* Add demonstrators for 4 code generation scenarios:

    * a : Program initialization :https://HybroLangDemos/Newton-SquareRoot-VariablePrecision/Newton.hl
    * b : Kernel initialization : https://CodeExamples/Array-Mult-Specialization.hl
    * c : Applicaiton controlled : https://HybroLangDemos/Newton-SquareRoot-VariablePrecision/Newton.hl
	* d : Heterogeneous target : https://HybroLangDemos/ImagePixelSum/CxRAM-ImagePixelAccumulation.hl

* Correct minor bugs (mul on cxram)

ChangeLog

@@ -1,15 +1,26 @@
-2023-04-19  Henri-Pierre CHARLES HC227932  <hc227932@gre056290.intra.cea.fr>
+2023-05-22  Henri-Pierre CHARLES HC227932  <Henri-Pierre.Charles@cea.fr>
+
+	* Add demonstrators for 4 code generation scenarios
+		* a : Program initialization :
+	https://HybroLangDemos/Newton-SquareRoot-VariablePrecision/Newton.hl
+		* b : Kernel initialization :
+	https://CodeExamples/Array-Mult-Specialization.hl
+		* c : Applicaiton controlled :
+	https://HybroLangDemos/Newton-SquareRoot-VariablePrecision/Newton.hl
+		* d : Heterogeneous target : https://HybroLangDemos/ImagePixelSum/CxRAM-ImagePixelAccumulation.hl
+
+	* Correct minor bugs (mul on cxram)
+
+2023-04-19  Henri-Pierre CHARLES HC227932  <Henri-Pierre.Charles@cea.fr>
 
 	* Release 3.0: Target platforms risc-v, power, aarch64, cxram
 	* More complex demonstrations
 	* Sync with release 1.0 of https://github.com/CEA-LIST/csram-qemu-plugin
 
 
-2022-07-14  Henri-Pierre Charles  <hc227932@gre061041>
+2022-07-14  Henri-Pierre Charles  <hc227932@gre056290.intra.cea.fr>
 
 	* Target platforms : partial part of the ISA of RISCV, Power,
 	Kalray, CxRAM platforms
 
 	* Release 2.0: First public release
-
-

CodeExamples/Array-Mult-Specialization.hl

@@ -0,0 +1,86 @@
+// -*- c -*-
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+
+typedef  int (*pifi)(int *, int);
+
+
+h2_insn_t * genMulArray(h2_insn_t * ptr, int  b)
+{
+  #[
+  int 32 1 mult (int[] 32 1 a, int 32 1 len)
+  {
+    int 32 1 r, i;
+    for (i = 0; i < len; i = i + 1)
+      {
+        a[i] = a[i] * #(b); // b values will be included in code generation
+      }
+  }
+  ]#
+  return (h2_insn_t *) ptr;
+}
+
+void printArray(int * array, int len)
+{
+  printf ("Array values (%d) : ", len);
+  for (int i = 0; i < len; i++)
+    printf ("%d ", array[i]);
+  printf ("\n");
+}
+
+void mulArray(int * array, int len, int value)
+{
+  for (int i = 0; i < len; i++)
+    array[i] = array[i] * value;
+}
+
+int main(int argc, char * argv[])
+{
+
+  h2_insn_t * ptr;
+  int arrayLen, mulValue, i, resultOK;
+  pifi fPtr;
+  int *array1, *array2;
+  clock_t startC, stopC, startHg, stopHg;
+
+  if (argc < 3)
+    {
+      printf("Give 2 values (array len & mult. coefficient )\n");
+      exit(-1);
+    }
+  arrayLen  = atoi (argv[1]);   // Get the
+  mulValue  = atoi (argv[2]);
+  array1 = calloc (arrayLen, sizeof (int));
+  array2 = calloc (arrayLen, sizeof (int));
+  for (i = 0; i < arrayLen; i++)
+	{
+      array1[i] = random () % 100;
+	  array2[i] = array1[i];
+	}
+  printArray (array1, arrayLen);
+  printf("// Multiply an array with a constant value\n");
+  printf("// Array len = %d\n", arrayLen);
+  printf("// Code specialization on value = %d\n", mulValue);
+  ptr  = h2_malloc (1024);  // Allocate memory for 1024 instructions
+  fPtr = (pifi) genMulArray (ptr, mulValue);
+  startHg = clock();
+  fPtr(array1, arrayLen);  // Call generated code
+  stopHg = clock();
+  printArray (array1, arrayLen);
+  resultOK = 0;
+  startC = clock();
+  mulArray (array2, arrayLen, mulValue);
+  stopC = clock();
+  printf ("C version %ld Hg version %ld\n\n", stopC - startC, stopHg - startHg);
+  for (int i = 0; i < arrayLen; i++)
+  {
+    if (array2[i] != array1[i])
+    {
+      printf("Erreur multiplication!\n");
+      resultOK = -1;
+      break;
+    }
+  }
+  return resultOK;
+}

