首页 / Adreno (TM) 750 / Qualcomm Technologies Inc. Adreno Vulkan Driver

报告 #442 驱动检测

已认证，设备已锁定有效

GPU: Adreno (TM) 750
驱动: Qualcomm Technologies Inc. Adreno Vulkan Driver Driver Build: f6b5df5188, I29aebc4401, 1758283784 Date: 09/19/25 Compiler Version: E031.45.02.26 Driver Branch:
设备: Samsung Galaxy Z Flip6
芯片: SM8650
CPU: 2 x 2.27 GHz + 3 x 3.15 GHz + 2 x 2.96 GHz + 1 x 3.40 GHz, arm64-v8a
内存: 10.9 GB
上报为: samsung / SM-F741U
收到时间: 2026-10-13 00:15 UTC
Android: 16

测试框架检测 (1)

通过 harness.baseline_draw

检测	像素	预期	实际	一致
drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

核心检测 (21)

通过 core.readback_same_command_buffer

检测	像素	预期	实际	一致
drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 push.max_combined_image_samplers

检测	像素	预期	实际	一致
combined image samplers pushed in one set (the driver's stated limit)	,		32	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: the pushed texture	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 core.resubmit_command_buffer

检测	像素	预期	实际	一致
first submit: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
second submit of the same command buffer: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
second submit of the same command buffer: cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 core.compute_udiv

检测	像素	预期	实际	一致
703787787 / 47721859	,	[14]	[14]	通过
703787787 % 47721859	,	[35681761]	[35681761]	通过
4294967295 / 2147483649	,	[1]	[1]	通过
4294967295 % 2147483649	,	[2147483646]	[2147483646]	通过
4294967295 / 3	,	[1431655765]	[1431655765]	通过
4294967295 % 3	,	[0]	[0]	通过
4000000000 / 3	,	[1333333333]	[1333333333]	通过
4000000000 % 3	,	[1]	[1]	通过
2147483648 / 65537	,	[32767]	[32767]	通过
2147483648 % 65537	,	[32769]	[32769]	通过
2147483647 / 46341	,	[46340]	[46340]	通过
2147483647 % 46341	,	[41707]	[41707]	通过
1000000007 / 31607	,	[31638]	[31638]	通过
1000000007 % 31607	,	[17741]	[17741]	通过
16777217 / 4097	,	[4095]	[4095]	通过
16777217 % 4097	,	[2]	[2]	通过
16777216 / 4096	,	[4096]	[4096]	通过
16777216 % 4096	,	[0]	[0]	通过
4294967295 / 4294967295	,	[1]	[1]	通过
4294967295 % 4294967295	,	[0]	[0]	通过
3000000000 / 1500000001	,	[1]	[1]	通过
3000000000 % 1500000001	,	[1499999999]	[1499999999]	通过
123456789 / 10007	,	[12337]	[12337]	通过
123456789 % 10007	,	[430]	[430]	通过
4294967294 / 2	,	[2147483647]	[2147483647]	通过
4294967294 % 2	,	[0]	[0]	通过
65536 / 65537	,	[0]	[0]	通过
65536 % 65537	,	[65536]	[65536]	通过
99999999 / 4294967295	,	[0]	[0]	通过
99999999 % 4294967295	,	[99999999]	[99999999]	通过
1 / 1	,	[1]	[1]	通过
1 % 1	,	[0]	[0]	通过

通过 core.texel_fetch_beside_ubo

检测	像素	预期	实际	一致
left half: texel (1, 0)	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
right half: texel (0, 1)	48,32	[0, 0, 255, 255]	[0, 0, 255, 255]	通过

通过 push.max_uniform_buffers

检测	像素	预期	实际	一致
uniform buffers pushed in one set (the driver's stated limit)	,		32	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: element 5 of the pushed buffer	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.uniform_buffer_range

检测	像素	预期	实际	一致
bytes in one uniform buffer binding (the driver's stated limit)	,		65536	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: element 127 of the block	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.dynamic_uniform_buffers

检测	像素	预期	实际	一致
dynamic uniform buffers in one set, bound with as many offsets (the driver's stated limit)	,		32	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: the block the dynamic offset selects	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 stress.multi_draw_indirect_large

检测	像素	预期	实际	一致
draws in one vkCmdDrawIndirect call (capped by this check, below the stated limit)	,		16384	通过
drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 stress.compute_after_render

检测	像素	预期	实际	一致
wrong answers in 3200 divisions over 200 rounds	,	[0]	[0]	通过
the last render pass: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过

通过 stress.submit_memory

检测	像素	预期	实际	一致
KiB of memory gained over 4000 submits (limit 16384)	,		0	通过
the last submit: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过

失败 core.copy_wide_row_length

检测	像素	预期	实际	一致
first copied row (the last green row)	16,0	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
second copied row, 4,210,688 bytes later (the first black row)	16,1	[0, 0, 0, 255]	[127, 127, 127, 127]	失败

