vkd3d-shader/ir: Introduce a simple control flow graph structurizer.

The structurizer is implemented along the lines of what is usually called
the "structured program theorem": the control flow is completely
virtualized by mean of an additional TEMP register which stores the
block index which is currently running. The whole program is then
converted to a huge switch construction enclosed in a loop, executing
at each iteration the appropriate block and updating the register
depending on block jump instruction.

The algorithm's generality is also its major weakness: it accepts any
input program, even if its CFG is not reducible, but the output
program lacks any useful convergence information. It satisfies the
letter of the SPIR-V requirements, but it is expected that it will
be very inefficient to run on a GPU (unless a downstream compiler is
able to devirtualize the control flow and do a proper convergence
analysis pass). The algorithm is however very simple, and good enough
to at least pass tests, enabling further development. A better
alternative is expected to be upstreamed incrementally.

Side note: the structured program theorem is often called the
Böhm-Jacopini theorem; Böhm and Jacopini did indeed prove a variation
of it, but their algorithm is different from what is commontly attributed
to them and implemented here, so I opted for not using their name.

libs/vkd3d-shader/ir.c

@@ -470,6 +470,25 @@ static void dst_param_init_ssa_bool(struct vkd3d_shader_dst_param *dst, unsigned
     dst->reg.idx[0].offset = idx;
 }
 
+static void dst_param_init_temp_uint(struct vkd3d_shader_dst_param *dst, unsigned int idx)
+{
+    vsir_dst_param_init(dst, VKD3DSPR_TEMP, VKD3D_DATA_UINT, 1);
+    dst->reg.idx[0].offset = idx;
+    dst->write_mask = VKD3DSP_WRITEMASK_0;
+}
+
+static void src_param_init_temp_uint(struct vkd3d_shader_src_param *src, unsigned int idx)
+{
+    vsir_src_param_init(src, VKD3DSPR_TEMP, VKD3D_DATA_UINT, 1);
+    src->reg.idx[0].offset = idx;
+}
+
+static void src_param_init_const_uint(struct vkd3d_shader_src_param *src, uint32_t value)
+{
+    vsir_src_param_init(src, VKD3DSPR_IMMCONST, VKD3D_DATA_UINT, 0);
+    src->reg.u.immconst_u32[0] = value;
+}
+
 void vsir_instruction_init(struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location,
         enum vkd3d_shader_opcode handler_idx)
 {
@@ -2871,6 +2890,125 @@ fail:
     return VKD3D_ERROR_OUT_OF_MEMORY;
 }
 
