From ce00048d8cd4ab880e699542f577f399b5f9dde9 Mon Sep 17 00:00:00 2001
From: Alejandro Saucedo <axsauze@gmail.com>
Date: Sat, 22 Aug 2020 15:21:20 +0100
Subject: [PATCH] Reverted back to cpp

---
 shaders/glsl/opmult.comp       |   5 +++++
 shaders/glsl/opmult.comp.spv   | Bin 1404 -> 1692 bytes
 src/{OpMult.tpp => OpMult.cpp} |  22 +++++++++++++++++++++-
 3 files changed, 26 insertions(+), 1 deletion(-)
 rename src/{OpMult.tpp => OpMult.cpp} (83%)

diff --git a/shaders/glsl/opmult.comp b/shaders/glsl/opmult.comp
index 2fa804a99..109b48cdd 100644
--- a/shaders/glsl/opmult.comp
+++ b/shaders/glsl/opmult.comp
@@ -12,6 +12,10 @@ layout(binding = 2) buffer tensorOutput {
    uint valuesOutput[ ];
 };
 
+layout(binding = 3) buffer tensorInvalid {
+   uint valuesInvalid[ ];
+};
+
 // TODO: Explore how to make layout inside shader dynamic
 layout (local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
 
@@ -24,6 +28,7 @@ void main()
     valuesOutput[index] = 100 + index;
     valuesRhs[index] = 100 + index;
     valuesLhs[index] = 100 + index;
+    valuesInvalid[index] = 100 + index;
 }
 
 
diff --git a/shaders/glsl/opmult.comp.spv b/shaders/glsl/opmult.comp.spv
index c446f2d71b9ed6e4ac8727662af7dce9f14c63c2..69ffb1c8ba52d2f3e09ae7142c04a77f7715686b 100755
GIT binary patch
delta 354
zcmYL@-%0{O7{$NYbyrLFLP7+==+C9QS-OfM1zlAS5f!vV5G%H(x4lAi+e^s4o<hG*
zW^k9one&@(&dlcUFN%EcyeeX(#y?s|Pi7KGD2a$Hvb>mmKIF@nkM~z{DAgTjDX_1s
z*zX16ORvmK-8ACes7M9Bmq3!=A9Ivwp^xt1J^T#m=kh6Qw?8cB-pV|70IuwqwN_|8
zq2dDfkg}=GTC1!}4)GKj6q63{p=HX`GOE0p=W9E9392ENuWbJ2{aV#WEu{Pw(#8i!
SuWTYcf5Q`%>Fkz^$LJp(h9B|(

delta 86
zcmbQk`-h8{nMs+Qfq{{M0|@mt^2Rf6Uc<=3$f&!Sk-3~vRhxl@fsKKIK?jH#!898K
YCy>^Kih;y+HeY2)XXN+`RA2$b0QF4@u>b%7

diff --git a/src/OpMult.tpp b/src/OpMult.cpp
similarity index 83%
rename from src/OpMult.tpp
rename to src/OpMult.cpp
index 0980ed809..43fcb7cf5 100644
--- a/src/OpMult.tpp
+++ b/src/OpMult.cpp
@@ -127,14 +127,34 @@ OpMult<tX, tY, tZ>::record()
         vk::AccessFlagBits::eTransferRead,
         vk::PipelineStageFlagBits::eComputeShader,
         vk::PipelineStageFlagBits::eTransfer);
+    this->mTensorLHS->recordBufferMemoryBarrier(
+        vk::AccessFlagBits::eShaderWrite,
+        vk::AccessFlagBits::eTransferRead,
+        vk::PipelineStageFlagBits::eComputeShader,
+        vk::PipelineStageFlagBits::eTransfer);
+    this->mTensorRHS->recordBufferMemoryBarrier(
+        vk::AccessFlagBits::eShaderWrite,
+        vk::AccessFlagBits::eTransferRead,
+        vk::PipelineStageFlagBits::eComputeShader,
+        vk::PipelineStageFlagBits::eTransfer);
 
-    this->mTensorOutputStaging->recordCopyFrom(this->mTensorOutput);
+    this->mTensorOutputStaging->recordCopyFrom(this->mTensorLHS);
 
     this->mTensorOutput->recordBufferMemoryBarrier(
         vk::AccessFlagBits::eTransferWrite,
         vk::AccessFlagBits::eHostRead,
         vk::PipelineStageFlagBits::eTransfer,
         vk::PipelineStageFlagBits::eHost);
+    this->mTensorLHS->recordBufferMemoryBarrier(
+        vk::AccessFlagBits::eTransferWrite,
+        vk::AccessFlagBits::eHostRead,
+        vk::PipelineStageFlagBits::eTransfer,
+        vk::PipelineStageFlagBits::eHost);
+    this->mTensorRHS->recordBufferMemoryBarrier(
+        vk::AccessFlagBits::eTransferWrite,
+        vk::AccessFlagBits::eHostRead,
+        vk::PipelineStageFlagBits::eTransfer,
+        vk::PipelineStageFlagBits::eHost);
 }
 
 template<uint32_t tX, uint32_t tY, uint32_t tZ>