Forums - Issue with Compute Shader Programs on Redmagic 9 Pro (Adreno 750)

4 posts / 0 new
or Register
Last post
Issue with Compute Shader Programs on Redmagic 9 Pro (Adreno 750)
smiyaxdev
Join Date: 11 Apr 16
Posts: 5
Posted: Sun, 2024-03-31 07:06
Top

Hello Community, I'm encountering a perplexing issue when attempting to deploy compute shader programs on the Redmagic 9 Pro, which is powered by an Adreno 750 GPU. My workflow involves compiling GLSL code via Shaderc. However, upon executing vkCreateComputePipelines, I'm consistently met with a VK_ERROR_UNKNOWN response. It's noteworthy that this issue is not present on the Redmagic 8, featuring the Adreno 740 GPU, where my shaders run without hitches. I'm reaching out to see if anyone else has faced similar challenges or if there are known workarounds or compatibility concerns with the Adreno 750 GPU and Vulkan's compute pipelines. Attached below is the minimal reproducible example of the code that leads to this error. I appreciate any insights or suggestions you might have.

#version 320 es
precision highp float;
precision highp int;
precision highp image2D;
layout(local_size_x = 16, local_size_y = 16, local_size_z = 1) in;
layout(rgba8, binding = 0)  uniform writeonly image2D outSurface;
layout(std430, binding = 1) readonly buffer VDP2C { uint cram[]; };
void main() {
  ivec2 texel = ivec2(gl_GlobalInvocationID.xy);
  uint kdata = cram[texel.x];
  kdata = ((kdata >> 8) & 0xFFu) | ((kdata << 8) & 0xFF00u);
  if ((kdata&0x8000u)!=0x0u) { return; }
  imageStore(outSurface,texel,vec4(float(kdata)));
}

 

When I change line 10th to "kdata = ((kdata >> 8) & 0xFFu) + ((kdata << 8) & 0xFF00u);" it returns VK_SUCCESS "kdata = ( ((kdata>>8)&0x00FFu) | (kdata<<8) ) & 0xFFFFu ;" is also returns VK_SUCCESS for more details check my GitHub page.

https://github.com/devmiyax/yabause/issues/998

 

Thank you in advance for your support and guidance.

  • Up0
  • Down0
Re: Issue with Compute Shader Programs on Redmagic 9 Pro ...
smiyaxdev
Join Date: 11 Apr 16
Posts: 5
Posted: Sat, 2024-04-20 20:36
Top

This code failed to vkCreateComputePipelines

; SPIR-V
; Version: 1.0
; Generator: Google Shaderc over Glslang; 10
; Bound: 65
; Schema: 0
OpCapability Shader
%1 = OpExtInstImport "GLSL.std.450"
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %4 "main" %gl_GlobalInvocationID
OpExecutionMode %4 LocalSize 16 16 1
OpDecorate %gl_GlobalInvocationID BuiltIn GlobalInvocationId
OpDecorate %_runtimearr_uint ArrayStride 4
OpMemberDecorate %_struct_21 0 NonWritable
OpMemberDecorate %_struct_21 0 Offset 0
OpDecorate %_struct_21 BufferBlock
OpDecorate %23 DescriptorSet 0
OpDecorate %23 Binding 1
OpDecorate %53 DescriptorSet 0
OpDecorate %53 Binding 0
OpDecorate %53 NonReadable
OpDecorate %gl_WorkGroupSize BuiltIn WorkgroupSize
%void = OpTypeVoid
%3 = OpTypeFunction %void
%int = OpTypeInt 32 1
%v2int = OpTypeVector %int 2
%uint = OpTypeInt 32 0
%v3uint = OpTypeVector %uint 3
%_ptr_Input_v3uint = OpTypePointer Input %v3uint
%gl_GlobalInvocationID = OpVariable %_ptr_Input_v3uint Input
%v2uint = OpTypeVector %uint 2
%_runtimearr_uint = OpTypeRuntimeArray %uint
%_struct_21 = OpTypeStruct %_runtimearr_uint
%_ptr_Uniform__struct_21 = OpTypePointer Uniform %_struct_21
%23 = OpVariable %_ptr_Uniform__struct_21 Uniform
%int_0 = OpConstant %int 0
%uint_0 = OpConstant %uint 0
%_ptr_Uniform_uint = OpTypePointer Uniform %uint
%int_8 = OpConstant %int 8
%uint_255 = OpConstant %uint 255
%uint_65280 = OpConstant %uint 65280
%uint_32768 = OpConstant %uint 32768
%bool = OpTypeBool
%float = OpTypeFloat 32
%51 = OpTypeImage %float 2D 0 0 0 2 Rgba8
%_ptr_UniformConstant_51 = OpTypePointer UniformConstant %51
%53 = OpVariable %_ptr_UniformConstant_51 UniformConstant
%v4float = OpTypeVector %float 4
%uint_16 = OpConstant %uint 16
%uint_1 = OpConstant %uint 1
%gl_WorkGroupSize = OpConstantComposite %v3uint %uint_16 %uint_16 %uint_1
%4 = OpFunction %void None %3
%5 = OpLabel
OpSelectionMerge %63 None
OpSwitch %uint_0 %64
%64 = OpLabel
%15 = OpLoad %v3uint %gl_GlobalInvocationID
%16 = OpVectorShuffle %v2uint %15 %15 0 1
%17 = OpBitcast %v2int %16
%28 = OpCompositeExtract %int %17 0
%30 = OpAccessChain %_ptr_Uniform_uint %23 %int_0 %28
%31 = OpLoad %uint %30
%34 = OpShiftRightLogical %uint %31 %int_8
%36 = OpBitwiseAnd %uint %34 %uint_255
%38 = OpShiftLeftLogical %uint %31 %int_8
%40 = OpBitwiseAnd %uint %38 %uint_65280
%41 = OpBitwiseOr %uint %36 %40
%44 = OpBitwiseAnd %uint %41 %uint_32768
%46 = OpINotEqual %bool %44 %uint_0
OpSelectionMerge %48 None
OpBranchConditional %46 %47 %48
%47 = OpLabel
OpBranch %63
%48 = OpLabel
%54 = OpLoad %51 %53
%57 = OpConvertUToF %float %41
%59 = OpCompositeConstruct %v4float %57 %57 %57 %57
OpImageWrite %54 %17 %59
OpBranch %63
%63 = OpLabel
OpReturn
OpFunctionEnd

