Daniel/master-thesis
1
0
Fork 0
Code Issues 3 Pull Requests Actions 1 Packages Projects 1 Releases Wiki Activity

benchmarking: fixed bugs; took initial_benchmark
Some checks are pending
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.10) (push) Waiting to run
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.6) (push) Waiting to run
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, pre) (push) Waiting to run
Details

Browse Source
This commit is contained in:
Daniel
2025-03-30 12:54:50 +02:00
parent ad175abac0
commit 1dc0c1898d
4 changed files with 63 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
package/src/Interpreter.jl
View File

@ -45,19 +45,23 @@ end
#TODO: Add @inbounds to all indexing after it is verified that all works https://cuda.juliagpu.org/stable/development/kernel/#Bounds-checking
const MAX_STACK_SIZE = 25 # The depth of the stack to store the values and intermediate results
function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, variables::CuDeviceArray{Float32}, parameters::CuDeviceArray{Float32}, results::CuDeviceArray{Float32}, stepsize::CuDeviceArray{Int}, exprIndex::Int)
index = (blockIdx().x - 1) * blockDim().x + threadIdx().x # ctaid.x * ntid.x + tid.x
stride = gridDim().x * blockDim().x # nctaid.x * ntid.x
varSetIndex = (blockIdx().x - 1) * blockDim().x + threadIdx().x # ctaid.x * ntid.x + tid.x (1-based)
# stride = gridDim().x * blockDim().x # nctaid.x * ntid.x
variableCols = length(variables) / stepsize[3]
if varSetIndex > variableCols
return
end
firstExprIndex = ((exprIndex - 1) * stepsize[1]) + 1 # Inclusive
lastExprIndex = firstExprIndex + stepsize[1] - 1 # Inclusive
firstParamIndex = ((exprIndex - 1) * stepsize[2]) # Exclusive
variableCols = length(variables) / stepsize[3]
operationStack = MVector{MAX_STACK_SIZE, Float32}(undef) # Try to get this to function with variable size too, to allow better memory usage
operationStackTop = 0 # stores index of the last defined/valid value
for varSetIndex in index:stride
firstVariableIndex = ((varSetIndex - 1) * stepsize[3]) # Exclusive
# for varSetIndex in index:stride
firstVariableIndex = ((varSetIndex-1) * stepsize[3]) # Exclusive
for i in firstExprIndex:lastExprIndex
if expressions[i].Type == EMPTY
@ -69,7 +73,7 @@ function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, var
if val > 0
operationStack[operationStackTop] = variables[firstVariableIndex + val]
else
val = -val
val = abs(val)
operationStack[operationStackTop] = parameters[firstParamIndex + val]
end
elseif expressions[i].Type == FLOAT32
@ -110,7 +114,7 @@ function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, var
# "+ varSetIndex" -> to get the row inside the column at which to insert the result of the variable set (variable set = row)
resultIndex = convert(Int, (exprIndex - 1) * variableCols + varSetIndex) # Inclusive
results[resultIndex] = operationStack[operationStackTop]
end
# end
return
end