[Mesa-dev] [PATCH v4 (part2) 15/59] i965/vec4: Implement unsized array's length calculation
Samuel Iglesias Gonsálvez
siglesias at igalia.com
Thu Aug 27 23:06:20 PDT 2015
On 28/08/15 02:21, Jordan Justen wrote:
> On 2015-08-05 01:30:12, Iago Toral Quiroga wrote:
>> From: Samuel Iglesias Gonsalvez <siglesias at igalia.com>
>>
>> Notice that Skylake needs to include a header in the sampler message
>> so it will need some tweaks to work there.
>>
>> Signed-off-by: Samuel Iglesias Gonsalvez <siglesias at igalia.com>
>> ---
>> src/mesa/drivers/dri/i965/brw_defines.h | 3 ++
>> src/mesa/drivers/dri/i965/brw_shader.cpp | 3 ++
>> src/mesa/drivers/dri/i965/brw_vec4.cpp | 1 +
>> src/mesa/drivers/dri/i965/brw_vec4.h | 6 ++++
>> src/mesa/drivers/dri/i965/brw_vec4_generator.cpp | 31 ++++++++++++++++
>
>> src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp | 46 ++++++++++++++++++++++--
>
> Shouldn't this be in brw_vec4_nir.cpp instead?
>
I wrote this patch before vec4 NIR backend was upstreamed. I have a
separate patch that adds a similar code to vec4 nir backend.
Or do you suggest to delete this patch?
Sam
> -Jordan
>
>> 6 files changed, 88 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
>> index f595366..3148849 100644
>> --- a/src/mesa/drivers/dri/i965/brw_defines.h
>> +++ b/src/mesa/drivers/dri/i965/brw_defines.h
>> @@ -1050,6 +1050,9 @@ enum opcode {
>> VS_OPCODE_PULL_CONSTANT_LOAD,
>> VS_OPCODE_PULL_CONSTANT_LOAD_GEN7,
>> VS_OPCODE_SET_SIMD4X2_HEADER_GEN9,
>> +
>> + VS_OPCODE_UNSIZED_ARRAY_LENGTH,
>> +
>> VS_OPCODE_UNPACK_FLAGS_SIMD4X2,
>>
>> /**
>> diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
>> index 819e4f2..7890835 100644
>> --- a/src/mesa/drivers/dri/i965/brw_shader.cpp
>> +++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
>> @@ -743,6 +743,9 @@ brw_instruction_name(enum opcode op)
>> case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9:
>> return "set_simd4x2_header_gen9";
>>
>> + case VS_OPCODE_UNSIZED_ARRAY_LENGTH:
>> + return "vs_unsized_array_length";
>> +
>> case VS_OPCODE_UNPACK_FLAGS_SIMD4X2:
>> return "unpack_flags_simd4x2";
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> index f18915a..7c9f362 100644
>> --- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> +++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> @@ -328,6 +328,7 @@ vec4_visitor::implied_mrf_writes(vec4_instruction *inst)
>> case SHADER_OPCODE_TXS:
>> case SHADER_OPCODE_TG4:
>> case SHADER_OPCODE_TG4_OFFSET:
>> + case VS_OPCODE_UNSIZED_ARRAY_LENGTH:
>> return inst->header_size;
>> default:
>> unreachable("not reached");
>> diff --git a/src/mesa/drivers/dri/i965/brw_vec4.h b/src/mesa/drivers/dri/i965/brw_vec4.h
>> index 985886d..3f73beb 100644
>> --- a/src/mesa/drivers/dri/i965/brw_vec4.h
>> +++ b/src/mesa/drivers/dri/i965/brw_vec4.h
>> @@ -566,6 +566,12 @@ private:
>> struct brw_reg offset);
>> void generate_set_simd4x2_header_gen9(vec4_instruction *inst,
>> struct brw_reg dst);
>> +
>> + void generate_unsized_array_length(vec4_instruction *inst,
>> + struct brw_reg dst,
>> + struct brw_reg src,
>> + struct brw_reg index);
>> +
>> void generate_unpack_flags(struct brw_reg dst);
>>
>> const struct brw_compiler *compiler;
>> diff --git a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
>> index 92050b9..1472f75 100644
>> --- a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
>> +++ b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
>> @@ -1029,6 +1029,32 @@ vec4_generator::generate_pull_constant_load(vec4_instruction *inst,
