Reland "Integrate SIMD optimisations for zlib"

third_party/zlib/x86.c

Index: third_party/zlib/x86.c
diff --git a/third_party/zlib/x86.c b/third_party/zlib/x86.c
new file mode 100644
index 0000000000000000000000000000000000000000..11998667854ccd30a3bc96737d740d4087d753cb
--- /dev/null
+++ b/third_party/zlib/x86.c
@@ -0,0 +1,97 @@
+/*
+ * x86 feature check
+ *
+ * Copyright (C) 2013 Intel Corporation. All rights reserved.
+ * Author:
+ *  Jim Kukunas
+ * 
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "x86.h"
+
+int x86_cpu_enable_simd;
+
+#ifndef _MSC_VER
+#include <pthread.h>
+
+pthread_once_t cpu_check_inited_once = PTHREAD_ONCE_INIT;
+static void _x86_check_features(void);
+
+void x86_check_features(void)
+{
+  pthread_once(&cpu_check_inited_once, _x86_check_features);
+}
+
+static void _x86_check_features(void)
+{
+    int x86_cpu_has_sse2;
+    int x86_cpu_has_sse42;
+    int x86_cpu_has_pclmulqdq;
+    unsigned eax, ebx, ecx, edx;
+
+    eax = 1;
+#ifdef __i386__
+    __asm__ __volatile__ (
+        "xchg %%ebx, %1\n\t"
+        "cpuid\n\t"
+        "xchg %1, %%ebx\n\t"
+    : "+a" (eax), "=S" (ebx), "=c" (ecx), "=d" (edx)
+    );
+#else
+    __asm__ __volatile__ (
+        "cpuid\n\t"
+    : "+a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
+    );
+#endif  /* (__i386__) */
+
+    x86_cpu_has_sse2 = edx & 0x4000000;
+    x86_cpu_has_sse42 = ecx & 0x100000;
+    x86_cpu_has_pclmulqdq = ecx & 0x2;
+
+    x86_cpu_enable_simd = x86_cpu_has_sse2 &&
+                          x86_cpu_has_sse42 &&
+                          x86_cpu_has_pclmulqdq;
+}
+#else
+#include <intrin.h>
+#include <windows.h>
+
+static BOOL CALLBACK _x86_check_features(PINIT_ONCE once, PVOID param, PVOID *context);
+static INIT_ONCE cpu_check_inited_once = INIT_ONCE_STATIC_INIT;
+
+void x86_check_features(void)
+{
+#if (_WIN32_WINNT >= 0x0600)

[agl, 2014/10/30 01:25:43]

I assume that this #if check was supposed to catch this? What's it doing if it
doesn't?

[robert.bradford, 2014/10/30 14:31:50]

It checked the presence in the header file, as we no longer reference the API
call from the header file. I've removed them. Thanks.

+    typedef BOOL (WINAPI *INIT_ONCE_FUNC)(PINIT_ONCE, PINIT_ONCE_FN, PVOID, LPVOID);
+    INIT_ONCE_FUNC once_func;
+    /* Windows XP / Server 2003 does not have InitOnceExecuteOnce */
+    once_func = (INIT_ONCE_FUNC)GetProcAddress(GetModuleHandle(TEXT("kernel32.dll")),

[agl, 2014/10/30 01:25:43]

It seems that GetProcAddress might be a really expensive function to call often.
Are you sure that this doesn't cause performance to tank?

[robert.bradford, 2014/10/30 14:31:50]

I've uploaded a revised version that saves the function pointer. I don't think
this is racey as the address for the symbol will always be the same.

+                                               "InitOnceExecuteOnce");
+    if (once_func) {
+        once_func(&cpu_check_inited_once, _x86_check_features, NULL, NULL);
+    }
+#endif  /* (_WIN32_WINNT >= 0x0600) */
+}
+
+#if (_WIN32_WINNT >= 0x0600)
+static BOOL CALLBACK _x86_check_features(PINIT_ONCE once, PVOID param, PVOID *context)
+{
+    int x86_cpu_has_sse2;
+    int x86_cpu_has_sse42;
+    int x86_cpu_has_pclmulqdq;
+    int regs[4];
+
+    __cpuid(regs, 1);
+
+    x86_cpu_has_sse2 = regs[3] & 0x4000000;
+    x86_cpu_has_sse42= regs[2] & 0x100000;
+    x86_cpu_has_pclmulqdq = regs[2] & 0x2;
+
+    x86_cpu_enable_simd = x86_cpu_has_sse2 &&
+                          x86_cpu_has_sse42 &&
+                          x86_cpu_has_pclmulqdq;
+    return TRUE;
+}
+#endif  /* (_WIN32_WINNT >= 0x0600) */
+#endif  /* _MSC_VER */