From: Giovanni Mascellani Subject: [PATCH vkd3d v3 7/7] vkd3d-shader/hlsl: Parse the mul() intrinsic. Message-Id: <20220602140609.3419134-7-gmascellani@codeweavers.com> Date: Thu, 2 Jun 2022 16:06:09 +0200 In-Reply-To: <20220602140609.3419134-1-gmascellani@codeweavers.com> References: <20220602140609.3419134-1-gmascellani@codeweavers.com> Signed-off-by: Giovanni Mascellani --- libs/vkd3d-shader/hlsl.y | 118 ++++++++++++++++++++++++ tests/hlsl-majority-pragma.shader_test | 2 +- tests/hlsl-majority-typedef.shader_test | 2 +- tests/hlsl-mul.shader_test | 36 ++++---- 4 files changed, 138 insertions(+), 20 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index 2aee5153..df5fda47 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -2046,6 +2046,123 @@ static bool intrinsic_min(struct hlsl_ctx *ctx, return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MIN, params->args[0], params->args[1], loc); } +static bool intrinsic_mul(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_ir_node *arg1 = params->args[0], *arg2 = params->args[1], *cast1, *cast2; + enum hlsl_base_type base = expr_common_base_type(arg1->data_type->base_type, arg2->data_type->base_type); + struct hlsl_type *cast_type1 = arg1->data_type, *cast_type2 = arg2->data_type, *matrix_type, *ret_type; + unsigned int i, j, k, vect_count = 0; + struct vkd3d_string_buffer *name; + static unsigned int counter = 0; + struct hlsl_ir_load *load; + struct hlsl_ir_var *var; + + if (arg1->data_type->type == HLSL_CLASS_SCALAR || arg2->data_type->type == HLSL_CLASS_SCALAR) + return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, arg1, arg2, loc); + + if (arg1->data_type->type == HLSL_CLASS_VECTOR) + { + vect_count++; + cast_type1 = hlsl_get_matrix_type(ctx, base, arg1->data_type->dimx, 1); + } + if (arg2->data_type->type == HLSL_CLASS_VECTOR) + { + vect_count++; + cast_type2 = hlsl_get_matrix_type(ctx, base, 1, arg2->data_type->dimx); + } + + matrix_type = hlsl_get_matrix_type(ctx, base, cast_type2->dimx, cast_type1->dimy); + + if (vect_count == 0) + { + ret_type = matrix_type; + } + else if (vect_count == 1) + { + assert(matrix_type->dimx == 1 || matrix_type->dimy == 1); + ret_type = hlsl_get_vector_type(ctx, base, matrix_type->dimx * matrix_type->dimy); + } + else + { + assert(matrix_type->dimx == 1 && matrix_type->dimy == 1); + ret_type = hlsl_get_scalar_type(ctx, base); + } + + if (!(cast1 = add_implicit_conversion(ctx, params->instrs, arg1, cast_type1, loc))) + return false; + + if (!(cast2 = add_implicit_conversion(ctx, params->instrs, arg2, cast_type2, loc))) + return false; + + name = vkd3d_string_buffer_get(&ctx->string_buffers); + vkd3d_string_buffer_printf(name, "", counter++); + var = hlsl_new_synthetic_var(ctx, name->buffer, matrix_type, *loc); + vkd3d_string_buffer_release(&ctx->string_buffers, name); + if (!var) + return false; + + for (i = 0; i < matrix_type->dimx; ++i) + for (j = 0; j < matrix_type->dimy; ++j) + { + struct hlsl_ir_node *node = NULL; + struct hlsl_type *scalar_type; + struct hlsl_ir_store *store; + struct hlsl_ir_constant *c; + unsigned int offset; + + for (k = 0; k < cast_type1->dimx && k < cast_type2->dimy; ++k) + { + struct hlsl_ir_load *value1, *value2; + struct hlsl_ir_node *mul; + + offset = hlsl_compute_component_offset(ctx, cast_type1, j * cast_type1->dimx + k, &scalar_type); + if (!(c = hlsl_new_uint_constant(ctx, offset, loc))) + return false; + list_add_tail(params->instrs, &c->node.entry); + + if (!(value1 = add_load(ctx, params->instrs, cast1, &c->node, scalar_type, *loc))) + return false; + + offset = hlsl_compute_component_offset(ctx, cast_type2, k * cast_type2->dimx + i, &scalar_type); + if (!(c = hlsl_new_uint_constant(ctx, offset, loc))) + return false; + list_add_tail(params->instrs, &c->node.entry); + + if (!(value2 = add_load(ctx, params->instrs, cast2, &c->node, scalar_type, *loc))) + return false; + + if (!(mul = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, &value1->node, &value2->node, loc))) + return false; + + if (node) + { + if (!