[PATCH v6 0/9] MR281: vkd3d-shader/tpf: Replace sm4 register structs with vkd3d-shader register structs, part 1.

[Francisco Casas (＠fcasas)]

Recap:

I implemented relative addressing in !229, but I was suggested to handle register indexes by moving all the registers to the heap, which I did in my [nonconst-offsets-6](https://gitlab.winehq.org/fcasas/vkd3d/-/commits/nonconst-offsets-6) branch. A part of this implementation required !269, but I was asked to try to turn the sm4 register structs into the vkd3d-shader register structs instead, to get closer to a common low level IR.

This is the first part of that transformation. The whole thing is in my [use_vkd3d_reg_4](https://gitlab.winehq.org/fcasas/vkd3d/-/commits/use_vkd3d_reg_4) branch.

~~This is built on top of !225 (the first two patches), so it may be good to get that upstream first.~~

---

This patch series aims to do the following replacements:
```
struct sm4_register -> struct vkd3d_shader_register
struct sm4_dst_register -> struct vkd3d_shader_dst_param
struct sm4_src_register -> struct vkd3d_shader_src_param
```
to get us closer to a common level IR and simplify the implementation of relative-addressing.

To achieve this, the fields in the sm4 register structs are replaced with fields in the vkd3d-shader structs or removed altogether, one at the time.

As can be seen when looking at the whole branch, it is possible to do this transformation without having to add additional fields to `struct vkd3d_shader_register`, by restricting each register type to a single `enum vkd3d_sm4_dimension` (and its src registers to a single `enum vkd3d_sm4_swizzle_type`) by default.

The only exception we need so far is for sampler registers: They always have dimension NONE, except when used as arguments of gather instructions, in which case they have dimension VEC4 and SCALAR swizzle. This, and similar exceptions we may find in the future, can be handled using the opcode_info, as in 7/8 [81e17506](https://gitlab.winehq.org/fcasas/vkd3d/-/commit/81e17506ba2cb1fbf4d021159ab1...).

--
  v6: vkd3d-shader/tpf: Get rid of sm4_register.dim.
      vkd3d-shader/tpf: Use 's' in src_info to expect sampler register with vec4 dimension.
      vkd3d-shader/tpf: Separate dst register write function.
      vkd3d-shader/tpf: Separate src register write function.
      vkd3d-shader/tpf: Make register_type_table an array of structs with lookup tables.
      vkd3d-shader/tpf: Introduce struct tpf_writer to group sm4 write arguments.

https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/281

[PATCH v6 1/9] vkd3d-shader/tpf: Use enum vkd3d_shader_register_type in sm4_register.type.

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/hlsl.h |  2 +-
 libs/vkd3d-shader/tpf.c  | 83 ++++++++++++++++++++++++----------------
 2 files changed, 50 insertions(+), 35 deletions(-)

diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h
index ab91974a..668690bd 100644
--- a/libs/vkd3d-shader/hlsl.h
+++ b/libs/vkd3d-shader/hlsl.h
@@ -1249,7 +1249,7 @@ int hlsl_sm1_write(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_fun
 bool hlsl_sm4_usage_from_semantic(struct hlsl_ctx *ctx,
         const struct hlsl_semantic *semantic, bool output, D3D_NAME *usage);
 bool hlsl_sm4_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semantic *semantic,
-        bool output, unsigned int *type, enum vkd3d_sm4_swizzle_type *swizzle_type, bool *has_idx);
+        bool output, enum vkd3d_shader_register_type *type, enum vkd3d_sm4_swizzle_type *swizzle_type, bool *has_idx);
 int hlsl_sm4_write(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, struct vkd3d_shader_code *out);

int hlsl_lexer_compile(struct hlsl_ctx *ctx, const struct vkd3d_shader_code *hlsl);
diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 4be2f8b8..049cde43 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -2519,7 +2519,7 @@ static bool type_is_integer(const struct hlsl_type *type)
 }

bool hlsl_sm4_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semantic *semantic,
-        bool output, unsigned int *type, enum vkd3d_sm4_swizzle_type *swizzle_type, bool *has_idx)
+        bool output, enum vkd3d_shader_register_type *type, enum vkd3d_sm4_swizzle_type *swizzle_type, bool *has_idx)
 {
     unsigned int i;

@@ -2529,24 +2529,24 @@ bool hlsl_sm4_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_sem
         bool output;
         enum vkd3d_shader_type shader_type;
         enum vkd3d_sm4_swizzle_type swizzle_type;
-        enum vkd3d_sm4_register_type type;
+        enum vkd3d_shader_register_type type;
         bool has_idx;
     }
     register_table[] =
     {
-        {"sv_dispatchthreadid", false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM5_RT_THREAD_ID,         false},
-        {"sv_groupid",          false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM5_RT_THREAD_GROUP_ID,   false},
-        {"sv_groupthreadid",    false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM5_RT_LOCAL_THREAD_ID,   false},
+        {"sv_dispatchthreadid", false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_THREADID,          false},
+        {"sv_groupid",          false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_THREADGROUPID,     false},
+        {"sv_groupthreadid",    false, VKD3D_SHADER_TYPE_COMPUTE,   VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_LOCALTHREADID,     false},

-        {"sv_primitiveid",      false, VKD3D_SHADER_TYPE_GEOMETRY,  VKD3D_SM4_SWIZZLE_NONE, VKD3D_SM4_RT_PRIMID,            false},
+        {"sv_primitiveid",      false, VKD3D_SHADER_TYPE_GEOMETRY,  VKD3D_SM4_SWIZZLE_NONE, VKD3DSPR_PRIMID,            false},

/* Put sv_target in this table, instead of letting it fall through to
          * default varying allocation, so that the register index matches the
          * usage index. */
-        {"color",               true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM4_RT_OUTPUT,            true},
-        {"depth",               true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM4_RT_DEPTHOUT,          false},
-        {"sv_depth",            true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM4_RT_DEPTHOUT,          false},
-        {"sv_target",           true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3D_SM4_RT_OUTPUT,            true},
+        {"color",               true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_OUTPUT,            true},
+        {"depth",               true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_DEPTHOUT,          false},
+        {"sv_depth",            true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_DEPTHOUT,          false},
+        {"sv_target",           true,  VKD3D_SHADER_TYPE_PIXEL,     VKD3D_SM4_SWIZZLE_VEC4, VKD3DSPR_OUTPUT,            true},
     };

for (i = 0; i < ARRAY_SIZE(register_table); ++i)
@@ -2652,7 +2652,7 @@ static void write_sm4_signature(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc,
     LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
     {
         unsigned int width = (1u << var->data_type->dimx) - 1, use_mask;
-        enum vkd3d_sm4_register_type type;
+        enum vkd3d_shader_register_type type;
         uint32_t usage_idx, reg_idx;
         D3D_NAME usage;
         bool has_idx;
@@ -3431,7 +3431,7 @@ static uint32_t sm4_encode_instruction_modifier(const struct sm4_instruction_mod

struct sm4_register
 {
-    enum vkd3d_sm4_register_type type;
+    enum vkd3d_shader_register_type type;
     uint32_t idx[2];
     unsigned int idx_count;
     enum vkd3d_sm4_dimension dim;
@@ -3480,7 +3480,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

if (regset == HLSL_REGSET_TEXTURES)
         {
-            reg->type = VKD3D_SM4_RT_RESOURCE;
+            reg->type = VKD3DSPR_RESOURCE;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
@@ -3492,7 +3492,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         }
         else if (regset == HLSL_REGSET_UAVS)
         {
-            reg->type = VKD3D_SM5_RT_UAV;
+            reg->type = VKD3DSPR_UAV;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
@@ -3504,7 +3504,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         }
         else if (regset == HLSL_REGSET_SAMPLERS)
         {
-            reg->type = VKD3D_SM4_RT_SAMPLER;
+            reg->type = VKD3DSPR_SAMPLER;
             reg->dim = VKD3D_SM4_DIMENSION_NONE;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
@@ -3519,7 +3519,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             unsigned int offset = hlsl_offset_from_deref_safe(ctx, deref) + var->buffer_offset;

assert(data_type->class <= HLSL_CLASS_VECTOR);
-            reg->type = VKD3D_SM4_RT_CONSTBUFFER;
+            reg->type = VKD3DSPR_CONSTBUFFER;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
@@ -3551,7 +3551,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             struct hlsl_reg hlsl_reg = hlsl_reg_from_deref(ctx, deref);

assert(hlsl_reg.allocated);
-            reg->type = VKD3D_SM4_RT_INPUT;
+            reg->type = VKD3DSPR_INPUT;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
@@ -3574,7 +3574,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
                 reg->idx_count = 1;
             }

-            if (reg->type == VKD3D_SM4_RT_DEPTHOUT)
+            if (reg->type == VKD3DSPR_DEPTHOUT)
                 reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
             else
                 reg->dim = VKD3D_SM4_DIMENSION_VEC4;
@@ -3585,7 +3585,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             struct hlsl_reg hlsl_reg = hlsl_reg_from_deref(ctx, deref);

assert(hlsl_reg.allocated);
-            reg->type = VKD3D_SM4_RT_OUTPUT;
+            reg->type = VKD3DSPR_OUTPUT;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             reg->idx[0] = hlsl_reg.id;
             reg->idx_count = 1;
@@ -3597,7 +3597,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         struct hlsl_reg hlsl_reg = hlsl_reg_from_deref(ctx, deref);

assert(hlsl_reg.allocated);
-        reg->type = VKD3D_SM4_RT_TEMP;
+        reg->type = VKD3DSPR_TEMP;
         reg->dim = VKD3D_SM4_DIMENSION_VEC4;
         if (swizzle_type)
             *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
@@ -3621,7 +3621,7 @@ static void sm4_register_from_node(struct sm4_register *reg, unsigned int *write
         enum vkd3d_sm4_swizzle_type *swizzle_type, const struct hlsl_ir_node *instr)
 {
     assert(instr->reg.allocated);
-    reg->type = VKD3D_SM4_RT_TEMP;
+    reg->type = VKD3DSPR_TEMP;
     reg->dim = VKD3D_SM4_DIMENSION_VEC4;
     *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
     reg->idx[0] = instr->reg.id;
@@ -3640,7 +3640,7 @@ static void sm4_src_from_constant_value(struct sm4_src_register *src,
         const struct hlsl_constant_value *value, unsigned int width, unsigned int map_writemask)
 {
     src->swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
-    src->reg.type = VKD3D_SM4_RT_IMMCONST;
+    src->reg.type = VKD3DSPR_IMMCONST;
     if (width == 1)
     {
         src->reg.dim = VKD3D_SM4_DIMENSION_SCALAR;
@@ -3679,7 +3679,22 @@ static void sm4_src_from_node(struct sm4_src_register *src,

static uint32_t sm4_encode_register(const struct sm4_register *reg)
 {
-    return (reg->type << VKD3D_SM4_REGISTER_TYPE_SHIFT)
+    uint32_t sm4_reg_type = ~0u;
+    unsigned int i;
+
+    /* Find sm4 register type from vkd3d_shader_register_type. */
+    for (i = 0; i < ARRAY_SIZE(register_type_table); ++i)
+    {
+        if (reg->type == register_type_table[i])
+        {
+            if (sm4_reg_type != ~0u)
+                ERR("Multiple maps for register_type %#x.\n", reg->type);
+            sm4_reg_type = i;
+        }
+    }
+    assert(sm4_reg_type != ~0u);
+
+    return (sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT)
             | (reg->idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT)
             | (reg->dim << VKD3D_SM4_DIMENSION_SHIFT);
 }
@@ -3687,7 +3702,7 @@ static uint32_t sm4_encode_register(const struct sm4_register *reg)
 static uint32_t sm4_register_order(const struct sm4_register *reg)
 {
     uint32_t order = 1;
-    if (reg->type == VKD3D_SM4_RT_IMMCONST)
+    if (reg->type == VKD3DSPR_IMMCONST)
         order += reg->dim == VKD3D_SM4_DIMENSION_VEC4 ? 4 : 1;
     order += reg->idx_count;
     if (reg->mod)
@@ -3750,7 +3765,7 @@ static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const st
         for (j = 0; j < instr->srcs[i].reg.idx_count; ++j)
             put_u32(buffer, instr->srcs[i].reg.idx[j]);

-        if (instr->srcs[i].reg.type == VKD3D_SM4_RT_IMMCONST)
+        if (instr->srcs[i].reg.type == VKD3DSPR_IMMCONST)
         {
             put_u32(buffer, instr->srcs[i].reg.immconst_uint[0]);
             if (instr->srcs[i].reg.dim == VKD3D_SM4_DIMENSION_VEC4)
@@ -3803,7 +3818,7 @@ static void write_sm4_dcl_constant_buffer(struct vkd3d_bytecode_buffer *buffer,
         .opcode = VKD3D_SM4_OP_DCL_CONSTANT_BUFFER,

.srcs[0].reg.dim = VKD3D_SM4_DIMENSION_VEC4,
-        .srcs[0].reg.type = VKD3D_SM4_RT_CONSTBUFFER,
+        .srcs[0].reg.type = VKD3DSPR_CONSTBUFFER,
         .srcs[0].reg.idx = {cbuffer->reg.id, (cbuffer->used_size + 3) / 4},
         .srcs[0].reg.idx_count = 2,
         .srcs[0].swizzle_type = VKD3D_SM4_SWIZZLE_VEC4,
@@ -3822,7 +3837,7 @@ static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
     {
         .opcode = VKD3D_SM4_OP_DCL_SAMPLER,

-        .dsts[0].reg.type = VKD3D_SM4_RT_SAMPLER,
+        .dsts[0].reg.type = VKD3DSPR_SAMPLER,
         .dsts[0].reg.idx_count = 1,
         .dst_count = 1,
     };
@@ -3863,7 +3878,7 @@ static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b

instr = (struct sm4_instruction)
         {
-            .dsts[0].reg.type = uav ? VKD3D_SM5_RT_UAV : VKD3D_SM4_RT_RESOURCE,
+            .dsts[0].reg.type = uav ? VKD3DSPR_UAV : VKD3DSPR_RESOURCE,
             .dsts[0].reg.idx = {resource->id + i},
             .dsts[0].reg.idx_count = 1,
             .dst_count = 1,
@@ -3929,13 +3944,13 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
     }
     else
     {
-        instr.dsts[0].reg.type = output ? VKD3D_SM4_RT_OUTPUT : VKD3D_SM4_RT_INPUT;
+        instr.dsts[0].reg.type = output ? VKD3DSPR_OUTPUT : VKD3DSPR_INPUT;
         instr.dsts[0].reg.idx[0] = var->regs[HLSL_REGSET_NUMERIC].id;
         instr.dsts[0].reg.idx_count = 1;
         instr.dsts[0].writemask = var->regs[HLSL_REGSET_NUMERIC].writemask;
     }

-    if (instr.dsts[0].reg.type == VKD3D_SM4_RT_DEPTHOUT)
+    if (instr.dsts[0].reg.type == VKD3DSPR_DEPTHOUT)
         instr.dsts[0].reg.dim = VKD3D_SM4_DIMENSION_SCALAR;

hlsl_sm4_usage_from_semantic(ctx, &var->semantic, output, &usage);
@@ -4067,7 +4082,7 @@ static void write_sm4_unary_op_with_two_destinations(struct vkd3d_bytecode_buffe
     assert(dst_idx < ARRAY_SIZE(instr.dsts));
     sm4_dst_from_node(&instr.dsts[dst_idx], dst);
     assert(1 - dst_idx >= 0);
-    instr.dsts[1 - dst_idx].reg.type = VKD3D_SM4_RT_NULL;
+    instr.dsts[1 - dst_idx].reg.type = VKD3DSPR_NULL;
     instr.dsts[1 - dst_idx].reg.dim = VKD3D_SM4_DIMENSION_NONE;
     instr.dsts[1 - dst_idx].reg.idx_count = 0;
     instr.dst_count = 2;
@@ -4127,7 +4142,7 @@ static void write_sm4_binary_op_with_two_destinations(struct vkd3d_bytecode_buff
     assert(dst_idx < ARRAY_SIZE(instr.dsts));
     sm4_dst_from_node(&instr.dsts[dst_idx], dst);
     assert(1 - dst_idx >= 0);
-    instr.dsts[1 - dst_idx].reg.type = VKD3D_SM4_RT_NULL;
+    instr.dsts[1 - dst_idx].reg.type = VKD3DSPR_NULL;
     instr.dsts[1 - dst_idx].reg.dim = VKD3D_SM4_DIMENSION_NONE;
     instr.dsts[1 - dst_idx].reg.idx_count = 0;
     instr.dst_count = 2;
@@ -4199,7 +4214,7 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf

memset(&instr.srcs[2], 0, sizeof(instr.srcs[2]));
             instr.srcs[2].swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
-            reg->type = VKD3D_SM4_RT_IMMCONST;
+            reg->type = VKD3DSPR_IMMCONST;
             reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
             reg->immconst_uint[0] = index->value.u[0].u;
         }
@@ -4318,7 +4333,7 @@ static void write_sm4_cast_from_bool(struct hlsl_ctx *ctx,

sm4_src_from_node(&instr.srcs[0], arg, instr.dsts[0].writemask);
     instr.srcs[1].swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
-    instr.srcs[1].reg.type = VKD3D_SM4_RT_IMMCONST;
+    instr.srcs[1].reg.type = VKD3DSPR_IMMCONST;
     instr.srcs[1].reg.dim = VKD3D_SM4_DIMENSION_SCALAR;
     instr.srcs[1].reg.immconst_uint[0] = mask;
     instr.src_count = 2;

[PATCH v6 2/9] vkd3d-shader/tpf: Allow passing NULL register type on hlsl_sm4_register_from_semantic().

