util: Remove usage of USE_**_ASM macros

Use DETECT_ARCH_X86 DETECT_ARCH_X86_64 DETECT_ARCH_AARCH64 DETECT_CC_GCC instead

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36635>

src/mesa/main/debug.c

@@ -134,7 +134,7 @@ void _mesa_print_info( struct gl_context *ctx )
     */
    _mesa_debug(NULL, "Mesa GL_EXTENSIONS = %s\n", ctx->Extensions.String);
 
-#if DETECT_ARCH_X86 || DETECT_ARCH_X86_64
+#if DETECT_ARCH_X86
    _mesa_debug(NULL, "Mesa x86-optimized: YES\n");
 #else
    _mesa_debug(NULL, "Mesa x86-optimized: NO\n");

src/util/bitscan.h

@@ -42,7 +42,9 @@
 #include <popcntintrin.h>
 #endif
 
-#include "macros.h"
+#include "util/detect_arch.h"
+#include "util/detect_cc.h"
+#include "util/macros.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -363,7 +365,7 @@ util_bitcount(unsigned n)
 static inline unsigned
 util_popcnt_inline_asm(unsigned n)
 {
-#if defined(USE_X86_64_ASM) || defined(USE_X86_ASM)
+#if (DETECT_ARCH_X86 || DETECT_ARCH_X86_64) && DETECT_CC_GCC
    uint32_t out;
    __asm volatile("popcnt %1, %0" : "=r"(out) : "r"(n));
    return out;

src/util/half_float.h

@@ -29,9 +29,11 @@
 #include <stdbool.h>
 #include <stdint.h>
 #include <string.h>
+#include "util/detect_arch.h"
+#include "util/detect_cc.h"
 #include "util/u_cpu_detect.h"
 
-#if defined(USE_X86_64_ASM)
+#if DETECT_ARCH_X86_64
 #include <xmmintrin.h>
 #endif
 
@@ -57,7 +59,7 @@ uint16_t _mesa_float_to_float16_rtz_slow(float val);
 static inline uint16_t
 _mesa_float_to_half(float val)
 {
-#if defined(USE_X86_64_ASM)
+#if DETECT_ARCH_X86_64 && DETECT_CC_GCC
    if (util_get_cpu_caps()->has_f16c) {
       __m128 in = {val};
       __m128i out;
@@ -73,7 +75,7 @@ _mesa_float_to_half(float val)
 static inline float
 _mesa_half_to_float(uint16_t val)
 {
-#if defined(USE_X86_64_ASM)
+#if DETECT_ARCH_X86_64 && DETECT_CC_GCC
    if (util_get_cpu_caps()->has_f16c) {
       __m128i in = {val};
       __m128 out;
@@ -81,7 +83,7 @@ _mesa_half_to_float(uint16_t val)
       __asm volatile("vcvtph2ps %1, %0" : "=v"(out) : "v"(in));
       return out[0];
    }
-#elif defined(USE_AARCH64_ASM)
+#elif DETECT_ARCH_AARCH64
    float result;
    uint16_t in = val;
 
@@ -98,7 +100,7 @@ _mesa_half_to_float(uint16_t val)
 static inline uint16_t
 _mesa_float_to_float16_rtz(float val)
 {
-#if defined(USE_X86_64_ASM)
+#if DETECT_ARCH_X86_64 && DETECT_CC_GCC
    if (util_get_cpu_caps()->has_f16c) {
       __m128 in = {val};
       __m128i out;

src/util/streaming-load-memcpy.c

@@ -26,6 +26,7 @@
  *
  */
 
+#include "util/detect_arch.h"
 #include "util/streaming-load-memcpy.h"
 #include "util/u_cpu_detect.h"
 #include "util/u_math.h"
@@ -42,7 +43,7 @@ util_streaming_load_memcpy(void *restrict dst, void *restrict src, size_t len)
    char *restrict d = dst;
    char *restrict s = src;
 
-#if defined(USE_SSE41) || defined(USE_AARCH64_ASM)
+#if defined(USE_SSE41) || DETECT_ARCH_AARCH64
    /* If dst and src are not co-aligned, or if non-temporal load instructions
     * are not present, fallback to memcpy(). */
    if (((uintptr_t)d & 15) != ((uintptr_t)s & 15)
@@ -91,7 +92,7 @@ util_streaming_load_memcpy(void *restrict dst, void *restrict src, size_t len)
       len -= 64;
    }
 
-#elif defined(USE_AARCH64_ASM)
+#elif DETECT_ARCH_AARCH64
    if (len >= 64) {
       __asm__ volatile(
          /* Memory barrier for loads completion in the non-shareable domain:

src/util/u_math.h

@@ -130,7 +130,7 @@ util_fast_log2(float x)
 static inline int
 util_ifloor(float f)
 {
-#if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
+#if DETECT_ARCH_X86 && DETECT_CC_GCC
    /*
     * IEEE floor for computers that round to nearest or even.
     * 'f' must be between -4194304 and 4194303.