eu07/maszyna
17
0
Fork 0
mirror of https://github.com/MaSzyna-EU07/maszyna.git synced 2026-11-06 05:48:35 +01:00
Code Issues 150 Wiki Activity

Vulkan: parity with GL (P1–P7), shared GPU culling records (N6c), threaded command recording and render thread (M1, M4)

Browse Source
This commit is contained in:
Hirek
2026-10-25 03:06:01 +01:00
parent 49f9df7d98
commit c48c58776a
146 changed files with 20748 additions and 4606 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
shaders/vk14/instance_cull_offsets.comp Normal file
View File

@@ -0,0 +1,62 @@
#version 460 core
#define STAGE_COMPUTE 1
// turns the counts of instances of the entries, written by stage 0 of instance_cull.comp, into the first slots of the
// entries in the output (cursors, which stages 1 and 2 advance as they write, and the same first slots kept for
// instance_cull_commands.comp), and puts the total in the header.
// a single work group: each invocation sums a run of the entries, then the sums are scanned in shared memory
#define LAYOUT_COMPUTE 1
#include "common.glsl"
layout(local_size_x = INSTANCE_OFFSETS_GROUP_SIZE) in;
layout(std430, set = COMPUTE_SET, binding = 0) buffer instance_counter_ssbo
{
uint counters[];
};
// the parameters of the kernel (push constants)
layout(push_constant) uniform kernel_constants
{
// x: entries
uvec4 counts;
};
shared uint sums[INSTANCE_OFFSETS_GROUP_SIZE];
void main()
{
uint entries = counts.x;
uint invocation = gl_LocalInvocationIndex;
uint run = (entries + INSTANCE_OFFSETS_GROUP_SIZE - 1u) / INSTANCE_OFFSETS_GROUP_SIZE;
uint begin = min(invocation * run, entries);
uint end = min(begin + run, entries);
uint sum = 0u;
for (uint entry = begin; entry < end; ++entry)
sum += counters[INSTANCE_COUNTER_HEADER + entry];
sums[invocation] = sum;
memoryBarrierShared();
barrier();
// inclusive scan of the sums
for (uint step = 1u; step < INSTANCE_OFFSETS_GROUP_SIZE; step *= 2u)
{
uint added = (invocation >= step ? sums[invocation - step] : 0u);
memoryBarrierShared();
barrier();
sums[invocation] += added;
memoryBarrierShared();
barrier();
}
uint offset = sums[invocation] - sum;
for (uint entry = begin; entry < end; ++entry)
{
counters[INSTANCE_COUNTER_HEADER + entries + entry] = offset;
counters[INSTANCE_COUNTER_HEADER + 2u * entries + entry] = offset;
offset += counters[INSTANCE_COUNTER_HEADER + entry];
}
if (invocation == INSTANCE_OFFSETS_GROUP_SIZE - 1u)
counters[1] = sums[invocation];
}