| /* | |
| * Copyright 2014 Advanced Micro Devices, Inc. | |
| * | |
| * SPDX-License-Identifier: MIT | |
| */ | |
| #include "ac_binary.h" | |
| #include "ac_gpu_info.h" | |
| #include "util/u_math.h" | |
| #include <sid.h> | |
| #include <stdio.h> | |
| #define SPILLED_SGPRS 0x4 | |
| #define SPILLED_VGPRS 0x8 | |
| /* Parse configuration data in .AMDGPU.config section format. */ | |
| void ac_parse_llvm_binary_config(const char *data, size_t nbytes, unsigned wave_size, | |
| const struct ac_compiler_info *compiler_info, | |
| struct ac_shader_config *conf) | |
| { | |
| unsigned lds_granularity = compiler_info->gfx_level >= GFX7 ? 512 : 256; | |
| unsigned ps_lds_granularity = compiler_info->gfx_level >= GFX11 ? 1024 : lds_granularity; | |
| for (size_t i = 0; i < nbytes; i += 8) { | |
| unsigned reg = util_le32_to_cpu(*(uint32_t *)(data + i)); | |
| unsigned value = util_le32_to_cpu(*(uint32_t *)(data + i + 4)); | |
| switch (reg) { | |
| case R_00B028_SPI_SHADER_PGM_RSRC1_PS: | |
| case R_00B128_SPI_SHADER_PGM_RSRC1_VS: | |
| case R_00B228_SPI_SHADER_PGM_RSRC1_GS: | |
| case R_00B848_COMPUTE_PGM_RSRC1: | |
| case R_00B428_SPI_SHADER_PGM_RSRC1_HS: | |
| conf->num_vgprs = MAX2(conf->num_vgprs, | |
| (G_00B028_VGPRS(value) + 1) * | |
| compiler_info->wave64_vgpr_encode_granularity * | |
| (wave_size == 32 ? 2 : 1)); | |
| conf->num_sgprs = MAX2(conf->num_sgprs, (G_00B028_SGPRS(value) + 1) * 8); | |
| /* TODO: LLVM doesn't set FLOAT_MODE or MEM_ORDERED for non-compute shaders, | |
| * and sets MEM_ORDERED on GFX12+. | |
| */ | |
| bool has_mem_ordered = | |
| compiler_info->gfx_level >= GFX10 && compiler_info->gfx_level < GFX12; | |
| conf->float_mode = G_00B028_FLOAT_MODE(value); | |
| conf->mem_ordered = (reg != R_00B848_COMPUTE_PGM_RSRC1 || | |
| G_00B848_MEM_ORDERED(value)) && | |
| has_mem_ordered; | |
| conf->rsrc1 = value; | |
| break; | |
| case R_00B02C_SPI_SHADER_PGM_RSRC2_PS: | |
| conf->lds_size = MAX2(conf->lds_size, G_00B02C_EXTRA_LDS_SIZE(value) * ps_lds_granularity); | |
| /* TODO: LLVM doesn't set SHARED_VGPR_CNT for all shader types */ | |
| conf->num_shared_vgprs = G_00B02C_SHARED_VGPR_CNT(value); | |
| conf->rsrc2 = value; | |
| break; | |
| case R_00B12C_SPI_SHADER_PGM_RSRC2_VS: | |
| conf->num_shared_vgprs = G_00B12C_SHARED_VGPR_CNT(value); | |
| conf->rsrc2 = value; | |
| break; | |
| case R_00B22C_SPI_SHADER_PGM_RSRC2_GS: | |
| conf->num_shared_vgprs = G_00B22C_SHARED_VGPR_CNT(value); | |
| conf->rsrc2 = value; | |
| break; | |
| case R_00B42C_SPI_SHADER_PGM_RSRC2_HS: | |
| conf->num_shared_vgprs = G_00B42C_SHARED_VGPR_CNT(value); | |
| conf->rsrc2 = value; | |
| break; | |
| case R_00B84C_COMPUTE_PGM_RSRC2: | |
| conf->lds_size = MAX2(conf->lds_size, G_00B84C_LDS_SIZE(value) * lds_granularity); | |
| conf->rsrc2 = value; | |
