math: p_round: Implement Round Function

README.md

@@ -233,7 +233,8 @@ FUNCTION     | NOTES
 [p_popcount()](src/math/p_popcount.c) | count the number of bits set  
 [p_pow()](src/math/p_pow.c)           | element raised to a power  
 [p_rand()](src/math/p_rand.c)         | random number generator  
-[p_randinit()](src/math/p_rand.c)     | init random number generator  
+[p_randinit()](src/math/p_rand.c)     | init random number generator
+[p_round()](src/math/p_round.c)       | round  
 [p_sort()](src/math/p_sort.c)         | heap sort  
 [p_sin()](src/math/p_sin.c)           | sine  
 [p_sinh()](src/math/p_sinh.c)         | hyperbolic sine  

include/pal_math.h

@@ -158,6 +158,9 @@ void p_log10_f32(const float *a, float *c, int n);
 /*element raised to a power: c = pow ( a , b ) */
 void p_pow_f32(const float *a, const float *b, float *c, int n);
 
+/*round: c = round ( a ) */
+void p_round_f32(const float *a, float *c, int n);
+
 /*sine: c = sin ( a ) */
 void p_sin_f32(const float *a, float *c, int n);
 

src/math/Makefile.am

@@ -38,6 +38,7 @@ libpal_math_la_SOURCES = \
     p_popcount.c \
     p_pow.c \
     p_rand.c \
+    p_round.c \
     p_sin.c \
     p_sincos.c \
     p_sinh.c \

src/math/p_round.c

@@ -0,0 +1,31 @@
+#include <pal.h>
+
+/*
+ *
+ * Compute the round of a
+ *
+ * @param a     Pointer to input vector
+ *
+ * @param c     Pointer to output vector
+ *
+ * @param n     Size of 'a' and 'c' vector.
+ *
+ * @return      None
+ *
+ */
+
+void p_round_f32(const float *a, float *c, int n)
+{
+    int i;int sign; float ff; float x;
+    for ( i = 0; i < n; i++)
+      {
+       const float *pa = (a+i);
+             float *pc = (c+i);
+       ff = *pa;
+       if (ff >= 0) sign = 1.0; else sign = -1.0;       // store sign of input variable
+       if (ff < 0) x = ff * (-1.0f); else x = ff;       // this is equivalent of x=abs(a)
+
+  if (x >= 0x1.0p23)  *pc = sign*x;
+  *pc = sign*((float) (unsigned int) (x + 0.49999997f));
+      }
+}

tests/math/Makefile.am

@@ -52,6 +52,7 @@ BUILT_SOURCES = \
     gold/p_mode_f32.gold.h \
     gold/p_mul_f32.gold.h \
     gold/p_pow_f32.gold.h \
+    gold/p_round_f32.gold.h \
     gold/p_sin_f32.gold.h \
     gold/p_sincos_f32.gold.h \
     gold/p_sinh_f32.gold.h \
@@ -114,6 +115,7 @@ check_PROGRAMS = \
 	check_p_popcount \
 	check_p_pow_f32 \
 	check_p_rand \
+        check_p_round_f32 \
 	check_p_sin_f32 \
 	check_p_sincos_f32 \
 	check_p_sinh_f32 \
@@ -159,6 +161,7 @@ check_p_mul_f32_SOURCES         = $(SIMPLE)
 check_p_popcount_SOURCES        = $(NOTEST)
 check_p_pow_f32_SOURCES         = $(SIMPLE) p_pow.c
 check_p_rand_SOURCES            = $(NOTEST)
+check_p_round_f32_SOURCES       = $(SIMPLE) p_round.c
 check_p_sin_f32_SOURCES         = $(SIMPLE) p_sin.c
 check_p_sincos_f32_SOURCES      = check_p_sincos.c
 check_p_sinh_f32_SOURCES        = $(SIMPLE) p_sinh.c
@@ -205,6 +208,7 @@ check_p_popcount_CFLAGS         = -DFUNCTION=p_popcount
 # Override fault tolerance for p_pow_f32
 check_p_pow_f32_CFLAGS          = -DFUNCTION=p_pow_f32     -DIS_BINARY -DEPSILON_MAX=0.003f
 check_p_rand_CFLAGS             = -DFUNCTION=p_rand                    -DSCALAR_OUTPUT
+check_p_round_f32_CFLAGS        = -DFUNCTION=p_round_f32   -DIS_UNARY
 check_p_sin_f32_CFLAGS          = -DFUNCTION=p_sin_f32     -DIS_UNARY
 check_p_sincos_f32_CFLAGS       = -DFUNCTION=p_sincos_f32
 check_p_sinh_f32_CFLAGS         = -DFUNCTION=p_sinh_f32    -DIS_UNARY
@@ -251,6 +255,7 @@ check_p_mul_f32_LDFLAGS         = $(CHECKLDFLAGS)
 check_p_popcount_LDFLAGS        = $(CHECKLDFLAGS)
 check_p_pow_f32_LDFLAGS         = $(CHECKLDFLAGS)
 check_p_rand_LDFLAGS            = $(CHECKLDFLAGS)
+check_p_round_f32_LDFLAGS       = $(CHECKLDFLAGS)
 check_p_sin_f32_LDFLAGS         = $(CHECKLDFLAGS)
 check_p_sincos_f32_LDFLAGS      = $(CHECKLDFLAGS)
 check_p_sinh_f32_LDFLAGS        = $(CHECKLDFLAGS)

