jacob/power_play
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

shade & blit to final target separately

Browse Source
This commit is contained in:
jacob 2025-07-20 10:40:53 -05:00
parent 762a7d83fe
commit c6f500ebf4
8 changed files with 266 additions and 176 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
res/sh/blit.hlsl
View File

@ -50,6 +50,17 @@ SH_ENTRY(ROOTSIG) struct vs_output vs(struct vs_input input)
return output;
}
/* ========================== *
* Tone map
* ========================== */
/* ACES approximation by Krzysztof Narkowicz
* https://knarkowicz.wordpress.com/2016/01/06/aces-filmic-tone-mapping-curve/ */
INLINE float3 tone_map(float3 v)
{
return saturate((v * (2.51f * v + 0.03f)) / (v * (2.43f * v + 0.59f) + 0.14f));
}
/* ========================== *
* Pixel shader
* ========================== */
@ -66,6 +77,18 @@ SH_ENTRY(ROOTSIG) struct ps_output ps(struct ps_input input)
{
struct ps_output output;
float4 color = g_textures[g_constants.tex_urid].Sample(g_sampler, input.vs.uv);
/* Apply tone map */
if (g_constants.flags & SH_BLIT_FLAG_TONE_MAP) {
/* TODO: Dynamic exposure based on average scene luminance */
color.rgb = tone_map(color.rgb) * g_constants.exposure;
}
/* Apply gamma correction */
if (g_constants.flags & SH_BLIT_FLAG_GAMMA_CORRECT) {
color = pow(abs(color), 1/g_constants.gamma);
}
output.SV_Target = color;
return output;
}

6
res/sh/flood.hlsl
View File

@ -39,7 +39,7 @@ SH_ENTRY(ROOTSIG) void cs(struct cs_input input)
if (id.x < tex_size.x && id.y < tex_size.y) {
Texture2D<float4> emittance_tex = g_emittance_textures[g_constants.emittance_tex_urid];
RWTexture2D<uint2> read_flood_tex = g_flood_textures[g_constants.read_flood_tex_urid];
RWTexture2D<uint2> write_flood_tex = g_flood_textures[g_constants.write_flood_tex_urid];
RWTexture2D<uint2> target_flood_tex = g_flood_textures[g_constants.target_flood_tex_urid];
int step_len = g_constants.step_len;
if (step_len == -1) {
/* Seed */
@ -48,7 +48,7 @@ SH_ENTRY(ROOTSIG) void cs(struct cs_input input)
if (emittance.a > 0) {
seed = id;
}
write_flood_tex[id] = seed;
target_flood_tex[id] = seed;
} else {
/* Flood */
int2 read_coords[9] = {
@ -76,7 +76,7 @@ SH_ENTRY(ROOTSIG) void cs(struct cs_input input)
}
}
}
write_flood_tex[id] = closest_seed;
target_flood_tex[id] = closest_seed;
}
}
}