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 049cde43..b9bafc08 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -2555,7 +2555,8 @@ bool hlsl_sm4_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_sem
                 && output == register_table[i].output
                 && ctx->profile->type == register_table[i].shader_type)
         {
-            *type = register_table[i].type;
+            if (type)
+                *type = register_table[i].type;
             if (swizzle_type)
                 *swizzle_type = register_table[i].swizzle_type;
             *has_idx = register_table[i].has_idx;
@@ -2652,7 +2653,6 @@ static void write_sm4_signature(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc,
     LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
     {
         unsigned int width = (1u << var->data_type->dimx) - 1, use_mask;
-        enum vkd3d_shader_register_type type;
         uint32_t usage_idx, reg_idx;
         D3D_NAME usage;
         bool has_idx;
@@ -2666,14 +2666,13 @@ static void write_sm4_signature(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc,
             continue;
         usage_idx = var->semantic.index;

-        if (hlsl_sm4_register_from_semantic(ctx, &var->semantic, output, &type, NULL, &has_idx))
+        if (hlsl_sm4_register_from_semantic(ctx, &var->semantic, output, NULL, NULL, &has_idx))
         {
             reg_idx = has_idx ? var->semantic.index : ~0u;
         }
         else
         {
             assert(var->regs[HLSL_REGSET_NUMERIC].allocated);
-            type = VKD3D_SM4_RT_INPUT;
             reg_idx = var->regs[HLSL_REGSET_NUMERIC].id;
         }

[PATCH v6 3/9] vkd3d-shader/tpf: Use struct vkd3d_shader_register_index in sm4_register.idx[].

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c | 55 ++++++++++++++++++++++++-----------------
 1 file changed, 32 insertions(+), 23 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index b9bafc08..f3adefd2 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -3431,7 +3431,7 @@ static uint32_t sm4_encode_instruction_modifier(const struct sm4_instruction_mod
 struct sm4_register
 {
     enum vkd3d_shader_register_type type;
-    uint32_t idx[2];
+    struct vkd3d_shader_register_index idx[2];
     unsigned int idx_count;
     enum vkd3d_sm4_dimension dim;
     uint32_t immconst_uint[4];
@@ -3483,8 +3483,8 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-            reg->idx[0] = var->regs[HLSL_REGSET_TEXTURES].id;
-            reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
+            reg->idx[0].offset = var->regs[HLSL_REGSET_TEXTURES].id;
+            reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
             assert(regset == HLSL_REGSET_TEXTURES);
             reg->idx_count = 1;
             *writemask = VKD3DSP_WRITEMASK_ALL;
@@ -3495,8 +3495,8 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-            reg->idx[0] = var->regs[HLSL_REGSET_UAVS].id;
-            reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
+            reg->idx[0].offset = var->regs[HLSL_REGSET_UAVS].id;
+            reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
             assert(regset == HLSL_REGSET_UAVS);
             reg->idx_count = 1;
             *writemask = VKD3DSP_WRITEMASK_ALL;
@@ -3507,8 +3507,8 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             reg->dim = VKD3D_SM4_DIMENSION_NONE;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
-            reg->idx[0] = var->regs[HLSL_REGSET_SAMPLERS].id;
-            reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
+            reg->idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].id;
+            reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
             assert(regset == HLSL_REGSET_SAMPLERS);
             reg->idx_count = 1;
             *writemask = VKD3DSP_WRITEMASK_ALL;
@@ -3522,8 +3522,8 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-            reg->idx[0] = var->buffer->reg.id;
-            reg->idx[1] = offset / 4;
+            reg->idx[0].offset = var->buffer->reg.id;
+            reg->idx[1].offset = offset / 4;
             reg->idx_count = 2;
             *writemask = ((1u << data_type->dimx) - 1) << (offset & 3);
         }
@@ -3538,7 +3538,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

if (has_idx)
             {
-                reg->idx[0] = var->semantic.index + offset / 4;
+                reg->idx[0].offset = var->semantic.index + offset / 4;
                 reg->idx_count = 1;
             }

@@ -3554,7 +3554,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-            reg->idx[0] = hlsl_reg.id;
+            reg->idx[0].offset = hlsl_reg.id;
             reg->idx_count = 1;
             *writemask = hlsl_reg.writemask;
         }
@@ -3569,7 +3569,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

if (has_idx)
             {
-                reg->idx[0] = var->semantic.index + offset / 4;
+                reg->idx[0].offset = var->semantic.index + offset / 4;
                 reg->idx_count = 1;
             }

@@ -3586,7 +3586,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
             assert(hlsl_reg.allocated);
             reg->type = VKD3DSPR_OUTPUT;
             reg->dim = VKD3D_SM4_DIMENSION_VEC4;
-            reg->idx[0] = hlsl_reg.id;
+            reg->idx[0].offset = hlsl_reg.id;
             reg->idx_count = 1;
             *writemask = hlsl_reg.writemask;
         }
@@ -3600,7 +3600,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         reg->dim = VKD3D_SM4_DIMENSION_VEC4;
         if (swizzle_type)
             *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-        reg->idx[0] = hlsl_reg.id;
+        reg->idx[0].offset = hlsl_reg.id;
         reg->idx_count = 1;
         *writemask = hlsl_reg.writemask;
     }
@@ -3623,7 +3623,7 @@ static void sm4_register_from_node(struct sm4_register *reg, unsigned int *write
     reg->type = VKD3DSPR_TEMP;
     reg->dim = VKD3D_SM4_DIMENSION_VEC4;
     *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
-    reg->idx[0] = instr->reg.id;
+    reg->idx[0].offset = instr->reg.id;
     reg->idx_count = 1;
     *writemask = instr->reg.writemask;
 }
@@ -3745,7 +3745,10 @@ static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const st
         put_u32(buffer, token);

for (j = 0; j < instr->dsts[i].reg.idx_count; ++j)
-            put_u32(buffer, instr->dsts[i].reg.idx[j]);
+        {
+            put_u32(buffer, instr->dsts[i].reg.idx[j].offset);
+            assert(!instr->dsts[i].reg.idx[j].rel_addr);
+        }
     }

for (i = 0; i < instr->src_count; ++i)
@@ -3762,7 +3765,10 @@ static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const st
                     | VKD3D_SM4_EXTENDED_OPERAND_MODIFIER);

for (j = 0; j < instr->srcs[i].reg.idx_count; ++j)
-            put_u32(buffer, instr->srcs[i].reg.idx[j]);
+        {
+            put_u32(buffer, instr->srcs[i].reg.idx[j].offset);
+            assert(!instr->srcs[i].reg.idx[j].rel_addr);
+        }

if (instr->srcs[i].reg.type == VKD3DSPR_IMMCONST)
         {
@@ -3818,7 +3824,8 @@ static void write_sm4_dcl_constant_buffer(struct vkd3d_bytecode_buffer *buffer,

.srcs[0].reg.dim = VKD3D_SM4_DIMENSION_VEC4,
         .srcs[0].reg.type = VKD3DSPR_CONSTBUFFER,
-        .srcs[0].reg.idx = {cbuffer->reg.id, (cbuffer->used_size + 3) / 4},
+        .srcs[0].reg.idx[0].offset = cbuffer->reg.id,
+        .srcs[0].reg.idx[1].offset = (cbuffer->used_size + 3) / 4,
         .srcs[0].reg.idx_count = 2,
         .srcs[0].swizzle_type = VKD3D_SM4_SWIZZLE_VEC4,
         .srcs[0].swizzle = HLSL_SWIZZLE(X, Y, Z, W),
@@ -3853,7 +3860,7 @@ static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
         if (resource->var && !resource->var->objects_usage[HLSL_REGSET_SAMPLERS][i].used)
             continue;

-        instr.dsts[0].reg.idx[0] = resource->id + i;
+        instr.dsts[0].reg.idx[0].offset = resource->id + i;
         write_sm4_instruction(buffer, &instr);
     }
 }
@@ -3878,7 +3885,7 @@ static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
         instr = (struct sm4_instruction)
         {
             .dsts[0].reg.type = uav ? VKD3DSPR_UAV : VKD3DSPR_RESOURCE,
-            .dsts[0].reg.idx = {resource->id + i},
+            .dsts[0].reg.idx[0].offset = resource->id + i,
             .dsts[0].reg.idx_count = 1,
             .dst_count = 1,

@@ -3932,7 +3939,7 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
     {
         if (has_idx)
         {
-            instr.dsts[0].reg.idx[0] = var->semantic.index;
+            instr.dsts[0].reg.idx[0].offset = var->semantic.index;
             instr.dsts[0].reg.idx_count = 1;
         }
         else
@@ -3944,7 +3951,7 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
     else
     {
         instr.dsts[0].reg.type = output ? VKD3DSPR_OUTPUT : VKD3DSPR_INPUT;
-        instr.dsts[0].reg.idx[0] = var->regs[HLSL_REGSET_NUMERIC].id;
+        instr.dsts[0].reg.idx[0].offset = var->regs[HLSL_REGSET_NUMERIC].id;
         instr.dsts[0].reg.idx_count = 1;
         instr.dsts[0].writemask = var->regs[HLSL_REGSET_NUMERIC].writemask;
     }
@@ -4034,7 +4041,9 @@ static void write_sm4_dcl_thread_group(struct vkd3d_bytecode_buffer *buffer, con
     {
         .opcode = VKD3D_SM5_OP_DCL_THREAD_GROUP,

-        .idx = {thread_count[0], thread_count[1], thread_count[2]},
+        .idx[0] = thread_count[0],
+        .idx[1] = thread_count[1],
+        .idx[2] = thread_count[2],
         .idx_count = 3,
     };

[PATCH v6 4/9] vkd3d-shader/tpf: Introduce struct tpf_writer to group sm4 write arguments.

[Francisco Casas]

From: Francisco Casas

We will add register information lookup tables on this struct later.
They will be used by sm4_encode_register(), and thus, they will be
required by write_sm4_instruction().
---
 libs/vkd3d-shader/tpf.c | 464 ++++++++++++++++++++--------------------
 1 file changed, 233 insertions(+), 231 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index f3adefd2..f83029b2 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -1524,6 +1524,12 @@ static const enum vkd3d_shader_register_precision register_precision_table[] =
     /* VKD3D_SM4_REGISTER_PRECISION_MIN_UINT_16 */  VKD3D_SHADER_REGISTER_PRECISION_MIN_UINT_16,
 };

+struct tpf_writer
+{
+    struct hlsl_ctx *ctx;
+    struct vkd3d_bytecode_buffer *buffer;
+};
+
 static const struct vkd3d_sm4_opcode_info *get_opcode_info(enum vkd3d_sm4_opcode opcode)
 {
     unsigned int i;
@@ -1536,6 +1542,12 @@ static const struct vkd3d_sm4_opcode_info *get_opcode_info(enum vkd3d_sm4_opcode
     return NULL;
 }

+static void tpf_writer_init(struct tpf_writer *tpf, struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer)
+{
+    tpf->ctx = ctx;
+    tpf->buffer = buffer;
+}
+
 static void map_register(const struct vkd3d_shader_sm4_parser *sm4, struct vkd3d_shader_register *reg)
 {
     switch (sm4->p.shader_version.type)
@@ -2502,7 +2514,7 @@ int vkd3d_shader_sm4_parser_create(const struct vkd3d_shader_compile_info *compi
     return sm4->p.failed ? VKD3D_ERROR_INVALID_SHADER : VKD3D_OK;
 }

-static void write_sm4_block(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_block *block);
+static void write_sm4_block(const struct tpf_writer *tpf, const struct hlsl_block *block);

static bool type_is_integer(const struct hlsl_type *type)
 {
@@ -3676,7 +3688,7 @@ static void sm4_src_from_node(struct sm4_src_register *src,
         src->swizzle = hlsl_map_swizzle(hlsl_swizzle_from_writemask(writemask), map_writemask);
 }

-static uint32_t sm4_encode_register(const struct sm4_register *reg)
+static uint32_t sm4_encode_register(const struct tpf_writer *tpf, const struct sm4_register *reg)
 {
     uint32_t sm4_reg_type = ~0u;
     unsigned int i;
@@ -3709,8 +3721,9 @@ static uint32_t sm4_register_order(const struct sm4_register *reg)
     return order;
 }

-static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const struct sm4_instruction *instr)
+static void write_sm4_instruction(const struct tpf_writer *tpf, const struct sm4_instruction *instr)
 {
+    struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
     uint32_t token = instr->opcode;
     unsigned int size = 1, i, j;

@@ -3739,7 +3752,7 @@ static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const st

for (i = 0; i < instr->dst_count; ++i)
     {
-        token = sm4_encode_register(&instr->dsts[i].reg);
+        token = sm4_encode_register(tpf, &instr->dsts[i].reg);
         if (instr->dsts[i].reg.dim == VKD3D_SM4_DIMENSION_VEC4)
             token |= instr->dsts[i].writemask << VKD3D_SM4_WRITEMASK_SHIFT;
         put_u32(buffer, token);
@@ -3753,7 +3766,7 @@ static void write_sm4_instruction(struct vkd3d_bytecode_buffer *buffer, const st

for (i = 0; i < instr->src_count; ++i)
     {
-        token = sm4_encode_register(&instr->srcs[i].reg);
+        token = sm4_encode_register(tpf, &instr->srcs[i].reg);
         token |= (uint32_t)instr->srcs[i].swizzle_type << VKD3D_SM4_SWIZZLE_TYPE_SHIFT;
         token |= instr->srcs[i].swizzle << VKD3D_SM4_SWIZZLE_SHIFT;
         if (instr->srcs[i].reg.mod)
@@ -3816,7 +3829,7 @@ static bool encode_texel_offset_as_aoffimmi(struct sm4_instruction *instr,
     return true;
 }

-static void write_sm4_dcl_constant_buffer(struct vkd3d_bytecode_buffer *buffer, const struct hlsl_buffer *cbuffer)
+static void write_sm4_dcl_constant_buffer(const struct tpf_writer *tpf, const struct hlsl_buffer *cbuffer)
 {
     const struct sm4_instruction instr =
     {
@@ -3831,11 +3844,10 @@ static void write_sm4_dcl_constant_buffer(struct vkd3d_bytecode_buffer *buffer,
         .srcs[0].swizzle = HLSL_SWIZZLE(X, Y, Z, W),
         .src_count = 1,
     };
-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct extern_resource *resource)
+static void write_sm4_dcl_samplers(const struct tpf_writer *tpf, const struct extern_resource *resource)
 {
     struct hlsl_type *component_type;
     unsigned int i;
@@ -3848,7 +3860,7 @@ static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
         .dst_count = 1,
     };