  • Up0
  • Down0
Re: Issue with Compute Shader Programs on Redmagic 9 Pro ...
smiyaxdev
Join Date: 11 Apr 16
Posts: 5
Posted: Sat, 2024-04-20 20:37
Top

This code success to vkCreateComputePipelines

; SPIR-V
; Version: 1.0
; Generator: Google Shaderc over Glslang; 10
; Bound: 65
; Schema: 0
OpCapability Shader
%1 = OpExtInstImport "GLSL.std.450"
OpMemoryModel Logical GLSL450
OpEntryPoint GLCompute %4 "main" %gl_GlobalInvocationID
OpExecutionMode %4 LocalSize 16 16 1
OpDecorate %gl_GlobalInvocationID BuiltIn GlobalInvocationId
OpDecorate %_runtimearr_uint ArrayStride 4
OpMemberDecorate %_struct_21 0 NonWritable
OpMemberDecorate %_struct_21 0 Offset 0
OpDecorate %_struct_21 BufferBlock
OpDecorate %23 DescriptorSet 0
OpDecorate %23 Binding 1
OpDecorate %53 DescriptorSet 0
OpDecorate %53 Binding 0
OpDecorate %53 NonReadable
OpDecorate %gl_WorkGroupSize BuiltIn WorkgroupSize
%void = OpTypeVoid
%3 = OpTypeFunction %void
%int = OpTypeInt 32 1
%v2int = OpTypeVector %int 2
%uint = OpTypeInt 32 0
%v3uint = OpTypeVector %uint 3
%_ptr_Input_v3uint = OpTypePointer Input %v3uint
%gl_GlobalInvocationID = OpVariable %_ptr_Input_v3uint Input
%v2uint = OpTypeVector %uint 2
%_runtimearr_uint = OpTypeRuntimeArray %uint
%_struct_21 = OpTypeStruct %_runtimearr_uint
%_ptr_Uniform__struct_21 = OpTypePointer Uniform %_struct_21
%23 = OpVariable %_ptr_Uniform__struct_21 Uniform
%int_0 = OpConstant %int 0
%uint_0 = OpConstant %uint 0
%_ptr_Uniform_uint = OpTypePointer Uniform %uint
%int_8 = OpConstant %int 8
%uint_255 = OpConstant %uint 255
%uint_65535 = OpConstant %uint 65535
%uint_32768 = OpConstant %uint 32768
%bool = OpTypeBool
%float = OpTypeFloat 32
%51 = OpTypeImage %float 2D 0 0 0 2 Rgba8
%_ptr_UniformConstant_51 = OpTypePointer UniformConstant %51
%53 = OpVariable %_ptr_UniformConstant_51 UniformConstant
%v4float = OpTypeVector %float 4
%uint_16 = OpConstant %uint 16
%uint_1 = OpConstant %uint 1
%gl_WorkGroupSize = OpConstantComposite %v3uint %uint_16 %uint_16 %uint_1
%4 = OpFunction %void None %3
%5 = OpLabel
OpSelectionMerge %63 None
OpSwitch %uint_0 %64
%64 = OpLabel
%15 = OpLoad %v3uint %gl_GlobalInvocationID
%16 = OpVectorShuffle %v2uint %15 %15 0 1
%17 = OpBitcast %v2int %16
%28 = OpCompositeExtract %int %17 0
%30 = OpAccessChain %_ptr_Uniform_uint %23 %int_0 %28
%31 = OpLoad %uint %30
%34 = OpShiftRightLogical %uint %31 %int_8
%36 = OpBitwiseAnd %uint %34 %uint_255
%38 = OpShiftLeftLogical %uint %31 %int_8
%39 = OpBitwiseOr %uint %36 %38
%41 = OpBitwiseAnd %uint %39 %uint_65535
%44 = OpBitwiseAnd %uint %41 %uint_32768
%46 = OpINotEqual %bool %44 %uint_0
OpSelectionMerge %48 None
OpBranchConditional %46 %47 %48
%47 = OpLabel
OpBranch %63
%48 = OpLabel
%54 = OpLoad %51 %53
%57 = OpConvertUToF %float %41
%59 = OpCompositeConstruct %v4float %57 %57 %57 %57
OpImageWrite %54 %17 %59
OpBranch %63
%63 = OpLabel
OpReturn
OpFunctionEnd

  • Up0
  • Down0
Re: Issue with Compute Shader Programs on Redmagic 9 Pro ...
smiyaxdev
Join Date: 11 Apr 16
Posts: 5
Posted: Sat, 2024-04-20 20:38
Top

Diffrence are

--bad--
%40 = OpBitwiseAnd %uint %38 %uint_65280
%41 = OpBitwiseOr %uint %36 %40

--good--
%39 = OpBitwiseOr %uint %36 %38
%41 = OpBitwiseAnd %uint %39 %uint_65535

  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.