fixed 8-bit int to float16 conversions

Author: davidssmith

c/ra.c

@@ -402,13 +402,13 @@ ra_convert (ra_t *r, const uint64_t eltype, const uint64_t elbyte)
     else if CASE(UINT,2,FLOAT,2)
         CONVERT_TO_F16(uint16_t,float16)
     else if CASE(FLOAT,2,INT,1)
-        CONVERT_TO_F16(float16,int8_t)
+        CONVERT_FROM_F16(float16,int8_t)
     else if CASE(FLOAT,2,UINT,1)
-        CONVERT_TO_F16(float16,uint8_t)
+        CONVERT_FROM_F16(float16,uint8_t)
     else if CASE(FLOAT,2,INT,2)
-        CONVERT_TO_F16(float16,int16_t)
+        CONVERT_FROM_F16(float16,int16_t)
     else if CASE(FLOAT,2,UINT,2)
-        CONVERT_TO_F16(float16,uint16_t)
+        CONVERT_FROM_F16(float16,uint16_t)
 
     else if CASE(FLOAT,4,UINT,1)
         CONVERT(float,uint8_t)

c/runtests.sh

@@ -61,7 +61,6 @@ dotest ../data/randu8.ra 3 8
 dotest ../data/randu32.ra 2 8
 dotest ../data/randf16.ra 3 4
 dotest ../data/randf16.ra 3 8
-#dotest ../data/randf32.ra 3 2
 dotest ../data/randf32.ra 3 4
 dotest ../data/randf32.ra 3 8
 dotest ../data/randc32.ra 4 8