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

benchmarking: improved performance with @inbounds. still slower in most cases
Some checks failed
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.10) (push) Has been cancelled
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.6) (push) Has been cancelled
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, pre) (push) Has been cancelled
Details

Browse Source
This commit is contained in:
Daniel
2025-04-01 21:48:59 +02:00
parent d9c83caad9
commit 2b9c394f1b
4 changed files with 36 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
package/src/Interpreter.jl
View File

@ -15,7 +15,7 @@ export interpret
function interpret(expressions::Vector{Expr}, variables::Matrix{Float32}, parameters::Vector{Vector{Float32}})::Matrix{Float32}
exprs = Vector{ExpressionProcessing.PostfixType}(undef, length(expressions))
for i in eachindex(expressions)
@inbounds for i in eachindex(expressions)
exprs[i] = ExpressionProcessing.expr_to_postfix(expressions[i])
end
@ -30,7 +30,7 @@ function interpret(expressions::Vector{Expr}, variables::Matrix{Float32}, parame
cudaResults = CuArray{Float32}(undef, variableCols, length(exprs))
# Start kernel for each expression to ensure that no warp is working on different expressions
for i in eachindex(exprs)
@inbounds for i in eachindex(exprs)
kernel = @cuda launch=false interpret_expression(cudaExprs, cudaVars, cudaParams, cudaResults, cudaStepsize, i)
# config = launch_configuration(kernel.fun)
threads = min(variableCols, 256)
@ -61,11 +61,11 @@ function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, var
firstVariableIndex = ((varSetIndex-1) * stepsize[3]) # Exclusive
for i in firstExprIndex:lastExprIndex
if expressions[i].Type == EMPTY
@inbounds for expr in expressions
if expr.Type == EMPTY
break
elseif expressions[i].Type == INDEX
val = expressions[i].Value
elseif expr.Type == INDEX
val = expr.Value
operationStackTop += 1
if val > 0
@ -74,11 +74,11 @@ function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, var
val = abs(val)
operationStack[operationStackTop] = parameters[firstParamIndex + val]
end
elseif expressions[i].Type == FLOAT32
elseif expr.Type == FLOAT32
operationStackTop += 1
operationStack[operationStackTop] = reinterpret(Float32, expressions[i].Value)
elseif expressions[i].Type == OPERATOR
type = reinterpret(Operator, expressions[i].Value)
operationStack[operationStackTop] = reinterpret(Float32, expr.Value)
elseif expr.Type == OPERATOR
type = reinterpret(Operator, expr.Value)
if type == ADD
operationStackTop -= 1
operationStack[operationStackTop] = operationStack[operationStackTop] + operationStack[operationStackTop + 1]
@ -108,10 +108,11 @@ function interpret_expression(expressions::CuDeviceArray{ExpressionElement}, var
break
end
end
# "(exprIndex - 1) * variableCols" -> calculates the column in which to insert the result (expression = column)
# "+ 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]
@inbounds results[resultIndex] = operationStack[operationStackTop]
return
end