
|
|
|
|
|
|
shader = shader_variant_compile(device, NULL, &b.shader, 1,
|
|
|
|
MESA_SHADER_COMPUTE, &info, &options,
|
|
|
|
false, true, true, false, &binary);
|
|
|
|
|
|
|
|
ralloc_free(b.shader);
|
|
|
|
free(binary);
|
|
|
|
|
|
|
|
return shader;
|
2017-09-01 10:41:18 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
radv_shader_variant_destroy(struct radv_device *device,
|
|
|
|
struct radv_shader_variant *variant)
|
|
|
|
{
|
|
|
|
if (!p_atomic_dec_zero(&variant->ref_count))
|
|
|
|
return;
|
|
|
|
|
|
|
|
mtx_lock(&device->shader_slab_mutex);
|
|
|
|
list_del(&variant->slab_list);
|
|
|
|
mtx_unlock(&device->shader_slab_mutex);
|
|
|
|
|
2019-10-28 15:56:15 +00:00
|
|
|
free(variant->spirv);
|
2019-06-01 19:54:35 +01:00
|
|
|
free(variant->nir_string);
|
2017-09-01 12:45:33 +01:00
|
|
|
free(variant->disasm_string);
|
2019-09-25 11:48:04 +01:00
|
|
|
free(variant->ir_string);
|
2019-12-04 14:46:31 +00:00
|
|
|
free(variant->statistics);
|
2017-09-01 10:41:18 +01:00
|
|
|
free(variant);
|
|
|
|
}
|
|
|
|
|
|
|
|
const char *
|
2019-09-03 16:39:23 +01:00
|
|
|
radv_get_shader_name(struct radv_shader_info *info,
|
2019-07-11 17:03:55 +01:00
|
|
|
gl_shader_stage stage)
|
2017-09-01 10:41:18 +01:00
|
|
|
{
|
|
|
|
switch (stage) {
|
2019-07-11 17:03:55 +01:00
|
|
|
case MESA_SHADER_VERTEX:
|
|
|
|
if (info->vs.as_ls)
|
|
|
|
return "Vertex Shader as LS";
|
|
|
|
else if (info->vs.as_es)
|
|
|
|
return "Vertex Shader as ES";
|
|
|
|
else if (info->is_ngg)
|
|
|
|
return "Vertex Shader as ESGS";
|
|
|
|
else
|
|
|
|
return "Vertex Shader as VS";
|
|
|
|
case MESA_SHADER_TESS_CTRL:
|
|
|
|
return "Tessellation Control Shader";
|
|
|
|
case MESA_SHADER_TESS_EVAL:
|
|
|
|
if (info->tes.as_es)
|
|
|
|
return "Tessellation Evaluation Shader as ES";
|
|
|
|
else if (info->is_ngg)
|
|
|
|
return "Tessellation Evaluation Shader as ESGS";
|
|
|
|
else
|
|
|
|
return "Tessellation Evaluation Shader as VS";
|
|
|
|
case MESA_SHADER_GEOMETRY:
|
|
|
|
return "Geometry Shader";
|
|
|
|
case MESA_SHADER_FRAGMENT:
|
|
|
|
return "Pixel Shader";
|
|
|
|
case MESA_SHADER_COMPUTE:
|
|
|
|
return "Compute Shader";
|
2017-09-01 10:41:18 +01:00
|
|
|
default:
|
|
|
|
return "Unknown shader";
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2019-06-01 19:25:47 +01:00
|
|
|
unsigned
|
|
|
|
radv_get_max_workgroup_size(enum chip_class chip_class,
|
|
|
|
gl_shader_stage stage,
|
|
|
|
const unsigned *sizes)
|
|
|
|
{
|
|
|
|
switch (stage) {
|
|
|
|
case MESA_SHADER_TESS_CTRL:
|
|
|
|
return chip_class >= GFX7 ? 128 : 64;
|
|
|
|
case MESA_SHADER_GEOMETRY:
|
|
|
|
return chip_class >= GFX9 ? 128 : 64;
|
|
|
|
case MESA_SHADER_COMPUTE:
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
