April 2024 - Wine-commits - list.winehq.org

Nikolay Sivov : vkd3d-shader/hlsl: Support refract() intrinsic.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: 8d78e3a821f3770a359ee29e806b80cfd6e4eb15 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/8d78e3a821f3770a359ee29e806b8… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Thu Sep 7 14:16:15 2023 +0200 vkd3d-shader/hlsl: Support refract() intrinsic. With some changes by Giovanni Mascellani. --- libs/vkd3d-shader/hlsl.y | 54 ++++++++++++++++++++++++++++++++++++++++++ tests/hlsl/refract.shader_test | 42 ++++++++++++++++---------------- 2 files changed, 75 insertions(+), 21 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index 03e14d84..7fc35d4e 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -3740,6 +3740,59 @@ static bool intrinsic_reflect(struct hlsl_ctx *ctx, return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_ADD, i, neg, loc); } +static bool intrinsic_refract(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_type *r_type = params->args[0]->data_type; + struct hlsl_type *n_type = params->args[1]->data_type; + struct hlsl_type *i_type = params->args[2]->data_type; + struct hlsl_type *res_type, *idx_type, *scal_type; + struct parse_initializer mut_params; + struct hlsl_ir_function_decl *func; + enum hlsl_base_type base; + char *body; + + static const char template[] = + "%s refract(%s r, %s n, %s i)\n" + "{\n" + " %s d, t;\n" + " d = dot(r, n);\n" + " t = 1 - i.x * i.x * (1 - d * d);\n" + " return t >= 0.0 ? i.x * r - (i.x * d + sqrt(t)) * n : 0;\n" + "}"; + + if (r_type->class == HLSL_CLASS_MATRIX + || n_type->class == HLSL_CLASS_MATRIX + || i_type->class == HLSL_CLASS_MATRIX) + { + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE, "Matrix arguments are not supported."); + return false; + } + + assert(params->args_count == 3); + mut_params = *params; + mut_params.args_count = 2; + if (!(res_type = elementwise_intrinsic_get_common_type(ctx, &mut_params, loc))) + return false; + + base = expr_common_base_type(res_type->base_type, i_type->base_type); + base = base == HLSL_TYPE_HALF ? HLSL_TYPE_HALF : HLSL_TYPE_FLOAT; + res_type = convert_numeric_type(ctx, res_type, base); + idx_type = convert_numeric_type(ctx, i_type, base); + scal_type = hlsl_get_scalar_type(ctx, base); + + if (!(body = hlsl_sprintf_alloc(ctx, template, res_type->name, res_type->name, + res_type->name, idx_type->name, scal_type->name))) + return false; + + func = hlsl_compile_internal_function(ctx, "refract", body); + vkd3d_free(body); + if (!func) + return false; + + return add_user_call(ctx, func, params, loc); +} + static bool intrinsic_round(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -4255,6 +4308,7 @@ intrinsic_functions[] = {"pow", 2, true, intrinsic_pow}, {"radians", 1, true, intrinsic_radians}, {"reflect", 2, true, intrinsic_reflect}, + {"refract", 3, true, intrinsic_refract}, {"round", 1, true, intrinsic_round}, {"rsqrt", 1, true, intrinsic_rsqrt}, {"saturate", 1, true, intrinsic_saturate}, diff --git a/tests/hlsl/refract.shader_test b/tests/hlsl/refract.shader_test index ecd554b0..53af811f 100644 --- a/tests/hlsl/refract.shader_test +++ b/tests/hlsl/refract.shader_test @@ -1,4 +1,4 @@ -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -12,13 +12,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (-0.550931, -0.453954, 0.3654653, -1.0248856) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -34,13 +34,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (-0.519904912, -0.4332699, 0.0, 0.0) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -55,13 +55,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (-0.519904912, -0.4332699, 0.0, 0.0) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -76,13 +76,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (-0.519904912, -0.4332699, 0.0, 0.0) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(sm<4 | glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -98,13 +98,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (-0.524978, 0.0, 0.0, 0.0) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader todo(sm<6)] +[pixel shader] float4 r; float4 n; float i; @@ -118,13 +118,13 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float 0.2 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (-0.5562381, -0.6762381, -0.6162381, -0.5962381) 32 uniform 8 float 100.0 -todo(sm<6 | glsl) draw quad +todo(glsl) draw quad probe all rgba (0.0, 0.0, 0.0, 0.0) -[pixel shader fail(sm>=6) todo(sm<6)] +[pixel shader fail(sm>=6)] float4 r; float4 n; float4 i; @@ -138,10 +138,10 @@ float4 main() : sv_target uniform 0 float4 0.5 -0.1 0.2 0.3 uniform 4 float4 0.6 0.4 -0.3 1.0 uniform 8 float4 0.2 0.3 0.4 0.5 -if(sm<6) todo(sm<6 | glsl) draw quad +if(sm<6) todo(glsl) draw quad if(sm<6) probe all rgba (-0.550931, -0.453954, 0.3654653, -1.0248856) 32 uniform 8 float 100.0 -if(sm<6) todo(sm<6 | glsl) draw quad +if(sm<6) todo(glsl) draw quad if(sm<6) probe all rgba (0.0, 0.0, 0.0, 0.0) [pixel shader fail]

