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..3593d5d6929b575d00719a71a003880a24b27106
--- /dev/null
+++ b/third_party/zlib/x86.c
@@ -0,0 +1,88 @@
+/*
+ * 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) {

[agl, 2014/10/17 00:45:35]

you've switched brace style and indentation here.

[robert.bradford, 2014/10/20 16:45:54]

Done.

+  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;
+    __asm__ __volatile__ (
+#ifdef __i386__

[agl, 2014/10/17 00:45:34]

I think it would be better to suffer a little duplication in order to reduce the
amount of preprocessor gunk here:

#if defined(__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

(check that I actually got that right, of course)

[robert.bradford, 2014/10/20 16:45:54]

Done.

+        "xchg %%ebx, %1\n\t"
+#endif
+        "cpuid\n\t"
+#ifdef __i386__
+        "xchg %1, %%ebx\n\t"
+    : "+a" (eax), "=S" (ebx), "=c" (ecx), "=d" (edx)
+#else
+    : "+a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
+#endif
+    );
+
+    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)
+  InitOnceExecuteOnce(&cpu_check_inited_once, _x86_check_features, NULL, NULL);
+#endif // (_WIN32_WINNT >= 0x0600)

[agl, 2014/10/17 00:45:35]

ditto about two spaces.

[robert.bradford, 2014/10/20 16:45:54]

Done.

+}
+
+#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)

[agl, 2014/10/17 00:45:35]

nit: two spaces before "//"

Also, does VC++ accept "//" comments in .c files now? I guess the try servers
will tell us.

[robert.bradford, 2014/10/20 16:45:54]

Well I did build it with the windows toolchain. But lets play it safe and switch
to C style.

+#endif

[agl, 2014/10/17 00:45:35]

#endif  // _MSC_VER

[robert.bradford, 2014/10/20 16:45:54]

Done.