unsigned max_workgroup_size = sizes[0] * sizes[1] * sizes[2];
|
|
|
|
return max_workgroup_size;
|
|
|
|
}
|
2019-06-01 17:46:21 +01:00
|
|
|
|
|
|
|
unsigned
|
|
|
|
radv_get_max_waves(struct radv_device *device,
|
|
|
|
struct radv_shader_variant *variant,
|
|
|
|
gl_shader_stage stage)
|
2017-09-05 14:34:07 +01:00
|
|
|
{
|
2019-02-01 11:04:39 +00:00
|
|
|
enum chip_class chip_class = device->physical_device->rad_info.chip_class;
|
2019-05-15 03:16:20 +01:00
|
|
|
unsigned lds_increment = chip_class >= GFX7 ? 512 : 256;
|
2019-09-03 16:39:23 +01:00
|
|
|
uint8_t wave_size = variant->info.wave_size;
|
2019-06-01 17:46:21 +01:00
|
|
|
struct ac_shader_config *conf = &variant->config;
|
2017-09-05 14:34:07 +01:00
|
|
|
unsigned max_simd_waves;
|
|
|
|
unsigned lds_per_wave = 0;
|
|
|
|
|
2019-09-13 00:39:02 +01:00
|
|
|
max_simd_waves = device->physical_device->rad_info.max_wave64_per_simd;
|
2017-09-05 14:34:07 +01:00
|
|
|
|
|
|
|
if (stage == MESA_SHADER_FRAGMENT) {
|
|
|
|
lds_per_wave = conf->lds_size * lds_increment +
|
2019-09-03 16:39:23 +01:00
|
|
|
align(variant->info.ps.num_interp * 48,
|
2017-09-05 14:34:07 +01:00
|
|
|
lds_increment);
|
2019-02-01 11:04:39 +00:00
|
|
|
} else if (stage == MESA_SHADER_COMPUTE) {
|
|
|
|
unsigned max_workgroup_size =
|
2019-06-01 19:25:47 +01:00
|
|
|
radv_get_max_workgroup_size(chip_class, stage, variant->info.cs.block_size);
|
2019-02-01 11:04:39 +00:00
|
|
|
lds_per_wave = (conf->lds_size * lds_increment) /
|
2019-07-30 17:32:42 +01:00
|
|
|
DIV_ROUND_UP(max_workgroup_size, wave_size);
|
2017-09-05 14:34:07 +01:00
|
|
|
}
|
|
|
|
|
2019-10-18 21:13:44 +01:00
|
|
|
if (conf->num_sgprs) {
|
|
|
|
unsigned sgprs = align(conf->num_sgprs, chip_class >= GFX8 ? 16 : 8);
|
2018-04-06 13:06:24 +01:00
|
|
|
max_simd_waves =
|
|
|
|
MIN2(max_simd_waves,
|
2019-09-13 00:46:02 +01:00
|
|
|
device->physical_device->rad_info.num_physical_sgprs_per_simd /
|
2019-10-18 21:13:44 +01:00
|
|
|
sgprs);
|
|
|
|
}
|
2017-09-05 14:34:07 +01:00
|
|
|
|
2019-10-18 21:13:44 +01:00
|
|
|
if (conf->num_vgprs) {
|
|
|
|
unsigned vgprs = align(conf->num_vgprs, wave_size == 32 ? 8 : 4);
|
2018-04-06 13:10:34 +01:00
|
|
|
max_simd_waves =
|
|
|
|
MIN2(max_simd_waves,
|
2020-02-20 08:26:03 +00:00
|
|
|
device->physical_device->rad_info.num_physical_wave64_vgprs_per_simd / vgprs);
|
2019-10-18 21:13:44 +01:00
|
|
|
}
|
2017-09-05 14:34:07 +01:00
|
|
|
|
2020-02-27 07:45:46 +00:00
|
|
|
unsigned max_lds_per_simd = device->physical_device->rad_info.lds_size_per_workgroup / device->physical_device->rad_info.num_simd_per_compute_unit;
|
2017-09-05 14:34:07 +01:00
|
|
|
if (lds_per_wave)
|
2020-02-20 07:50:29 +00:00
|
|