1 0

Giovanni Mascellani : tests: Check that side effects do not happen after discard.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: e29706d3666e3d29e84e76d4deca3c07190d63da URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/e29706d3666e3d29e84e76d4deca3… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Mar 20 13:29:39 2024 +0100 tests: Check that side effects do not happen after discard. --- tests/hlsl/discard.shader_test | 50 +++++++++++++++++++++++++++++++++++++++++- 1 file changed, 49 insertions(+), 1 deletion(-) diff --git a/tests/hlsl/discard.shader_test b/tests/hlsl/discard.shader_test index a5c210f0..cf53bf48 100644 --- a/tests/hlsl/discard.shader_test +++ b/tests/hlsl/discard.shader_test @@ -3,7 +3,8 @@ uniform float4 x; float4 main() : sv_target { - if (x.x == 9.0f) discard; + if (x.x == 9.0f) + discard; return x; } @@ -14,3 +15,50 @@ probe all rgba (1, 2, 3, 4) uniform 0 float4 9 8 7 6 todo(sm<4 | glsl) draw quad probe all rgba (1, 2, 3, 4) + +[require] +shader model >= 5.0 + +[uav 0] +format r32 float +size (2d, 2, 1) + +0.0 0.0 + +[uav 1] +format r32 float +size (2d, 2, 1) + +0.0 0.0 + +% Check that side effects stop happening after discard +[pixel shader] +uniform float4 x; +RWTexture2D<float> y; + +float4 main(float4 pos : sv_position) : sv_target +{ + if (pos.x == 0.5 && pos.y == 0.5) + y[uint2(0, 0)] += 1; + if (x.x == 9.0f) + discard; + if (pos.x == 0.5 && pos.y == 0.5) + y[uint2(1, 0)] += 1; + return x; +} + +[test] +uniform 0 float4 1 2 3 4 +todo(glsl) draw quad +probe all rgba (1, 2, 3, 4) +if(sm<6) probe uav 1 (0, 0) r (1.0) +if(sm<6) probe uav 1 (1, 0) r (1.0) +if(sm>=6) probe uav 0 (0, 0) r (1.0) +if(sm>=6) probe uav 0 (1, 0) r (1.0) +uniform 0 float4 9 8 7 6 +todo(glsl) draw quad +probe all rgba (1, 2, 3, 4) +if(sm<6) probe uav 1 (0, 0) r (2.0) +if(sm<6) probe uav 1 (1, 0) r (1.0) +if(sm>=6) probe uav 0 (0, 0) r (2.0) +if(sm>=6) probe uav 0 (1, 0) r (1.0)

