I have a question about computeheadless example. here it creates a buffer memory barrier for devicebuffer from HOST_WRITE to SHADER_READ
After copy from staging buffer to device buffer, there is a fence to wait for the finish of copy. So is this barrier still necessary ?
if it's necessary, why srcAccessMask isVK_ACCESS_HOST_WRITE_BIT. Isn't host_write for the staging buffer, not device buffer? After the copy, shouldn't the srcAccessMask of device buffer be VK_ACCESS_TRANSFER_WRITE_BIT?
Hi Willems,
I have a question about computeheadless example. here it creates a buffer memory barrier for
devicebuffer
fromHOST_WRITE
toSHADER_READ
After copy from staging buffer to device buffer, there is a fence to wait for the finish of copy. So is this barrier still necessary ?
if it's necessary, why
srcAccessMask
isVK_ACCESS_HOST_WRITE_BIT
. Isn'thost_write
for the staging buffer, not device buffer? After the copy, shouldn't the srcAccessMask of device buffer beVK_ACCESS_TRANSFER_WRITE_BIT
?And thank you for this great tutorial!