Add hybrid detect for Intel laptop cpus

- Add +i8mm build option for sve ARGBToUV which uses usdot
- util/cpuid Get cpu count (windows, macos, linux)
- For each x86 cpu, detect hybrid (e-core)
- Includes a comment fix for ubsan unittest
- Bump version
- Apply clang format to util/*.c as well as all *.cc/*.h

Bug: 424637372
Change-Id: I08310e18051fff62c9e4e4a10d1e4361871119ac
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6635640
Reviewed-by: Wan-Teh Chang <wtc@google.com>

BUILD.gn

@@ -254,7 +254,7 @@ if (libyuv_use_sve) {
     public_configs = [ ":libyuv_config" ]
 
     # SVE2 is an Armv9-A feature.
-    cflags = [ "-march=armv9-a+sve2" ]
+    cflags = [ "-march=armv9-a+sve2+i8mm" ]
   }
 }
 

README.chromium

@@ -1,6 +1,6 @@
 Name: libyuv
 URL: https://chromium.googlesource.com/libyuv/libyuv/
-Version: 1911
+Version: 1912
 License: BSD-3-Clause
 License File: LICENSE
 Shipped: yes

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 1911
+#define LIBYUV_VERSION 1912
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_

source/compare_neon64.cc

@@ -116,8 +116,7 @@ uint32_t HashDjb2_NEON(const uint8_t* src, int count, uint32_t seed) {
   uint32_t hash = seed;
   const uint32_t c16 = 0x92d9e201;  // 33^16
   uint32_t tmp, tmp2;
-      asm(
+      asm("ld1         {v16.4s, v17.4s, v18.4s, v19.4s}, [%[kIdx]] \n"
-      "ld1         {v16.4s, v17.4s, v18.4s, v19.4s}, [%[kIdx]] \n"
       "ld1         {v4.4s, v5.4s, v6.4s, v7.4s}, [%[kMuls]] \n"
 
       // count is always a multiple of 16.

source/row_gcc.cc

@@ -1734,9 +1734,13 @@ void ARGBToUV444MatrixRow_AVX2(const uint8_t* src_argb,
 
 #ifdef HAS_ARGBTOUVROW_SSSE3
 
-void OMITFP ARGBToUVMatrixRow_SSSE3(
+void OMITFP
-    const uint8_t* src_argb, int src_stride_argb, uint8_t* dst_u,
+ARGBToUVMatrixRow_SSSE3(const uint8_t* src_argb,
-    uint8_t* dst_v, int width, const struct RgbUVConstants* rgbuvconstants) {
+                        int src_stride_argb,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width,
+                        const struct RgbUVConstants* rgbuvconstants) {
   asm volatile(
       "movdqa      %0,%%xmm3                     \n"
       "movdqa      %1,%%xmm4                     \n"
@@ -1821,9 +1825,13 @@ static const UVMatrixConstants kShufARGBToUV_AVX = {
     0, 128, 0, 128, 0, 128, 0,  128, 0, 128, 0,  128, 0, 128, 0,  128,
     0, 128, 0, 128, 0, 128, 0,  128, 0, 128, 0,  128, 0, 128, 0,  128};
 
-void OMITFP ARGBToUVMatrixRow_AVX2(
+void OMITFP
-    const uint8_t* src_argb, int src_stride_argb, uint8_t* dst_u,
+ARGBToUVMatrixRow_AVX2(const uint8_t* src_argb,
-    uint8_t* dst_v, int width, const struct RgbUVConstants* rgbuvconstants) {
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width,
+                       const struct RgbUVConstants* rgbuvconstants) {
   asm volatile(
       "vbroadcastf128 %0,%%ymm6                  \n"
       "vbroadcastf128 %1,%%ymm7                  \n"

unit_test/convert_argb_test.cc

@@ -2744,7 +2744,7 @@ TEST_F(LibYUVConvertTest, TestI400LargeSize) {
   free_aligned_buffer_page_end(dest_argb);
   free_aligned_buffer_page_end(orig_i400);
 }
-#endif  // defined(_M_X64) || defined(_M_X64) || defined(__aarch64__)
+#endif  // defined(__x86_64__) || defined(_M_X64) || defined(__aarch64__)
 
 #endif  // !defined(LEAN_TESTS)
 

util/cpuid.c

@@ -15,6 +15,13 @@
 #ifdef __linux__
 #include <ctype.h>
 #include <sys/utsname.h>
+#include <unistd.h>  // for sysconf
+#endif
+#if defined(_WIN32)
+#include <windows.h>  // for GetSystemInfo
+#endif
+#if defined(__APPLE__)
+#include <sys/sysctl.h>  // for sysctlbyname
 #endif
 
 #include "libyuv/cpu_id.h"
@@ -51,6 +58,23 @@ int main(int argc, const char* argv[]) {
     printf("Kernel Version %d.%d\n", kernelversion[0], kernelversion[1]);
   }
 #endif  // defined(__linux__)
+#if defined(_WIN32)
+  SYSTEM_INFO sysInfo;
+  GetSystemInfo(&sysInfo);
+  int num_cpus = (int)sysInfo.dwNumberOfProcessors;
+#elif defined(__linux__)
+  int num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+#elif defined(__APPLE__)
+  int num_cpus = 0;
+  size_t num_cpus_len = sizeof(num_cpus);
+  // Get the number of logical CPU cores
+  if (sysctlbyname("hw.logicalcpu", &num_cpus, &num_cpus_len, NULL, 0) == -1) {
+    printf("sysctlbyname failed to get hw.logicalcpu\n");
+  }
+#else
+  int num_cpus = 0;  // unknown OS
+#endif
+  printf("Number of cpus: %d\n", num_cpus);
 