1 0

Giovanni Mascellani : vkd3d-shader/ir: Move `break's out of selection constructs when possible.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: f5d1b5d263cda3c4db0195075733bb5eac30eb9d URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/f5d1b5d263cda3c4db0195075733b… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Mar 13 13:09:51 2024 +0100 vkd3d-shader/ir: Move `break's out of selection constructs when possible. --- libs/vkd3d-shader/ir.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 68 insertions(+) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 9d5b7904..450f753c 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4115,6 +4115,71 @@ static void vsir_cfg_remove_trailing_continue(struct vsir_cfg_structure_list *li --list->count; } +static struct vsir_cfg_structure *vsir_cfg_get_trailing_break(struct vsir_cfg_structure_list *list) +{ + struct vsir_cfg_structure *structure; + size_t count = list->count; + + if (count == 0) + return NULL; + + structure = &list->structures[count - 1]; + + if (structure->type != STRUCTURE_TYPE_JUMP || structure->u.jump.type != JUMP_BREAK + || structure->u.jump.condition) + return NULL; + + return structure; +} + +/* When the last instruction in both branches of a selection construct + * is an unconditional break, any of them can be moved after the + * selection construct. If they break the same loop both of them can + * be moved out, otherwise we can choose which one: we choose the one + * that breaks the innermost loop, because we hope to eventually + * remove the loop itself. + * + * In principle a similar movement could be done when the last + * instructions are continue and continue, or continue and break. But + * in practice I don't think those situations can happen given the + * previous passes we do on the program, so we don't care. */ +static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg_structure_list *list) +{ + struct vsir_cfg_structure *selection, *if_break, *else_break, *new_break; + unsigned int if_target, else_target, max_target; + size_t pos = list->count - 1; + + selection = &list->structures[pos]; + assert(selection->type == STRUCTURE_TYPE_SELECTION); + + if_break = vsir_cfg_get_trailing_break(&selection->u.selection.if_body); + else_break = vsir_cfg_get_trailing_break(&selection->u.selection.else_body); + + if (!if_break || !else_break) + return VKD3D_OK; + + if_target = if_break->u.jump.target; + else_target = else_break->u.jump.target; + max_target = max(if_target, else_target); + + if (!(new_break = vsir_cfg_structure_list_append(list, STRUCTURE_TYPE_JUMP))) + return VKD3D_ERROR_OUT_OF_MEMORY; + new_break->u.jump.type = JUMP_BREAK; + new_break->u.jump.target = max_target; + + /* Pointer `selection' could have been invalidated by the append + * operation. */ + selection = &list->structures[pos]; + assert(selection->type == STRUCTURE_TYPE_SELECTION); + + if (if_target == max_target) + --selection->u.selection.if_body.count; + if (else_target == max_target) + --selection->u.selection.else_body.count; + + return VKD3D_OK; +} + static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg_structure_list *list) { enum vkd3d_result ret; @@ -4155,6 +4220,9 @@ static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg_structur if ((ret = vsir_cfg_synthesize_selections(&structure->u.selection.else_body)) < 0) return ret; + if ((ret = vsir_cfg_move_breaks_out_of_selections(list)) < 0) + return ret; + break; }