-    component_type = hlsl_type_get_component_type(ctx, resource->data_type, 0);
+    component_type = hlsl_type_get_component_type(tpf->ctx, resource->data_type, 0);

if (component_type->sampler_dim == HLSL_SAMPLER_DIM_COMPARISON)
         instr.opcode |= VKD3D_SM4_SAMPLER_COMPARISON << VKD3D_SM4_SAMPLER_MODE_SHIFT;
@@ -3861,12 +3873,12 @@ static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
             continue;

instr.dsts[0].reg.idx[0].offset = resource->id + i;
-        write_sm4_instruction(buffer, &instr);
+        write_sm4_instruction(tpf, &instr);
     }
 }

-static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct extern_resource *resource, bool uav)
+static void write_sm4_dcl_textures(const struct tpf_writer *tpf, const struct extern_resource *resource,
+        bool uav)
 {
     enum hlsl_regset regset = uav ? HLSL_REGSET_UAVS : HLSL_REGSET_TEXTURES;
     struct hlsl_type *component_type;
@@ -3875,7 +3887,7 @@ static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b

assert(resource->regset == regset);

-    component_type = hlsl_type_get_component_type(ctx, resource->data_type, 0);
+    component_type = hlsl_type_get_component_type(tpf->ctx, resource->data_type, 0);

for (i = 0; i < resource->bind_count; ++i)
     {
@@ -3918,13 +3930,13 @@ static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
             instr.opcode |= component_type->sample_count << VKD3D_SM4_RESOURCE_SAMPLE_COUNT_SHIFT;
         }

-        write_sm4_instruction(buffer, &instr);
+        write_sm4_instruction(tpf, &instr);
     }
 }

-static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_var *var)
+static void write_sm4_dcl_semantic(const struct tpf_writer *tpf, const struct hlsl_ir_var *var)
 {
-    const struct hlsl_profile_info *profile = ctx->profile;
+    const struct hlsl_profile_info *profile = tpf->ctx->profile;
     const bool output = var->is_output_semantic;
     D3D_NAME usage;
     bool has_idx;
@@ -3935,7 +3947,7 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
         .dst_count = 1,
     };

-    if (hlsl_sm4_register_from_semantic(ctx, &var->semantic, output, &instr.dsts[0].reg.type, NULL, &has_idx))
+    if (hlsl_sm4_register_from_semantic(tpf->ctx, &var->semantic, output, &instr.dsts[0].reg.type, NULL, &has_idx))
     {
         if (has_idx)
         {
@@ -3959,7 +3971,7 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
     if (instr.dsts[0].reg.type == VKD3DSPR_DEPTHOUT)
         instr.dsts[0].reg.dim = VKD3D_SM4_DIMENSION_SCALAR;

-    hlsl_sm4_usage_from_semantic(ctx, &var->semantic, output, &usage);
+    hlsl_sm4_usage_from_semantic(tpf->ctx, &var->semantic, output, &usage);
     if (usage == ~0u)
         usage = D3D_NAME_UNDEFINED;

@@ -4019,10 +4031,10 @@ static void write_sm4_dcl_semantic(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
             break;
     }

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_dcl_temps(struct vkd3d_bytecode_buffer *buffer, uint32_t temp_count)
+static void write_sm4_dcl_temps(const struct tpf_writer *tpf, uint32_t temp_count)
 {
     struct sm4_instruction instr =
     {
@@ -4032,10 +4044,10 @@ static void write_sm4_dcl_temps(struct vkd3d_bytecode_buffer *buffer, uint32_t t
         .idx_count = 1,
     };

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_dcl_thread_group(struct vkd3d_bytecode_buffer *buffer, const uint32_t thread_count[3])
+static void write_sm4_dcl_thread_group(const struct tpf_writer *tpf, const uint32_t thread_count[3])
 {
     struct sm4_instruction instr =
     {
@@ -4047,20 +4059,20 @@ static void write_sm4_dcl_thread_group(struct vkd3d_bytecode_buffer *buffer, con
         .idx_count = 3,
     };

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_ret(struct vkd3d_bytecode_buffer *buffer)
+static void write_sm4_ret(const struct tpf_writer *tpf)
 {
     struct sm4_instruction instr =
     {
         .opcode = VKD3D_SM4_OP_RET,
     };

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_unary_op(struct vkd3d_bytecode_buffer *buffer, enum vkd3d_sm4_opcode opcode,
+static void write_sm4_unary_op(const struct tpf_writer *tpf, enum vkd3d_sm4_opcode opcode,
         const struct hlsl_ir_node *dst, const struct hlsl_ir_node *src, unsigned int src_mod)
 {
     struct sm4_instruction instr;
@@ -4075,12 +4087,11 @@ static void write_sm4_unary_op(struct vkd3d_bytecode_buffer *buffer, enum vkd3d_
     instr.srcs[0].reg.mod = src_mod;
     instr.src_count = 1;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_unary_op_with_two_destinations(struct vkd3d_bytecode_buffer *buffer,
-        enum vkd3d_sm4_opcode opcode, const struct hlsl_ir_node *dst, unsigned dst_idx,
-        const struct hlsl_ir_node *src)
+static void write_sm4_unary_op_with_two_destinations(const struct tpf_writer *tpf, enum vkd3d_sm4_opcode opcode,
+        const struct hlsl_ir_node *dst, unsigned dst_idx, const struct hlsl_ir_node *src)
 {
     struct sm4_instruction instr;

@@ -4098,10 +4109,10 @@ static void write_sm4_unary_op_with_two_destinations(struct vkd3d_bytecode_buffe
     sm4_src_from_node(&instr.srcs[0], src, instr.dsts[dst_idx].writemask);
     instr.src_count = 1;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_binary_op(struct vkd3d_bytecode_buffer *buffer, enum vkd3d_sm4_opcode opcode,
+static void write_sm4_binary_op(const struct tpf_writer *tpf, enum vkd3d_sm4_opcode opcode,
         const struct hlsl_ir_node *dst, const struct hlsl_ir_node *src1, const struct hlsl_ir_node *src2)
 {
     struct sm4_instruction instr;
@@ -4116,11 +4127,11 @@ static void write_sm4_binary_op(struct vkd3d_bytecode_buffer *buffer, enum vkd3d
     sm4_src_from_node(&instr.srcs[1], src2, instr.dsts[0].writemask);
     instr.src_count = 2;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

/* dp# instructions don't map the swizzle. */
-static void write_sm4_binary_op_dot(struct vkd3d_bytecode_buffer *buffer, enum vkd3d_sm4_opcode opcode,
+static void write_sm4_binary_op_dot(const struct tpf_writer *tpf, enum vkd3d_sm4_opcode opcode,
         const struct hlsl_ir_node *dst, const struct hlsl_ir_node *src1, const struct hlsl_ir_node *src2)
 {
     struct sm4_instruction instr;
@@ -4135,10 +4146,10 @@ static void write_sm4_binary_op_dot(struct vkd3d_bytecode_buffer *buffer, enum v
     sm4_src_from_node(&instr.srcs[1], src2, VKD3DSP_WRITEMASK_ALL);
     instr.src_count = 2;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_binary_op_with_two_destinations(struct vkd3d_bytecode_buffer *buffer,
+static void write_sm4_binary_op_with_two_destinations(const struct tpf_writer *tpf,
         enum vkd3d_sm4_opcode opcode, const struct hlsl_ir_node *dst, unsigned dst_idx,
         const struct hlsl_ir_node *src1, const struct hlsl_ir_node *src2)
 {
@@ -4159,15 +4170,15 @@ static void write_sm4_binary_op_with_two_destinations(struct vkd3d_bytecode_buff
     sm4_src_from_node(&instr.srcs[1], src2, instr.dsts[dst_idx].writemask);
     instr.src_count = 2;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct hlsl_ir_node *dst, const struct hlsl_deref *resource,
-        const struct hlsl_ir_node *coords, const struct hlsl_ir_node *sample_index,
-        const struct hlsl_ir_node *texel_offset, enum hlsl_sampler_dim dim)
+static void write_sm4_ld(const struct tpf_writer *tpf, const struct hlsl_ir_node *dst,
+        const struct hlsl_deref *resource, const struct hlsl_ir_node *coords,
+        const struct hlsl_ir_node *sample_index, const struct hlsl_ir_node *texel_offset,
+        enum hlsl_sampler_dim dim)
 {
-    const struct hlsl_type *resource_type = hlsl_deref_get_type(ctx, resource);
+    const struct hlsl_type *resource_type = hlsl_deref_get_type(tpf->ctx, resource);
     bool multisampled = resource_type->base_type == HLSL_TYPE_TEXTURE
             && (resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMS || resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMSARRAY);
     bool uav = (hlsl_type_get_regset(resource_type) == HLSL_REGSET_UAVS);
@@ -4184,7 +4195,7 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
     {
         if (!encode_texel_offset_as_aoffimmi(&instr, texel_offset))
         {
-            hlsl_error(ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
+            hlsl_error(tpf->ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
                     "Offset must resolve to integer literal in the range -8 to 7.");
             return;
         }
@@ -4207,7 +4218,7 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf

sm4_src_from_node(&instr.srcs[0], coords, coords_writemask);

-    sm4_src_from_deref(ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);
+    sm4_src_from_deref(tpf->ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);

instr.src_count = 2;

@@ -4226,9 +4237,9 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
             reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
             reg->immconst_uint[0] = index->value.u[0].u;
         }
-        else if (ctx->profile->major_version == 4 && ctx->profile->minor_version == 0)
+        else if (tpf->ctx->profile->major_version == 4 && tpf->ctx->profile->minor_version == 0)
         {
-            hlsl_error(ctx, &sample_index->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE, "Expected literal sample index.");
+            hlsl_error(tpf->ctx, &sample_index->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE, "Expected literal sample index.");
         }
         else
         {
@@ -4238,11 +4249,10 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
         ++instr.src_count;
     }

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct hlsl_ir_resource_load *load)
+static void write_sm4_sample(const struct tpf_writer *tpf, const struct hlsl_ir_resource_load *load)
 {
     const struct hlsl_ir_node *texel_offset = load->texel_offset.node;
     const struct hlsl_ir_node *coords = load->coords.node;
@@ -4286,7 +4296,7 @@ static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
     {
         if (!encode_texel_offset_as_aoffimmi(&instr, texel_offset))
         {
-            hlsl_error(ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
+            hlsl_error(tpf->ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
                     "Offset must resolve to integer literal in the range -8 to 7.");
             return;
         }
@@ -4296,8 +4306,8 @@ static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
     instr.dst_count = 1;

sm4_src_from_node(&instr.srcs[0], coords, VKD3DSP_WRITEMASK_ALL);
-    sm4_src_from_deref(ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);
-    sm4_src_from_deref(ctx, &instr.srcs[2], sampler, VKD3DSP_WRITEMASK_ALL);
+    sm4_src_from_deref(tpf->ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);
+    sm4_src_from_deref(tpf->ctx, &instr.srcs[2], sampler, VKD3DSP_WRITEMASK_ALL);
     instr.src_count = 3;

if (load->load_type == HLSL_RESOURCE_SAMPLE_LOD
@@ -4319,7 +4329,7 @@ static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
         ++instr.src_count;
     }

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

static bool type_is_float(const struct hlsl_type *type)
@@ -4327,8 +4337,7 @@ static bool type_is_float(const struct hlsl_type *type)
     return type->base_type == HLSL_TYPE_FLOAT || type->base_type == HLSL_TYPE_HALF;
 }

-static void write_sm4_cast_from_bool(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_expr *expr,
+static void write_sm4_cast_from_bool(const struct tpf_writer *tpf, const struct hlsl_ir_expr *expr,
         const struct hlsl_ir_node *arg, uint32_t mask)
 {
     struct sm4_instruction instr;
@@ -4346,11 +4355,10 @@ static void write_sm4_cast_from_bool(struct hlsl_ctx *ctx,
     instr.srcs[1].reg.immconst_uint[0] = mask;
     instr.src_count = 2;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_cast(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_expr *expr)
+static void write_sm4_cast(const struct tpf_writer *tpf, const struct hlsl_ir_expr *expr)
 {
     static const union
     {
@@ -4372,23 +4380,23 @@ static void write_sm4_cast(struct hlsl_ctx *ctx,
             {
                 case HLSL_TYPE_HALF:
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_INT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_ITOF, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_ITOF, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_UINT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_UTOF, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_UTOF, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_BOOL:
-                    write_sm4_cast_from_bool(ctx, buffer, expr, arg1, one.u);
+                    write_sm4_cast_from_bool(tpf, expr, arg1, one.u);
                     break;

case HLSL_TYPE_DOUBLE:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 cast from double to float.");
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 cast from double to float.");
                     break;

default:
@@ -4401,20 +4409,20 @@ static void write_sm4_cast(struct hlsl_ctx *ctx,
             {
                 case HLSL_TYPE_HALF:
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_FTOI, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_FTOI, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_BOOL:
-                    write_sm4_cast_from_bool(ctx, buffer, expr, arg1, 1);
+                    write_sm4_cast_from_bool(tpf, expr, arg1, 1);
                     break;

case HLSL_TYPE_DOUBLE:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 cast from double to int.");
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 cast from double to int.");
                     break;

default:
@@ -4427,20 +4435,20 @@ static void write_sm4_cast(struct hlsl_ctx *ctx,
             {
                 case HLSL_TYPE_HALF:
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_FTOU, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_FTOU, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
                     break;

case HLSL_TYPE_BOOL:
-                    write_sm4_cast_from_bool(ctx, buffer, expr, arg1, 1);
+                    write_sm4_cast_from_bool(tpf, expr, arg1, 1);
                     break;

case HLSL_TYPE_DOUBLE:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 cast from double to uint.");
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 cast from double to uint.");
                     break;

default:
@@ -4449,7 +4457,7 @@ static void write_sm4_cast(struct hlsl_ctx *ctx,
             break;

case HLSL_TYPE_DOUBLE:
-            hlsl_fixme(ctx, &expr->node.loc, "SM4 cast to double.");
+            hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 cast to double.");
             break;

case HLSL_TYPE_BOOL:
@@ -4459,26 +4467,25 @@ static void write_sm4_cast(struct hlsl_ctx *ctx,
     }
 }

-static void write_sm4_store_uav_typed(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct hlsl_deref *dst, const struct hlsl_ir_node *coords, const struct hlsl_ir_node *value)
+static void write_sm4_store_uav_typed(const struct tpf_writer *tpf, const struct hlsl_deref *dst,
+        const struct hlsl_ir_node *coords, const struct hlsl_ir_node *value)
 {
     struct sm4_instruction instr;

memset(&instr, 0, sizeof(instr));
     instr.opcode = VKD3D_SM5_OP_STORE_UAV_TYPED;

-    sm4_register_from_deref(ctx, &instr.dsts[0].reg, &instr.dsts[0].writemask, NULL, dst);
+    sm4_register_from_deref(tpf->ctx, &instr.dsts[0].reg, &instr.dsts[0].writemask, NULL, dst);
     instr.dst_count = 1;

sm4_src_from_node(&instr.srcs[0], coords, VKD3DSP_WRITEMASK_ALL);
     sm4_src_from_node(&instr.srcs[1], value, VKD3DSP_WRITEMASK_ALL);
     instr.src_count = 2;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_expr(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_expr *expr)
+static void write_sm4_expr(const struct tpf_writer *tpf, const struct hlsl_ir_expr *expr)
 {
     const struct hlsl_ir_node *arg1 = expr->operands[0].node;
     const struct hlsl_ir_node *arg2 = expr->operands[1].node;
@@ -4487,7 +4494,7 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,

assert(expr->node.reg.allocated);

