From 5689bb00229a422267ada6aef770dacc81e3319e Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 15 Nov 2017 09:51:36 +1000 Subject: r600/shader: implement getting thread id. We need the thread id to use the immediate buffer readback mechanism, so add support for calculating it. Tested-By: Gert Wollny Signed-off-by: Dave Airlie --- src/gallium/drivers/r600/r600_shader.c | 74 ++++++++++++++++++++++++++++++++++ 1 file changed, 74 insertions(+) (limited to 'src/gallium/drivers/r600/r600_shader.c') diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c index 919685ca6a2..b351c8904c5 100644 --- a/src/gallium/drivers/r600/r600_shader.c +++ b/src/gallium/drivers/r600/r600_shader.c @@ -353,6 +353,7 @@ struct r600_shader_ctx { unsigned enabled_stream_buffers_mask; unsigned tess_input_info; /* temp with tess input offsets */ unsigned tess_output_info; /* temp with tess input offsets */ + unsigned thread_id_gpr; /* temp with thread id calculated for images */ }; struct r600_shader_tgsi_instruction { @@ -2935,6 +2936,69 @@ static int r600_emit_tess_factor(struct r600_shader_ctx *ctx) return 0; } +/* + * We have to work out the thread ID for load and atomic + * operations, which store the returned value to an index + * in an intermediate buffer. + * The index is calculated by taking the thread id, + * calculated from the MBCNT instructions. + * Then the shader engine ID is multiplied by 256, + * and the wave id is added. + * Then the result is multipled by 64 and thread id is + * added. + */ +static int load_thread_id_gpr(struct r600_shader_ctx *ctx) +{ + struct r600_bytecode_alu alu; + int r; + + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); + alu.op = ALU_OP1_MBCNT_32LO_ACCUM_PREV_INT; + alu.dst.sel = ctx->temp_reg; + alu.dst.chan = 0; + alu.src[0].sel = V_SQ_ALU_SRC_LITERAL; + alu.src[0].value = 0xffffffff; + alu.dst.write = 1; + r = r600_bytecode_add_alu(ctx->bc, &alu); + if (r) + return r; + + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); + alu.op = ALU_OP1_MBCNT_32HI_INT; + alu.dst.sel = ctx->temp_reg; + alu.dst.chan = 1; + alu.src[0].sel = V_SQ_ALU_SRC_LITERAL; + alu.src[0].value = 0xffffffff; + alu.dst.write = 1; + r = r600_bytecode_add_alu(ctx->bc, &alu); + if (r) + return r; + + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); + alu.op = ALU_OP3_MULADD_UINT24; + alu.dst.sel = ctx->temp_reg; + alu.dst.chan = 2; + alu.src[0].sel = EG_V_SQ_ALU_SRC_SE_ID; + alu.src[1].sel = V_SQ_ALU_SRC_LITERAL; + alu.src[1].value = 256; + alu.src[2].sel = EG_V_SQ_ALU_SRC_HW_WAVE_ID; + alu.dst.write = 1; + alu.is_op3 = 1; + alu.last = 1; + r = r600_bytecode_add_alu(ctx->bc, &alu); + if (r) + return r; + + r = single_alu_op3(ctx, ALU_OP3_MULADD_UINT24, + ctx->thread_id_gpr, 1, + ctx->temp_reg, 2, + V_SQ_ALU_SRC_LITERAL, 0x40, + ctx->temp_reg, 0); + if (r) + return r; + return 0; +} + static int r600_shader_from_tgsi(struct r600_context *rctx, struct r600_pipe_shader *pipeshader, union r600_shader_key key) @@ -3138,6 +3202,12 @@ static int r600_shader_from_tgsi(struct r600_context *rctx, ctx.temp_reg = ctx.bc->ar_reg + 3; } + if (shader->uses_images && ctx.type == PIPE_SHADER_FRAGMENT) { + ctx.thread_id_gpr = ctx.temp_reg; + ctx.temp_reg++; + } else + ctx.thread_id_gpr = 0; + shader->max_arrays = 0; shader->num_arrays = 0; if (indirect_gprs) { @@ -3281,6 +3351,10 @@ static int r600_shader_from_tgsi(struct r600_context *rctx, } } + if (ctx.thread_id_gpr) { + load_thread_id_gpr(&ctx); + } + if (ctx.type == PIPE_SHADER_GEOMETRY) { struct r600_bytecode_alu alu; int r; -- cgit v1.2.3