1 0

Giovanni Mascellani : vkd3d-shader/ir: Count how many jumps target each loop.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: 105ccc4769b966f019d9e04b16cf75cf6f1a1e81 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/105ccc4769b966f019d9e04b16cf7… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Mar 13 13:39:24 2024 +0100 vkd3d-shader/ir: Count how many jumps target each loop. --- libs/vkd3d-shader/ir.c | 67 ++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 60 insertions(+), 7 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 450f753c..52e1a9c2 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -3056,6 +3056,9 @@ struct vsir_cfg * block), but we still try to keep `begin' as forward as * possible, to keep the loop scope as small as possible. */ bool synthetic; + /* The number of jump instructions (both conditional and + * unconditional) that target this loop. */ + unsigned int target_count; } *loop_intervals; size_t loop_interval_count, loop_interval_capacity; @@ -3104,6 +3107,7 @@ static enum vkd3d_result vsir_cfg_add_loop_interval(struct vsir_cfg *cfg, unsign interval->begin = begin; interval->end = end; interval->synthetic = synthetic; + interval->target_count = 0; return VKD3D_OK; } @@ -4106,13 +4110,18 @@ fail: return VKD3D_ERROR_OUT_OF_MEMORY; } -static void vsir_cfg_remove_trailing_continue(struct vsir_cfg_structure_list *list, unsigned int target) +static void vsir_cfg_remove_trailing_continue(struct vsir_cfg *cfg, + struct vsir_cfg_structure_list *list, unsigned int target) { struct vsir_cfg_structure *last = &list->structures[list->count - 1]; if (last->type == STRUCTURE_TYPE_JUMP && last->u.jump.type == JUMP_CONTINUE && !last->u.jump.condition && last->u.jump.target == target) + { --list->count; + assert(cfg->loop_intervals[target].target_count > 0); + --cfg->loop_intervals[target].target_count; + } } static struct vsir_cfg_structure *vsir_cfg_get_trailing_break(struct vsir_cfg_structure_list *list) @@ -4143,7 +4152,8 @@ static struct vsir_cfg_structure *vsir_cfg_get_trailing_break(struct vsir_cfg_st * instructions are continue and continue, or continue and break. But * in practice I don't think those situations can happen given the * previous passes we do on the program, so we don't care. */ -static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg_structure_list *list) +static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg *cfg, + struct vsir_cfg_structure_list *list) { struct vsir_cfg_structure *selection, *if_break, *else_break, *new_break; unsigned int if_target, else_target, max_target; @@ -4166,6 +4176,7 @@ static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg_ return VKD3D_ERROR_OUT_OF_MEMORY; new_break->u.jump.type = JUMP_BREAK; new_break->u.jump.target = max_target; + ++cfg->loop_intervals[max_target].target_count; /* Pointer `selection' could have been invalidated by the append * operation. */ @@ -4173,14 +4184,24 @@ static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg_ assert(selection->type == STRUCTURE_TYPE_SELECTION); if (if_target == max_target) + { --selection->u.selection.if_body.count; + assert(cfg->loop_intervals[if_target].target_count > 0); + --cfg->loop_intervals[if_target].target_count; + } + if (else_target == max_target) + { --selection->u.selection.else_body.count; + assert(cfg->loop_intervals[else_target].target_count > 0); + --cfg->loop_intervals[else_target].target_count; + } return VKD3D_OK; } -static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg_structure_list *list) +static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg *cfg, + struct vsir_cfg_structure_list *list) { enum vkd3d_result ret; size_t i; @@ -4217,10 +4238,10 @@ static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg_structur *structure = new_selection; list->count = i + 1; - if ((ret = vsir_cfg_synthesize_selections(&structure->u.selection.else_body)) < 0) + if ((ret = vsir_cfg_synthesize_selections(cfg, &structure->u.selection.else_body)) < 0) return ret; - if ((ret = vsir_cfg_move_breaks_out_of_selections(list)) < 0) + if ((ret = vsir_cfg_move_breaks_out_of_selections(cfg, list)) < 0) return ret; break; @@ -4247,22 +4268,54 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct if (loop_body->count == 0) continue; - vsir_cfg_remove_trailing_continue(loop_body, loop->u.loop.idx); + vsir_cfg_remove_trailing_continue(cfg, loop_body, loop->u.loop.idx); if ((ret = vsir_cfg_optimize_recurse(cfg, loop_body)) < 0) return ret; - if ((ret = vsir_cfg_synthesize_selections(loop_body)) < 0) + if ((ret = vsir_cfg_synthesize_selections(cfg, loop_body)) < 0) return ret; } return VKD3D_OK; } +static void vsir_cfg_count_targets(struct vsir_cfg *cfg, struct vsir_cfg_structure_list *list) +{ + size_t i; + + for (i = 0; i < list->count; ++i) + { + struct vsir_cfg_structure *structure = &list->structures[i]; + + switch (structure->type) + { + case STRUCTURE_TYPE_BLOCK: + break; + + case STRUCTURE_TYPE_LOOP: + vsir_cfg_count_targets(cfg, &structure->u.loop.body); + break; + + case STRUCTURE_TYPE_SELECTION: + vsir_cfg_count_targets(cfg, &structure->u.selection.if_body); + vsir_cfg_count_targets(cfg, &structure->u.selection.else_body); + break; + + case STRUCTURE_TYPE_JUMP: + if (structure->u.jump.type == JUMP_BREAK || structure->u.jump.type == JUMP_CONTINUE) + ++cfg->loop_intervals[structure->u.jump.target].target_count; + break; + } + } +} + static enum vkd3d_result vsir_cfg_optimize(struct vsir_cfg *cfg) { enum vkd3d_result ret; + vsir_cfg_count_targets(cfg, &cfg->structured_program); + ret = vsir_cfg_optimize_recurse(cfg, &cfg->structured_program); if (TRACE_ON())