+static enum vkd3d_result simple_structurizer_run(struct vkd3d_shader_parser *parser)
+{
+    const unsigned int block_temp_idx = parser->program.temp_count;
+    struct vkd3d_shader_instruction *instructions = NULL;
+    const struct vkd3d_shader_location no_loc = {0};
+    size_t ins_capacity = 0, ins_count = 0, i;
+    bool first_label_found = false;
+
+    if (!reserve_instructions(&instructions, &ins_capacity, parser->program.instructions.count))
+        goto fail;
+
+    for (i = 0; i < parser->program.instructions.count; ++i)
+    {
+        struct vkd3d_shader_instruction *ins = &parser->program.instructions.elements[i];
+
+        switch (ins->handler_idx)
+        {
+            case VKD3DSIH_PHI:
+            case VKD3DSIH_SWITCH_MONOLITHIC:
+                vkd3d_unreachable();
+
+            case VKD3DSIH_LABEL:
+                if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 4))
+                    goto fail;
+
+                if (!first_label_found)
+                {
+                    first_label_found = true;
+
+                    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_MOV, 1, 1))
+                        goto fail;
+                    dst_param_init_temp_uint(&instructions[ins_count].dst[0], block_temp_idx);
+                    src_param_init_const_uint(&instructions[ins_count].src[0], label_from_src_param(&ins->src[0]));
+                    ins_count++;
+
+                    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_LOOP, 0, 0))
+                        goto fail;
+                    ins_count++;
+
+                    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_SWITCH, 0, 1))
+                        goto fail;
+                    src_param_init_temp_uint(&instructions[ins_count].src[0], block_temp_idx);
+                    ins_count++;
+                }
+
+                if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_CASE, 0, 1))
+                    goto fail;
+                src_param_init_const_uint(&instructions[ins_count].src[0], label_from_src_param(&ins->src[0]));
+                ins_count++;
+                break;
+
+            case VKD3DSIH_BRANCH:
+                if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 2))
+                    goto fail;
+
+                if (vsir_register_is_label(&ins->src[0].reg))
+                {
+                    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_MOV, 1, 1))
+                        goto fail;
+                    dst_param_init_temp_uint(&instructions[ins_count].dst[0], block_temp_idx);
+                    src_param_init_const_uint(&instructions[ins_count].src[0], label_from_src_param(&ins->src[0]));
+                    ins_count++;
+                }
+                else
+                {
+                    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_MOVC, 1, 3))
+                        goto fail;
+                    dst_param_init_temp_uint(&instructions[ins_count].dst[0], block_temp_idx);
+                    instructions[ins_count].src[0] = ins->src[0];
+                    src_param_init_const_uint(&instructions[ins_count].src[1], label_from_src_param(&ins->src[1]));
+                    src_param_init_const_uint(&instructions[ins_count].src[2], label_from_src_param(&ins->src[2]));
+                    ins_count++;
+                }
+
+                if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_BREAK, 0, 0))
+                    goto fail;
+                ins_count++;
+                break;
+
+            case VKD3DSIH_RET:
+            default:
+                if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 1))
+                    goto fail;
+
+                instructions[ins_count++] = *ins;
+                break;
+        }
+    }
+
+    assert(first_label_found);
+
+    if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 3))
+        goto fail;
+
+    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_ENDSWITCH, 0, 0))
+        goto fail;
+    ins_count++;
+
+    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_ENDLOOP, 0, 0))
+        goto fail;
+    ins_count++;
+
+    if (!vsir_instruction_init_with_params(&parser->program, &instructions[ins_count], &no_loc, VKD3DSIH_RET, 0, 0))
+        goto fail;
+    ins_count++;
+
+    vkd3d_free(parser->program.instructions.elements);
+    parser->program.instructions.elements = instructions;
+    parser->program.instructions.capacity = ins_capacity;
+    parser->program.instructions.count = ins_count;
+    parser->program.temp_count += 1;
+
+    return VKD3D_OK;
+
+fail:
+    vkd3d_free(instructions);
+    return VKD3D_ERROR_OUT_OF_MEMORY;
+}
+
 enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser,
         const struct vkd3d_shader_compile_info *compile_info)
 {
@@ -2889,6 +3027,9 @@ enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser,
 
         if ((result = materialize_ssas_to_temps(parser)) < 0)
             return result;
+
+        if ((result = simple_structurizer_run(parser)) < 0)
+            return result;
     }
     else
     {
@@ -2916,13 +3057,13 @@ enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser,
 
         remove_dead_code(&parser->program);
 
-        if ((result = flatten_control_flow_constructs(parser)) < 0)
-            return result;
-
         if ((result = normalise_combined_samplers(parser)) < 0)
             return result;
     }
 
+    if ((result = flatten_control_flow_constructs(parser)) < 0)
+        return result;
+
     if (TRACE_ON())
         vkd3d_shader_trace(&parser->program);
 

tests/hlsl/conditional.shader_test

@@ -88,7 +88,7 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float4 0.0 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.9, 0.8, 0.7, 0.6)
 
 [pixel shader]

tests/hlsl/for.shader_test

@@ -54,7 +54,7 @@ float4 main(float tex : texcoord) : sv_target
 
 [test]
 uniform 0 uint4 10 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe (  0, 0, 159, 480) rgba (10.0, 35.0, 0.0, 0.0)
 probe (161, 0, 479, 480) rgba (10.0, 38.0, 0.0, 0.0)
 probe (481, 0, 640, 480) rgba ( 5.0, 10.0, 0.0, 0.0)

tests/hlsl/function-return.shader_test

@@ -79,16 +79,16 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.3, 0.2, 0.6, 0.3) 1
 uniform 0 float 0.4
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.6, 0.5, 0.6, 0.3) 1
 uniform 0 float 0.6
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.6, 0.5, 0.4, 0.5) 1
 uniform 0 float 0.8
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.8, 0.7, 0.4, 0.5) 1
 
