GithubStar/ncnn
1
0
Fork 0
mirror of https://github.com/Tencent/ncnn.git synced 2026-10-29 12:39:40 +08:00
Code Issues Packages Projects Releases Wiki Activity

radv crash with large local group size, workaround

Browse Source
This commit is contained in:
nihuini
2019-09-12 19:49:48 +08:00
parent 8166584b54
commit 8a7b4b035e
1 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/layer/vulkan/instancenorm_vulkan.cpp
+6 -6
View File
@@ -49,14 +49,14 @@ int InstanceNorm_vulkan::create_pipeline(const Option& opt)
if (channels % 4 != 0)
{
pipeline_instancenorm_reduce_sum4_fp16_to_fp32 = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp16_to_fp32->set_optimal_local_size_xyz(64, 1, channels);
pipeline_instancenorm_reduce_sum4_fp16_to_fp32->set_optimal_local_size_xyz(16, 1, channels);
pipeline_instancenorm_reduce_sum4_fp16_to_fp32->create("instancenorm_reduce_sum4_fp16_to_fp32", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_sum4_fp32[0] = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp32[0]->set_optimal_local_size_xyz(64, 1, channels);
pipeline_instancenorm_reduce_sum4_fp32[0]->set_optimal_local_size_xyz(16, 1, channels);
pipeline_instancenorm_reduce_sum4_fp32[0]->create("instancenorm_reduce_sum4_fp32", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_sum4_fp32[1] = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp32[1]->set_optimal_local_size_xyz(64, 1, channels);
pipeline_instancenorm_reduce_sum4_fp32[1]->set_optimal_local_size_xyz(16, 1, channels);
pipeline_instancenorm_reduce_sum4_fp32[1]->create("instancenorm_reduce_sum4_fp32", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_mean = new Pipeline(vkdev);
@@ -80,14 +80,14 @@ int InstanceNorm_vulkan::create_pipeline(const Option& opt)
if (channels % 4 == 0)
{
pipeline_instancenorm_reduce_sum4_fp16_to_fp32_pack4 = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp16_to_fp32_pack4->set_optimal_local_size_xyz(64, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp16_to_fp32_pack4->set_optimal_local_size_xyz(16, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp16_to_fp32_pack4->create("instancenorm_reduce_sum4_fp16_to_fp32_pack4", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_sum4_fp32_pack4[0] = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp32_pack4[0]->set_optimal_local_size_xyz(64, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp32_pack4[0]->set_optimal_local_size_xyz(16, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp32_pack4[0]->create("instancenorm_reduce_sum4_fp32_pack4", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_sum4_fp32_pack4[1] = new Pipeline(vkdev);
pipeline_instancenorm_reduce_sum4_fp32_pack4[1]->set_optimal_local_size_xyz(64, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp32_pack4[1]->set_optimal_local_size_xyz(16, 1, std::max(1, channels / 4));
pipeline_instancenorm_reduce_sum4_fp32_pack4[1]->create("instancenorm_reduce_sum4_fp32_pack4", opt, std::vector<vk_specialization_type>(), 2, 6);
pipeline_instancenorm_reduce_mean_pack4 = new Pipeline(vkdev);