From 32f8632cdd4f6906a6b45a8a0397a5f2b352efc2 Mon Sep 17 00:00:00 2001
From: swdee <shanon@bokumail.com>
Date: Sat, 19 Apr 2025 15:07:18 +1200
Subject: [PATCH] removed old matmul code

---
 postprocess/matmul.go | 92 ++-----------------------------------------
 1 file changed, 4 insertions(+), 88 deletions(-)

diff --git a/postprocess/matmul.go b/postprocess/matmul.go
index b714ecc..635f2f3 100644
--- a/postprocess/matmul.go
+++ b/postprocess/matmul.go
@@ -5,94 +5,6 @@ import (
 	"sync"
 )
 
-// matmulUint8 performs matrix multiplication using the CPU
-func matmulUint8Org(data *strideData, boxesNum int, protoChannel int,
-	protoHeight int, protoWeight int) []uint8 {
-
-	A := data.filterSegmentsByNMS
-	B := data.proto
-	// C is matmulOut
-	C := make([]uint8, boxesNum*protoHeight*protoWeight)
-
-	rowsA := boxesNum
-	colsA := protoChannel
-	colsB := protoHeight * protoWeight
-
-	var temp float32
-
-	for i := 0; i < rowsA; i++ {
-		for j := 0; j < colsB; j++ {
-			temp = 0
-			for k := 0; k < colsA; k++ {
-				temp += A[i*colsA+k] * B[k*colsB+j]
-			}
-			if temp > 0 {
-				C[i*colsB+j] = 4 // an object
-			} else {
-				C[i*colsB+j] = 0 // background
-			}
-		}
-	}
-
-	return C
-}
-
-// matmulUint8Parallel performs matrix multiplication using the CPU with
-// optimization to process in parallel across goroutines
-func matmulUint8ParallelOrg(data *strideData, boxesNum int, protoChannel int,
-	protoHeight int, protoWeight int) []uint8 {
-
-	A := data.filterSegmentsByNMS
-	B := data.proto
-	C := make([]uint8, boxesNum*protoHeight*protoWeight)
-
-	rowsA := boxesNum
-	colsA := protoChannel
-	colsB := protoHeight * protoWeight
-
-	// use a worker pool based on available CPU cores
-	numWorkers := runtime.NumCPU()
-	rowCh := make(chan int, rowsA)
-
-	// worker function for performing the matrix multiplication on a row
-	worker := func() {
-		for i := range rowCh {
-			for j := 0; j < colsB; j++ {
-				var temp float32
-				for k := 0; k < colsA; k++ {
-					temp += A[i*colsA+k] * B[k*colsB+j]
-				}
-				if temp > 0 {
-					C[i*colsB+j] = 4 // an object
-				} else {
-					C[i*colsB+j] = 0 // background
-				}
-			}
-		}
-	}
-
-	// start the workers
-	var wg sync.WaitGroup
-	wg.Add(numWorkers)
-	for w := 0; w < numWorkers; w++ {
-		go func() {
-			defer wg.Done()
-			worker()
-		}()
-	}
-
-	// distribute rows to workers
-	for i := 0; i < rowsA; i++ {
-		rowCh <- i
-	}
-	close(rowCh)
-
-	// wait for all workers to complete
-	wg.Wait()
-
-	return C
-}
-
 // matmulUint8 is the straight‑line version: no channels or goroutines.
 func matmulUint8(data *strideData, boxesNum, protoChannel,
 	protoHeight, protoWeight int, C []uint8) {
@@ -147,14 +59,18 @@ func matmulUint8Parallel(data *strideData, boxesNum, protoChannel,
 	for w := 0; w < numWorkers; w++ {
 		go func(w int) {
 			defer wg.Done()
+
 			for i := w; i < rows; i += numWorkers {
 				baseA := i * colsA
 				baseC := i * colsB
+
 				for j := 0; j < colsB; j++ {
 					var sum float32
+
 					for k := 0; k < colsA; k++ {
 						sum += A[baseA+k] * B[k*colsB+j]
 					}
+
 					if sum > 0 {
 						C[baseC+j] = 4
 					}