100
res/sh/sh_common.h
View File

@ -4,8 +4,9 @@
#define SH_DECL(t, n) struct CAT(sh_, t) n
#define SH_DECLS(t, n) SH_DECL(t, n)
#define SH_ENTRY(rootsig) static
#define SH_ASSERT_32BIT(s, n) STATIC_ASSERT(sizeof(s) % 16 == 0); /* Root constant structs should pad to 16 byte alignment */ \
STATIC_ASSERT((sizeof(s) / 4) == n) /* Verify that struct size matches supplied 32 bit count */
#define SH_ASSERT_ROOT_CONST(s, n) STATIC_ASSERT(sizeof(s) % 16 == 0); /* Root constant struct should pad to 16 byte alignment */ \
STATIC_ASSERT((sizeof(s) / 4) == n); /* Root constant struct size should match the specified 32-bit-constant count */ \
STATIC_ASSERT((sizeof(s) <= 256)) /* Root constant struct can only fit 64 DWORDS */
struct sh_uint { u32 v; };
INLINE struct sh_uint sh_uint_from_u32(u32 v)
@ -20,15 +21,21 @@ INLINE struct sh_int sh_int_from_i32(i32 v)
}
struct sh_uint2 { u32 v[2]; };
INLINE struct sh_uint2 sh_uint2_from_u32(u32 a, u32 b)
INLINE struct sh_uint2 sh_uint2_from_u32(u32 x, u32 y)
{
return (struct sh_uint2) { .v[0] = a, .v[1] = b };
return (struct sh_uint2) { .v[0] = x, .v[1] = y };
}
struct sh_uint3 { u32 v[3]; };
INLINE struct sh_uint3 sh_uint3_from_u32(u32 a, u32 b, u32 c)
INLINE struct sh_uint3 sh_uint3_from_u32(u32 x, u32 y, u32 z)
{
return (struct sh_uint3) { .v[0] = a, .v[1] = b, .v[2] = c };
return (struct sh_uint3) { .v[0] = x, .v[1] = y, .v[2] = z };
}
struct sh_uint4 { u32 v[4]; };
INLINE struct sh_uint4 sh_uint4_from_u32(u32 x, u32 y, u32 z, u32 w)
{
return (struct sh_uint4) { .v[0] = x, .v[1] = y, .v[2] = z, .v[3] = w };
}
struct sh_float { f32 v; };
@ -73,7 +80,7 @@ INLINE struct sh_float2x3 sh_float2x3_from_xform(struct xform v)
#define SH_DECL(t, n) t n
#define SH_DECLS(t, n) t n : n
#define SH_ENTRY(rootsig) [RootSignature(rootsig)]
#define SH_ASSERT_32BIT(s, n)
#define SH_ASSERT_ROOT_CONST(s, n)
#endif
@ -92,11 +99,11 @@ INLINE struct sh_float2x3 sh_float2x3_from_xform(struct xform v)
* ========================== */
SH_STRUCT(sh_material_constants {
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(float4x4, projection); /* 16 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_material_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_material_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_STRUCT(sh_material_instance {
SH_DECL(int, tex_nurid);
@ -125,55 +132,52 @@ SH_STRUCT(sh_material_grid {
* ========================== */
SH_STRUCT(sh_flood_constants {
/* ==================================================== */
SH_DECL(int, step_len); /* 04 consts */
SH_DECL(uint, emittance_tex_urid); /* 04 consts */
SH_DECL(uint, read_flood_tex_urid); /* 04 consts */
SH_DECL(uint, write_flood_tex_urid); /* 04 consts */
/* ==================================================== */
SH_DECL(uint, tex_width); /* 04 consts */
SH_DECL(uint, tex_height); /* 04 consts */
SH_DECL(uint2, _pad0); /* 08 consts (padding) */
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(int, step_len); /* 01 consts */
SH_DECL(uint, emittance_tex_urid); /* 01 consts */
SH_DECL(uint, read_flood_tex_urid); /* 01 consts */
SH_DECL(uint, target_flood_tex_urid); /* 01 consts */
/* ---------------------------------------------------- */
SH_DECL(uint, tex_width); /* 01 consts */
SH_DECL(uint, tex_height); /* 01 consts */
SH_DECL(uint2, _pad0); /* 02 consts (padding) */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_flood_constants, 8); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_flood_constants, 8); /* Expected to match num32BitConstants in shader's root signature */
/* ========================== *
* Shade shader structures
* ========================== */
SH_STRUCT(sh_shade_constants {
/* ==================================================== */
SH_DECL(uint, frame_index); /* 01 consts */
SH_DECL(uint3, frame_seed); /* 03 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(uint4, frame_seed); /* 04 consts */
/* ---------------------------------------------------- */
SH_DECL(float2, camera_offset); /* 02 consts */
SH_DECL(uint, frame_index); /* 01 consts */
SH_DECL(uint, albedo_tex_urid); /* 01 consts */
SH_DECL(uint, _pad0); /* 01 consts (padding) */
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(uint, emittance_tex_urid); /* 01 consts */
SH_DECL(uint, emittance_flood_tex_urid); /* 01 consts */
SH_DECL(uint, write_tex_urid); /* 01 consts */
SH_DECL(uint, target_tex_urid); /* 01 consts */
SH_DECL(uint, tex_width); /* 01 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(uint, tex_height); /* 01 consts */
SH_DECL(float, exposure); /* 01 consts */
SH_DECL(float, gamma); /* 01 consts */
SH_DECL(uint, _pad1); /* 01 consts (padding) */
/* ==================================================== */
SH_DECL(uint3, _pad0); /* 03 consts (padding) */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_shade_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_shade_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
/* ========================== *
* Shape shader structures
* ========================== */
SH_STRUCT(sh_shape_constants {
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(float4x4, projection); /* 16 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_shape_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_shape_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_STRUCT(sh_shape_vert {
SH_DECLS(float2, pos);
@ -185,11 +189,11 @@ SH_STRUCT(sh_shape_vert {
* ========================== */
SH_STRUCT(sh_ui_constants {
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(float4x4, projection); /* 16 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_ui_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_ui_constants, 16); /* Expected to match num32BitConstants in shader's root signature */
SH_STRUCT(sh_ui_instance {
SH_DECL(int, tex_nurid);
@ -204,12 +208,18 @@ SH_STRUCT(sh_ui_instance {
* Blit shader structures
* ========================== */
#define SH_BLIT_FLAG_NONE (0 << 0)
#define SH_BLIT_FLAG_TONE_MAP (1 << 0)
#define SH_BLIT_FLAG_GAMMA_CORRECT (1 << 1)
SH_STRUCT(sh_blit_constants {
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(float4x4, projection); /* 16 consts */
/* ==================================================== */
/* ---------------------------------------------------- */
SH_DECL(uint, flags); /* 01 consts */
SH_DECL(uint, tex_urid); /* 01 consts */
SH_DECL(uint3, _pad0); /* 03 consts (padding) */
/* ==================================================== */
SH_DECL(float, exposure); /* 01 consts */
SH_DECL(float, gamma); /* 01 consts */
/* ---------------------------------------------------- */
});
SH_ASSERT_32BIT(struct sh_blit_constants, 20); /* Expected to match num32BitConstants in shader's root signature */
SH_ASSERT_ROOT_CONST(struct sh_blit_constants, 20); /* Expected to match num32BitConstants in shader's root signature */