1 0

Giovanni Mascellani : vkd3d-shader/ir: Remove loops that terminate with a `break'.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: aad040af05139ab0335b90a0112871ac818747c5 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/aad040af05139ab0335b90a011287… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Mar 13 22:43:50 2024 +0100 vkd3d-shader/ir: Remove loops that terminate with a `break'. --- libs/vkd3d-shader/ir.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 79 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 52e1a9c2..e0928ff7 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4250,8 +4250,64 @@ static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg *cfg, return VKD3D_OK; } +static enum vkd3d_result vsir_cfg_append_loop(struct vsir_cfg *cfg, + struct vsir_cfg_structure_list *new_list, struct vsir_cfg_structure *loop) +{ + struct vsir_cfg_structure_list *loop_body = &loop->u.loop.body; + unsigned int target, loop_idx = loop->u.loop.idx; + struct vsir_cfg_structure *trailing_break; + enum vkd3d_result ret; + + trailing_break = vsir_cfg_get_trailing_break(loop_body); + + /* If the loop's last instruction is not a break, we cannot remove + * the loop itself. */ + if (!trailing_break) + { + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, loop, 1)) < 0) + return ret; + memset(loop, 0, sizeof(*loop)); + return VKD3D_OK; + } + + target = trailing_break->u.jump.target; + assert(cfg->loop_intervals[target].target_count > 0); + + /* If the loop is not targeted by any jump, we can remove it. The + * trailing `break' then targets another loop, so we have to keep + * it. */ + if (cfg->loop_intervals[loop_idx].target_count == 0) + { + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, + &loop_body->structures[0], loop_body->count)) < 0) + return ret; + loop_body->count = 0; + return VKD3D_OK; + } + + /* If the loop is targeted only by its own trailing `break' + * instruction, then we can remove it together with the `break' + * itself. */ + if (target == loop_idx && cfg->loop_intervals[loop_idx].target_count == 1) + { + --cfg->loop_intervals[loop_idx].target_count; + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, + &loop_body->structures[0], loop_body->count - 1)) < 0) + return ret; + loop_body->count = 0; + return VKD3D_OK; + } + + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, loop, 1)) < 0) + return ret; + memset(loop, 0, sizeof(*loop)); + + return VKD3D_OK; +} + static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct vsir_cfg_structure_list *list) { + struct vsir_cfg_structure_list new_list = {0}; enum vkd3d_result ret; size_t i; @@ -4261,23 +4317,44 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct struct vsir_cfg_structure_list *loop_body; if (loop->type != STRUCTURE_TYPE_LOOP) + { + if ((ret = vsir_cfg_structure_list_append_from_region(&new_list, loop, 1)) < 0) + goto fail; + memset(loop, 0, sizeof(*loop)); continue; + } loop_body = &loop->u.loop.body; if (loop_body->count == 0) + { + if ((ret = vsir_cfg_structure_list_append_from_region(&new_list, loop, 1)) < 0) + goto fail; + memset(loop, 0, sizeof(*loop)); continue; + } vsir_cfg_remove_trailing_continue(cfg, loop_body, loop->u.loop.idx); if ((ret = vsir_cfg_optimize_recurse(cfg, loop_body)) < 0) - return ret; + goto fail; if ((ret = vsir_cfg_synthesize_selections(cfg, loop_body)) < 0) - return ret; + goto fail; + + if ((ret = vsir_cfg_append_loop(cfg, &new_list, loop)) < 0) + goto fail; } + vsir_cfg_structure_list_cleanup(list); + *list = new_list; + return VKD3D_OK; + +fail: + vsir_cfg_structure_list_cleanup(list); + + return ret; } static void vsir_cfg_count_targets(struct vsir_cfg *cfg, struct vsir_cfg_structure_list *list)

1 0

Henri Verbeet : configure: Use AC_CHECK_TOOL to check for widl.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: 2b7d9790d6da8055aefd1ab395335cabc002e71f URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/2b7d9790d6da8055aefd1ab395335… Author: Henri Verbeet <hverbeet(a)codeweavers.com> Date: Wed Apr 10 20:36:40 2024 +0200 configure: Use AC_CHECK_TOOL to check for widl. In particular, this will automatically pick up x86_64-w64-mingw32-widl/i686-w64-mingw32-widl (provided by mingw-w64-tools on Debian) when present and building for the corresponding hosts. --- m4/widl.m4 | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/m4/widl.m4 b/m4/widl.m4 index 6d226064..2311b832 100644 --- a/m4/widl.m4 +++ b/m4/widl.m4 @@ -1,7 +1,7 @@ dnl VKD3D_PROG_WIDL(major, minor) AC_DEFUN([VKD3D_PROG_WIDL], -[AC_CHECK_PROG([WIDL], [widl], [widl], [no]) +[AC_CHECK_TOOL([WIDL], [widl], [no]) AS_IF([test "x$WIDL" != "xno"], [AC_MSG_CHECKING([checking whether widl version >= $1.$2])