 #if defined(__arm__) || defined(__aarch64__)
   int has_arm = TestCpuFlag(kCpuHasARM);
@@ -75,7 +99,8 @@ int main(int argc, const char* argv[]) {
     // Read and print the SVE and SME vector lengths.
     if (has_sve) {
       int sve_vl;
-      __asm__(".inst 0x04bf5020    \n"  // rdvl x0, #1
+      __asm__(
+          ".inst 0x04bf5020    \n"  // rdvl x0, #1
           "mov %w[sve_vl], w0  \n"
           : [sve_vl] "=r"(sve_vl)  // %[sve_vl]
           :
@@ -84,7 +109,8 @@ int main(int argc, const char* argv[]) {
     }
     if (has_sme) {
       int sme_vl;
-      __asm__(".inst 0x04bf5820    \n"  // rdsvl x0, #1
+      __asm__(
+          ".inst 0x04bf5820    \n"  // rdsvl x0, #1
           "mov %w[sme_vl], w0  \n"
           : [sme_vl] "=r"(sme_vl)  // %[sme_vl]
           :
@@ -131,8 +157,8 @@ int main(int argc, const char* argv[]) {
   }
 #endif  // defined(__loongarch__)
 
-#if defined(__i386__) || defined(__x86_64__) || \
+#if defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) || \
-    defined(_M_IX86) || defined(_M_X64)
+    defined(_M_X64)
   int has_x86 = TestCpuFlag(kCpuHasX86);
   if (has_x86) {
     int family, model, cpu_info[4];
@@ -153,6 +179,13 @@ int main(int argc, const char* argv[]) {
     cpu_info[3] = 0;
     printf("Cpu Vendor: %s\n", (char*)(&cpu_info[0]));
 
+    for (int n = 0; n < num_cpus; ++n) {
+      // Check EDX bit 15 for hybrid design indication
+      CpuId(7, n, &cpu_info[0]);
+      int hybrid = (cpu_info[3] >> 15) & 1;
+      printf("  Cpu %d Hybrid %d\n", n, hybrid);
+    }
+
     // CPU Family and Model
     // 3:0 - Stepping
     // 7:4 - Model
@@ -163,8 +196,8 @@ int main(int argc, const char* argv[]) {
     CpuId(1, 0, &cpu_info[0]);
     family = ((cpu_info[0] >> 8) & 0x0f) | ((cpu_info[0] >> 16) & 0xff0);
     model = ((cpu_info[0] >> 4) & 0x0f) | ((cpu_info[0] >> 12) & 0xf0);
-    printf("Cpu Family %d (0x%x), Model %d (0x%x)\n", family, family,
+    printf("Cpu Family %d (0x%x), Model %d (0x%x)\n", family, family, model,
-           model, model);
+           model);
 
     int has_sse2 = TestCpuFlag(kCpuHasSSE2);
     int has_ssse3 = TestCpuFlag(kCpuHasSSSE3);
@@ -210,7 +243,7 @@ int main(int argc, const char* argv[]) {
     printf("Has AVXVNNIINT8 0x%x\n", has_avxvnniint8);
     printf("Has AMXINT8 0x%x\n", has_amxint8);
   }
-#endif  // defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) || defined(_M_X64)
+#endif  // defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) ||
+        // defined(_M_X64)
   return 0;
 }
-

util/yuvconstants.c

@@ -76,10 +76,14 @@ int main(int argc, const char* argv[]) {
   //
   // // U and V contributions to R,G,B.
 
-  printf("UB %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ub * 64), ub, ub * 64);
+  printf("UB %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ub * 64), ub,
-  printf("UG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ug * 64), ug, ug * 64);
+         ub * 64);
-  printf("VG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vg * 64), vg, vg * 64);
+  printf("UG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ug * 64), ug,
-  printf("VR %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vr * 64), vr, vr * 64);
+         ug * 64);
+  printf("VG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vg * 64), vg,
+         vg * 64);
+  printf("VR %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vr * 64), vr,
+         vr * 64);
 
   vr = 255.f / 224.f * 2 * (1 - kr);
   ug = 255.f / 224.f * 2 * ((1 - kb) * kb / kg);
@@ -97,10 +101,14 @@ int main(int argc, const char* argv[]) {
   //
   // // U and V contributions to R,G,B.
 
-  printf("UB %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ub * 64), ub, ub * 64);
+  printf("UB %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ub * 64), ub,
-  printf("UG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ug * 64), ug, ug * 64);
+         ub * 64);
-  printf("VG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vg * 64), vg, vg * 64);
+  printf("UG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(ug * 64), ug,
-  printf("VR %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vr * 64), vr, vr * 64);
+         ug * 64);
+  printf("VG %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vg * 64), vg,
+         vg * 64);
+  printf("VR %-3.0f /* round(%f * 64 = %8.4f) */\n", round(vr * 64), vr,
+         vr * 64);
 
   return 0;
 }