24
res/sh/shade.hlsl
View File

@ -15,7 +15,7 @@ ConstantBuffer<struct sh_shade_constants> g_constants : register(b0);
Texture2D<float4> g_textures_float4[] : register(t0, space0);
Texture2D<uint2> g_textures_uint2[] : register(t0, space1);
Texture3D<uint> g_noise_textures[] : register(t0, space2);
RWTexture2D<float4> g_write_textures[]: register(u0, space3);
RWTexture2D<float4> g_target_textures[]: register(u0, space3);
struct cs_input {
DECLS(uint3, SV_DispatchThreadID);
@ -82,17 +82,6 @@ INLINE float4 get_light_at_pos(int2 pos)
return result;
}
/* ========================== *
* Tone map
* ========================== */
/* ACES approximation by Krzysztof Narkowicz
* https://knarkowicz.wordpress.com/2016/01/06/aces-filmic-tone-mapping-curve/ */
INLINE float3 tone_map(float3 v)
{
return saturate((v * (2.51f * v + 0.03f)) / (v * (2.43f * v + 0.59f) + 0.14f));
}
/* ========================== *
* Entry point
* ========================== */
@ -103,7 +92,7 @@ SH_ENTRY(ROOTSIG) void cs(struct cs_input input)
uint2 id = input.SV_DispatchThreadID.xy;
if (id.x < g_constants.tex_width && id.y < g_constants.tex_height) {
Texture2D<float4> albedo_tex = g_textures_float4[g_constants.albedo_tex_urid];
RWTexture2D<float4> write_tex = g_write_textures[g_constants.write_tex_urid];
RWTexture2D<float4> target_tex = g_target_textures[g_constants.target_tex_urid];
float4 color = float4(1, 1, 1, 1);
/* Apply albedo */
@ -112,13 +101,6 @@ SH_ENTRY(ROOTSIG) void cs(struct cs_input input)
/* Apply lighting */
color *= get_light_at_pos(id);
/* Apply tone map */
/* TODO: Dynamic exposure based on average scene luminance */
color.rgb = tone_map(color.rgb) * g_constants.exposure;
/* Apply gamma correction */
color = pow(abs(color), 1/g_constants.gamma);
write_tex[id] = color;
target_tex[id] = color;
}
}

1
src/gp.h
View File

@ -107,7 +107,6 @@ struct gp_render_params {
struct v2i32 draw_target_size;
struct rect draw_target_viewport;
struct xform draw_target_view;
b32 clear_target;
};
struct gp_render_sig *gp_render_sig_alloc(void);