>> }
>>
>> void
>> +vec4_generator::generate_unsized_array_length(vec4_instruction *inst,
>> + struct brw_reg dst,
>> + struct brw_reg src,
>> + struct brw_reg surf_index)
>> +{
>> + assert(devinfo->gen >= 7);
>> + assert(surf_index.type == BRW_REGISTER_TYPE_UD &&
>> + surf_index.file == BRW_IMMEDIATE_VALUE);
>> +
>> + brw_SAMPLE(p,
>> + dst,
>> + inst->base_mrf,
>> + src,
>> + surf_index.dw1.ud,
>> + 0,
>> + GEN5_SAMPLER_MESSAGE_SAMPLE_RESINFO,
>> + 1, /* response length */
>> + inst->mlen,
>> + inst->header_size > 0,
>> + BRW_SAMPLER_SIMD_MODE_SIMD4X2,
>> + BRW_SAMPLER_RETURN_FORMAT_SINT32);
>> +
>> + brw_mark_surface_used(&prog_data->base, surf_index.dw1.ud);
>> +}
>> +
>> +void
>> vec4_generator::generate_pull_constant_load_gen7(vec4_instruction *inst,
>> struct brw_reg dst,
>> struct brw_reg surf_index,
>> @@ -1401,6 +1427,11 @@ vec4_generator::generate_code(const cfg_t *cfg)
>> generate_set_simd4x2_header_gen9(inst, dst);
>> break;
>>
>> +
>> + case VS_OPCODE_UNSIZED_ARRAY_LENGTH:
>> + generate_unsized_array_length(inst, dst, src[0], src[1]);
>> + break;
>> +
>> case GS_OPCODE_URB_WRITE:
>> generate_gs_urb_write(inst);
>> break;
>> diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
>> index 1bf2e1b..c7a0d4a 100644
>> --- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
>> +++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
>> @@ -1940,9 +1940,51 @@ vec4_visitor::visit(ir_expression *ir)
>> emit(BFE(result_dst, op[2], op[1], op[0]));
>> break;
>>
>> - case ir_triop_ssbo_unsized_array_length:
>> - unreachable("not reached: not implemented");
>> + case ir_triop_ssbo_unsized_array_length: {
>> + ir_constant *const_uniform_block = ir->operands[0]->as_constant();
>> + unsigned ubo_index = const_uniform_block->value.u[0];
>> + ir_constant *const_offset_ir = ir->operands[1]->as_constant();
>> + int const_offset = const_offset_ir ? const_offset_ir->value.u[0] : 0;
>> + ir_constant *const_stride_ir = ir->operands[2]->as_constant();
>> + int unsized_array_stride = const_stride_ir ? const_stride_ir->value.u[0] : 1;
>> +
>> + assert(shader->base.UniformBlocks[ubo_index].IsShaderStorage);
>> +
>> + src_reg surf_index = src_reg(prog_data->base.binding_table.ubo_start +
>> + ubo_index);
>> +
>> + dst_reg buffer_size = dst_reg(this, ir->type);
>> +
>> + vec4_instruction *inst = new(mem_ctx) vec4_instruction(
>> + VS_OPCODE_UNSIZED_ARRAY_LENGTH, buffer_size);
>> +
>> + inst->base_mrf = 2;
>> + inst->mlen = 1; /* always at least one */
>> + inst->src[1] = src_reg(surf_index);
>> +
>> + /* MRF for the first parameter */
>> + src_reg lod = src_reg(0);
>> + int param_base = inst->base_mrf;
>> + int writemask = WRITEMASK_X;
>> + emit(MOV(dst_reg(MRF, param_base, glsl_type::int_type, writemask), lod));
>> +
>> + emit(inst);
>> +
>> + /* array.length() =
>> + max((buffer_object_size - offset_of_array) / stride_of_array, 0) */
>> + emit(ADD(buffer_size, src_reg(buffer_size), brw_imm_d(-const_offset)));
>> +
>> + assert(unsized_array_stride > 0);
>> +
>> + src_reg stride = src_reg(unsized_array_stride);
>> + dst_reg temp = dst_reg(this, glsl_type::int_type);
>> + emit_math(SHADER_OPCODE_INT_QUOTIENT,
>> + temp,
>> + src_reg(buffer_size),
>> + stride);
>> + emit_minmax(BRW_CONDITIONAL_GE, result_dst, src_reg(temp), brw_imm_d(0));
>> break;
>> + }
>>
>> case ir_triop_vector_insert:
>> unreachable("should have been lowered by lower_vector_insert");
>> --
>> 1.9.1
>>
>
More information about the mesa-dev
mailing list