 [pixel shader todo(sm<4)]
@@ -134,13 +134,13 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.2, 0.1, 0.2, 0.1) 1
 uniform 0 float 0.5
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.5, 0.4, 1.0, 0.9) 1
 uniform 0 float 0.9
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (1.0, 0.9, 1.0, 0.6) 1
 
 [pixel shader todo(sm<4)]
@@ -235,23 +235,23 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.3, 0.2, 0.3, 0.3) 1
 
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.3, 0.3, 0.3, 0.3) 1
 
 uniform 0 float 0.3
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.3, 0.5, 0.3, 0.3) 1
 
 uniform 0 float 0.7
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.3, 0.9, 0.7, 0.6) 1
 
 uniform 0 float 0.9
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.4, 0.1, 0.7, 0.6) 1
 
 [pixel shader todo(sm<4)]
@@ -291,21 +291,21 @@ float4 main() : sv_target
 uniform 0 float4 0.3 0.0 0.0 0.0
 uniform 4 float4 0.0 0.0 0.0 0.0
 uniform 8 float4 0.1 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
-todo(sm>=6) probe all rgba (0.3, 0.2, 0.6, 0.6) 1
+probe all rgba (0.3, 0.2, 0.6, 0.6) 1
 
 uniform 4 float4 0.35 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
-todo(sm>=6) probe all rgba (0.3, 0.3, 0.6, 0.6) 1
+probe all rgba (0.3, 0.3, 0.6, 0.6) 1
 
 uniform 8 float4 0.5 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
-todo(sm>=6) probe all rgba (0.3, 0.5, 0.6, 0.6) 1
+probe all rgba (0.3, 0.5, 0.6, 0.6) 1
 
 uniform 0 float4 1.0 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
-todo(sm>=6) probe all rgba (0.3, 0.5, 0.6, 0.6) 1
+probe all rgba (0.3, 0.5, 0.6, 0.6) 1
 
 uniform 4 float4 2.0 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
-todo(sm>=6) probe all rgba (0.4, 0.1, 0.6, 0.6) 1
+probe all rgba (0.4, 0.1, 0.6, 0.6) 1

tests/hlsl/loop.shader_test

@@ -118,7 +118,7 @@ float4 main() : sv_target
 }
 
 [test]
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (10.0, 10.0, 10.0, 10.0)
 
 [pixel shader todo(sm<4)]
@@ -137,7 +137,7 @@ float4 main() : sv_target
 }
 
 [test]
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (10.0, 10.0, 10.0, 10.0)
 
 [pixel shader todo(sm<4)]
@@ -156,7 +156,7 @@ float4 main() : sv_target
 }
 
 [test]
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (10.0, 10.0, 10.0, 10.0)
 
 % unroll can't be used with fastopt or loop

tests/hlsl/return.shader_test

@@ -89,13 +89,13 @@ void main(out float4 ret : sv_target)
 
 [test]
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.1, 0.2, 0.3, 0.4) 1
 uniform 0 float 0.5
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.2, 0.3, 0.4, 0.5) 1
 uniform 0 float 0.9
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.5, 0.6, 0.7, 0.8) 1
 
 [pixel shader todo(sm<4)]
@@ -115,13 +115,13 @@ void main(out float4 ret : sv_target)
 
 [test]
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.1, 0.2, 0.3, 0.4) 1
 uniform 0 float 0.5
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.5, 0.6, 0.7, 0.8) 1
 uniform 0 float 0.9
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.4, 0.5, 0.6, 0.7) 1
 
 [pixel shader todo(sm<4)]
@@ -160,23 +160,23 @@ void main(out float4 ret : sv_target)
 