253
src/gp_dx12.c
View File

@ -1869,14 +1869,14 @@ struct dx12_resource_barrier_desc {
enum D3D12_RESOURCE_STATES new_state; /* 0 if type != D3D12_RESOURCE_BARRIER_TYPE_TRANSITION */
};
INTERNAL void dx12_resource_barriers(ID3D12GraphicsCommandList *cl, i32 num_barriers, struct dx12_resource_barrier_desc *descs)
INTERNAL void dx12_resource_barriers(ID3D12GraphicsCommandList *cl, i32 num_descs, struct dx12_resource_barrier_desc *descs)
{
__prof;
struct arena_temp scratch = scratch_begin_no_conflict();
i32 num_rbs = 0;
struct D3D12_RESOURCE_BARRIER *rbs = arena_push_array_no_zero(scratch.arena, struct D3D12_RESOURCE_BARRIER, num_barriers);
for (i32 i = 0; i < num_barriers; ++i) {
struct D3D12_RESOURCE_BARRIER *rbs = arena_push_array_no_zero(scratch.arena, struct D3D12_RESOURCE_BARRIER, num_descs);
for (i32 i = 0; i < num_descs; ++i) {
struct dx12_resource_barrier_desc *desc = &descs[i];
struct dx12_resource *resource = desc->resource;
enum D3D12_RESOURCE_BARRIER_TYPE type = desc->type;
@ -1901,7 +1901,7 @@ INTERNAL void dx12_resource_barriers(ID3D12GraphicsCommandList *cl, i32 num_barr
rb->Flags = 0;
rb->UAV.pResource = resource->resource;
} else {
/* Unknown barrier */
/* Unknown barrier type */
ASSERT(0);
}
}
@ -2787,11 +2787,13 @@ struct render_sig {
struct arena *material_grid_descs_arena;
/* Resources */
struct dx12_resource *final_target;
struct dx12_resource *albedo;
struct dx12_resource *emittance;
struct dx12_resource *emittance_flood_a;
struct dx12_resource *emittance_flood_b;
struct dx12_resource *emittance_flood_read;
struct dx12_resource *emittance_flood_target;
struct dx12_resource *shade_read;
struct dx12_resource *shade_target;
struct dx12_resource *final_target;
};
struct material_instance_desc {
@ -2947,32 +2949,35 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
__prof;
struct arena_temp scratch = scratch_begin_no_conflict();
struct render_sig *sig = (struct render_sig *)render_sig;
++sig->frame_index;
struct v2i32 final_target_size = params.draw_target_size;
final_target_size.x = max_i32(final_target_size.x, 1);
final_target_size.y = max_i32(final_target_size.y, 1);
++sig->frame_index;
/* Release sig resources if size changed */
/* Release buffers if size changed */
if (sig->final_target && !v2i32_eq(final_target_size, sig->final_target->texture_size)) {
__profn("Release sig resources");
fenced_release(sig->final_target, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->albedo, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance_flood_a, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance_flood_b, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->albedo, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance_flood_read, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->emittance_flood_target, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->shade_read, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->shade_target, FENCED_RELEASE_KIND_RESOURCE);
fenced_release(sig->final_target, FENCED_RELEASE_KIND_RESOURCE);
sig->final_target = 0;
}
/* Allocate sig resources */
/* Allocate buffers */
if (!sig->final_target) {
__profn("Allocate sig resources");
sig->final_target = gbuff_alloc(DXGI_FORMAT_R8G8B8A8_UNORM, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
sig->albedo = gbuff_alloc(DXGI_FORMAT_R8G8B8A8_UNORM, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
sig->emittance = gbuff_alloc(DXGI_FORMAT_R16G16B16A16_FLOAT, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
sig->emittance_flood_a = gbuff_alloc(DXGI_FORMAT_R16G16_UINT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->emittance_flood_b = gbuff_alloc(DXGI_FORMAT_R16G16_UINT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->albedo = gbuff_alloc(DXGI_FORMAT_R8G8B8A8_UNORM, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
sig->emittance = gbuff_alloc(DXGI_FORMAT_R16G16B16A16_FLOAT, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
sig->emittance_flood_read = gbuff_alloc(DXGI_FORMAT_R16G16_UINT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->emittance_flood_target = gbuff_alloc(DXGI_FORMAT_R16G16_UINT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->shade_read = gbuff_alloc(DXGI_FORMAT_R16G16B16A16_FLOAT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->shade_target = gbuff_alloc(DXGI_FORMAT_R16G16B16A16_FLOAT, final_target_size, D3D12_RESOURCE_STATE_UNORDERED_ACCESS);
sig->final_target = gbuff_alloc(DXGI_FORMAT_R8G8B8A8_UNORM, final_target_size, D3D12_RESOURCE_STATE_RENDER_TARGET);
}
struct sprite_scope *sprite_scope = sprite_scope_begin();
@ -2980,6 +2985,7 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