-    if (!(dst_type_string = hlsl_type_to_string(ctx, dst_type)))
+    if (!(dst_type_string = hlsl_type_to_string(tpf->ctx, dst_type)))
         return;

switch (expr->op)
@@ -4496,181 +4503,181 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, VKD3D_SM4_REGISTER_MODIFIER_ABS);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, VKD3D_SM4_REGISTER_MODIFIER_ABS);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s absolute value expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s absolute value expression.", dst_type_string->buffer);
             }
             break;

case HLSL_OP1_BIT_NOT:
             assert(type_is_integer(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_NOT, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_NOT, &expr->node, arg1, 0);
             break;

case HLSL_OP1_CAST:
-            write_sm4_cast(ctx, buffer, expr);
+            write_sm4_cast(tpf, expr);
             break;

case HLSL_OP1_COS:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op_with_two_destinations(buffer, VKD3D_SM4_OP_SINCOS, &expr->node, 1, arg1);
+            write_sm4_unary_op_with_two_destinations(tpf, VKD3D_SM4_OP_SINCOS, &expr->node, 1, arg1);
             break;

case HLSL_OP1_DSX:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_DERIV_RTX, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_DERIV_RTX, &expr->node, arg1, 0);
             break;

case HLSL_OP1_DSX_COARSE:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTX_COARSE, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM5_OP_DERIV_RTX_COARSE, &expr->node, arg1, 0);
             break;

case HLSL_OP1_DSX_FINE:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTX_FINE, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM5_OP_DERIV_RTX_FINE, &expr->node, arg1, 0);
             break;

case HLSL_OP1_DSY:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_DERIV_RTY, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_DERIV_RTY, &expr->node, arg1, 0);
             break;

case HLSL_OP1_DSY_COARSE:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTY_COARSE, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM5_OP_DERIV_RTY_COARSE, &expr->node, arg1, 0);
             break;

case HLSL_OP1_DSY_FINE:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTY_FINE, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM5_OP_DERIV_RTY_FINE, &expr->node, arg1, 0);
             break;

case HLSL_OP1_EXP2:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_EXP, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_EXP, &expr->node, arg1, 0);
             break;

case HLSL_OP1_FLOOR:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_ROUND_NI, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_ROUND_NI, &expr->node, arg1, 0);
             break;

case HLSL_OP1_FRACT:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_FRC, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_FRC, &expr->node, arg1, 0);
             break;

case HLSL_OP1_LOG2:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_LOG, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_LOG, &expr->node, arg1, 0);
             break;

case HLSL_OP1_LOGIC_NOT:
             assert(dst_type->base_type == HLSL_TYPE_BOOL);
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_NOT, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_NOT, &expr->node, arg1, 0);
             break;

case HLSL_OP1_NEG:
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, VKD3D_SM4_REGISTER_MODIFIER_NEGATE);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, VKD3D_SM4_REGISTER_MODIFIER_NEGATE);
                     break;

case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_unary_op(buffer, VKD3D_SM4_OP_INEG, &expr->node, arg1, 0);
+                    write_sm4_unary_op(tpf, VKD3D_SM4_OP_INEG, &expr->node, arg1, 0);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s negation expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s negation expression.", dst_type_string->buffer);
             }
             break;

case HLSL_OP1_REINTERPRET:
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV, &expr->node, arg1, 0);
             break;

case HLSL_OP1_ROUND:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_ROUND_NE, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_ROUND_NE, &expr->node, arg1, 0);
             break;

case HLSL_OP1_RSQ:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_RSQ, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_RSQ, &expr->node, arg1, 0);
             break;

case HLSL_OP1_SAT:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_MOV
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV
                     | (VKD3D_SM4_INSTRUCTION_FLAG_SATURATE << VKD3D_SM4_INSTRUCTION_FLAGS_SHIFT),
                     &expr->node, arg1, 0);
             break;

case HLSL_OP1_SIN:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op_with_two_destinations(buffer, VKD3D_SM4_OP_SINCOS, &expr->node, 0, arg1);
+            write_sm4_unary_op_with_two_destinations(tpf, VKD3D_SM4_OP_SINCOS, &expr->node, 0, arg1);
             break;

case HLSL_OP1_SQRT:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_SQRT, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_SQRT, &expr->node, arg1, 0);
             break;

case HLSL_OP1_TRUNC:
             assert(type_is_float(dst_type));
-            write_sm4_unary_op(buffer, VKD3D_SM4_OP_ROUND_Z, &expr->node, arg1, 0);
+            write_sm4_unary_op(tpf, VKD3D_SM4_OP_ROUND_Z, &expr->node, arg1, 0);
             break;

case HLSL_OP2_ADD:
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_ADD, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_ADD, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_IADD, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_IADD, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s addition expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s addition expression.", dst_type_string->buffer);
             }
             break;

case HLSL_OP2_BIT_AND:
             assert(type_is_integer(dst_type));
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_AND, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_AND, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_BIT_OR:
             assert(type_is_integer(dst_type));
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_OR, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_OR, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_BIT_XOR:
             assert(type_is_integer(dst_type));
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_XOR, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_XOR, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_DIV:
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_DIV, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_DIV, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_UINT:
-                    write_sm4_binary_op_with_two_destinations(buffer, VKD3D_SM4_OP_UDIV, &expr->node, 0, arg1, arg2);
+                    write_sm4_binary_op_with_two_destinations(tpf, VKD3D_SM4_OP_UDIV, &expr->node, 0, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s division expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s division expression.", dst_type_string->buffer);
             }
             break;

@@ -4681,15 +4688,15 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
                     switch (arg1->data_type->dimx)
                     {
                         case 4:
-                            write_sm4_binary_op_dot(buffer, VKD3D_SM4_OP_DP4, &expr->node, arg1, arg2);
+                            write_sm4_binary_op_dot(tpf, VKD3D_SM4_OP_DP4, &expr->node, arg1, arg2);
                             break;

case 3:
-                            write_sm4_binary_op_dot(buffer, VKD3D_SM4_OP_DP3, &expr->node, arg1, arg2);
+                            write_sm4_binary_op_dot(tpf, VKD3D_SM4_OP_DP3, &expr->node, arg1, arg2);
                             break;

case 2:
-                            write_sm4_binary_op_dot(buffer, VKD3D_SM4_OP_DP2, &expr->node, arg1, arg2);
+                            write_sm4_binary_op_dot(tpf, VKD3D_SM4_OP_DP2, &expr->node, arg1, arg2);
                             break;

case 1:
@@ -4699,7 +4706,7 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s dot expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s dot expression.", dst_type_string->buffer);
             }
             break;

@@ -4712,18 +4719,18 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (src_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_EQ, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_EQ, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_BOOL:
                 case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_IEQ, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_IEQ, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 equality between "%s" operands.",
-                            debug_hlsl_type(ctx, src_type));
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 equality between "%s" operands.",
+                            debug_hlsl_type(tpf->ctx, src_type));
                     break;
             }
             break;
@@ -4738,21 +4745,21 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (src_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_GE, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_GE, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_IGE, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_IGE, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_BOOL:
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_UGE, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_UGE, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 greater-than-or-equal between "%s" operands.",
-                            debug_hlsl_type(ctx, src_type));
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 greater-than-or-equal between "%s" operands.",
+                            debug_hlsl_type(tpf->ctx, src_type));
                     break;
             }
             break;
@@ -4767,21 +4774,21 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (src_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_LT, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_LT, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_ILT, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_ILT, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_BOOL:
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_ULT, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_ULT, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 less-than between "%s" operands.",
-                            debug_hlsl_type(ctx, src_type));
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 less-than between "%s" operands.",
+                            debug_hlsl_type(tpf->ctx, src_type));
                     break;
             }
             break;
@@ -4789,37 +4796,37 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,

case HLSL_OP2_LOGIC_AND:
             assert(dst_type->base_type == HLSL_TYPE_BOOL);
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_AND, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_AND, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_LOGIC_OR:
             assert(dst_type->base_type == HLSL_TYPE_BOOL);
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_OR, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_OR, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_LSHIFT:
             assert(type_is_integer(dst_type));
             assert(dst_type->base_type != HLSL_TYPE_BOOL);
-            write_sm4_binary_op(buffer, VKD3D_SM4_OP_ISHL, &expr->node, arg1, arg2);
+            write_sm4_binary_op(tpf, VKD3D_SM4_OP_ISHL, &expr->node, arg1, arg2);
             break;

case HLSL_OP2_MAX:
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_MAX, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_MAX, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_IMAX, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_IMAX, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_UMAX, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_UMAX, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s maximum expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s maximum expression.", dst_type_string->buffer);
             }
             break;

@@ -4827,19 +4834,19 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_MIN, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_MIN, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_IMIN, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_IMIN, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_UMIN, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_UMIN, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s minimum expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s minimum expression.", dst_type_string->buffer);
             }
             break;

@@ -4847,11 +4854,11 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op_with_two_destinations(buffer, VKD3D_SM4_OP_UDIV, &expr->node, 1, arg1, arg2);
+                    write_sm4_binary_op_with_two_destinations(tpf, VKD3D_SM4_OP_UDIV, &expr->node, 1, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s modulus expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s modulus expression.", dst_type_string->buffer);
             }
             break;

@@ -4859,18 +4866,18 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (dst_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_MUL, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_MUL, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
                     /* Using IMUL instead of UMUL because we're taking the low
                      * bits, and the native compiler generates IMUL. */
-                    write_sm4_binary_op_with_two_destinations(buffer, VKD3D_SM4_OP_IMUL, &expr->node, 1, arg1, arg2);
+                    write_sm4_binary_op_with_two_destinations(tpf, VKD3D_SM4_OP_IMUL, &expr->node, 1, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 %s multiplication expression.", dst_type_string->buffer);
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s multiplication expression.", dst_type_string->buffer);
             }
             break;

@@ -4883,18 +4890,18 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
             switch (src_type->base_type)
             {
                 case HLSL_TYPE_FLOAT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_NE, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_NE, &expr->node, arg1, arg2);
                     break;

case HLSL_TYPE_BOOL:
                 case HLSL_TYPE_INT:
                 case HLSL_TYPE_UINT:
-                    write_sm4_binary_op(buffer, VKD3D_SM4_OP_INE, &expr->node, arg1, arg2);
+                    write_sm4_binary_op(tpf, VKD3D_SM4_OP_INE, &expr->node, arg1, arg2);
                     break;

default:
-                    hlsl_fixme(ctx, &expr->node.loc, "SM4 inequality between "%s" operands.",
-                            debug_hlsl_type(ctx, src_type));
+                    hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 inequality between "%s" operands.",
+                            debug_hlsl_type(tpf->ctx, src_type));
                     break;
             }
             break;
@@ -4903,18 +4910,18 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
         case HLSL_OP2_RSHIFT:
             assert(type_is_integer(dst_type));
             assert(dst_type->base_type != HLSL_TYPE_BOOL);
-            write_sm4_binary_op(buffer, dst_type->base_type == HLSL_TYPE_INT ? VKD3D_SM4_OP_ISHR : VKD3D_SM4_OP_USHR,
+            write_sm4_binary_op(tpf, dst_type->base_type == HLSL_TYPE_INT ? VKD3D_SM4_OP_ISHR : VKD3D_SM4_OP_USHR,
                     &expr->node, arg1, arg2);
             break;

default:
-            hlsl_fixme(ctx, &expr->node.loc, "SM4 %s expression.", debug_hlsl_expr_op(expr->op));
+            hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s expression.", debug_hlsl_expr_op(expr->op));
     }

-    hlsl_release_string_buffer(ctx, dst_type_string);
+    hlsl_release_string_buffer(tpf->ctx, dst_type_string);
 }

-static void write_sm4_if(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_if *iff)
+static void write_sm4_if(const struct tpf_writer *tpf, const struct hlsl_ir_if *iff)
 {
     struct sm4_instruction instr =
     {
@@ -4925,26 +4932,25 @@ static void write_sm4_if(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
     assert(iff->condition.node->data_type->dimx == 1);

sm4_src_from_node(&instr.srcs[0], iff->condition.node, VKD3DSP_WRITEMASK_ALL);
-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);

-    write_sm4_block(ctx, buffer, &iff->then_block);
+    write_sm4_block(tpf, &iff->then_block);

if (!list_empty(&iff->else_block.instrs))
     {
         instr.opcode = VKD3D_SM4_OP_ELSE;
         instr.src_count = 0;
-        write_sm4_instruction(buffer, &instr);
+        write_sm4_instruction(tpf, &instr);

-        write_sm4_block(ctx, buffer, &iff->else_block);
+        write_sm4_block(tpf, &iff->else_block);
     }

instr.opcode = VKD3D_SM4_OP_ENDIF;
     instr.src_count = 0;
-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_jump(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_jump *jump)
+static void write_sm4_jump(const struct tpf_writer *tpf, const struct hlsl_ir_jump *jump)
 {
     struct sm4_instruction instr = {0};

@@ -4968,11 +4974,11 @@ static void write_sm4_jump(struct hlsl_ctx *ctx,
             vkd3d_unreachable();

default:
-            hlsl_fixme(ctx, &jump->node.loc, "Jump type %s.", hlsl_jump_type_to_string(jump->type));
+            hlsl_fixme(tpf->ctx, &jump->node.loc, "Jump type %s.", hlsl_jump_type_to_string(jump->type));
             return;
     }

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

/* Does this variable's data come directly from the API user, rather than being
@@ -4986,8 +4992,7 @@ static bool var_is_user_input(struct hlsl_ctx *ctx, const struct hlsl_ir_var *va
     return var->is_input_semantic && ctx->profile->type == VKD3D_SHADER_TYPE_VERTEX;
 }

-static void write_sm4_load(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_load *load)
+static void write_sm4_load(const struct tpf_writer *tpf, const struct hlsl_ir_load *load)
 {
     const struct hlsl_type *type = load->node.data_type;
     struct sm4_instruction instr;
@@ -4998,7 +5003,7 @@ static void write_sm4_load(struct hlsl_ctx *ctx,
     instr.dst_count = 1;

assert(type->class <= HLSL_CLASS_LAST_NUMERIC);
-    if (type->base_type == HLSL_TYPE_BOOL && var_is_user_input(ctx, load->src.var))
+    if (type->base_type == HLSL_TYPE_BOOL && var_is_user_input(tpf->ctx, load->src.var))
     {
         struct hlsl_constant_value value;

@@ -5007,7 +5012,7 @@ static void write_sm4_load(struct hlsl_ctx *ctx,

instr.opcode = VKD3D_SM4_OP_MOVC;

-        sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);
+        sm4_src_from_deref(tpf->ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);

memset(&value, 0xff, sizeof(value));
         sm4_src_from_constant_value(&instr.srcs[1], &value, type->dimx, instr.dsts[0].writemask);
@@ -5019,32 +5024,31 @@ static void write_sm4_load(struct hlsl_ctx *ctx,
     {
         instr.opcode = VKD3D_SM4_OP_MOV;

-        sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);
+        sm4_src_from_deref(tpf->ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);
         instr.src_count = 1;
     }

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_loop(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_loop *loop)
+static void write_sm4_loop(const struct tpf_writer *tpf, const struct hlsl_ir_loop *loop)
 {
     struct sm4_instruction instr =
     {
         .opcode = VKD3D_SM4_OP_LOOP,
     };

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);

