fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-06 05:08:08 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: add a helper to dump SGPR to memory for the trap handler

Browse source 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32026>
This commit is contained in:
Samuel Pitoiset 2024-11-07 15:11:19 +01:00 • committed by Marge Bot
parent 8c6f2fef1b
commit 494050d2ea
1 changed files with 22 additions and 24 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

46
src/amd/compiler/aco_instruction_selection.cpp
View file

@ -12405,6 +12405,26 @@ select_program(Program* program, unsigned shader_count, struct nir_shader* const
}
}
void
dump_sgpr_to_mem(isel_context* ctx, Operand rsrc, Operand data, uint32_t offset)
{
Builder bld(ctx->program, ctx->block);
ac_hw_cache_flags cache_glc;
cache_glc.value = ac_glc;
if (ctx->program->gfx_level >= GFX9) {
bld.copy(Definition(PhysReg{256}, v1) /* v0 */, data);
bld.mubuf(aco_opcode::buffer_store_dword, Operand(rsrc), Operand(v1), Operand::c32(0u),
Operand(PhysReg{256}, v1) /* v0 */, offset, false /* offen */, false /* idxen */,
/* addr64 */ false, /* disable_wqm */ false, cache_glc);
} else {
bld.smem(aco_opcode::s_buffer_store_dword, Operand(rsrc), Operand::c32(offset), data,
memory_sync_info(), cache_glc);
}
}
void
select_trap_handler_shader(Program* program, struct nir_shader* shader, ac_shader_config* config,
const struct aco_compiler_options* options,
@ -12501,35 +12521,13 @@ select_trap_handler_shader(Program* program, struct nir_shader* shader, ac_shade
bld.sopk(aco_opcode::s_getreg_b32, Definition(ttmp0_reg, s1),
((32 - 1) << 11) | hw_regs_idx[i]);
if (ctx.program->gfx_level >= GFX9) {
bld.copy(Definition(PhysReg{256}, v1) /* v0 */, Operand(ttmp0_reg, s1));
bld.mubuf(aco_opcode::buffer_store_dword, Operand(tma_rsrc, s4), Operand(v1),
Operand::c32(offset), Operand(PhysReg{256}, v1) /* v0 */, 0 /* offset */,
false /* offen */, false /* idxen */, /* addr64 */ false,
/* disable_wqm */ false, cache_glc);
} else {
bld.smem(aco_opcode::s_buffer_store_dword, Operand(tma_rsrc, s4), Operand::c32(offset),
Operand(ttmp0_reg, s1), memory_sync_info(), cache_glc);
}
dump_sgpr_to_mem(&ctx, Operand(tma_rsrc, s4), Operand(ttmp0_reg, s1), offset);
offset += 4;
}
/* Dump all SGPRs. */
for (uint32_t i = 0; i < program->dev.sgpr_limit; i++) {
if (ctx.program->gfx_level >= GFX9) {
bld.copy(Definition(PhysReg{256}, v1) /* v0 */, Operand(PhysReg{i}, s1));
bld.mubuf(aco_opcode::buffer_store_dword, Operand(tma_rsrc, s4), Operand(v1),
Operand::c32(0u), Operand(PhysReg{256}, v1) /* v0 */, offset, false /* offen */,
false /* idxen */, /* addr64 */ false,
/* disable_wqm */ false, cache_glc);
} else {
bld.smem(aco_opcode::s_buffer_store_dword, Operand(tma_rsrc, s4), Operand::c32(offset),
Operand(PhysReg{i}, s1), memory_sync_info(), cache_glc);
}
dump_sgpr_to_mem(&ctx, Operand(tma_rsrc, s4), Operand(PhysReg{i}, s1), offset);
offset += 4;
}