通过 core.mediump_mul_extended

检测	像素	预期	实际	一致
low half of 255 * 3	,	[765]	[765]	通过
low half of 242 * 19	,	[4598]	[4598]	通过
low half of 229 * 35	,	[8015]	[8015]	通过
low half of 216 * 51	,	[11016]	[11016]	通过
low half of 203 * 67	,	[13601]	[13601]	通过
low half of 190 * 83	,	[15770]	[15770]	通过
low half of 177 * 99	,	[17523]	[17523]	通过
low half of 164 * 115	,	[18860]	[18860]	通过
low half of 151 * 131	,	[19781]	[19781]	通过
low half of 138 * 147	,	[20286]	[20286]	通过
low half of 125 * 163	,	[20375]	[20375]	通过
low half of 112 * 179	,	[20048]	[20048]	通过
low half of 99 * 195	,	[19305]	[19305]	通过
low half of 86 * 211	,	[18146]	[18146]	通过
low half of 73 * 227	,	[16571]	[16571]	通过
low half of 60 * 243	,	[14580]	[14580]	通过
high half of the first product	,	[0]	[0]	通过

通过 core.msaa_resolve

检测	像素	预期	实际	一致
resolved image: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
resolved image: cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.color_attachments

检测	像素	预期	实际	一致
colour attachments in one render pass (the driver's stated limit)	,		8	通过
bytes changed in memory next to the driver's allocations	,		0	通过
first attachment: drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
first attachment: cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.vertex_input

检测	像素	预期	实际	一致
vertex buffer bindings in one pipeline, bound in one call (the driver's stated limit)	,		32	通过
vertex attributes in one pipeline (the driver's stated limit)	,		32	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.push_constants_size

检测	像素	预期	实际	一致
push constant bytes pushed in one call (the driver's stated limit)	,		256	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: the colour in the first 16 bytes	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.per_stage_samplers

检测	像素	预期	实际	一致
combined image samplers in one descriptor set, bound and drawn with (capped by this check, below the stated limit)	,		4096	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: the texture in element 0	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 limit.bound_descriptor_sets

检测	像素	预期	实际	一致
descriptor sets bound in one call (the driver's stated limit)	,		7	通过
bytes changed in memory next to the driver's allocations	,		0	通过
drawn half: the texture in set 0	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 stress.pipeline_cache_threads

检测	像素	预期	实际	一致
every vkCreateGraphicsPipelines call returned VK_SUCCESS	,		0	通过
pipelines asked for in 8 seconds, at least one from each thread	,		1024	通过
pipelines built, the same number	,		1024	通过
every save made while compiling returned VK_SUCCESS or VK_INCOMPLETE	,		0	通过
saves made while compiling	,		207	通过
bytes a save reported beyond the buffer size it was given	,		0	通过
bytes a save wrote past the end of the buffer it was given	,		0	通过
final save size in bytes holds a header	,		1308471	通过
header names this device	,		32	通过
drawn half, after every pipeline drew it	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

通过 stress.pipeline_threads_no_cache

检测	像素	预期	实际	一致
every vkCreateGraphicsPipelines call returned VK_SUCCESS	,		0	通过
pipelines asked for in 8 seconds, at least one from each thread	,		1024	通过
pipelines built, the same number	,		1024	通过
drawn half, after every pipeline drew it	16,32	[0, 255, 0, 255]	[0, 255, 0, 255]	通过
cleared half	48,32	[0, 0, 0, 255]	[0, 0, 0, 255]	通过

功能结果

VkPhysicalDeviceFeatures（已测试 9 个）

geometryShader 通过tessellationShader 通过multiDrawIndirect 通过drawIndirectFirstInstance 通过multiViewport 通过textureCompressionETC2 通过textureCompressionASTC_LDR 通过textureCompressionBC 通过occlusionQueryPrecise 通过

VK_EXT_extended_dynamic_state（已测试 1 个）

extendedDynamicState 通过

VK_EXT_extended_dynamic_state2（已测试 2 个）

extendedDynamicState2 通过extendedDynamicState2LogicOp 通过

VK_EXT_vertex_input_dynamic_state（已测试 1 个）

vertexInputDynamicState 通过

VK_EXT_custom_border_color（已测试 1 个）

customBorderColors 通过

VK_EXT_border_color_swizzle（已测试 1 个）

borderColorSwizzle 通过

VK_EXT_depth_clip_control（已测试 1 个）

depthClipControl 通过

VK_EXT_transform_feedback（已测试 1 个）

transformFeedback 通过

VK_EXT_color_write_enable（已测试 1 个）

colorWriteEnable 通过

VK_EXT_inline_uniform_block（已测试 1 个）

inlineUniformBlock 通过

VK_KHR_timeline_semaphore（已测试 1 个）

timelineSemaphore 通过

VK_KHR_shader_atomic_int64（已测试 1 个）

shaderBufferInt64Atomics 通过

VK_EXT_robustness2（已测试 2 个）

robustBufferAccess2 通过robustImageAccess2 通过

不支持或没有检测项的功能不在此列出。