1 0

Conor McCarthy : tests/hlsl: Add tests for CalculateLevelOfDetail().
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: bdfa14d219c68aab54a3a8a5ddcb8a99698332e7 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/bdfa14d219c68aab54a3a8a5ddcb8… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Thu Mar 21 11:21:12 2024 +1000 tests/hlsl: Add tests for CalculateLevelOfDetail(). Includes improvements made by Giovanni Mascellani. --- Makefile.am | 1 + tests/hlsl/calculate-lod.shader_test | 84 ++++++++++++++++++++++++++++++++++++ tests/shader_runner_d3d11.c | 2 +- tests/shader_runner_d3d12.c | 2 +- 4 files changed, 87 insertions(+), 2 deletions(-) diff --git a/Makefile.am b/Makefile.am index ead3c1f5..995f70b2 100644 --- a/Makefile.am +++ b/Makefile.am @@ -65,6 +65,7 @@ vkd3d_shader_tests = \ tests/hlsl/bitwise.shader_test \ tests/hlsl/bool-cast.shader_test \ tests/hlsl/bool-semantics.shader_test \ + tests/hlsl/calculate-lod.shader_test \ tests/hlsl/cast-64-bit.shader_test \ tests/hlsl/cast-broadcast.shader_test \ tests/hlsl/cast-componentwise-compatible.shader_test \ diff --git a/tests/hlsl/calculate-lod.shader_test b/tests/hlsl/calculate-lod.shader_test new file mode 100644 index 00000000..807776ee --- /dev/null +++ b/tests/hlsl/calculate-lod.shader_test @@ -0,0 +1,84 @@ +[require] +shader model >= 4.1 + + +[sampler 0] +filter linear linear linear +address clamp clamp clamp + +[srv 0] +size (2d, 4, 4) +levels 3 + +0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 +0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 +0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 +0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 + + +[pixel shader todo] +sampler s; +Texture2D t; + +float4 main(float4 pos : sv_position) : sv_target +{ + return float4(t.CalculateLevelOfDetail(s, pos.xy), + t.CalculateLevelOfDetail(s, float2(0.5 * pos.x, 0.5 * pos.y)), + t.CalculateLevelOfDetail(s, float2(0.25 * pos.x, 0.25 * pos.y)), + floor(10.0 * t.CalculateLevelOfDetail(s, float2(0.1 * pos.x, 0.7 * pos.y)))); +} + +[test] +todo draw quad +probe all rgba (2.0, 1.0, 0.0, 14.0) + + +[pixel shader todo] +sampler s; +Texture2D t; + +float4 main(float4 pos : sv_position) : sv_target +{ + return float4(t.CalculateLevelOfDetail(s, float2(3.0 * pos.x, 5.0 * pos.y)), + t.CalculateLevelOfDetail(s, float2(10.0 * pos.x, 10.0 * pos.y)), + t.CalculateLevelOfDetail(s, float2(0.1 * pos.x, 0.02 * pos.y)), + t.CalculateLevelOfDetail(s, float2(0.01 * pos.x, 0.01 * pos.y))); +} + +[test] +todo draw quad +probe all rgba (2.0, 2.0, 0.0, 0.0) + + +[pixel shader todo] +sampler s; +Texture2D t; + +float4 main(float4 pos : sv_position) : sv_target +{ + return float4(t.CalculateLevelOfDetailUnclamped(s, pos.xy), + t.CalculateLevelOfDetailUnclamped(s, float2(0.5 * pos.x, 0.5 * pos.y)), + t.CalculateLevelOfDetailUnclamped(s, float2(0.25 * pos.x, 0.25 * pos.y)), + floor(10.0 * t.CalculateLevelOfDetailUnclamped(s, float2(0.1 * pos.x, 0.7 * pos.y)))); +} + +[test] +todo draw quad +probe all rgba (2.0, 1.0, 0.0, 14.0) + + +[pixel shader todo] +sampler s; +Texture2D t; + +float4 main(float4 pos : sv_position) : sv_target +{ + return float4(floor(10.0 * t.CalculateLevelOfDetailUnclamped(s, float2(3.0 * pos.x, 6.0 * pos.y))), + floor(10.0 * t.CalculateLevelOfDetailUnclamped(s, float2(8.0 * pos.x, 8.0 * pos.y))), + floor(10.0 * t.CalculateLevelOfDetailUnclamped(s, float2(0.1 * pos.x, 0.02 * pos.y))), + floor(10.0 * t.CalculateLevelOfDetailUnclamped(s, float2(0.01 * pos.x, 0.01 * pos.y)))); +} + +[test] +todo draw quad +probe all rgba (45.0, 50.0, -14.0, -47.0) diff --git a/tests/shader_runner_d3d11.c b/tests/shader_runner_d3d11.c index 1285ca41..b26d6a36 100644 --- a/tests/shader_runner_d3d11.c +++ b/tests/shader_runner_d3d11.c @@ -358,7 +358,7 @@ static ID3D11Buffer *create_buffer(ID3D11Device *device, unsigned int bind_flags static void init_resource_2d(struct d3d11_shader_runner *runner, struct d3d11_resource *resource, const struct resource_params *params) { - D3D11_SUBRESOURCE_DATA resource_data[2]; + D3D11_SUBRESOURCE_DATA resource_data[3]; ID3D11Device *device = runner->device; D3D11_TEXTURE2D_DESC desc = {0}; HRESULT hr; diff --git a/tests/shader_runner_d3d12.c b/tests/shader_runner_d3d12.c index 9bf7bb24..56844ae3 100644 --- a/tests/shader_runner_d3d12.c +++ b/tests/shader_runner_d3d12.c @@ -104,7 +104,7 @@ static struct resource *d3d12_runner_create_resource(struct shader_runner *r, co struct d3d12_shader_runner *runner = d3d12_shader_runner(r); struct test_context *test_context = &runner->test_context; ID3D12Device *device = test_context->device; - D3D12_SUBRESOURCE_DATA resource_data[2] = {0}; + D3D12_SUBRESOURCE_DATA resource_data[3] = {0}; struct d3d12_resource *resource; unsigned int buffer_offset = 0; D3D12_RESOURCE_STATES state;