| break; | |
| case R_00B8A0_COMPUTE_PGM_RSRC3: | |
| conf->num_shared_vgprs = G_00B8A0_SHARED_VGPR_CNT(value); | |
| conf->rsrc3 = value; | |
| break; | |
| case R_02865C_SPI_PS_INPUT_ENA: | |
| case R_0286CC_SPI_PS_INPUT_ENA: | |
| conf->spi_ps_input_ena = value; | |
| break; | |
| case R_028660_SPI_PS_INPUT_ADDR: | |
| case R_0286D0_SPI_PS_INPUT_ADDR: | |
| conf->spi_ps_input_addr = value; | |
| break; | |
| case R_0286E8_SPI_TMPRING_SIZE: | |
| case R_00B860_COMPUTE_TMPRING_SIZE: | |
| if (compiler_info->gfx_level >= GFX11) | |
| conf->scratch_bytes_per_wave = G_00B860_WAVESIZE(value) * 256; | |
| else | |
| conf->scratch_bytes_per_wave = G_00B860_WAVESIZE(value) * 1024; | |
| break; | |
| case SPILLED_SGPRS: | |
| conf->spilled_sgprs = value; | |
| break; | |
| case SPILLED_VGPRS: | |
| conf->spilled_vgprs = value; | |
| break; | |
| default: { | |
| static bool printed; | |
| if (!printed) { | |
| fprintf(stderr, | |
| "Warning: LLVM emitted unknown " | |
| "config register: 0x%x\n", | |
| reg); | |
| printed = true; | |
| } | |
| } break; | |
| } | |
| } | |
| if (!conf->spi_ps_input_addr) | |
| conf->spi_ps_input_addr = conf->spi_ps_input_ena; | |
| } | |
| unsigned ac_align_shader_binary_for_prefetch(enum amd_gfx_level gfx_level, | |
| unsigned prefetch_distance, | |
| unsigned size) | |
| { | |
| /* The SQ fetches up to N cache lines of 16 dwords | |
| * ahead of the PC, configurable by SH_MEM_CONFIG and | |
| * S_INST_PREFETCH. This can cause two issues: | |
| * | |
| * (1) Crossing a page boundary to an unmapped page. The logic | |
| * does not distinguish between a required fetch and a "mere" | |
| * prefetch and will fault. | |
| * | |
| * (2) Prefetching instructions that will be changed for a | |
| * different shader. | |
| * | |
| * (2) is not currently an issue because we flush the I$ at IB | |
| * boundaries, but (1) needs to be addressed. Due to buffer | |
| * suballocation, we just play it safe. | |
| */ | |
| if (prefetch_distance) { | |
| if (gfx_level >= GFX11) | |
| size = align(size + prefetch_distance * 64, 128); | |
| else | |
| size = align(size + prefetch_distance * 64, 64); | |
| } | |
| return size; | |
| } | |
| unsigned ac_get_instr_prefetch_size(enum amd_gfx_level gfx_level, | |
| unsigned prefetch_distance, | |
| unsigned size) | |
| { | |
| assert(gfx_level >= GFX11); | |
| unsigned aligned_size = ac_align_shader_binary_for_prefetch(gfx_level, | |
| prefetch_distance, | |
| size); | |
| unsigned inst_pref_size = DIV_ROUND_UP(aligned_size, 128); | |
| /* GFX12 allows 255, but that means one shader can thrash the entire cache. | |
| * Limit to half of the 32KiB WGP instruction cache size. | |
| */ | |
| unsigned max_pref_size = gfx_level >= GFX12 ? 128 : 63; | |
| return MIN2(inst_pref_size, max_pref_size); | |
| } |