struct pipeline *material_pipeline = pipeline_from_name(pipeline_scope, LIT("material"));
struct pipeline *flood_pipeline = pipeline_from_name(pipeline_scope, LIT("flood"));
struct pipeline *shade_pipeline = pipeline_from_name(pipeline_scope, LIT("shade"));
struct pipeline *blit_pipeline = pipeline_from_name(pipeline_scope, LIT("blit"));
struct pipeline *ui_pipeline = pipeline_from_name(pipeline_scope, LIT("ui"));
struct pipeline *shape_pipeline = pipeline_from_name(pipeline_scope, LIT("shape"));
struct command_queue *cq = G.command_queues[DX12_QUEUE_DIRECT];
@ -2988,6 +2994,7 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
__profn("Run render");
__profnc_dx12(cl->cq->prof, cl->cl, "Run render", RGB32_F(0.5, 0.2, 0.2));
struct mat4x4 world_vp_matrix = calculate_vp(params.draw_target_view, params.draw_target_viewport.width, params.draw_target_viewport.height);
struct mat4x4 blit_vp_matrix = calculate_vp(XFORM_TRS(.t = v2_mul(V2(final_target_size.x, final_target_size.y), 0.5), .s = V2(final_target_size.x, final_target_size.y)), params.draw_target_viewport.width, params.draw_target_viewport.height);
struct mat4x4 ui_vp_matrix = calculate_vp(XFORM_IDENT, params.draw_target_viewport.width, params.draw_target_viewport.height);
/* Upload dummmy vert & index buffer */
@ -3085,27 +3092,29 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
ID3D12DescriptorHeap *heaps[] = { descriptor_heap->heap };
ID3D12GraphicsCommandList_SetDescriptorHeaps(cl->cl, countof(heaps), heaps);
/* Bind gbuffers */
/* Prep for material pass */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->albedo, D3D12_RESOURCE_STATE_RENDER_TARGET },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_RENDER_TARGET }
};
D3D12_CPU_DESCRIPTOR_HANDLE rtvs[] = {
sig->albedo->rtv_descriptor->handle,
sig->emittance->rtv_descriptor->handle,
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
ID3D12GraphicsCommandList_OMSetRenderTargets(cl->cl, countof(rtvs), rtvs, 0, 0);
}
/* Clear gbuffers */
{
__profn("Clear gbuffers");
__profnc_dx12(cl->cq->prof, cl->cl, "Clear gbuffers", RGB32_F(0.5, 0.2, 0.2));
f32 clear_color[] = { 0.0f, 0.0f, 0.0f, 0.0f };
ID3D12GraphicsCommandList_ClearRenderTargetView(cl->cl, sig->albedo->rtv_descriptor->handle, clear_color, 0, 0);
ID3D12GraphicsCommandList_ClearRenderTargetView(cl->cl, sig->emittance->rtv_descriptor->handle, clear_color, 0, 0);
/* Barrier */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->albedo, D3D12_RESOURCE_STATE_RENDER_TARGET },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_RENDER_TARGET }
};
D3D12_CPU_DESCRIPTOR_HANDLE rtvs[] = {
sig->albedo->rtv_descriptor->handle,
sig->emittance->rtv_descriptor->handle,
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
ID3D12GraphicsCommandList_OMSetRenderTargets(cl->cl, countof(rtvs), rtvs, 0, 0);
}
/* Clear */
{
__profn("Clear gbuffers");
__profnc_dx12(cl->cq->prof, cl->cl, "Clear gbuffers", RGB32_F(0.5, 0.2, 0.2));
f32 clear_color[] = { 0.0f, 0.0f, 0.0f, 0.0f };
ID3D12GraphicsCommandList_ClearRenderTargetView(cl->cl, sig->albedo->rtv_descriptor->handle, clear_color, 0, 0);
ID3D12GraphicsCommandList_ClearRenderTargetView(cl->cl, sig->emittance->rtv_descriptor->handle, clear_color, 0, 0);
}
}
/* Material pass */
@ -3143,16 +3152,20 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
ID3D12GraphicsCommandList_DrawIndexedInstanced(cl->cl, 6, instance_count, 0, 0, 0);
}
/* Transition emittance & emittance flood */
struct dx12_resource *emittance_flood_read = sig->emittance_flood_a;
struct dx12_resource *emittance_flood_write = sig->emittance_flood_a;
/* Prep for flood pass */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, emittance_flood_read, D3D12_RESOURCE_STATE_UNORDERED_ACCESS },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, emittance_flood_write, D3D12_RESOURCE_STATE_UNORDERED_ACCESS }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
/* Barrier */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance_flood_read, D3D12_RESOURCE_STATE_UNORDERED_ACCESS },
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->emittance_flood_read, 0 },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance_flood_target, D3D12_RESOURCE_STATE_UNORDERED_ACCESS }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
}
}
/* Flood pass */
@ -3176,7 +3189,7 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
/* UAV barrier */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, emittance_flood_read, 0 }
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->emittance_flood_read, 0 }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
}
@ -3185,8 +3198,8 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