-    write_sm4_block(ctx, buffer, &loop->body);
+    write_sm4_block(tpf, &loop->body);

instr.opcode = VKD3D_SM4_OP_ENDLOOP;
-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct hlsl_ir_node *dst, const struct hlsl_deref *resource, const struct hlsl_deref *sampler,
-        const struct hlsl_ir_node *coords, unsigned int swizzle, const struct hlsl_ir_node *texel_offset)
+static void write_sm4_gather(const struct tpf_writer *tpf, const struct hlsl_ir_node *dst,
+        const struct hlsl_deref *resource, const struct hlsl_deref *sampler,  const struct hlsl_ir_node *coords,
+        unsigned int swizzle, const struct hlsl_ir_node *texel_offset)
 {
     struct sm4_src_register *src;
     struct sm4_instruction instr;
@@ -5062,9 +5066,9 @@ static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
     {
         if (!encode_texel_offset_as_aoffimmi(&instr, texel_offset))
         {
-            if (ctx->profile->major_version < 5)
+            if (tpf->ctx->profile->major_version < 5)
             {
-                hlsl_error(ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
+                hlsl_error(tpf->ctx, &texel_offset->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TEXEL_OFFSET,
                     "Offset must resolve to integer literal in the range -8 to 7 for profiles < 5.");
                 return;
             }
@@ -5073,19 +5077,18 @@ static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
         }
     }

-    sm4_src_from_deref(ctx, &instr.srcs[instr.src_count++], resource, instr.dsts[0].writemask);
+    sm4_src_from_deref(tpf->ctx, &instr.srcs[instr.src_count++], resource, instr.dsts[0].writemask);

src = &instr.srcs[instr.src_count++];
-    sm4_src_from_deref(ctx, src, sampler, VKD3DSP_WRITEMASK_ALL);
+    sm4_src_from_deref(tpf->ctx, src, sampler, VKD3DSP_WRITEMASK_ALL);
     src->reg.dim = VKD3D_SM4_DIMENSION_VEC4;
     src->swizzle_type = VKD3D_SM4_SWIZZLE_SCALAR;
     src->swizzle = swizzle;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_resource_load(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_resource_load *load)
+static void write_sm4_resource_load(const struct tpf_writer *tpf, const struct hlsl_ir_resource_load *load)
 {
     const struct hlsl_ir_node *texel_offset = load->texel_offset.node;
     const struct hlsl_ir_node *sample_index = load->sample_index.node;
@@ -5093,20 +5096,20 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx,

if (load->sampler.var && !load->sampler.var->is_uniform)
     {
-        hlsl_fixme(ctx, &load->node.loc, "Sample using non-uniform sampler variable.");
+        hlsl_fixme(tpf->ctx, &load->node.loc, "Sample using non-uniform sampler variable.");
         return;
     }

if (!load->resource.var->is_uniform)
     {
-        hlsl_fixme(ctx, &load->node.loc, "Load from non-uniform resource variable.");
+        hlsl_fixme(tpf->ctx, &load->node.loc, "Load from non-uniform resource variable.");
         return;
     }

switch (load->load_type)
     {
         case HLSL_RESOURCE_LOAD:
-            write_sm4_ld(ctx, buffer, &load->node, &load->resource,
+            write_sm4_ld(tpf, &load->node, &load->resource,
                     coords, sample_index, texel_offset, load->sampling_dim);
             break;

@@ -5118,53 +5121,51 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx,
         case HLSL_RESOURCE_SAMPLE_GRAD:
             /* Combined sample expressions were lowered. */
             assert(load->sampler.var);
-            write_sm4_sample(ctx, buffer, load);
+            write_sm4_sample(tpf, load);
             break;

case HLSL_RESOURCE_GATHER_RED:
-            write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+            write_sm4_gather(tpf, &load->node, &load->resource, &load->sampler, coords,
                     HLSL_SWIZZLE(X, X, X, X), texel_offset);
             break;

case HLSL_RESOURCE_GATHER_GREEN:
-            write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+            write_sm4_gather(tpf, &load->node, &load->resource, &load->sampler, coords,
                     HLSL_SWIZZLE(Y, Y, Y, Y), texel_offset);
             break;

case HLSL_RESOURCE_GATHER_BLUE:
-            write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+            write_sm4_gather(tpf, &load->node, &load->resource, &load->sampler, coords,
                     HLSL_SWIZZLE(Z, Z, Z, Z), texel_offset);
             break;

case HLSL_RESOURCE_GATHER_ALPHA:
-            write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+            write_sm4_gather(tpf, &load->node, &load->resource, &load->sampler, coords,
                     HLSL_SWIZZLE(W, W, W, W), texel_offset);
             break;
     }
 }

-static void write_sm4_resource_store(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_resource_store *store)
+static void write_sm4_resource_store(const struct tpf_writer *tpf, const struct hlsl_ir_resource_store *store)
 {
-    struct hlsl_type *resource_type = hlsl_deref_get_type(ctx, &store->resource);
+    struct hlsl_type *resource_type = hlsl_deref_get_type(tpf->ctx, &store->resource);

if (!store->resource.var->is_uniform)
     {
-        hlsl_fixme(ctx, &store->node.loc, "Store to non-uniform resource variable.");
+        hlsl_fixme(tpf->ctx, &store->node.loc, "Store to non-uniform resource variable.");
         return;
     }

if (resource_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER)
     {
-        hlsl_fixme(ctx, &store->node.loc, "Structured buffers store is not implemented.");
+        hlsl_fixme(tpf->ctx, &store->node.loc, "Structured buffers store is not implemented.");
         return;
     }

-    write_sm4_store_uav_typed(ctx, buffer, &store->resource, store->coords.node, store->value.node);
+    write_sm4_store_uav_typed(tpf, &store->resource, store->coords.node, store->value.node);
 }

-static void write_sm4_store(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_store *store)
+static void write_sm4_store(const struct tpf_writer *tpf, const struct hlsl_ir_store *store)
 {
     const struct hlsl_ir_node *rhs = store->rhs.node;
     struct sm4_instruction instr;
@@ -5173,18 +5174,17 @@ static void write_sm4_store(struct hlsl_ctx *ctx,
     memset(&instr, 0, sizeof(instr));
     instr.opcode = VKD3D_SM4_OP_MOV;

-    sm4_register_from_deref(ctx, &instr.dsts[0].reg, &writemask, NULL, &store->lhs);
+    sm4_register_from_deref(tpf->ctx, &instr.dsts[0].reg, &writemask, NULL, &store->lhs);
     instr.dsts[0].writemask = hlsl_combine_writemasks(writemask, store->writemask);
     instr.dst_count = 1;

sm4_src_from_node(&instr.srcs[0], rhs, instr.dsts[0].writemask);
     instr.src_count = 1;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_swizzle(struct hlsl_ctx *ctx,
-        struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_swizzle *swizzle)
+static void write_sm4_swizzle(const struct tpf_writer *tpf, const struct hlsl_ir_swizzle *swizzle)
 {
     struct sm4_instruction instr;
     unsigned int writemask;
@@ -5200,11 +5200,10 @@ static void write_sm4_swizzle(struct hlsl_ctx *ctx,
             swizzle->swizzle, swizzle->node.data_type->dimx), instr.dsts[0].writemask);
     instr.src_count = 1;

-    write_sm4_instruction(buffer, &instr);
+    write_sm4_instruction(tpf, &instr);
 }

-static void write_sm4_block(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
-        const struct hlsl_block *block)
+static void write_sm4_block(const struct tpf_writer *tpf, const struct hlsl_block *block)
 {
     const struct hlsl_ir_node *instr;

@@ -5214,12 +5213,12 @@ static void write_sm4_block(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *
         {
             if (instr->data_type->class == HLSL_CLASS_MATRIX)
             {
-                hlsl_fixme(ctx, &instr->loc, "Matrix operations need to be lowered.");
+                hlsl_fixme(tpf->ctx, &instr->loc, "Matrix operations need to be lowered.");
                 break;
             }
             else if (instr->data_type->class == HLSL_CLASS_OBJECT)
             {
-                hlsl_fixme(ctx, &instr->loc, "Object copy.");
+                hlsl_fixme(tpf->ctx, &instr->loc, "Object copy.");
                 break;
             }

@@ -5239,43 +5238,43 @@ static void write_sm4_block(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *
                 vkd3d_unreachable();

case HLSL_IR_EXPR:
-                write_sm4_expr(ctx, buffer, hlsl_ir_expr(instr));
+                write_sm4_expr(tpf, hlsl_ir_expr(instr));
                 break;

case HLSL_IR_IF:
-                write_sm4_if(ctx, buffer, hlsl_ir_if(instr));
+                write_sm4_if(tpf, hlsl_ir_if(instr));
                 break;

case HLSL_IR_JUMP:
-                write_sm4_jump(ctx, buffer, hlsl_ir_jump(instr));
+                write_sm4_jump(tpf, hlsl_ir_jump(instr));
                 break;

case HLSL_IR_LOAD:
-                write_sm4_load(ctx, buffer, hlsl_ir_load(instr));
+                write_sm4_load(tpf, hlsl_ir_load(instr));
                 break;

case HLSL_IR_RESOURCE_LOAD:
-                write_sm4_resource_load(ctx, buffer, hlsl_ir_resource_load(instr));
+                write_sm4_resource_load(tpf, hlsl_ir_resource_load(instr));
                 break;

case HLSL_IR_RESOURCE_STORE:
-                write_sm4_resource_store(ctx, buffer, hlsl_ir_resource_store(instr));
+                write_sm4_resource_store(tpf, hlsl_ir_resource_store(instr));
                 break;

case HLSL_IR_LOOP:
-                write_sm4_loop(ctx, buffer, hlsl_ir_loop(instr));
+                write_sm4_loop(tpf, hlsl_ir_loop(instr));
                 break;

case HLSL_IR_STORE:
-                write_sm4_store(ctx, buffer, hlsl_ir_store(instr));
+                write_sm4_store(tpf, hlsl_ir_store(instr));
                 break;

case HLSL_IR_SWIZZLE:
-                write_sm4_swizzle(ctx, buffer, hlsl_ir_swizzle(instr));
+                write_sm4_swizzle(tpf, hlsl_ir_swizzle(instr));
                 break;

default:
-                hlsl_fixme(ctx, &instr->loc, "Instruction type %s.", hlsl_node_type_to_string(instr->type));
+                hlsl_fixme(tpf->ctx, &instr->loc, "Instruction type %s.", hlsl_node_type_to_string(instr->type));
         }
     }
 }
@@ -5290,6 +5289,7 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
     const struct hlsl_buffer *cbuffer;
     const struct hlsl_ir_var *var;
     size_t token_count_position;
+    struct tpf_writer tpf;

static const uint16_t shader_types[VKD3D_SHADER_TYPE_COUNT] =
     {
@@ -5304,6 +5304,8 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
         VKD3D_SM4_LIB,
     };

+    tpf_writer_init(&tpf, ctx, &buffer);
+
     extern_resources = sm4_get_extern_resources(ctx, &extern_resources_count);

put_u32(&buffer, vkd3d_make_u32((profile->major_version << 4) | profile->minor_version, shader_types[profile->type]));
@@ -5312,7 +5314,7 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
     LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry)
     {
         if (cbuffer->reg.allocated)
-            write_sm4_dcl_constant_buffer(&buffer, cbuffer);
+            write_sm4_dcl_constant_buffer(&tpf, cbuffer);
     }

for (i = 0; i < extern_resources_count; ++i)
@@ -5320,28 +5322,28 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
         const struct extern_resource *resource = &extern_resources[i];

if (resource->regset == HLSL_REGSET_SAMPLERS)
-            write_sm4_dcl_samplers(ctx, &buffer, resource);
+            write_sm4_dcl_samplers(&tpf, resource);
         else if (resource->regset == HLSL_REGSET_TEXTURES)
-            write_sm4_dcl_textures(ctx, &buffer, resource, false);
+            write_sm4_dcl_textures(&tpf, resource, false);
         else if (resource->regset == HLSL_REGSET_UAVS)
-            write_sm4_dcl_textures(ctx, &buffer, resource, true);
+            write_sm4_dcl_textures(&tpf, resource, true);
     }

LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
     {
         if ((var->is_input_semantic && var->last_read) || (var->is_output_semantic && var->first_write))
-            write_sm4_dcl_semantic(ctx, &buffer, var);
+            write_sm4_dcl_semantic(&tpf, var);
     }

if (profile->type == VKD3D_SHADER_TYPE_COMPUTE)
-        write_sm4_dcl_thread_group(&buffer, ctx->thread_count);
+        write_sm4_dcl_thread_group(&tpf, ctx->thread_count);

if (ctx->temp_count)
-        write_sm4_dcl_temps(&buffer, ctx->temp_count);
+        write_sm4_dcl_temps(&tpf, ctx->temp_count);

-    write_sm4_block(ctx, &buffer, &entry_func->body);
+    write_sm4_block(&tpf, &entry_func->body);

-    write_sm4_ret(&buffer);
+    write_sm4_ret(&tpf);

set_u32(&buffer, token_count_position, bytecode_get_size(&buffer) / sizeof(uint32_t));

[PATCH v6 5/9] vkd3d-shader/tpf: Make register_type_table an array of structs with lookup tables.

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c                  | 165 +++++++++++++++--------
 libs/vkd3d-shader/vkd3d_shader_private.h |   2 +
 2 files changed, 108 insertions(+), 59 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index f83029b2..248fe047 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -430,6 +430,8 @@ enum vkd3d_sm4_register_type
     VKD3D_SM5_RT_DEPTHOUT_GREATER_EQUAL  = 0x26,
     VKD3D_SM5_RT_DEPTHOUT_LESS_EQUAL     = 0x27,
     VKD3D_SM5_RT_OUTPUT_STENCIL_REF      = 0x29,
+
+    VKD3D_SM4_REGISTER_TYPE_COUNT,
 };

enum vkd3d_sm4_extended_operand_type
@@ -571,6 +573,12 @@ struct sm4_index_range_array
     struct sm4_index_range ranges[MAX_REG_OUTPUT * 2];
 };

+struct vkd3d_sm4_lookup_tables
+{
+    const struct vkd3d_sm4_register_type_info *register_type_info_from_sm4[VKD3D_SM4_REGISTER_TYPE_COUNT];
+    const struct vkd3d_sm4_register_type_info *register_type_info_from_vkd3d[VKD3DSPR_COUNT];
+};
+
 struct vkd3d_shader_sm4_parser
 {
     const uint32_t *start, *end, *ptr;
@@ -587,6 +595,8 @@ struct vkd3d_shader_sm4_parser
     struct sm4_index_range_array output_index_ranges;
     struct sm4_index_range_array patch_constant_index_ranges;

+    struct vkd3d_sm4_lookup_tables lookup;
+
     struct vkd3d_shader_parser p;
 };

@@ -1468,50 +1478,10 @@ static const struct vkd3d_sm4_opcode_info opcode_table[] =
     {VKD3D_SM5_OP_CHECK_ACCESS_FULLY_MAPPED,        VKD3DSIH_CHECK_ACCESS_FULLY_MAPPED,        "u",    "u"},
 };

