diff --git a/cmd/v/internal/help/build.txt b/cmd/v/internal/help/build.txt
index 94cec5896f..7c89477728 100644
--- a/cmd/v/internal/help/build.txt
+++ b/cmd/v/internal/help/build.txt
@@ -65,8 +65,15 @@ The build flags are shared by the build and run commands:
    -prod
       Compile the executable in production mode where most optimizations are enabled.
 
-   -prof, -profile
+   -prof, -profile <file.txt>
       Compile the executable with all functions profiled.
+      The profile results will be stored in `file.txt`.
+      The format is 4 fields, separated by a space, for each v function:
+         a) how many times it was called
+         b) how much *nanoseconds in total* it took
+         c) an average for each function (i.e. (b) / (a) )
+         d) the function name
+      NB: if you want to output the profile info to stdout, use `-profile -`.
 
    -stats
       Enable more detailed statistics reporting, while compiling test files.
diff --git a/cmd/v/v.v b/cmd/v/v.v
index c4784633dc..b6b410f1e9 100644
--- a/cmd/v/v.v
+++ b/cmd/v/v.v
@@ -146,7 +146,9 @@ fn parse_args(args []string) (&pref.Preferences, string) {
 				res.is_bare = true
 			}
 			'-prof', '-profile' {
+				res.profile_file = cmdline.option(current_args, '-profile', '-')
 				res.is_prof = true
+				i++
 			}
 			'-prod' {
 				res.is_prod = true
diff --git a/vlib/v/gen/fn.v b/vlib/v/gen/fn.v
index 3cac3a779d..9379771289 100644
--- a/vlib/v/gen/fn.v
+++ b/vlib/v/gen/fn.v
@@ -141,8 +141,17 @@ fn (mut g Gen) gen_fn_decl(it ast.FnDecl) {
 	if is_main {
 		if g.pref.is_prof {
 			g.pcs_declarations.writeln('void vprint_profile_stats(){')
-			for pfn_name, pcounter_name in g.pcs {
-				g.pcs_declarations.writeln('\tif (${pcounter_name}_calls) printf("%llu %f %f ${pfn_name} \\n", ${pcounter_name}_calls, $pcounter_name, $pcounter_name / ${pcounter_name}_calls );')
+			if g.pref.profile_file == '-' {
+				for pfn_name, pcounter_name in g.pcs {
+					g.pcs_declarations.writeln('\tif (${pcounter_name}_calls) printf("%llu %f %f ${pfn_name} \\n", ${pcounter_name}_calls, $pcounter_name, $pcounter_name / ${pcounter_name}_calls );')
+				}
+			}else{
+				g.pcs_declarations.writeln('\tFILE * fp;')
+				g.pcs_declarations.writeln('\tfp = fopen ("${g.pref.profile_file}", "w+");')
+				for pfn_name, pcounter_name in g.pcs {
+					g.pcs_declarations.writeln('\tif (${pcounter_name}_calls) fprintf(fp, "%llu %f %f ${pfn_name} \\n", ${pcounter_name}_calls, $pcounter_name, $pcounter_name / ${pcounter_name}_calls );')
+				}
+				g.pcs_declarations.writeln('\tfclose(fp);')
 			}
 			g.pcs_declarations.writeln('}')
 		}
diff --git a/vlib/v/pref/pref.v b/vlib/v/pref/pref.v
index a84850d553..ddf6233236 100644
--- a/vlib/v/pref/pref.v
+++ b/vlib/v/pref/pref.v
@@ -31,6 +31,7 @@ pub mut:
 	is_live             bool // main program that contains live/hot code
 	is_shared               bool // an ordinary shared library, -shared, no matter if it is live or not
 	is_prof             bool // benchmark every function
+	profile_file        string // the profile results will be stored inside profile_file
 	translated          bool // `v translate doom.v` are we running V code translated from C? allow globals, ++ expressions, etc
 	is_prod             bool // use "-O2"
 	obfuscate           bool // `v -obf program.v`, renames functions to "f_XXX"