|
max_simd_waves = MIN2(max_simd_waves, max_lds_per_simd / lds_per_wave);
|
2017-09-05 14:34:07 +01:00
|
|
|
|
2019-06-01 17:46:21 +01:00
|
|
|
return max_simd_waves;
|
|
|
|
}
|
|
|
|
|
2017-10-27 14:25:05 +01:00
|
|
|
VkResult
|
|
|
|
radv_GetShaderInfoAMD(VkDevice _device,
|
|
|
|
VkPipeline _pipeline,
|
|
|
|
VkShaderStageFlagBits shaderStage,
|
|
|
|
VkShaderInfoTypeAMD infoType,
|
|
|
|
size_t* pInfoSize,
|
|
|
|
void* pInfo)
|
|
|
|
{
|
|
|
|
RADV_FROM_HANDLE(radv_device, device, _device);
|
|
|
|
RADV_FROM_HANDLE(radv_pipeline, pipeline, _pipeline);
|
|
|
|
gl_shader_stage stage = vk_to_mesa_shader_stage(shaderStage);
|
|
|
|
struct radv_shader_variant *variant = pipeline->shaders[stage];
|
|
|
|
VkResult result = VK_SUCCESS;
|
|
|
|
|
|
|
|
/* Spec doesn't indicate what to do if the stage is invalid, so just
|
|
|
|
* return no info for this. */
|
|
|
|
if (!variant)
|
2018-05-31 00:06:41 +01:00
|
|
|
return vk_error(device->instance, VK_ERROR_FEATURE_NOT_PRESENT);
|
2017-10-27 14:25:05 +01:00
|
|
|
|
|
|
|
switch (infoType) {
|
|
|
|
case VK_SHADER_INFO_TYPE_STATISTICS_AMD:
|
|
|
|
if (!pInfo) {
|
|
|
|
*pInfoSize = sizeof(VkShaderStatisticsInfoAMD);
|
|
|
|
} else {
|
2019-05-15 03:16:20 +01:00
|
|
|
unsigned lds_multiplier = device->physical_device->rad_info.chip_class >= GFX7 ? 512 : 256;
|
2017-10-27 14:25:05 +01:00
|
|
|
struct ac_shader_config *conf = &variant->config;
|
|
|
|
|
|
|
|
VkShaderStatisticsInfoAMD statistics = {};
|
|
|
|
statistics.shaderStageMask = shaderStage;
|
2020-02-20 08:26:03 +00:00
|
|
|
statistics.numPhysicalVgprs = device->physical_device->rad_info.num_physical_wave64_vgprs_per_simd;
|
2019-09-13 00:46:02 +01:00
|
|
|
statistics.numPhysicalSgprs = device->physical_device->rad_info.num_physical_sgprs_per_simd;
|
2017-10-27 14:25:05 +01:00
|
|
|
statistics.numAvailableSgprs = statistics.numPhysicalSgprs;
|
|
|
|
|
|
|
|
if (stage == MESA_SHADER_COMPUTE) {
|
2019-06-01 19:25:47 +01:00
|
|
|
unsigned *local_size = variant->info.cs.block_size;
|
2017-10-27 14:25:05 +01:00
|
|
|
unsigned workgroup_size = local_size[0] * local_size[1] * local_size[2];
|
|
|
|
|
|
|
|
statistics.numAvailableVgprs = statistics.numPhysicalVgprs /
|
2018-06-15 17:49:08 +01:00
|
|
|
ceil((double)workgroup_size / statistics.numPhysicalVgprs);
|
2017-10-27 14:25:05 +01:00
|
|
|
|
|
|
|
statistics.computeWorkGroupSize[0] = local_size[0];
|
|
|
|
statistics.computeWorkGroupSize[1] = local_size[1];
|
|
|
|
statistics.computeWorkGroupSize[2] = local_size[2];
|
|
|
|
} else {
|
|
|
|
statistics.numAvailableVgprs = statistics.numPhysicalVgprs;
|
|
|
|
}
|
|
|
|
|
|
|
|
statistics.resourceUsage.numUsedVgprs = conf->num_vgprs;
|
|
|
|