struct sh_flood_constants constants = ZI;
constants.step_len = sh_int_from_i32(step_length);
constants.emittance_tex_urid = sh_uint_from_u32(sig->emittance->srv_descriptor->index);
constants.read_flood_tex_urid = sh_uint_from_u32(emittance_flood_read->uav_descriptor->index);
constants.write_flood_tex_urid = sh_uint_from_u32(emittance_flood_write->uav_descriptor->index);
constants.read_flood_tex_urid = sh_uint_from_u32(sig->emittance_flood_read->uav_descriptor->index);
constants.target_flood_tex_urid = sh_uint_from_u32(sig->emittance_flood_target->uav_descriptor->index);
constants.tex_width = sh_uint_from_u32(final_target_size.x);
constants.tex_height = sh_uint_from_u32(final_target_size.y);
@ -3199,14 +3212,13 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
ID3D12GraphicsCommandList_Dispatch(cl->cl, (final_target_size.x + 7) / 8, (final_target_size.y + 7) / 8, 1);
/* Swap buffers */
struct dx12_resource *swp = emittance_flood_read;
emittance_flood_read = emittance_flood_write;
emittance_flood_write = swp;
struct dx12_resource *swp = sig->emittance_flood_read;
sig->emittance_flood_read = sig->emittance_flood_target;
sig->emittance_flood_target = swp;
/* Update step */
if (step_length == -1) {
step_length = max_i32(final_target_size.x, final_target_size.y) / 2;
//step_length = 16;
} else {
step_length /= 2;
}
@ -3214,24 +3226,31 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
}
}
/* Transition gbuffers & final target */
/* Prep for shade pass */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->albedo, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, emittance_flood_read, 0 },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, emittance_flood_read, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->final_target, D3D12_RESOURCE_STATE_UNORDERED_ACCESS }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
}
/* Barrier */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->albedo, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance, D3D12_RESOURCE_STATE_ALL_SHADER_RESOURCE },
/* Clear final target */
if (params.clear_target) {
__profn("Clear target");
__profnc_dx12(cl->cq->prof, cl->cl, "Clear target", RGB32_F(0.5, 0.2, 0.2));
f32 clear_color[] = { 0.0f, 0.0f, 0.0f, 0.0f };
ID3D12GraphicsCommandList_ClearUnorderedAccessViewFloat(cl->cl, gpu_handle_from_descriptor(sig->final_target->uav_descriptor, descriptor_heap), sig->final_target->uav_descriptor->handle, sig->final_target->resource, clear_color, 0, 0);
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->emittance_flood_read, D3D12_RESOURCE_STATE_UNORDERED_ACCESS },
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->emittance_flood_read, 0 },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->shade_read, D3D12_RESOURCE_STATE_UNORDERED_ACCESS },
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->shade_read, 0 },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->shade_target, D3D12_RESOURCE_STATE_UNORDERED_ACCESS }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
}
/* Clear */
{
__profn("Clear shade target");
__profnc_dx12(cl->cq->prof, cl->cl, "Clear shade target", RGB32_F(0.5, 0.2, 0.2));
f32 clear_color[] = { 0.0f, 0.0f, 0.0f, 0.0f };
ID3D12GraphicsCommandList_ClearUnorderedAccessViewFloat(cl->cl, gpu_handle_from_descriptor(sig->shade_target->uav_descriptor, descriptor_heap), sig->shade_target->uav_descriptor->handle, sig->shade_target->resource, clear_color, 0, 0);
}
}
/* Shade pass */
@ -3245,18 +3264,18 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
/* Set constants */
struct sh_shade_constants constants = ZI;
/* TODO: Remove this */
constants.frame_seed = sh_uint4_from_u32((u32)rand_u64_from_state(&sig->rand),
(u32)rand_u64_from_state(&sig->rand),
(u32)rand_u64_from_state(&sig->rand),
(u32)rand_u64_from_state(&sig->rand));
constants.frame_index = sh_uint_from_u32(sig->frame_index);
constants.frame_seed = sh_uint3_from_u32((u32)rand_u64_from_state(&sig->rand), (u32)rand_u64_from_state(&sig->rand), (u32)rand_u64_from_state(&sig->rand));
constants.camera_offset = sh_float2_from_v2(params.draw_target_view.og);
constants.albedo_tex_urid = sh_uint_from_u32(sig->albedo->srv_descriptor->index);
constants.emittance_tex_urid = sh_uint_from_u32(sig->emittance->srv_descriptor->index);
constants.emittance_flood_tex_urid = sh_uint_from_u32(emittance_flood_read->srv_descriptor->index);
constants.write_tex_urid = sh_uint_from_u32(sig->final_target->uav_descriptor->index);
constants.emittance_flood_tex_urid = sh_uint_from_u32(sig->emittance_flood_read->srv_descriptor->index);
constants.target_tex_urid = sh_uint_from_u32(sig->shade_target->uav_descriptor->index);
