From 688e5c6d3895a8f6b5aec06977c9dcb4b00e0a76 Mon Sep 17 00:00:00 2001
From: Lukas Stockner <lukasstockner97>
Date: Mon, 11 Jan 2021 20:42:47 +0100
Subject: Fix T82351: Cycles: Tile stealing glitches with adaptive sampling

In my testing this works, but it requires me to remove the min(start_sample...) part in the
adaptive sampling kernel, and I assume there's a reason why it was there?

Reviewed By: brecht

Maniphest Tasks: T82351

Differential Revision: https://developer.blender.org/D9445
---
 intern/cycles/device/cuda/device_cuda_impl.cpp | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

(limited to 'intern/cycles/device/cuda/device_cuda_impl.cpp')

diff --git a/intern/cycles/device/cuda/device_cuda_impl.cpp b/intern/cycles/device/cuda/device_cuda_impl.cpp
index 48151365e5d..307d1469ba5 100644
--- a/intern/cycles/device/cuda/device_cuda_impl.cpp
+++ b/intern/cycles/device/cuda/device_cuda_impl.cpp
@@ -1929,18 +1929,19 @@ void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkT
   }
 
   uint step_samples = divide_up(min_blocks * num_threads_per_block, wtile->w * wtile->h);
-  if (task.adaptive_sampling.use) {
-    step_samples = task.adaptive_sampling.align_static_samples(step_samples);
-  }
 
   /* Render all samples. */
   int start_sample = rtile.start_sample;
   int end_sample = rtile.start_sample + rtile.num_samples;
 
-  for (int sample = start_sample; sample < end_sample; sample += step_samples) {
+  for (int sample = start_sample; sample < end_sample;) {
     /* Setup and copy work tile to device. */
     wtile->start_sample = sample;
-    wtile->num_samples = min(step_samples, end_sample - sample);
+    wtile->num_samples = step_samples;
+    if (task.adaptive_sampling.use) {
+      wtile->num_samples = task.adaptive_sampling.align_samples(sample, step_samples);
+    }
+    wtile->num_samples = min(wtile->num_samples, end_sample - sample);
     work_tiles.copy_to_device();
 
     CUdeviceptr d_work_tiles = (CUdeviceptr)work_tiles.device_pointer;
@@ -1962,7 +1963,8 @@ void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkT
     cuda_assert(cuCtxSynchronize());
 
     /* Update progress. */
-    rtile.sample = sample + wtile->num_samples;
+    sample += wtile->num_samples;
+    rtile.sample = sample;
     task.update_progress(&rtile, rtile.w * rtile.h * wtile->num_samples);
 
     if (task.get_cancel()) {
-- 
cgit v1.2.3