1 0

Conor McCarthy : vkd3d-shader/dxil: Implement DX intrinsic CalculateLOD.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: cb5411408a1399e6bb27c6cf104edc5381ef6742 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/cb5411408a1399e6bb27c6cf104ed… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Thu Jun 15 00:11:56 2023 +1000 vkd3d-shader/dxil: Implement DX intrinsic CalculateLOD. --- libs/vkd3d-shader/dxil.c | 36 ++++++++++++++++++++++++++++++++++++ tests/hlsl/calculate-lod.shader_test | 8 ++++---- 2 files changed, 40 insertions(+), 4 deletions(-) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index 03c5458b..da8ba662 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -404,6 +404,7 @@ enum dx_intrinsic_opcode DX_ATOMIC_BINOP = 78, DX_ATOMIC_CMP_XCHG = 79, DX_BARRIER = 80, + DX_CALCULATE_LOD = 81, DX_DISCARD = 82, DX_DERIV_COARSEX = 83, DX_DERIV_COARSEY = 84, @@ -4486,6 +4487,40 @@ static void sm6_parser_emit_dx_buffer_update_counter(struct sm6_parser *sm6, enu instruction_dst_param_init_ssa_scalar(ins, sm6); } +static void sm6_parser_emit_dx_calculate_lod(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, + const struct sm6_value **operands, struct function_emission_state *state) +{ + const struct sm6_value *resource, *sampler; + struct vkd3d_shader_src_param *src_params; + struct vkd3d_shader_instruction *ins; + struct vkd3d_shader_register coord; + unsigned int clamp; + + resource = operands[0]; + sampler = operands[1]; + if (!sm6_value_validate_is_texture_handle(resource, op, sm6) + || !sm6_value_validate_is_sampler_handle(sampler, op, sm6)) + { + return; + } + + if (!sm6_parser_emit_coordinate_construct(sm6, &operands[2], 3, NULL, state, &coord)) + return; + + clamp = sm6_value_get_constant_uint(operands[5]); + + ins = state->ins; + vsir_instruction_init(ins, &sm6->p.location, VKD3DSIH_LOD); + if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) + return; + src_param_init_vector_from_reg(&src_params[0], &coord); + src_params[1].reg = resource->u.handle.reg; + src_param_init_scalar(&src_params[1], !clamp); + src_param_init_vector_from_reg(&src_params[2], &sampler->u.handle.reg); + + instruction_dst_param_init_ssa_scalar(ins, sm6); +} + static void sm6_parser_emit_dx_cbuffer_load(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, const struct sm6_value **operands, struct function_emission_state *state) { @@ -5514,6 +5549,7 @@ static const struct sm6_dx_opcode_info sm6_dx_op_table[] = [DX_BUFFER_LOAD ] = {"o", "Hii", sm6_parser_emit_dx_buffer_load}, [DX_BUFFER_STORE ] = {"v", "Hiiooooc", sm6_parser_emit_dx_buffer_store}, [DX_BUFFER_UPDATE_COUNTER ] = {"i", "H8", sm6_parser_emit_dx_buffer_update_counter}, + [DX_CALCULATE_LOD ] = {"f", "HHfffb", sm6_parser_emit_dx_calculate_lod}, [DX_CBUFFER_LOAD_LEGACY ] = {"o", "Hi", sm6_parser_emit_dx_cbuffer_load}, [DX_COS ] = {"g", "R", sm6_parser_emit_dx_sincos}, [DX_COUNT_BITS ] = {"i", "m", sm6_parser_emit_dx_unary}, diff --git a/tests/hlsl/calculate-lod.shader_test b/tests/hlsl/calculate-lod.shader_test index 807776ee..88325af6 100644 --- a/tests/hlsl/calculate-lod.shader_test +++ b/tests/hlsl/calculate-lod.shader_test @@ -29,7 +29,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +todo(sm<6 | glsl) draw quad probe all rgba (2.0, 1.0, 0.0, 14.0) @@ -46,7 +46,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +todo(sm<6 | glsl) draw quad probe all rgba (2.0, 2.0, 0.0, 0.0) @@ -63,7 +63,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +todo(sm<6 | glsl) draw quad probe all rgba (2.0, 1.0, 0.0, 14.0) @@ -80,5 +80,5 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo draw quad +todo(sm<6 | glsl) draw quad probe all rgba (45.0, 50.0, -14.0, -47.0)