-static const enum vkd3d_shader_register_type register_type_table[] =
-{
-    /* VKD3D_SM4_RT_TEMP */                    VKD3DSPR_TEMP,
-    /* VKD3D_SM4_RT_INPUT */                   VKD3DSPR_INPUT,
-    /* VKD3D_SM4_RT_OUTPUT */                  VKD3DSPR_OUTPUT,
-    /* VKD3D_SM4_RT_INDEXABLE_TEMP */          VKD3DSPR_IDXTEMP,
-    /* VKD3D_SM4_RT_IMMCONST */                VKD3DSPR_IMMCONST,
-    /* VKD3D_SM4_RT_IMMCONST64 */              VKD3DSPR_IMMCONST64,
-    /* VKD3D_SM4_RT_SAMPLER */                 VKD3DSPR_SAMPLER,
-    /* VKD3D_SM4_RT_RESOURCE */                VKD3DSPR_RESOURCE,
-    /* VKD3D_SM4_RT_CONSTBUFFER */             VKD3DSPR_CONSTBUFFER,
-    /* VKD3D_SM4_RT_IMMCONSTBUFFER */          VKD3DSPR_IMMCONSTBUFFER,
-    /* UNKNOWN */                              ~0u,
-    /* VKD3D_SM4_RT_PRIMID */                  VKD3DSPR_PRIMID,
-    /* VKD3D_SM4_RT_DEPTHOUT */                VKD3DSPR_DEPTHOUT,
-    /* VKD3D_SM4_RT_NULL */                    VKD3DSPR_NULL,
-    /* VKD3D_SM4_RT_RASTERIZER */              VKD3DSPR_RASTERIZER,
-    /* VKD3D_SM4_RT_OMASK */                   VKD3DSPR_SAMPLEMASK,
-    /* VKD3D_SM5_RT_STREAM */                  VKD3DSPR_STREAM,
-    /* VKD3D_SM5_RT_FUNCTION_BODY */           VKD3DSPR_FUNCTIONBODY,
-    /* UNKNOWN */                              ~0u,
-    /* VKD3D_SM5_RT_FUNCTION_POINTER */        VKD3DSPR_FUNCTIONPOINTER,
-    /* UNKNOWN */                              ~0u,
-    /* UNKNOWN */                              ~0u,
-    /* VKD3D_SM5_RT_OUTPUT_CONTROL_POINT_ID */ VKD3DSPR_OUTPOINTID,
-    /* VKD3D_SM5_RT_FORK_INSTANCE_ID */        VKD3DSPR_FORKINSTID,
-    /* VKD3D_SM5_RT_JOIN_INSTANCE_ID */        VKD3DSPR_JOININSTID,
-    /* VKD3D_SM5_RT_INPUT_CONTROL_POINT */     VKD3DSPR_INCONTROLPOINT,
-    /* VKD3D_SM5_RT_OUTPUT_CONTROL_POINT */    VKD3DSPR_OUTCONTROLPOINT,
-    /* VKD3D_SM5_RT_PATCH_CONSTANT_DATA */     VKD3DSPR_PATCHCONST,
-    /* VKD3D_SM5_RT_DOMAIN_LOCATION */         VKD3DSPR_TESSCOORD,
-    /* UNKNOWN */                              ~0u,
-    /* VKD3D_SM5_RT_UAV */                     VKD3DSPR_UAV,
-    /* VKD3D_SM5_RT_SHARED_MEMORY */           VKD3DSPR_GROUPSHAREDMEM,
-    /* VKD3D_SM5_RT_THREAD_ID */               VKD3DSPR_THREADID,
-    /* VKD3D_SM5_RT_THREAD_GROUP_ID */         VKD3DSPR_THREADGROUPID,
-    /* VKD3D_SM5_RT_LOCAL_THREAD_ID */         VKD3DSPR_LOCALTHREADID,
-    /* VKD3D_SM5_RT_COVERAGE */                VKD3DSPR_COVERAGE,
-    /* VKD3D_SM5_RT_LOCAL_THREAD_INDEX */      VKD3DSPR_LOCALTHREADINDEX,
-    /* VKD3D_SM5_RT_GS_INSTANCE_ID */          VKD3DSPR_GSINSTID,
-    /* VKD3D_SM5_RT_DEPTHOUT_GREATER_EQUAL */  VKD3DSPR_DEPTHOUTGE,
-    /* VKD3D_SM5_RT_DEPTHOUT_LESS_EQUAL */     VKD3DSPR_DEPTHOUTLE,
-    /* VKD3D_SM5_RT_CYCLE_COUNTER */           ~0u,
-    /* VKD3D_SM5_RT_OUTPUT_STENCIL_REF */      VKD3DSPR_OUTSTENCILREF,
+struct vkd3d_sm4_register_type_info
+{
+    enum vkd3d_sm4_register_type sm4_type;
+    enum vkd3d_shader_register_type vkd3d_type;
 };

static const enum vkd3d_shader_register_precision register_precision_table[] =
@@ -1528,6 +1498,7 @@ struct tpf_writer
 {
     struct hlsl_ctx *ctx;
     struct vkd3d_bytecode_buffer *buffer;
+    struct vkd3d_sm4_lookup_tables lookup;
 };

static const struct vkd3d_sm4_opcode_info *get_opcode_info(enum vkd3d_sm4_opcode opcode)
@@ -1536,16 +1507,89 @@ static const struct vkd3d_sm4_opcode_info *get_opcode_info(enum vkd3d_sm4_opcode

for (i = 0; i < sizeof(opcode_table) / sizeof(*opcode_table); ++i)
     {
-        if (opcode == opcode_table[i].opcode) return &opcode_table[i];
+        if (opcode == opcode_table[i].opcode)
+            return &opcode_table[i];
     }

return NULL;
 }

+static void init_sm4_lookup_tables(struct vkd3d_sm4_lookup_tables *lookup)
+{
+    const struct vkd3d_sm4_register_type_info *info;
+    unsigned int i;
+
+    static const struct vkd3d_sm4_register_type_info register_type_table[] =
+    {
+        {VKD3D_SM4_RT_TEMP,                    VKD3DSPR_TEMP},
+        {VKD3D_SM4_RT_INPUT,                   VKD3DSPR_INPUT},
+        {VKD3D_SM4_RT_OUTPUT,                  VKD3DSPR_OUTPUT},
+        {VKD3D_SM4_RT_INDEXABLE_TEMP,          VKD3DSPR_IDXTEMP},
+        {VKD3D_SM4_RT_IMMCONST,                VKD3DSPR_IMMCONST},
+        {VKD3D_SM4_RT_IMMCONST64,              VKD3DSPR_IMMCONST64},
+        {VKD3D_SM4_RT_SAMPLER,                 VKD3DSPR_SAMPLER},
+        {VKD3D_SM4_RT_RESOURCE,                VKD3DSPR_RESOURCE},
+        {VKD3D_SM4_RT_CONSTBUFFER,             VKD3DSPR_CONSTBUFFER},
+        {VKD3D_SM4_RT_IMMCONSTBUFFER,          VKD3DSPR_IMMCONSTBUFFER},
+        {VKD3D_SM4_RT_PRIMID,                  VKD3DSPR_PRIMID},
+        {VKD3D_SM4_RT_DEPTHOUT,                VKD3DSPR_DEPTHOUT},
+        {VKD3D_SM4_RT_NULL,                    VKD3DSPR_NULL},
+        {VKD3D_SM4_RT_RASTERIZER,              VKD3DSPR_RASTERIZER},
+        {VKD3D_SM4_RT_OMASK,                   VKD3DSPR_SAMPLEMASK},
+        {VKD3D_SM5_RT_STREAM,                  VKD3DSPR_STREAM},
+        {VKD3D_SM5_RT_FUNCTION_BODY,           VKD3DSPR_FUNCTIONBODY},
+        {VKD3D_SM5_RT_FUNCTION_POINTER,        VKD3DSPR_FUNCTIONPOINTER},
+        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT_ID, VKD3DSPR_OUTPOINTID},
+        {VKD3D_SM5_RT_FORK_INSTANCE_ID,        VKD3DSPR_FORKINSTID},
+        {VKD3D_SM5_RT_JOIN_INSTANCE_ID,        VKD3DSPR_JOININSTID},
+        {VKD3D_SM5_RT_INPUT_CONTROL_POINT,     VKD3DSPR_INCONTROLPOINT},
+        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT,    VKD3DSPR_OUTCONTROLPOINT},
+        {VKD3D_SM5_RT_PATCH_CONSTANT_DATA,     VKD3DSPR_PATCHCONST},
+        {VKD3D_SM5_RT_DOMAIN_LOCATION,         VKD3DSPR_TESSCOORD},
+        {VKD3D_SM5_RT_UAV,                     VKD3DSPR_UAV},
+        {VKD3D_SM5_RT_SHARED_MEMORY,           VKD3DSPR_GROUPSHAREDMEM},
+        {VKD3D_SM5_RT_THREAD_ID,               VKD3DSPR_THREADID},
+        {VKD3D_SM5_RT_THREAD_GROUP_ID,         VKD3DSPR_THREADGROUPID},
+        {VKD3D_SM5_RT_LOCAL_THREAD_ID,         VKD3DSPR_LOCALTHREADID},
+        {VKD3D_SM5_RT_COVERAGE,                VKD3DSPR_COVERAGE},
+        {VKD3D_SM5_RT_LOCAL_THREAD_INDEX,      VKD3DSPR_LOCALTHREADINDEX},
+        {VKD3D_SM5_RT_GS_INSTANCE_ID,          VKD3DSPR_GSINSTID},
+        {VKD3D_SM5_RT_DEPTHOUT_GREATER_EQUAL,  VKD3DSPR_DEPTHOUTGE},
+        {VKD3D_SM5_RT_DEPTHOUT_LESS_EQUAL,     VKD3DSPR_DEPTHOUTLE},
+        {VKD3D_SM5_RT_OUTPUT_STENCIL_REF,      VKD3DSPR_OUTSTENCILREF},
+    };
+
+    memset(lookup, 0, sizeof(*lookup));
+
+    for (i = 0; i < ARRAY_SIZE(register_type_table); ++i)
+    {
+        info = &register_type_table[i];
+        lookup->register_type_info_from_sm4[info->sm4_type] = info;
+        lookup->register_type_info_from_vkd3d[info->vkd3d_type] = info;
+    }
+}
+
 static void tpf_writer_init(struct tpf_writer *tpf, struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer)
 {
     tpf->ctx = ctx;
     tpf->buffer = buffer;
+    init_sm4_lookup_tables(&tpf->lookup);
+}
+
+static const struct vkd3d_sm4_register_type_info *get_info_from_sm4_register_type(
+        const struct vkd3d_sm4_lookup_tables *lookup, enum vkd3d_sm4_register_type sm4_type)
+{
+    if (sm4_type >= VKD3D_SM4_REGISTER_TYPE_COUNT)
+        return NULL;
+    return lookup->register_type_info_from_sm4[sm4_type];
+}
+
+static const struct vkd3d_sm4_register_type_info *get_info_from_vkd3d_register_type(
+        const struct vkd3d_sm4_lookup_tables *lookup, enum vkd3d_shader_register_type vkd3d_type)
+{
+    if (vkd3d_type >= VKD3DSPR_COUNT)
+        return NULL;
+    return lookup->register_type_info_from_vkd3d[vkd3d_type];
 }

static void map_register(const struct vkd3d_shader_sm4_parser *sm4, struct vkd3d_shader_register *reg)
@@ -1654,6 +1698,7 @@ static bool sm4_register_is_descriptor(enum vkd3d_sm4_register_type register_typ
 static bool shader_sm4_read_param(struct vkd3d_shader_sm4_parser *priv, const uint32_t **ptr, const uint32_t *end,
         enum vkd3d_data_type data_type, struct vkd3d_shader_register *param, enum vkd3d_shader_src_modifier *modifier)
 {
+    const struct vkd3d_sm4_register_type_info *register_type_info;
     enum vkd3d_sm4_register_precision precision;
     enum vkd3d_sm4_register_type register_type;
     enum vkd3d_sm4_extended_operand_type type;
@@ -1668,15 +1713,15 @@ static bool shader_sm4_read_param(struct vkd3d_shader_sm4_parser *priv, const ui
     token = *(*ptr)++;

register_type = (token & VKD3D_SM4_REGISTER_TYPE_MASK) >> VKD3D_SM4_REGISTER_TYPE_SHIFT;
-    if (register_type >= ARRAY_SIZE(register_type_table)
-            || register_type_table[register_type] == VKD3DSPR_INVALID)
+    register_type_info = get_info_from_sm4_register_type(&priv->lookup, register_type);
+    if (!register_type_info)
     {
         FIXME("Unhandled register type %#x.\n", register_type);
         param->type = VKD3DSPR_TEMP;
     }
     else
     {
-        param->type = register_type_table[register_type];
+        param->type = register_type_info->vkd3d_type;
     }
     param->precision = VKD3D_SHADER_REGISTER_PRECISION_DEFAULT;
     param->non_uniform = false;
@@ -2376,6 +2421,8 @@ static bool shader_sm4_init(struct vkd3d_shader_sm4_parser *sm4, const uint32_t
         sm4->output_map[e->register_index] = e->semantic_index;
     }

+    init_sm4_lookup_tables(&sm4->lookup);
+
     return true;
 }

@@ -3690,20 +3737,20 @@ static void sm4_src_from_node(struct sm4_src_register *src,

static uint32_t sm4_encode_register(const struct tpf_writer *tpf, const struct sm4_register *reg)
 {
-    uint32_t sm4_reg_type = ~0u;
-    unsigned int i;
+    const struct vkd3d_sm4_register_type_info *register_type_info;
+    uint32_t sm4_reg_type;

-    /* Find sm4 register type from vkd3d_shader_register_type. */
-    for (i = 0; i < ARRAY_SIZE(register_type_table); ++i)
+    register_type_info = get_info_from_vkd3d_register_type(&tpf->lookup, reg->type);
+
+    if (!register_type_info)
     {
-        if (reg->type == register_type_table[i])
-        {
-            if (sm4_reg_type != ~0u)
-                ERR("Multiple maps for register_type %#x.\n", reg->type);
-            sm4_reg_type = i;
-        }
+        FIXME("Unhandled vkd3d-shader register type %#x.\n", reg->type);
+        sm4_reg_type = VKD3D_SM4_RT_TEMP;
+    }
+    else
+    {
+        sm4_reg_type = register_type_info->sm4_type;
     }
-    assert(sm4_reg_type != ~0u);

return (sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT)
             | (reg->idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT)
diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h
index 49ad7b3e..c40690e3 100644
--- a/libs/vkd3d-shader/vkd3d_shader_private.h
+++ b/libs/vkd3d-shader/vkd3d_shader_private.h
@@ -508,6 +508,8 @@ enum vkd3d_shader_register_type
     VKD3DSPR_RASTERIZER,
     VKD3DSPR_OUTSTENCILREF,

+    VKD3DSPR_COUNT,
+
     VKD3DSPR_INVALID = ~0u,
 };

[PATCH v6 6/9] vkd3d-shader/tpf: Separate src register write function.