 [test]
 uniform 0 float 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.1, 0.1, 0.1, 0.1) 1
 
 uniform 0 float 0.1
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.2, 0.2, 0.2, 0.2) 1
 
 uniform 0 float 0.3
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.4, 0.4, 0.4, 0.4) 1
 
 uniform 0 float 0.7
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.8, 0.8, 0.8, 0.8) 1
 
 uniform 0 float 0.9
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.9, 0.9, 0.9, 0.9) 1
 
 [pixel shader todo(sm<4)]
@@ -236,21 +236,21 @@ void main(out float4 ret : sv_target)
 uniform 0 float4 0.3 0.0 0.0 0.0
 uniform 4 float4 0.0 0.0 0.0 0.0
 uniform 8 float4 0.1 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.1, 0.1, 0.1, 0.1) 1
 
 uniform 4 float4 0.35 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.2, 0.2, 0.2, 0.2) 1
 
 uniform 8 float4 0.5 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.4, 0.4, 0.4, 0.4) 1
 
 uniform 0 float4 1.0 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.4, 0.4, 0.4, 0.4) 1
 
 uniform 4 float4 2.0 0.0 0.0 0.0
-todo(sm<4 | sm>=6) draw quad
+todo(sm<4) draw quad
 probe all rgba (0.9, 0.9, 0.9, 0.9) 1

tests/hlsl/sm6-ternary.shader_test

@@ -14,10 +14,10 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float4 2.0 3.0 4.0 5.0
-todo draw quad
+draw quad
 probe all rgba (2.0, 3.0, 4.0, 5.0)
 uniform 0 float4 0.0 10.0 11.0 12.0
-todo draw quad
+draw quad
 probe all rgba (-1.0, 9.0, 10.0, 11.0)
 
 

tests/hlsl/switch.shader_test

@@ -116,10 +116,10 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 2 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.1, 2.1, 3.1, 4.1)
 uniform 0 uint4 1 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 
 % floats are accepted
@@ -145,10 +145,10 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 2 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.1, 2.1, 3.1, 4.1)
 uniform 0 uint4 1 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 
 [pixel shader fail(sm>=6)]
@@ -173,10 +173,10 @@ float4 main() : sv_target
 
 [test]
 uniform 0 float4 2.0 0.0 0.0 0.0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.1, 2.1, 3.1, 4.1)
 uniform 0 float4 1.0 0.0 0.0 0.0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 
 [pixel shader fail]
@@ -374,13 +374,13 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 2 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.1, 2.1, 3.1, 4.1)
 uniform 0 uint4 1 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.2, 2.2, 3.2, 4.2)
 uniform 0 uint4 0 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 
 % switch breaks within a loop
@@ -412,7 +412,7 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 2 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (5.0, 6.0, 7.0, 8.0)
 
 % default case placement
@@ -443,13 +443,13 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 0 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (4.0, 5.0, 6.0, 7.0)
 uniform 0 uint4 2 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (2.0, 3.0, 4.0, 5.0)
 uniform 0 uint4 3 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (4.0, 5.0, 6.0, 7.0)
 
 [pixel shader]
@@ -480,13 +480,13 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 3 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 uniform 0 uint4 0 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (4.0, 5.0, 6.0, 7.0)
 uniform 0 uint4 5 0 0 0
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (1.0, 2.0, 3.0, 4.0)
 
 % 'continue' is not supported in switches
@@ -546,10 +546,10 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 0 0 3 1
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (10.0, 11.0, 12.0, 13.0)
 uniform 0 uint4 1 0 3 1
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (7.0, 8.0, 9.0, 10.0)
 
 % return from a switch nested in a loop
@@ -580,8 +580,8 @@ float4 main() : sv_target
 
 [test]
 uniform 0 uint4 0 0 3 1
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (304.0, 305.0, 306.0, 307.0)
 uniform 0 uint4 1 0 3 1
-todo(sm>=6) draw quad
+draw quad
 probe all rgba (3.0, 4.0, 5.0, 6.0)