statistics.resourceUsage.numUsedSgprs = conf->num_sgprs;
|
|
|
|
statistics.resourceUsage.ldsSizePerLocalWorkGroup = 32768;
|
|
|
|
statistics.resourceUsage.ldsUsageSizeInBytes = conf->lds_size * lds_multiplier;
|
|
|
|
statistics.resourceUsage.scratchMemUsageInBytes = conf->scratch_bytes_per_wave;
|
|
|
|
|
|
|
|
size_t size = *pInfoSize;
|
|
|
|
*pInfoSize = sizeof(statistics);
|
|
|
|
|
|
|
|
memcpy(pInfo, &statistics, MIN2(size, *pInfoSize));
|
|
|
|
|
|
|
|
if (size < *pInfoSize)
|
|
|
|
result = VK_INCOMPLETE;
|
|
|
|
}
|
|
|
|
|
|
|
|
break;
|
2020-08-20 15:03:36 +01:00
|
|
|
case VK_SHADER_INFO_TYPE_DISASSEMBLY_AMD: {
|
|
|
|
char *out;
|
|
|
|
size_t outsize;
|
|
|
|
FILE *memf = open_memstream(&out, &outsize);
|
2017-10-27 14:25:05 +01:00
|
|
|
|
2020-08-20 15:03:36 +01:00
|
|
|
fprintf(memf, "%s:\n", radv_get_shader_name(&variant->info, stage));
|
|
|
|
fprintf(memf, "%s\n\n", variant->ir_string);
|
|
|
|
fprintf(memf, "%s\n\n", variant->disasm_string);
|
|
|
|
radv_dump_shader_stats(device, pipeline, stage, memf);
|
|
|
|
fclose(memf);
|
2017-10-27 14:25:05 +01:00
|
|
|
|
|
|
|
/* Need to include the null terminator. */
|
2020-08-20 15:03:36 +01:00
|
|
|
size_t length = outsize + 1;
|
2017-10-27 14:25:05 +01:00
|
|
|
|
|
|
|
if (!pInfo) {
|
|
|
|
*pInfoSize = length;
|
|
|
|
} else {
|
|
|
|
size_t size = *pInfoSize;
|
|
|
|
*pInfoSize = length;
|
|
|
|
|
2020-08-20 15:03:36 +01:00
|
|
|
memcpy(pInfo, out, MIN2(size, length));
|
2017-10-27 14:25:05 +01:00
|
|
|
|
|
|
|
if (size < length)
|
|
|
|
result = VK_INCOMPLETE;
|
|
|
|
}
|
|
|
|
|
2020-08-20 15:03:36 +01:00
|
|
|
free(out);
|
2017-10-27 14:25:05 +01:00
|
|
|
break;
|
2020-08-20 15:03:36 +01:00
|
|
|
}
|
2017-10-27 14:25:05 +01:00
|
|
|
default:
|
|
|
|
/* VK_SHADER_INFO_TYPE_BINARY_AMD unimplemented for now. */
|
|
|
|
result = VK_ERROR_FEATURE_NOT_PRESENT;
|
|
|
|
break;
|
2017-09-05 14:34:07 +01:00
|
|
|
}
|
|
|
|
|
2017-10-27 14:25:05 +01:00
|
|
|
return result;
|
2017-09-05 14:34:07 +01:00
|
|
|
}
|
2020-08-20 15:03:36 +01:00
|
|
|
|
|
|
|
VkResult
|
|
|
|
radv_dump_shader_stats(struct radv_device *device,
|
|
|
|
struct radv_pipeline *pipeline,
|
|
|
|
gl_shader_stage stage, FILE *output)
|
|
|
|
{
|
|
|
|
struct radv_shader_variant *shader = pipeline->shaders[stage];
|
|
|
|
VkPipelineExecutablePropertiesKHR *props = NULL;
|
|
|
|
uint32_t prop_count = 0;
|
|
|
|
VkResult result;
|
|
|
|
|
|
|
|
VkPipelineInfoKHR pipeline_info = {};
|
|
|
|
pipeline_info.sType = VK_STRUCTURE_TYPE_PIPELINE_INFO_KHR;
|
|
|
|
pipeline_info.pipeline = radv_pipeline_to_handle(pipeline);
|
|
|
|
|
|
|
|
result = radv_GetPipelineExecutablePropertiesKHR(radv_device_to_handle(device),
|
|
|
|
&pipeline_info,
|
|
|
|
