pref: support -fast-math, passing either -ffast-math or /fp:fast (f…

…or msvc) to the C backend, and `$if fast_math {` to detect it at comptime

doc/docs.md

@@ -5613,7 +5613,7 @@ Full list of builtin options:
 | `mac`, `darwin`, `ios`,        | `clang`, `mingw` | `i386`, `arm32`               | `js`, `glibc`, `prealloc`                     |
 | `android`, `mach`, `dragonfly` | `msvc`           | `x64`, `x32`                  | `no_bounds_checking`, `freestanding`          |
 | `gnu`, `hpux`, `haiku`, `qnx`  | `cplusplus`      | `little_endian`, `big_endian` | `no_segfault_handler`, `no_backtrace`         |
-| `solaris`, `termux`            |                  |                               | `no_main`                                     |
+| `solaris`, `termux`            |                  |                               | `no_main`, 'fast_math'                        |
 
 #### `$embed_file`
 

vlib/v/ast/comptime_valid_idents.v

@@ -4,14 +4,14 @@ pub const (
 	valid_comptime_if_os             = ['windows', 'ios', 'macos', 'mach', 'darwin', 'hpux', 'gnu',
 		'qnx', 'linux', 'freebsd', 'openbsd', 'netbsd', 'bsd', 'dragonfly', 'android', 'termux',
 		'solaris', 'haiku', 'serenity', 'vinix']
-	valid_comptime_compression_types = ['none', 'zlib']
 	valid_comptime_if_compilers      = ['gcc', 'tinyc', 'clang', 'mingw', 'msvc', 'cplusplus']
 	valid_comptime_if_platforms      = ['amd64', 'i386', 'aarch64', 'arm64', 'arm32', 'rv64', 'rv32']
 	valid_comptime_if_cpu_features   = ['x64', 'x32', 'little_endian', 'big_endian']
 	valid_comptime_if_other          = ['apk', 'js', 'debug', 'prod', 'test', 'glibc', 'prealloc',
 		'no_bounds_checking', 'freestanding', 'threads', 'js_node', 'js_browser', 'js_freestanding',
-		'interpreter', 'es5', 'profile', 'wasm32', 'wasm32_emscripten', 'wasm32_wasi']
+		'interpreter', 'es5', 'profile', 'wasm32', 'wasm32_emscripten', 'wasm32_wasi', 'fast_math']
 	valid_comptime_not_user_defined  = all_valid_comptime_idents()
+	valid_comptime_compression_types = ['none', 'zlib']
 )
 
 fn all_valid_comptime_idents() []string {

vlib/v/gen/c/comptime.v

@@ -1157,6 +1157,14 @@ fn (mut g Gen) comptime_if_to_ifdef(name string, is_comptime_option bool) !strin
 		'big_endian' {
 			return 'TARGET_ORDER_IS_BIG'
 		}
+		'fast_math' {
+			if g.pref.ccompiler_type == .msvc {
+				// turned on by: `-cflags /fp:fast`
+				return '_M_FP_FAST'
+			}
+			// turned on by: `-cflags -ffast-math`
+			return '__FAST_MATH__'
+		}
 		else {
 			if is_comptime_option
 				|| (g.pref.compile_defines_all.len > 0 && name in g.pref.compile_defines_all) {

vlib/v/help/build/build-c.txt

@@ -318,3 +318,11 @@ see also `v help build`.
       It may be decreased, to reduce the memory footprint of programs that launch
       hundreds/thousands of threads, but where each of the threads does not need
       a big stack.
+
+   -fast-math
+      When present, pass either -ffast-math or /fp:fast (for msvc) to the C compiler.
+      See https://en.wikipedia.org/wiki/Floating-point_arithmetic#%22Fast_math%22_optimization
+          https://learn.microsoft.com/en-us/cpp/build/reference/fp-specify-floating-point-behavior?view=msvc-170&redirectedfrom=MSDN
+          https://clang.llvm.org/docs/UsersManual.html#cmdoption-ffast-math
+          https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-ffast-math
+

vlib/v/pref/pref.v

@@ -227,11 +227,13 @@ pub mut:
 	message_limit       int = 150 // the maximum amount of warnings/errors/notices that will be accumulated
 	nofloat             bool // for low level code, like kernels: replaces f32 with u32 and f64 with u64
 	use_coroutines      bool // experimental coroutines
+	fast_math           bool // -fast-math will pass either -ffast-math or /fp:fast (for msvc) to the C backend
 	// checker settings:
 	checker_match_exhaustive_cutoff_limit int = 12
 	thread_stack_size                     int = 8388608 // Change with `-thread-stack-size 4194304`. Note: on macos it was 524288, which is too small for more complex programs with many nested callexprs.
-	wasm_stack_top                        int = 1024 + (16 * 1024) // stack size for webassembly backend
-	wasm_validate                         bool // validate webassembly code, by calling `wasm-validate`
+	// wasm settings:
+	wasm_stack_top int = 1024 + (16 * 1024) // stack size for webassembly backend
+	wasm_validate  bool // validate webassembly code, by calling `wasm-validate`
 }
 
 pub fn parse_args(known_external_commands []string, args []string) (&Preferences, string) {
@@ -395,6 +397,9 @@ pub fn parse_args_and_show_errors(known_external_commands []string, args []strin
 				res.nofloat = true
 				res.compile_defines_all << 'nofloat' // so that `$if nofloat? {` works
 			}
+			'-fast-math' {
+				res.fast_math = true
+			}
 			'-e' {
 				res.is_eval_argument = true
 				res.eval_argument = cmdline.option(current_args, '-e', '')
@@ -903,7 +908,13 @@ pub fn parse_args_and_show_errors(known_external_commands []string, args []strin
 		eprintln('Cannot save output binary in a .v file.')
 		exit(1)
 	}
-
+	if res.fast_math {
+		if res.ccompiler_type == .msvc {
+			res.cflags += ' /fp:fast'
+		} else {
+			res.cflags += ' -ffast-math'
+		}
+	}
 	if res.is_eval_argument {
 		// `v -e "println(2+5)"`
 		run_code_in_tmp_vfile_and_exit(args, mut res, '-e', 'vsh', res.eval_argument)