tests/math/gold/p_round_f32.dat

@@ -0,0 +1,100 @@
+4.625243,0.000000,0.000000,5.000000
+10.712888,0.000000,0.000000,11.000000
+2.252235,0.000000,0.000000,2.000000
+2.280635,0.000000,0.000000,2.000000
+10.588324,0.000000,0.000000,11.000000
+6.065655,0.000000,0.000000,6.000000
+1.591502,0.000000,0.000000,2.000000
+10.343284,0.000000,0.000000,10.000000
+5.718785,0.000000,0.000000,6.000000
+5.982005,0.000000,0.000000,6.000000
+10.705152,0.000000,0.000000,11.000000
+7.696878,0.000000,0.000000,8.000000
+8.047289,0.000000,0.000000,8.000000
+10.776376,0.000000,0.000000,11.000000
+9.148800,0.000000,0.000000,9.000000
+3.082119,0.000000,0.000000,3.000000
+3.194645,0.000000,0.000000,3.000000
+7.935914,0.000000,0.000000,8.000000
+7.476189,0.000000,0.000000,7.000000
+7.470324,0.000000,0.000000,7.000000
+6.958097,0.000000,0.000000,7.000000
+8.072808,0.000000,0.000000,8.000000
+8.835902,0.000000,0.000000,9.000000
+4.629804,0.000000,0.000000,5.000000
+8.474785,0.000000,0.000000,8.000000
+6.133525,0.000000,0.000000,6.000000
+3.945584,0.000000,0.000000,4.000000
+9.232484,0.000000,0.000000,9.000000
+6.067547,0.000000,0.000000,6.000000
+8.317123,0.000000,0.000000,8.000000
+10.442951,0.000000,0.000000,10.000000
+10.692790,0.000000,0.000000,11.000000
+5.546363,0.000000,0.000000,6.000000
+10.211538,0.000000,0.000000,10.000000
+10.489777,0.000000,0.000000,10.000000
+5.134687,0.000000,0.000000,5.000000
+5.277193,0.000000,0.000000,5.000000
+1.081279,0.000000,0.000000,1.000000
+1.994324,0.000000,0.000000,2.000000
+8.512331,0.000000,0.000000,9.000000
+7.063284,0.000000,0.000000,7.000000
+1.699476,0.000000,0.000000,2.000000
+5.209209,0.000000,0.000000,5.000000
+1.626925,0.000000,0.000000,2.000000
+1.475853,0.000000,0.000000,1.000000
+0.874361,0.000000,0.000000,1.000000
+2.225396,0.000000,0.000000,2.000000
+2.186850,0.000000,0.000000,2.000000
+8.810275,0.000000,0.000000,9.000000
+7.217938,0.000000,0.000000,7.000000
+-9.657174,0.000000,0.000000,-10.000000
+-2.284724,0.000000,0.000000,-2.000000
+-4.290746,0.000000,0.000000,-4.000000
+-5.009429,0.000000,0.000000,-5.000000
+-4.430880,0.000000,0.000000,-4.000000
+-1.765531,0.000000,0.000000,-2.000000
+-0.142954,0.000000,0.000000,-0.000000
+-8.376464,0.000000,0.000000,-8.000000
+-10.998015,0.000000,0.000000,-11.000000
+-6.210501,0.000000,0.000000,-6.000000
+-3.209939,0.000000,0.000000,-3.000000
+-7.957318,0.000000,0.000000,-8.000000
+-3.419643,0.000000,0.000000,-3.000000
+-8.756303,0.000000,0.000000,-9.000000
+-7.168857,0.000000,0.000000,-7.000000
+-2.909420,0.000000,0.000000,-3.000000
+-0.407342,0.000000,0.000000,-0.000000
+-9.962402,0.000000,0.000000,-10.000000
+-3.990700,0.000000,0.000000,-4.000000
+-2.401666,0.000000,0.000000,-2.000000
+-4.991085,0.000000,0.000000,-5.000000
+-8.570336,0.000000,0.000000,-9.000000
+-1.617495,0.000000,0.000000,-2.000000
+-7.716646,0.000000,0.000000,-8.000000
+-7.713614,0.000000,0.000000,-8.000000
+-0.609700,0.000000,0.000000,-1.000000
+-8.591007,0.000000,0.000000,-9.000000
+-9.939010,0.000000,0.000000,-10.000000
+-2.796550,0.000000,0.000000,-3.000000
+-3.917634,0.000000,0.000000,-4.000000
+-6.156948,0.000000,0.000000,-6.000000
+-9.970076,0.000000,0.000000,-10.000000
+-6.202358,0.000000,0.000000,-6.000000
+-7.964046,0.000000,0.000000,-8.000000
+-3.979505,0.000000,0.000000,-4.000000
+-10.633238,0.000000,0.000000,-11.000000
+-7.245930,0.000000,0.000000,-7.000000
+-4.122460,0.000000,0.000000,-4.000000
+-5.526055,0.000000,0.000000,-6.000000
+-4.760297,0.000000,0.000000,-5.000000
+-7.849313,0.000000,0.000000,-8.000000
+-6.252346,0.000000,0.000000,-6.000000
+-10.233968,0.000000,0.000000,-10.000000
+-0.268957,0.000000,0.000000,-0.000000
+-4.008649,0.000000,0.000000,-4.000000
+-6.402825,0.000000,0.000000,-6.000000
+-3.178377,0.000000,0.000000,-3.000000
+-1.932344,0.000000,0.000000,-2.000000
+-2.881579,0.000000,0.000000,-3.000000
+-4.685429,0.000000,0.000000,-5.000000

tests/math/p_round.c

@@ -0,0 +1,10 @@
+#include <math.h>
+#include "simple.h"
+
+void generate_ref(float *out, size_t n)
+{
+    size_t i;
+
+    for (i = 0; i < n; i++)
+        out[i] = roundf(ai[i]);
+}