CodeExamples/CelciusFarenheit.hl

@@ -23,7 +23,8 @@ h2_insn_t * genF2C(h2_insn_t * ptr)
   int 32 1 F2C (int 32 1 a)
   {
     int 32 1 r;
-	r = a - 32 * 5 / 9;
+	r = a - 32;
+    r = r * 5 / 9;
     return r;
   }
   ]#
@@ -56,7 +57,7 @@ int main(int argc, char * argv[])
   fF2C = (pifi) genF2C (ptr); //
   res  = fF2C(in0);  // Call generated code
   printf("%d (farenheit)   -> (celcius)%d\n", in0, res);
-  if (res != (in0 - 32 * 5 / 9))
+  if (res != ((in0 - 32) * 5 / 9))
 	resultOK += 1;
 
   if (resultOK == 0)

CodeExamples/CxRAM-SimpleAdd16.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t * vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -57,11 +57,10 @@ int main(int argc, char * argv[])
 	printVec (in1, CXRAM_VECLEN/2, " ");
 	printVec (in2, CXRAM_VECLEN/2, "+");
 	printVec (res, CXRAM_VECLEN/2, "=");
-	resultOK = 0;
 	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i16[0]); i++)
 	  {
 		if (res->i16[i] != (in1->i16[i] + in2->i16[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleAdd32.hl

@@ -36,7 +36,7 @@ void printVec(int * vectorLine, int len, char * mesg)
 int main(int argc, char * argv[])
 {
 #define VECLEN 4
-  int i, resultOK;
+  int i;
   pifiii code;
 #ifdef CXRAM
    	h2_cxram_line_t *in1, *in2, *res;
@@ -66,24 +66,23 @@ int main(int argc, char * argv[])
 	  in1[i] = atoi(argv[i+1]);
 	  in2[i] = 542;
 	  res[i] = 0;
-#endif      
+#endif
 	}
 	ptr = h2_malloc (1024);
 	code = genAddImages(ptr);
 	code (in1, in2, res);
 	printVec (in1, VECLEN/4, " ");
 	printVec (in2, VECLEN/4, "+");
 	printVec (res, VECLEN/4, "=");
-	resultOK = 0;
-	for (i=0; i < VECLEN / sizeof(in1[0]); i++)
+	for (i=0; i < VECLEN; i++)
 	  {
 #ifdef CXRAM
 		if (res->i32[i] != (in1->i32[i] + in2->i32[i]))
-		  resultOK = 1;
+		  return 1;
 #else
 		if (res[i] != (in1[i] + in2[i]))
-		  resultOK = 1;
+		  return 1;
 #endif
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleAdd8.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t * vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -56,11 +56,10 @@ int main(int argc, char * argv[])
 	printVec (in1, CXRAM_VECLEN, " ");
 	printVec (in2, CXRAM_VECLEN, "+");
 	printVec (res, CXRAM_VECLEN, "=");
-	resultOK = 0;
-	for (i=0; i < CXRAM_VECLEN / sizeof(in1[0]); i++)
+	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i8[0]); i++)
 	  {
 		if (res->i8[i] != (in1->i8[i] + in2->i8[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleAnd.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t * vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -56,11 +56,11 @@ int main(int argc, char * argv[])
 	printVec (in1, CXRAM_VECLEN, " ");
 	printVec (in2, CXRAM_VECLEN, "&");
 	printVec (res, CXRAM_VECLEN, "=");
-	resultOK = 0;
-	for (i=0; i < CXRAM_VECLEN / sizeof(in1[0]); i++)
+	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i8[0]); i++)
 	  {
 		if (res->i8[i] != (in1->i8[i] & in2->i8[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	printf ("\n");
+	return 0;
 }