constants.tex_width = sh_uint_from_u32(final_target_size.x);
constants.tex_height = sh_uint_from_u32(final_target_size.y);
constants.exposure = sh_float_from_f32(1.0);
constants.gamma = sh_float_from_f32(2.2);
/* Set parameters */
command_list_set_compute_root_constant(cl, &constants, sizeof(constants));
@ -3267,16 +3286,73 @@ struct gp_resource *gp_run_render(struct gp_render_sig *render_sig, struct gp_re
/* Dispatch */
ID3D12GraphicsCommandList_Dispatch(cl->cl, (final_target_size.x + 7) / 8, (final_target_size.y + 7) / 8, 1);
/* Swap */
struct dx12_resource *swp = sig->shade_read;
sig->shade_read = sig->shade_target;
sig->shade_target = swp;
}
/* Bind final target as RTV */
/* Prep for UI pass */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->final_target, 0},
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->final_target, D3D12_RESOURCE_STATE_RENDER_TARGET },
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
ID3D12GraphicsCommandList_OMSetRenderTargets(cl->cl, 1, &sig->final_target->rtv_descriptor->handle, 0, 0);
/* Barrier */
{
struct dx12_resource_barrier_desc barriers[] = {
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->shade_read, D3D12_RESOURCE_STATE_UNORDERED_ACCESS },
{ D3D12_RESOURCE_BARRIER_TYPE_UAV, sig->shade_read, 0 },
{ D3D12_RESOURCE_BARRIER_TYPE_TRANSITION, sig->final_target, D3D12_RESOURCE_STATE_RENDER_TARGET }
};
dx12_resource_barriers(cl->cl, countof(barriers), barriers);
ID3D12GraphicsCommandList_OMSetRenderTargets(cl->cl, 1, &sig->final_target->rtv_descriptor->handle, 0, 0);
}
/* Clear */
{
__profn("Clear final target");
__profnc_dx12(cl->cq->prof, cl->cl, "Clear final target", RGB32_F(0.5, 0.2, 0.2));
f32 clear_color[] = { 0.0f, 0.0f, 0.0f, 0.0f };
ID3D12GraphicsCommandList_ClearRenderTargetView(cl->cl, sig->final_target->rtv_descriptor->handle, clear_color, 0, 0);
}
}
/* UI blit pass */
if (blit_pipeline->success) {
__profn("UI blit pass");
__profnc_dx12(cl->cq->prof, cl->cl, "UI blit pass", RGB32_F(0.5, 0.2, 0.2));
/* Bind pipeline */
ID3D12GraphicsCommandList_SetPipelineState(cl->cl, blit_pipeline->pso);
ID3D12GraphicsCommandList_SetGraphicsRootSignature(cl->cl, blit_pipeline->rootsig);
/* Set Rasterizer State */
#if 1
D3D12_VIEWPORT viewport = viewport_from_rect(params.draw_target_viewport);
D3D12_RECT scissor = scissor_from_rect(params.draw_target_viewport);
#else
D3D12_VIEWPORT viewport = viewport_from_rect(RECT_FROM_V2(V2(0, 0), V2(final_target_size.x, final_target_size.y)));
D3D12_RECT scissor = scissor_from_rect(params.draw_target_viewport);
#endif
ID3D12GraphicsCommandList_RSSetViewports(cl->cl, 1, &viewport);
ID3D12GraphicsCommandList_RSSetScissorRects(cl->cl, 1, &scissor);
/* Set constants */
struct sh_blit_constants constants = ZI;
constants.projection = sh_float4x4_from_mat4x4(blit_vp_matrix);
constants.flags = sh_uint_from_u32(SH_BLIT_FLAG_TONE_MAP | SH_BLIT_FLAG_GAMMA_CORRECT);
constants.exposure = sh_float_from_f32(2.0);
constants.gamma = sh_float_from_f32(2.2);
constants.tex_urid = sh_uint_from_u32(sig->shade_read->uav_descriptor->index);
/* Set parameters */
command_list_set_graphics_root_constant(cl, &constants, sizeof(constants));
ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable(cl->cl, 1, descriptor_heap->start_gpu_handle);
/* Draw */
D3D12_VERTEX_BUFFER_VIEW vbv = vbv_from_command_buffer(dummy_vertex_buffer, 0);
D3D12_INDEX_BUFFER_VIEW ibv = ibv_from_command_buffer(quad_index_buffer, DXGI_FORMAT_R16_UINT);
ID3D12GraphicsCommandList_IASetPrimitiveTopology(cl->cl, D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST);
ID3D12GraphicsCommandList_IASetVertexBuffers(cl->cl, 0, 1, &vbv);
ID3D12GraphicsCommandList_IASetIndexBuffer(cl->cl, &ibv);
ID3D12GraphicsCommandList_DrawIndexedInstanced(cl->cl, 6, 1, 0, 0, 0);
}
/* UI rect pass */
@ -3552,8 +3628,8 @@ INTERNAL void present_blit(struct swapchain_buffer *dst, struct dx12_resource *s
struct command_queue *cq = G.command_queues[DX12_QUEUE_DIRECT];
struct command_list *cl = command_list_open(cq->cl_pool);
{
__profn("Blit");
__profnc_dx12(cl->cq->prof, cl->cl, "Blit", RGB32_F(0.5, 0.2, 0.2));
__profn("Present blit");
__profnc_dx12(cl->cq->prof, cl->cl, "Present blit", RGB32_F(0.5, 0.2, 0.2));
struct swapchain *swapchain = dst->swapchain;
/* Upload dummmy vert & index buffer */
@ -3604,6 +3680,7 @@ INTERNAL void present_blit(struct swapchain_buffer *dst, struct dx12_resource *s
/* Set constants */
struct sh_blit_constants constants = ZI;
constants.projection = sh_float4x4_from_mat4x4(vp_matrix);
constants.flags = sh_uint_from_u32(SH_BLIT_FLAG_NONE);
constants.tex_urid = sh_uint_from_u32(src->srv_descriptor->index);
/* Set parameters */