1 0

Giovanni Mascellani : vkd3d-shader/d3d-asm: Reset indent level on HS phases.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: a2bb4ef3014088cd5d5c285555640979cf431ee8 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/a2bb4ef3014088cd5d5c285555640… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Mon Apr 8 23:45:11 2024 +0200 vkd3d-shader/d3d-asm: Reset indent level on HS phases. --- libs/vkd3d-shader/d3d_asm.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d-shader/d3d_asm.c index 2b5feb94..b65b2fe2 100644 --- a/libs/vkd3d-shader/d3d_asm.c +++ b/libs/vkd3d-shader/d3d_asm.c @@ -2354,6 +2354,10 @@ enum vkd3d_result d3d_asm_compile(const struct vsir_program *program, break; case VKD3DSIH_LABEL: + case VKD3DSIH_HS_DECLS: + case VKD3DSIH_HS_CONTROL_POINT_PHASE: + case VKD3DSIH_HS_FORK_PHASE: + case VKD3DSIH_HS_JOIN_PHASE: indent = 0; break;

1 0

Giovanni Mascellani : tests: Mark some test failures as todo on llvmpipe.
by Alexandre Julliard 11 Apr '24

11 Apr '24

Module: vkd3d Branch: master Commit: 0b2ce5c13a1d967bfaffe7db9407262af27a0623 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/0b2ce5c13a1d967bfaffe7db94072… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Tue Mar 12 11:37:00 2024 +0100 tests: Mark some test failures as todo on llvmpipe. --- tests/d3d12.c | 16 ++++++++++++++++ tests/d3d12_crosstest.h | 13 +++++++++++++ 2 files changed, 29 insertions(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index ccebf510..1222ed05 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -19214,8 +19214,16 @@ static void test_descriptors_visibility(void) root_signature_desc.NumStaticSamplers = 2; root_signature_desc.pStaticSamplers = sampler_desc; hr = create_root_signature(device, &root_signature_desc, &context.root_signature); + /* llvmpipe (from Mesa 24.0.4) has maxBoundDescriptorSet == 8, which is less than we need. */ + todo_if(is_llvmpipe_device(device)) ok(SUCCEEDED(hr), "Failed to create root signature, hr %#x.\n", hr); + if (FAILED(hr)) + { + destroy_test_context(&context); + return; + } + context.pipeline_state = create_pipeline_state(device, context.root_signature, context.render_target_desc.Format, &vs, &ps, NULL); @@ -20858,8 +20866,16 @@ static void test_depth_stencil_sampling(void) root_signature_desc.NumStaticSamplers = 2; root_signature_desc.pStaticSamplers = sampler_desc; hr = create_root_signature(device, &root_signature_desc, &context.root_signature); + /* llvmpipe (from Mesa 24.0.4) has maxBoundDescriptorSet == 8, which is less than we need. */ + todo_if(is_llvmpipe_device(device)) ok(SUCCEEDED(hr), "Failed to create root signature, hr %#x.\n", hr); + if (FAILED(hr)) + { + destroy_test_context(&context); + return; + } + pso_compare = create_pipeline_state(device, context.root_signature, context.render_target_desc.Format, NULL, &ps_compare, NULL); pso_depth = create_pipeline_state(device, diff --git a/tests/d3d12_crosstest.h b/tests/d3d12_crosstest.h index 6c0a4ef1..e2401645 100644 --- a/tests/d3d12_crosstest.h +++ b/tests/d3d12_crosstest.h @@ -401,6 +401,11 @@ static inline bool is_mesa_intel_device(ID3D12Device *device) return false; } +static inline bool is_llvmpipe_device(ID3D12Device *device) +{ + return false; +} + static inline bool is_nvidia_device(ID3D12Device *device) { return false; @@ -649,6 +654,14 @@ static inline bool is_mesa_intel_device(ID3D12Device *device) return properties.driverID == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA_KHR; } +static inline bool is_llvmpipe_device(ID3D12Device *device) +{ + VkPhysicalDeviceDriverPropertiesKHR properties; + + get_driver_properties(device, &properties); + return properties.driverID == VK_DRIVER_ID_MESA_LLVMPIPE; +} + static inline bool is_nvidia_device(ID3D12Device *device) { VkPhysicalDeviceDriverPropertiesKHR properties;

1 0