CodeExamples/CxRAM-SimpleMul16.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t *vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -60,11 +60,10 @@ int main(int argc, char * argv[])
 	printVec (in2, CXRAM_VECLEN/2, "*");
 	printVec (res, CXRAM_VECLEN/2, "=");
 #endif
-	resultOK = 0;
 	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i16[0]); i++)
 	  {
 		if (res->i16[i] != (in1->i16[i] * in2->i16[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleMul32.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t *vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -60,11 +60,10 @@ int main(int argc, char * argv[])
 	printVec (in2, CXRAM_VECLEN/4, "*");
 	printVec (res, CXRAM_VECLEN/4, "=");
 #endif
-	resultOK = 0;
 	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i32[0]); i++)
 	  {
 		if (res->i32[i] != (in1->i32[i] * in2->i32[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleMul8.hl

@@ -30,7 +30,7 @@ void printVec(h2_cxram_line_t *vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
 	h2_insn_t * ptr;
 	pifiii code;
 	h2_cxram_line_t *in1, *in2, *res;
@@ -60,11 +60,10 @@ int main(int argc, char * argv[])
 	printVec (in2, CXRAM_VECLEN, "*");
 	printVec (res, CXRAM_VECLEN, "=");
 #endif
-	resultOK = 0;
 	for (i=0; i < CXRAM_VECLEN / sizeof(in1->i8[0]); i++)
 	  {
 		if (res->i8[i] != (in1->i8[i] * in2->i8[i]))
-		  resultOK = 1;
+		  return 1;
 	  }
-	return resultOK;
+	return 0;
 }

CodeExamples/CxRAM-SimpleSub8.hl

@@ -30,7 +30,7 @@ void printVec(int8_t *vectorLine, int len, char * mesg)
 
 int main(int argc, char * argv[])
 {
-  int i, resultOK;
+  int i;
   h2_insn_t * ptr;
   pifiii code;
   int8_t *in1, *in2, *res;
@@ -60,15 +60,14 @@ int main(int argc, char * argv[])
   printVec (in2, CXRAM_VECLEN, "-");
   printVec (res, CXRAM_VECLEN, "=");
 #endif
-  resultOK = 0;
   for (i=0; i < CXRAM_VECLEN / sizeof(in1[0]); i++)
 	{
 	  int tmp = in1[i] - in2[i];
 	  int resultat = tmp; // La soustraction est signée (tmp < 0)?0:tmp;
 	  if (res[i] != resultat)
 		{
-		  resultOK = -1;
+		  return 1;
 		}
 	}
-  return resultOK;
+  return 0;
 }

CodeExamples/Regression.py

@@ -45,8 +45,8 @@ def cmd(cmdAndArgs, Verbose, doPrint = True, wdir = None, doExec = True):
     "CxRAM-ImageDiff",
     "CxRAM-ImagePixelAccumulation",
     "CxRAM-Broadcast32",
-    "CxRAM-Convolution8",
-    "CxRAM-Convolution16",
+#    "CxRAM-Convolution8", # Convolution8 & 16 are not bit compatible
+#    "CxRAM-Convolution16", # thus can not be used as regression test
     "CxRAM-Convolution32",
     "CxRAM-MatrixMultiplication"
     )