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c | 85 +++++++++++++++++++++++++++--------------
 1 file changed, 56 insertions(+), 29 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 248fe047..70bb7c1c 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -3757,6 +3757,61 @@ static uint32_t sm4_encode_register(const struct tpf_writer *tpf, const struct s
             | (reg->dim << VKD3D_SM4_DIMENSION_SHIFT);
 }

+static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm4_src_register *src)
+{
+    const struct vkd3d_sm4_register_type_info *register_type_info;
+    struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
+    uint32_t sm4_reg_type, reg_dim;
+    uint32_t token = 0;
+    unsigned int j;
+
+    register_type_info = get_info_from_vkd3d_register_type(&tpf->lookup, src->reg.type);
+    if (!register_type_info)
+    {
+        FIXME("Unhandled vkd3d-shader register type %#x.\n", src->reg.type);
+        sm4_reg_type = VKD3D_SM4_RT_TEMP;
+    }
+    else
+    {
+        sm4_reg_type = register_type_info->sm4_type;
+    }
+
+    reg_dim = src->reg.dim;
+
+    token |= sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT;
+    token |= src->reg.idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT;
+    token |= reg_dim << VKD3D_SM4_DIMENSION_SHIFT;
+    if (reg_dim == VKD3D_SM4_DIMENSION_VEC4)
+    {
+        token |= (uint32_t)src->swizzle_type << VKD3D_SM4_SWIZZLE_TYPE_SHIFT;
+        token |= src->swizzle << VKD3D_SM4_SWIZZLE_SHIFT;
+    }
+    if (src->reg.mod)
+        token |= VKD3D_SM4_EXTENDED_OPERAND;
+    put_u32(buffer, token);
+
+    if (src->reg.mod)
+        put_u32(buffer, (src->reg.mod << VKD3D_SM4_REGISTER_MODIFIER_SHIFT)
+                | VKD3D_SM4_EXTENDED_OPERAND_MODIFIER);
+
+    for (j = 0; j < src->reg.idx_count; ++j)
+    {
+        put_u32(buffer, src->reg.idx[j].offset);
+        assert(!src->reg.idx[j].rel_addr);
+    }
+
+    if (src->reg.type == VKD3DSPR_IMMCONST)
+    {
+        put_u32(buffer, src->reg.immconst_uint[0]);
+        if (reg_dim == VKD3D_SM4_DIMENSION_VEC4)
+        {
+            put_u32(buffer, src->reg.immconst_uint[1]);
+            put_u32(buffer, src->reg.immconst_uint[2]);
+            put_u32(buffer, src->reg.immconst_uint[3]);
+        }
+    }
+}
+
 static uint32_t sm4_register_order(const struct sm4_register *reg)
 {
     uint32_t order = 1;
@@ -3812,35 +3867,7 @@ static void write_sm4_instruction(const struct tpf_writer *tpf, const struct sm4
     }

for (i = 0; i < instr->src_count; ++i)
-    {
-        token = sm4_encode_register(tpf, &instr->srcs[i].reg);
-        token |= (uint32_t)instr->srcs[i].swizzle_type << VKD3D_SM4_SWIZZLE_TYPE_SHIFT;
-        token |= instr->srcs[i].swizzle << VKD3D_SM4_SWIZZLE_SHIFT;
-        if (instr->srcs[i].reg.mod)
-            token |= VKD3D_SM4_EXTENDED_OPERAND;
-        put_u32(buffer, token);
-
-        if (instr->srcs[i].reg.mod)
-            put_u32(buffer, (instr->srcs[i].reg.mod << VKD3D_SM4_REGISTER_MODIFIER_SHIFT)
-                    | VKD3D_SM4_EXTENDED_OPERAND_MODIFIER);
-
-        for (j = 0; j < instr->srcs[i].reg.idx_count; ++j)
-        {
-            put_u32(buffer, instr->srcs[i].reg.idx[j].offset);
-            assert(!instr->srcs[i].reg.idx[j].rel_addr);
-        }
-
-        if (instr->srcs[i].reg.type == VKD3DSPR_IMMCONST)
-        {
-            put_u32(buffer, instr->srcs[i].reg.immconst_uint[0]);
-            if (instr->srcs[i].reg.dim == VKD3D_SM4_DIMENSION_VEC4)
-            {
-                put_u32(buffer, instr->srcs[i].reg.immconst_uint[1]);
-                put_u32(buffer, instr->srcs[i].reg.immconst_uint[2]);
-                put_u32(buffer, instr->srcs[i].reg.immconst_uint[3]);
-            }
-        }
-    }
+        sm4_write_src_register(tpf, &instr->srcs[i]);

if (instr->byte_stride)
         put_u32(buffer, instr->byte_stride);

[PATCH v6 7/9] vkd3d-shader/tpf: Separate dst register write function.

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c | 42 +++++++++++++++++++++--------------------
 1 file changed, 22 insertions(+), 20 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 70bb7c1c..6adee197 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -3735,16 +3735,18 @@ static void sm4_src_from_node(struct sm4_src_register *src,
         src->swizzle = hlsl_map_swizzle(hlsl_swizzle_from_writemask(writemask), map_writemask);
 }

-static uint32_t sm4_encode_register(const struct tpf_writer *tpf, const struct sm4_register *reg)
+static void sm4_write_dst_register(const struct tpf_writer *tpf, const struct sm4_dst_register *dst)
 {
     const struct vkd3d_sm4_register_type_info *register_type_info;
-    uint32_t sm4_reg_type;
-
-    register_type_info = get_info_from_vkd3d_register_type(&tpf->lookup, reg->type);
+    struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
+    uint32_t sm4_reg_type, reg_dim;
+    uint32_t token = 0;
+    unsigned int j;

+    register_type_info = get_info_from_vkd3d_register_type(&tpf->lookup, dst->reg.type);
     if (!register_type_info)
     {
-        FIXME("Unhandled vkd3d-shader register type %#x.\n", reg->type);
+        FIXME("Unhandled vkd3d-shader register type %#x.\n", dst->reg.type);
         sm4_reg_type = VKD3D_SM4_RT_TEMP;
     }
     else
@@ -3752,9 +3754,20 @@ static uint32_t sm4_encode_register(const struct tpf_writer *tpf, const struct s
         sm4_reg_type = register_type_info->sm4_type;
     }

-    return (sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT)
-            | (reg->idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT)
-            | (reg->dim << VKD3D_SM4_DIMENSION_SHIFT);
+    reg_dim = dst->reg.dim;
+
+    token |= sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT;
+    token |= dst->reg.idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT;
+    token |= reg_dim << VKD3D_SM4_DIMENSION_SHIFT;
+    if (reg_dim == VKD3D_SM4_DIMENSION_VEC4)
+        token |= dst->writemask << VKD3D_SM4_WRITEMASK_SHIFT;
+    put_u32(buffer, token);
+
+    for (j = 0; j < dst->reg.idx_count; ++j)
+    {
+        put_u32(buffer, dst->reg.idx[j].offset);
+        assert(!dst->reg.idx[j].rel_addr);
+    }
 }

static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm4_src_register *src)
@@ -3853,18 +3866,7 @@ static void write_sm4_instruction(const struct tpf_writer *tpf, const struct sm4
     }

for (i = 0; i < instr->dst_count; ++i)
-    {
-        token = sm4_encode_register(tpf, &instr->dsts[i].reg);
-        if (instr->dsts[i].reg.dim == VKD3D_SM4_DIMENSION_VEC4)
-            token |= instr->dsts[i].writemask << VKD3D_SM4_WRITEMASK_SHIFT;
-        put_u32(buffer, token);
-
-        for (j = 0; j < instr->dsts[i].reg.idx_count; ++j)
-        {
-            put_u32(buffer, instr->dsts[i].reg.idx[j].offset);
-            assert(!instr->dsts[i].reg.idx[j].rel_addr);
-        }
-    }
+        sm4_write_dst_register(tpf, &instr->dsts[i]);

for (i = 0; i < instr->src_count; ++i)
         sm4_write_src_register(tpf, &instr->srcs[i]);

[PATCH v6 8/9] vkd3d-shader/tpf: Use 's' in src_info to expect sampler register with vec4 dimension.

[Francisco Casas]

From: Francisco Casas

So far, for every register type we write we only use a single dimension type.
The only exception being the sampler register for gather instructions,
where the register dimension must be VEC4 (and the swizzle_type SCALAR)
instead of NONE.

To get rid of sm4_src_register.dim, we need a way to handle this
nasty exception, which is provided by this patch.
---
 libs/vkd3d-shader/tpf.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 6adee197..9359e4ff 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -1206,6 +1206,7 @@ static void shader_sm5_read_sync(struct vkd3d_shader_instruction *ins, uint32_t
  * O -> VKD3D_DATA_OPAQUE
  * R -> VKD3D_DATA_RESOURCE
  * S -> VKD3D_DATA_SAMPLER
+ * s -> VKD3D_DATA_SAMPLER with scalar swizzle
  * U -> VKD3D_DATA_UAV
  */
 static const struct vkd3d_sm4_opcode_info opcode_table[] =
@@ -1337,7 +1338,7 @@ static const struct vkd3d_sm4_opcode_info opcode_table[] =
     {VKD3D_SM4_OP_DCL_GLOBAL_FLAGS,                 VKD3DSIH_DCL_GLOBAL_FLAGS,                 "",     "",
             shader_sm4_read_dcl_global_flags},
     {VKD3D_SM4_OP_LOD,                              VKD3DSIH_LOD,                              "f",    "fRS"},
-    {VKD3D_SM4_OP_GATHER4,                          VKD3DSIH_GATHER4,                          "u",    "fRS"},
+    {VKD3D_SM4_OP_GATHER4,                          VKD3DSIH_GATHER4,                          "u",    "fRs"},
     {VKD3D_SM4_OP_SAMPLE_POS,                       VKD3DSIH_SAMPLE_POS,                       "f",    "Ru"},
     {VKD3D_SM4_OP_SAMPLE_INFO,                      VKD3DSIH_SAMPLE_INFO,                      "f",    "R"},
     {VKD3D_SM5_OP_HS_DECLS,                         VKD3DSIH_HS_DECLS,                         "",     ""},
@@ -1353,9 +1354,9 @@ static const struct vkd3d_sm4_opcode_info opcode_table[] =
     {VKD3D_SM5_OP_DERIV_RTX_FINE,                   VKD3DSIH_DSX_FINE,                         "f",    "f"},
     {VKD3D_SM5_OP_DERIV_RTY_COARSE,                 VKD3DSIH_DSY_COARSE,                       "f",    "f"},
     {VKD3D_SM5_OP_DERIV_RTY_FINE,                   VKD3DSIH_DSY_FINE,                         "f",    "f"},
-    {VKD3D_SM5_OP_GATHER4_C,                        VKD3DSIH_GATHER4_C,                        "f",    "fRSf"},
-    {VKD3D_SM5_OP_GATHER4_PO,                       VKD3DSIH_GATHER4_PO,                       "f",    "fiRS"},
-    {VKD3D_SM5_OP_GATHER4_PO_C,                     VKD3DSIH_GATHER4_PO_C,                     "f",    "fiRSf"},
+    {VKD3D_SM5_OP_GATHER4_C,                        VKD3DSIH_GATHER4_C,                        "f",    "fRsf"},
+    {VKD3D_SM5_OP_GATHER4_PO,                       VKD3DSIH_GATHER4_PO,                       "f",    "fiRs"},
+    {VKD3D_SM5_OP_GATHER4_PO_C,                     VKD3DSIH_GATHER4_PO_C,                     "f",    "fiRsf"},
     {VKD3D_SM5_OP_RCP,                              VKD3DSIH_RCP,                              "f",    "f"},
     {VKD3D_SM5_OP_F32TOF16,                         VKD3DSIH_F32TOF16,                         "u",    "f"},
     {VKD3D_SM5_OP_F16TOF32,                         VKD3DSIH_F16TOF32,                         "f",    "u"},
@@ -1634,6 +1635,7 @@ static enum vkd3d_data_type map_data_type(char t)
         case 'R':
             return VKD3D_DATA_RESOURCE;
         case 'S':
+        case 's':
             return VKD3D_DATA_SAMPLER;
         case 'U':
             return VKD3D_DATA_UAV;
@@ -3735,7 +3737,8 @@ static void sm4_src_from_node(struct sm4_src_register *src,
         src->swizzle = hlsl_map_swizzle(hlsl_swizzle_from_writemask(writemask), map_writemask);
 }

-static void sm4_write_dst_register(const struct tpf_writer *tpf, const struct sm4_dst_register *dst)
+static void sm4_write_dst_register(const struct tpf_writer *tpf, const struct sm4_dst_register *dst,
+        char dst_info)
 {
     const struct vkd3d_sm4_register_type_info *register_type_info;
     struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
@@ -3770,7 +3773,7 @@ static void sm4_write_dst_register(const struct tpf_writer *tpf, const struct sm
     }
 }

-static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm4_src_register *src)
+static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm4_src_register *src, char src_info)
 {
     const struct vkd3d_sm4_register_type_info *register_type_info;
     struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
@@ -3790,6 +3793,8 @@ static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm
     }

reg_dim = src->reg.dim;
+    if (src_info == 's')
+        reg_dim = VKD3D_SM4_DIMENSION_VEC4;

token |= sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT;
     token |= src->reg.idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT;
@@ -3838,6 +3843,7 @@ static uint32_t sm4_register_order(const struct sm4_register *reg)

static void write_sm4_instruction(const struct tpf_writer *tpf, const struct sm4_instruction *instr)
 {
+    const struct vkd3d_sm4_opcode_info *opcode_info = get_opcode_info(instr->opcode);
     struct vkd3d_bytecode_buffer *buffer = tpf->buffer;
     uint32_t token = instr->opcode;
     unsigned int size = 1, i, j;
@@ -3866,10 +3872,10 @@ static void write_sm4_instruction(const struct tpf_writer *tpf, const struct sm4
     }

for (i = 0; i < instr->dst_count; ++i)
-        sm4_write_dst_register(tpf, &instr->dsts[i]);
+        sm4_write_dst_register(tpf, &instr->dsts[i], opcode_info ? opcode_info->dst_info[i] : ' ');

for (i = 0; i < instr->src_count; ++i)
-        sm4_write_src_register(tpf, &instr->srcs[i]);
+        sm4_write_src_register(tpf, &instr->srcs[i], opcode_info ? opcode_info->src_info[i] : ' ');

if (instr->byte_stride)
         put_u32(buffer, instr->byte_stride);
@@ -5157,7 +5163,6 @@ static void write_sm4_gather(const struct tpf_writer *tpf, const struct hlsl_ir_

src = &instr.srcs[instr.src_count++];
     sm4_src_from_deref(tpf->ctx, src, sampler, VKD3DSP_WRITEMASK_ALL);
-    src->reg.dim = VKD3D_SM4_DIMENSION_VEC4;
     src->swizzle_type = VKD3D_SM4_SWIZZLE_SCALAR;
     src->swizzle = swizzle;

[PATCH v6 9/9] vkd3d-shader/tpf: Get rid of sm4_register.dim.

[Francisco Casas]

From: Francisco Casas

---
 libs/vkd3d-shader/tpf.c | 114 +++++++++++++++++-----------------------
 1 file changed, 49 insertions(+), 65 deletions(-)

diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c
index 9359e4ff..67e853b1 100644
--- a/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d-shader/tpf.c
@@ -1483,6 +1483,7 @@ struct vkd3d_sm4_register_type_info
 {
     enum vkd3d_sm4_register_type sm4_type;
     enum vkd3d_shader_register_type vkd3d_type;
+    enum vkd3d_sm4_dimension default_dimension;
 };

static const enum vkd3d_shader_register_precision register_precision_table[] =
@@ -1522,42 +1523,42 @@ static void init_sm4_lookup_tables(struct vkd3d_sm4_lookup_tables *lookup)