&prop_count, NULL);
|
|
|
|
if (result != VK_SUCCESS)
|
|
|
|
return result;
|
|
|
|
|
|
|
|
props = calloc(prop_count, sizeof(*props));
|
|
|
|
if (!props)
|
|
|
|
return VK_ERROR_OUT_OF_HOST_MEMORY;
|
|
|
|
|
|
|
|
result = radv_GetPipelineExecutablePropertiesKHR(radv_device_to_handle(device),
|
|
|
|
&pipeline_info,
|
|
|
|
&prop_count, props);
|
|
|
|
if (result != VK_SUCCESS)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
for (unsigned i = 0; i < prop_count; i++) {
|
|
|
|
if (!(props[i].stages & mesa_to_vk_shader_stage(stage)))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
VkPipelineExecutableStatisticKHR *stats = NULL;
|
|
|
|
uint32_t stat_count = 0;
|
|
|
|
VkResult result;
|
|
|
|
|
|
|
|
VkPipelineExecutableInfoKHR exec_info = {};
|
|
|
|
exec_info.pipeline = radv_pipeline_to_handle(pipeline);
|
|
|
|
exec_info.executableIndex = i;
|
|
|
|
|
|
|
|
result = radv_GetPipelineExecutableStatisticsKHR(radv_device_to_handle(device),
|
|
|
|
&exec_info,
|
|
|
|
&stat_count, NULL);
|
|
|
|
if (result != VK_SUCCESS)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
stats = calloc(stat_count, sizeof(*stats));
|
|
|
|
if (!stats) {
|
|
|
|
result = VK_ERROR_OUT_OF_HOST_MEMORY;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
result = radv_GetPipelineExecutableStatisticsKHR(radv_device_to_handle(device),
|
|
|
|
&exec_info,
|
|
|
|
&stat_count, stats);
|
|
|
|
if (result != VK_SUCCESS) {
|
|
|
|
free(stats);
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(output, "\n%s:\n",
|
|
|
|
radv_get_shader_name(&shader->info, stage));
|
|
|
|
fprintf(output, "*** SHADER STATS ***\n");
|
|
|
|
|
|
|
|
for (unsigned i = 0; i < stat_count; i++) {
|
|
|
|
fprintf(output, "%s: ", stats[i].name);
|
|
|
|
switch (stats[i].format) {
|
|
|
|
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_BOOL32_KHR:
|
|
|
|
fprintf(output, "%s", stats[i].value.b32 == VK_TRUE ? "true" : "false");
|
|
|
|
break;
|
|
|
|
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_INT64_KHR:
|
|
|
|
fprintf(output, "%"PRIi64, stats[i].value.i64);
|
|
|
|
break;
|
|
|
|
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_UINT64_KHR:
|
|
|
|
fprintf(output, "%"PRIu64, stats[i].value.u64);
|
|
|
|
break;
|
|
|
|
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_FLOAT64_KHR:
|
|
|
|
fprintf(output, "%f", stats[i].value.f64);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
unreachable("Invalid pipeline statistic format");
|
|
|
|
}
|
|
|
|
fprintf(output, "\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(output, "********************\n\n\n");
|
|
|
|
|
|
|
|
free(stats);
|
|
|
|
}
|
|
|
|
|
|
|
|
fail:
|
|
|
|
free(props);
|
|
|
|
return result;
|
|
|
|
}
|