Support for extended feature flags enumeration leaf in CPUID instruction

This CL fixes an overcite with the AVX2 support CL previously
merged (Change-Id: Idc03f3fca4bf2d0afd33631ea1d3caf8fc34ec29) that
prevented runtime execution of AVX2 code in WebM.

Background:
Starting with the Sandybridge processor, the CPUID instruction was
enhanced to add various extended feature flag enumeration leaves.
Reading these leaves requires an additional input value for the CPUID
instruction which is stored in ECX. This change adds this second input
value for all ARCH_X86 and ARCH_x86_64 targets to the CPUID macros,
allowing checks of EBX bit 5 for AVX2 support. This capability will be
required moving forward to check for future processor features.

Change-Id: Ie9d872bc9ff68dad4b6578e4544e4dfd0ae26c36

diff --git a/vpx_ports/x86.h b/vpx_ports/x86.h
index 29905834772b37baceef8d182c52e7f8513d10b4..344a5739b09e7042dd3703d9c2fb90902b98677d 100644 (file)
--- a/vpx_ports/x86.h
+++ b/vpx_ports/x86.h
@@ -35,51 +35,53 @@ typedef enum {
 
 #if defined(__GNUC__) && __GNUC__ || defined(__ANDROID__)
 #if ARCH_X86_64
-#define cpuid(func,ax,bx,cx,dx)\
+#define cpuid(func, func2, ax, bx, cx, dx)\
   __asm__ __volatile__ (\
                         "cpuid           \n\t" \
                         : "=a" (ax), "=b" (bx), "=c" (cx), "=d" (dx) \
-                        : "a"  (func));
+                        : "a" (func), "c" (func2));
 #else
-#define cpuid(func,ax,bx,cx,dx)\
+#define cpuid(func, func2, ax, bx, cx, dx)\
   __asm__ __volatile__ (\
                         "mov %%ebx, %%edi   \n\t" \
                         "cpuid              \n\t" \
                         "xchg %%edi, %%ebx  \n\t" \
                         : "=a" (ax), "=D" (bx), "=c" (cx), "=d" (dx) \
-                        : "a" (func));
+                        : "a" (func), "c" (func2));
 #endif
 #elif defined(__SUNPRO_C) || defined(__SUNPRO_CC) /* end __GNUC__ or __ANDROID__*/
 #if ARCH_X86_64
-#define cpuid(func,ax,bx,cx,dx)\
+#define cpuid(func, func2, ax, bx, cx, dx)\
   asm volatile (\
                 "xchg %rsi, %rbx \n\t" \
                 "cpuid           \n\t" \
                 "movl %ebx, %edi \n\t" \
                 "xchg %rsi, %rbx \n\t" \
                 : "=a" (ax), "=D" (bx), "=c" (cx), "=d" (dx) \
-                : "a"  (func));
+                : "a" (func), "c" (func2));
 #else
-#define cpuid(func,ax,bx,cx,dx)\
+#define cpuid(func, func2, ax, bx, cx, dx)\
   asm volatile (\
                 "pushl %ebx       \n\t" \
                 "cpuid            \n\t" \
                 "movl %ebx, %edi  \n\t" \
                 "popl %ebx        \n\t" \
                 : "=a" (ax), "=D" (bx), "=c" (cx), "=d" (dx) \
-                : "a" (func));
+                : "a" (func), "c" (func2));
 #endif
 #else /* end __SUNPRO__ */
 #if ARCH_X86_64
 void __cpuid(int CPUInfo[4], int info_type);
 #pragma intrinsic(__cpuid)
-#define cpuid(func,a,b,c,d) do{\
+#define cpuid(func, func2, a, b, c, d) do {\
     int regs[4];\
-    __cpuid(regs,func); a=regs[0];  b=regs[1];  c=regs[2];  d=regs[3];\
+    __cpuid(regs, func, func2);
+    a = regs[0];  b = regs[1];  c = regs[2];  d = regs[3];\
   } while(0)
 #else
-#define cpuid(func,a,b,c,d)\
+#define cpuid(func, func2, a, b, c, d)\
   __asm mov eax, func\
+  __asm mov ecx, func2\
   __asm cpuid\
   __asm mov a, eax\
   __asm mov b, ebx\
@@ -120,13 +122,13 @@ x86_simd_caps(void) {
     mask = strtol(env, NULL, 0);
 
   /* Ensure that the CPUID instruction supports extended features */
-  cpuid(0, reg_eax, reg_ebx, reg_ecx, reg_edx);
+  cpuid(0, 0, reg_eax, reg_ebx, reg_ecx, reg_edx);
 
   if (reg_eax < 1)
     return 0;
 
   /* Get the standard feature flags */
-  cpuid(1, reg_eax, reg_ebx, reg_ecx, reg_edx);
+  cpuid(1, 0, reg_eax, reg_ebx, reg_ecx, reg_edx);
 
   if (reg_edx & BIT(23)) flags |= HAS_MMX;
 
@@ -142,6 +144,11 @@ x86_simd_caps(void) {
 
   if (reg_ecx & BIT(28)) flags |= HAS_AVX;
 
+  /* Get the leaf 7 feature flags. Needed to check for AVX2 support */
+  reg_eax = 7;
+  reg_ecx = 0;
+  cpuid(7, 0, reg_eax, reg_ebx, reg_ecx, reg_edx);
+
   if (reg_ebx & BIT(5)) flags |= HAS_AVX2;
 
   return flags & mask;

diff --git a/vpx_ports/x86_cpuid.c b/vpx_ports/x86_cpuid.c
index fe86cfc7b8a0645c7cf342ee80bbd6a40166db55..02d382c5db32105450c70e692ee844a33fd67321 100644 (file)
--- a/vpx_ports/x86_cpuid.c
+++ b/vpx_ports/x86_cpuid.c
@@ -38,7 +38,7 @@ vpx_cpu_t vpx_x86_vendor(void) {
   int i;
 
   /* Get the Vendor String from the CPU */
-  cpuid(0, reg_eax, vs[0], vs[2], vs[1]);
+  cpuid(0, 0, reg_eax, vs[0], vs[2], vs[1]);
 
   for (i = 0; i < VPX_CPU_LAST; i++) {
     if (strncmp((const char *)vs, cpuid_vendor_list[i].vendor_string, 12) == 0)