static const struct vkd3d_sm4_register_type_info register_type_table[] =
     {
-        {VKD3D_SM4_RT_TEMP,                    VKD3DSPR_TEMP},
-        {VKD3D_SM4_RT_INPUT,                   VKD3DSPR_INPUT},
-        {VKD3D_SM4_RT_OUTPUT,                  VKD3DSPR_OUTPUT},
-        {VKD3D_SM4_RT_INDEXABLE_TEMP,          VKD3DSPR_IDXTEMP},
-        {VKD3D_SM4_RT_IMMCONST,                VKD3DSPR_IMMCONST},
-        {VKD3D_SM4_RT_IMMCONST64,              VKD3DSPR_IMMCONST64},
-        {VKD3D_SM4_RT_SAMPLER,                 VKD3DSPR_SAMPLER},
-        {VKD3D_SM4_RT_RESOURCE,                VKD3DSPR_RESOURCE},
-        {VKD3D_SM4_RT_CONSTBUFFER,             VKD3DSPR_CONSTBUFFER},
-        {VKD3D_SM4_RT_IMMCONSTBUFFER,          VKD3DSPR_IMMCONSTBUFFER},
-        {VKD3D_SM4_RT_PRIMID,                  VKD3DSPR_PRIMID},
-        {VKD3D_SM4_RT_DEPTHOUT,                VKD3DSPR_DEPTHOUT},
-        {VKD3D_SM4_RT_NULL,                    VKD3DSPR_NULL},
-        {VKD3D_SM4_RT_RASTERIZER,              VKD3DSPR_RASTERIZER},
-        {VKD3D_SM4_RT_OMASK,                   VKD3DSPR_SAMPLEMASK},
-        {VKD3D_SM5_RT_STREAM,                  VKD3DSPR_STREAM},
-        {VKD3D_SM5_RT_FUNCTION_BODY,           VKD3DSPR_FUNCTIONBODY},
-        {VKD3D_SM5_RT_FUNCTION_POINTER,        VKD3DSPR_FUNCTIONPOINTER},
-        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT_ID, VKD3DSPR_OUTPOINTID},
-        {VKD3D_SM5_RT_FORK_INSTANCE_ID,        VKD3DSPR_FORKINSTID},
-        {VKD3D_SM5_RT_JOIN_INSTANCE_ID,        VKD3DSPR_JOININSTID},
-        {VKD3D_SM5_RT_INPUT_CONTROL_POINT,     VKD3DSPR_INCONTROLPOINT},
-        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT,    VKD3DSPR_OUTCONTROLPOINT},
-        {VKD3D_SM5_RT_PATCH_CONSTANT_DATA,     VKD3DSPR_PATCHCONST},
-        {VKD3D_SM5_RT_DOMAIN_LOCATION,         VKD3DSPR_TESSCOORD},
-        {VKD3D_SM5_RT_UAV,                     VKD3DSPR_UAV},
-        {VKD3D_SM5_RT_SHARED_MEMORY,           VKD3DSPR_GROUPSHAREDMEM},
-        {VKD3D_SM5_RT_THREAD_ID,               VKD3DSPR_THREADID},
-        {VKD3D_SM5_RT_THREAD_GROUP_ID,         VKD3DSPR_THREADGROUPID},
-        {VKD3D_SM5_RT_LOCAL_THREAD_ID,         VKD3DSPR_LOCALTHREADID},
-        {VKD3D_SM5_RT_COVERAGE,                VKD3DSPR_COVERAGE},
-        {VKD3D_SM5_RT_LOCAL_THREAD_INDEX,      VKD3DSPR_LOCALTHREADINDEX},
-        {VKD3D_SM5_RT_GS_INSTANCE_ID,          VKD3DSPR_GSINSTID},
-        {VKD3D_SM5_RT_DEPTHOUT_GREATER_EQUAL,  VKD3DSPR_DEPTHOUTGE},
-        {VKD3D_SM5_RT_DEPTHOUT_LESS_EQUAL,     VKD3DSPR_DEPTHOUTLE},
-        {VKD3D_SM5_RT_OUTPUT_STENCIL_REF,      VKD3DSPR_OUTSTENCILREF},
+        {VKD3D_SM4_RT_TEMP,                    VKD3DSPR_TEMP,            VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_INPUT,                   VKD3DSPR_INPUT,           VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_OUTPUT,                  VKD3DSPR_OUTPUT,          VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_INDEXABLE_TEMP,          VKD3DSPR_IDXTEMP,         VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_IMMCONST,                VKD3DSPR_IMMCONST,        0},
+        {VKD3D_SM4_RT_IMMCONST64,              VKD3DSPR_IMMCONST64,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_SAMPLER,                 VKD3DSPR_SAMPLER,         VKD3D_SM4_DIMENSION_NONE},
+        {VKD3D_SM4_RT_RESOURCE,                VKD3DSPR_RESOURCE,        VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_CONSTBUFFER,             VKD3DSPR_CONSTBUFFER,     VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_IMMCONSTBUFFER,          VKD3DSPR_IMMCONSTBUFFER,  VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_PRIMID,                  VKD3DSPR_PRIMID,          VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_DEPTHOUT,                VKD3DSPR_DEPTHOUT,        VKD3D_SM4_DIMENSION_SCALAR},
+        {VKD3D_SM4_RT_NULL,                    VKD3DSPR_NULL,            VKD3D_SM4_DIMENSION_NONE},
+        {VKD3D_SM4_RT_RASTERIZER,              VKD3DSPR_RASTERIZER,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM4_RT_OMASK,                   VKD3DSPR_SAMPLEMASK,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_STREAM,                  VKD3DSPR_STREAM,          VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_FUNCTION_BODY,           VKD3DSPR_FUNCTIONBODY,    VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_FUNCTION_POINTER,        VKD3DSPR_FUNCTIONPOINTER, VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT_ID, VKD3DSPR_OUTPOINTID,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_FORK_INSTANCE_ID,        VKD3DSPR_FORKINSTID,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_JOIN_INSTANCE_ID,        VKD3DSPR_JOININSTID,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_INPUT_CONTROL_POINT,     VKD3DSPR_INCONTROLPOINT,  VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_OUTPUT_CONTROL_POINT,    VKD3DSPR_OUTCONTROLPOINT, VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_PATCH_CONSTANT_DATA,     VKD3DSPR_PATCHCONST,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_DOMAIN_LOCATION,         VKD3DSPR_TESSCOORD,       VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_UAV,                     VKD3DSPR_UAV,             VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_SHARED_MEMORY,           VKD3DSPR_GROUPSHAREDMEM,  VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_THREAD_ID,               VKD3DSPR_THREADID,        VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_THREAD_GROUP_ID,         VKD3DSPR_THREADGROUPID,   VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_LOCAL_THREAD_ID,         VKD3DSPR_LOCALTHREADID,   VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_COVERAGE,                VKD3DSPR_COVERAGE,        VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_LOCAL_THREAD_INDEX,      VKD3DSPR_LOCALTHREADINDEX,VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_GS_INSTANCE_ID,          VKD3DSPR_GSINSTID,        VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_DEPTHOUT_GREATER_EQUAL,  VKD3DSPR_DEPTHOUTGE,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_DEPTHOUT_LESS_EQUAL,     VKD3DSPR_DEPTHOUTLE,      VKD3D_SM4_DIMENSION_VEC4},
+        {VKD3D_SM5_RT_OUTPUT_STENCIL_REF,      VKD3DSPR_OUTSTENCILREF,   VKD3D_SM4_DIMENSION_VEC4},
     };

memset(lookup, 0, sizeof(*lookup));
@@ -3494,7 +3495,7 @@ struct sm4_register
     enum vkd3d_shader_register_type type;
     struct vkd3d_shader_register_index idx[2];
     unsigned int idx_count;
-    enum vkd3d_sm4_dimension dim;
+    enum vkd3d_immconst_type immconst_type;
     uint32_t immconst_uint[4];
     unsigned int mod;
 };
@@ -3541,7 +3542,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         if (regset == HLSL_REGSET_TEXTURES)
         {
             reg->type = VKD3DSPR_RESOURCE;
-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
             reg->idx[0].offset = var->regs[HLSL_REGSET_TEXTURES].id;
@@ -3553,7 +3553,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         else if (regset == HLSL_REGSET_UAVS)
         {
             reg->type = VKD3DSPR_UAV;
-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
             reg->idx[0].offset = var->regs[HLSL_REGSET_UAVS].id;
@@ -3565,7 +3564,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
         else if (regset == HLSL_REGSET_SAMPLERS)
         {
             reg->type = VKD3DSPR_SAMPLER;
-            reg->dim = VKD3D_SM4_DIMENSION_NONE;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
             reg->idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].id;
@@ -3580,7 +3578,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

assert(data_type->class <= HLSL_CLASS_VECTOR);
             reg->type = VKD3DSPR_CONSTBUFFER;
-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
             reg->idx[0].offset = var->buffer->reg.id;
@@ -3603,7 +3600,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
                 reg->idx_count = 1;
             }

-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             *writemask = ((1u << data_type->dimx) - 1) << (offset % 4);
         }
         else
@@ -3612,7 +3608,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

assert(hlsl_reg.allocated);
             reg->type = VKD3DSPR_INPUT;
-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             if (swizzle_type)
                 *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
             reg->idx[0].offset = hlsl_reg.id;
@@ -3634,10 +3629,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
                 reg->idx_count = 1;
             }

-            if (reg->type == VKD3DSPR_DEPTHOUT)
-                reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
-            else
-                reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             *writemask = ((1u << data_type->dimx) - 1) << (offset % 4);
         }
         else
@@ -3646,7 +3637,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

assert(hlsl_reg.allocated);
             reg->type = VKD3DSPR_OUTPUT;
-            reg->dim = VKD3D_SM4_DIMENSION_VEC4;
             reg->idx[0].offset = hlsl_reg.id;
             reg->idx_count = 1;
             *writemask = hlsl_reg.writemask;
@@ -3658,7 +3648,6 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r

assert(hlsl_reg.allocated);
         reg->type = VKD3DSPR_TEMP;
-        reg->dim = VKD3D_SM4_DIMENSION_VEC4;
         if (swizzle_type)
             *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
         reg->idx[0].offset = hlsl_reg.id;
@@ -3682,7 +3671,6 @@ static void sm4_register_from_node(struct sm4_register *reg, unsigned int *write
 {
     assert(instr->reg.allocated);
     reg->type = VKD3DSPR_TEMP;
-    reg->dim = VKD3D_SM4_DIMENSION_VEC4;
     *swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
     reg->idx[0].offset = instr->reg.id;
     reg->idx_count = 1;
@@ -3703,14 +3691,14 @@ static void sm4_src_from_constant_value(struct sm4_src_register *src,
     src->reg.type = VKD3DSPR_IMMCONST;
     if (width == 1)
     {
-        src->reg.dim = VKD3D_SM4_DIMENSION_SCALAR;
+        src->reg.immconst_type = VKD3D_IMMCONST_SCALAR;
         src->reg.immconst_uint[0] = value->u[0].u;
     }
     else
     {
         unsigned int i, j = 0;

-        src->reg.dim = VKD3D_SM4_DIMENSION_VEC4;
+        src->reg.immconst_type = VKD3D_IMMCONST_VEC4;
         for (i = 0; i < 4; ++i)
         {
             if (map_writemask & (1u << i))
@@ -3751,14 +3739,14 @@ static void sm4_write_dst_register(const struct tpf_writer *tpf, const struct sm
     {
         FIXME("Unhandled vkd3d-shader register type %#x.\n", dst->reg.type);
         sm4_reg_type = VKD3D_SM4_RT_TEMP;
+        reg_dim = VKD3D_SM4_DIMENSION_VEC4;
     }
     else
     {
         sm4_reg_type = register_type_info->sm4_type;
+        reg_dim = register_type_info->default_dimension;
     }

-    reg_dim = dst->reg.dim;
-
     token |= sm4_reg_type << VKD3D_SM4_REGISTER_TYPE_SHIFT;
     token |= dst->reg.idx_count << VKD3D_SM4_REGISTER_ORDER_SHIFT;
     token |= reg_dim << VKD3D_SM4_DIMENSION_SHIFT;
@@ -3786,13 +3774,16 @@ static void sm4_write_src_register(const struct tpf_writer *tpf, const struct sm
     {
         FIXME("Unhandled vkd3d-shader register type %#x.\n", src->reg.type);
         sm4_reg_type = VKD3D_SM4_RT_TEMP;
+        reg_dim = VKD3D_SM4_DIMENSION_VEC4;
     }
     else
     {
         sm4_reg_type = register_type_info->sm4_type;
+        reg_dim = register_type_info->default_dimension;
     }

-    reg_dim = src->reg.dim;
+    if (src->reg.type == VKD3DSPR_IMMCONST)
+        reg_dim = src->reg.immconst_type == VKD3D_IMMCONST_VEC4 ? VKD3D_SM4_DIMENSION_VEC4 : VKD3D_SM4_DIMENSION_SCALAR;
     if (src_info == 's')
         reg_dim = VKD3D_SM4_DIMENSION_VEC4;

@@ -3834,7 +3825,7 @@ static uint32_t sm4_register_order(const struct sm4_register *reg)
 {
     uint32_t order = 1;
     if (reg->type == VKD3DSPR_IMMCONST)
-        order += reg->dim == VKD3D_SM4_DIMENSION_VEC4 ? 4 : 1;
+        order += reg->immconst_type == VKD3D_IMMCONST_VEC4 ? 4 : 1;
     order += reg->idx_count;
     if (reg->mod)
         ++order;
@@ -3917,7 +3908,6 @@ static void write_sm4_dcl_constant_buffer(const struct tpf_writer *tpf, const st
     {
         .opcode = VKD3D_SM4_OP_DCL_CONSTANT_BUFFER,

-        .srcs[0].reg.dim = VKD3D_SM4_DIMENSION_VEC4,
         .srcs[0].reg.type = VKD3DSPR_CONSTBUFFER,
         .srcs[0].reg.idx[0].offset = cbuffer->reg.id,
         .srcs[0].reg.idx[1].offset = (cbuffer->used_size + 3) / 4,
@@ -4025,7 +4015,6 @@ static void write_sm4_dcl_semantic(const struct tpf_writer *tpf, const struct hl

struct sm4_instruction instr =
     {
-        .dsts[0].reg.dim = VKD3D_SM4_DIMENSION_VEC4,
         .dst_count = 1,
     };

@@ -4050,9 +4039,6 @@ static void write_sm4_dcl_semantic(const struct tpf_writer *tpf, const struct hl
         instr.dsts[0].writemask = var->regs[HLSL_REGSET_NUMERIC].writemask;
     }

-    if (instr.dsts[0].reg.type == VKD3DSPR_DEPTHOUT)
-        instr.dsts[0].reg.dim = VKD3D_SM4_DIMENSION_SCALAR;
-
     hlsl_sm4_usage_from_semantic(tpf->ctx, &var->semantic, output, &usage);
     if (usage == ~0u)
         usage = D3D_NAME_UNDEFINED;
@@ -4184,7 +4170,6 @@ static void write_sm4_unary_op_with_two_destinations(const struct tpf_writer *tp
     sm4_dst_from_node(&instr.dsts[dst_idx], dst);
     assert(1 - dst_idx >= 0);
     instr.dsts[1 - dst_idx].reg.type = VKD3DSPR_NULL;
-    instr.dsts[1 - dst_idx].reg.dim = VKD3D_SM4_DIMENSION_NONE;
     instr.dsts[1 - dst_idx].reg.idx_count = 0;
     instr.dst_count = 2;

@@ -4244,7 +4229,6 @@ static void write_sm4_binary_op_with_two_destinations(const struct tpf_writer *t
     sm4_dst_from_node(&instr.dsts[dst_idx], dst);
     assert(1 - dst_idx >= 0);
     instr.dsts[1 - dst_idx].reg.type = VKD3DSPR_NULL;
-    instr.dsts[1 - dst_idx].reg.dim = VKD3D_SM4_DIMENSION_NONE;
     instr.dsts[1 - dst_idx].reg.idx_count = 0;
     instr.dst_count = 2;

@@ -4316,7 +4300,7 @@ static void write_sm4_ld(const struct tpf_writer *tpf, const struct hlsl_ir_node
             memset(&instr.srcs[2], 0, sizeof(instr.srcs[2]));
             instr.srcs[2].swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
             reg->type = VKD3DSPR_IMMCONST;
-            reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
+            reg->immconst_type = VKD3D_IMMCONST_SCALAR;
             reg->immconst_uint[0] = index->value.u[0].u;
         }
         else if (tpf->ctx->profile->major_version == 4 && tpf->ctx->profile->minor_version == 0)
@@ -4433,7 +4417,7 @@ static void write_sm4_cast_from_bool(const struct tpf_writer *tpf, const struct
     sm4_src_from_node(&instr.srcs[0], arg, instr.dsts[0].writemask);
     instr.srcs[1].swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
     instr.srcs[1].reg.type = VKD3DSPR_IMMCONST;
-    instr.srcs[1].reg.dim = VKD3D_SM4_DIMENSION_SCALAR;
+    instr.srcs[1].reg.immconst_type = VKD3D_IMMCONST_SCALAR;
     instr.srcs[1].reg.immconst_uint[0] = mask;
     instr.src_count = 2;

[Francisco Casas (＠fcasas)]

> This essentially works for me, but note that it doesn't currently apply cleanly.

Yep, I have been told to leave the rebases for after the MR is agreed upon.
> I generally prefer doing changes like these function by function; start in write_sm4_shdr(), and then push down from there. That would also make potential conflicts easier to handle/avoid. I think this change is straightforward enough that I'm not going to require splitting the commit, but please take it into future consideration.

Okay, it didn't occur to me doing this from the top-down.
> That works, but it might be nice to introduce a tpf_writer_init() function as well, particularly once the lookup tables are introduced.

Done.
> This works too, but perhaps something like the following is nicer?
>    const struct vkd3d_sm4_register_type_info *info;
>    ...
>    for (i = 0; i < ARRAY_SIZE(register_type_table); ++i)
>    {
>        info = &register_type_table[i];
>        lookup->register_type_info_from_sm4[info->sm4_type] = info;
>        lookup->register_type_info_from_vkd3d[info->vkd3d_type] = info;
>    }

Done.

[Francisco Casas (＠fcasas)]

:arrow_up: Applied requested changes.