lee/renderdoc
1
0
Fork 0
mirror of https://github.com/baldurk/renderdoc.git synced 2026-09-21 13:15:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix generation of NaNs in vulkan shader debug zoo test

Browse Source
This commit is contained in:
baldurk
2022-02-02 15:00:44 +00:00
parent e42b0ff2ca
commit e4fd52c367
1 changed files with 16 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
util/test/demos/vk/vk_shader_debug_zoo.cpp
+16 -4
View File
@@ -22,6 +22,7 @@
* THE SOFTWARE.
******************************************************************************/
#include <limits>
#include "3rdparty/fmt/core.h"
#include "vk_test.h"
@@ -110,7 +111,7 @@ layout(set = 0, binding = 10, std140) uniform constsbuf
vec4 first;
uint uniformIndex;
vec4 second;
vec4 pad2;
vec4 nan;
vec4 third;
vec4 pad3;
vec4 fourth;
@@ -317,6 +318,7 @@ void main()
float posinf = linearData.oneVal/linearData.zeroVal.x;
float neginf = linearData.negoneVal/linearData.zeroVal.x;
float nan = linearData.zeroVal.x/linearData.zeroVal.y;
nan *= cbuf.nan.x;
float negone = linearData.negoneVal;
float posone = linearData.oneVal;
@@ -507,7 +509,7 @@ void main()
case 30:
{
Color = cbuf.first + cbuf.second + cbuf.third + cbuf.fourth +
cbuf.pad2 + cbuf.pad3;
cbuf.pad3;
break;
}
case 31:
@@ -3163,7 +3165,7 @@ OpDecorate %cbuffer Binding 10
OpMemberDecorate %cbuffer_struct 0 Offset 0 ; vec4 first
OpMemberDecorate %cbuffer_struct 1 Offset 16 ; vec4 pad1
OpMemberDecorate %cbuffer_struct 2 Offset 32 ; vec4 second
OpMemberDecorate %cbuffer_struct 3 Offset 48 ; vec4 pad2
OpMemberDecorate %cbuffer_struct 3 Offset 48 ; vec4 nan
OpMemberDecorate %cbuffer_struct 4 Offset 64 ; vec4 third
OpMemberDecorate %cbuffer_struct 5 Offset 80 ; vec4 pad3
OpMemberDecorate %cbuffer_struct 6 Offset 96 ; vec4 fourth
@@ -3278,7 +3280,16 @@ OpMemberDecorate %cbuffer_struct 17 Offset 216 ; double doublePackSource
%posinf = OpFDiv %float %oneVal %zerof
%neginf = OpFDiv %float %negoneVal %zerof
%nan = OpFDiv %float %zerof %zerof
; NaN generation is hard and we want to avoid compilers compiling it out, so generate
; it in shader and multiply by one from a UBO so we get NaN either way
; (since NaN * anything = NaN)
%nan_shad = OpFDiv %float %zerof %zerof
%nan_ptr = OpAccessChain %ptr_Uniform_float4 %cbuffer %uint_3
%nan_ubo = OpLoad %float4 %nan_ptr
%nan_ubo_x = OpCompositeExtract %float %nan_ubo 0
%nan = OpFMul %float %nan_shad %nan_ubo_x
%intval_ptr = OpAccessChain %ptr_Input_uint %flatData %flatv2f_intval_idx
%intval = OpLoad %uint %intval_ptr
@@ -3952,6 +3963,7 @@ OpMemberDecorate %cbuffer_struct 17 Offset 216 ; double doublePackSource
cbufferdata[1] = Vec4f(1.1f, 2.2f, 3.3f, 4.4f);
cbufferdata[2] = Vec4f(5.5f, 6.6f, 7.7f, 8.8f);
cbufferdata[3] = Vec4f(std::numeric_limits<float>::quiet_NaN());
cbufferdata[4] = Vec4f(9.9f, 9.99f, 9.999f, 9.999f);
cbufferdata[6] = Vec4f(100.0f, 200.0f, 300.0f, 400.0f);