34
src/sys_win32.c
View File

@ -178,44 +178,44 @@ struct yield_param {
};
struct alignas(64) fiber {
/* ==================================================== */
/* ---------------------------------------------------- */
void *addr; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
char *name_cstr; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
struct atomic32 wake_lock; /* 04 bytes (4 byte alignment) */
i16 id; /* 02 bytes */
i16 parent_id; /* 02 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
u64 wait_addr; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
u64 wait_time; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
i16 next_addr_waiter; /* 02 bytes */
i16 prev_addr_waiter; /* 02 bytes */
i16 next_time_waiter; /* 02 bytes */
i16 prev_time_waiter; /* 02 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
u8 _pad1[8]; /* 08 bytes (padding) */
/* ==================================================== */
/* ---------------------------------------------------- */
u8 _pad2[8]; /* 08 bytes (padding) */
/* ==================================================== */
/* ==================== Cache line ==================== */
/* ==================================================== */
/* ---------------------------------------------------- */
/* -------------------- Cache line -------------------- */
/* ---------------------------------------------------- */
struct sys_scratch_ctx scratch_ctx; /* 16 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
sys_job_func *job_func; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
void *job_sig; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
i32 job_id; /* 04 bytes */
i16 job_pool; /* 02 bytes */
i16 job_priority; /* 02 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
struct snc_counter *job_counter; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
struct yield_param *yield_param; /* 08 bytes */
/* ==================================================== */
/* ---------------------------------------------------- */
u8 _pad3[8]; /* 08 bytes (padding) */
};

1
src/user.c
View File

@ -2027,7 +2027,6 @@ INTERNAL void user_update(struct sys_window *window)
params.draw_target_size = user_resolution;
params.draw_target_viewport = user_viewport;
params.draw_target_view = G.world_to_user_xf;
params.clear_target = 1;
render_texture = gp_run_render(G.render_sig, params);
}