(node = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_ADD, node, mul, loc))) + return false; + } + else + { + node = mul; + } + } + + offset = hlsl_compute_component_offset(ctx, matrix_type, j * matrix_type->dimx + i, &scalar_type); + if (!(c = hlsl_new_uint_constant(ctx, offset, loc))) + return false; + list_add_tail(params->instrs, &c->node.entry); + + if (!(store = hlsl_new_store(ctx, var, &c->node, node, 0, *loc))) + return false; + list_add_tail(params->instrs, &store->node.entry); + } + + if (!(load = hlsl_new_load(ctx, var, NULL, matrix_type, *loc))) + return false; + list_add_tail(params->instrs, &load->node.entry); + + return !!add_implicit_conversion(ctx, params->instrs, &load->node, ret_type, loc); +} + static bool intrinsic_pow(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -2106,6 +2223,7 @@ intrinsic_functions[] = {"floor", 1, true, intrinsic_floor}, {"max", 2, true, intrinsic_max}, {"min", 2, true, intrinsic_min}, + {"mul", 2, true, intrinsic_mul}, {"pow", 2, true, intrinsic_pow}, {"round", 1, true, intrinsic_round}, {"saturate", 1, true, intrinsic_saturate}, diff --git a/tests/hlsl-majority-pragma.shader_test b/tests/hlsl-majority-pragma.shader_test index a9f917ff..e7fc75cd 100644 --- a/tests/hlsl-majority-pragma.shader_test +++ b/tests/hlsl-majority-pragma.shader_test @@ -17,5 +17,5 @@ uniform 0 float4 0.1 0.2 0.0 0.0 uniform 4 float4 0.3 0.4 0.0 0.0 uniform 8 float4 0.1 0.3 0.0 0.0 uniform 12 float4 0.2 0.4 0.0 0.0 -todo draw quad +draw quad probe all rgba (0.17, 0.39, 0.17, 0.39) 1 diff --git a/tests/hlsl-majority-typedef.shader_test b/tests/hlsl-majority-typedef.shader_test index 192c96db..1460e9a0 100644 --- a/tests/hlsl-majority-typedef.shader_test +++ b/tests/hlsl-majority-typedef.shader_test @@ -18,5 +18,5 @@ uniform 0 float4 0.1 0.2 0.0 0.0 uniform 4 float4 0.3 0.4 0.0 0.0 uniform 8 float4 0.1 0.3 0.0 0.0 uniform 12 float4 0.2 0.4 0.0 0.0 -todo draw quad +draw quad probe all rgba (0.17, 0.39, 0.17, 0.39) 1 diff --git a/tests/hlsl-mul.shader_test b/tests/hlsl-mul.shader_test index 1d137e69..7b453187 100644 --- a/tests/hlsl-mul.shader_test +++ b/tests/hlsl-mul.shader_test @@ -12,7 +12,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (30.0, 70.0, 110.0, 150.0) [pixel shader] @@ -28,7 +28,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (90.0, 100.0, 110.0, 120.0) [pixel shader] @@ -44,7 +44,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (14.0, 38.0, 62.0, 86.0) [pixel shader] @@ -60,7 +60,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (38.0, 44.0, 50.0, 56.0) [pixel shader] @@ -75,7 +75,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (14.0, 32.0, 50.0, 0.0) [pixel shader] @@ -90,7 +90,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (30.0, 36.0, 42.0, 0.0) [pixel shader] @@ -106,7 +106,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (50.0, 60.0, 70.0, 80.0) [pixel shader] @@ -122,7 +122,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (50.0, 60.0, 70.0, 80.0) [pixel shader] @@ -138,7 +138,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (10.0, 20.0, 30.0, 40.0) [pixel shader] @@ -154,7 +154,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (10.0, 50.0, 90.0, 130.0) [pixel shader] @@ -170,7 +170,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (10.0, 20.0, 30.0, 40.0) [pixel shader] @@ -186,7 +186,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (10.0, 50.0, 90.0, 130.0) [pixel shader] @@ -202,7 +202,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (90.0, 100.0, 110.0, 120.0) [pixel shader] @@ -218,7 +218,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (5.0, 10.0, 15.0, 20.0) [pixel shader] @@ -234,7 +234,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (2.0, 4.0, 6.0, 8.0) [pixel shader] @@ -250,7 +250,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (30.0, 70.0, 110.0, 150.0) [pixel shader] @@ -268,7 +268,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (83.0, 98.0, 113.0, 128.0) [pixel shader] @@ -286,5 +286,5 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +draw quad probe all rgba (78.0, 96.0, 114.0, 0.0) -- 2.36.1