v2-0003-Refactor-the-detection-of-ZMM-registers.patch

From 8a61c21ad94bc6241989b6cd6237e75b16137ea1 Mon Sep 17 00:00:00 2001
From: John Naylor <john.naylor@postgresql.org>
Date: Fri, 13 Feb 2026 18:11:39 +0700
Subject: [PATCH v2 3/4] Refactor the detection of ZMM registers

- Call _xgetbv within x86_set_runtime_features rather than in a
  separate function
- Use symbols for XCR mask bits rather than a magic constant

A future commit will build on this to detect YMM registers without
code duplication.
---
 src/port/pg_cpu_x86.c      | 42 +++++++++++++++++++++-----------------
 src/port/pg_crc32c_sse42.c |  2 +-
 2 files changed, 24 insertions(+), 20 deletions(-)

diff --git a/src/port/pg_cpu_x86.c b/src/port/pg_cpu_x86.c
index 249283725c8..468972a3d12 100644
--- a/src/port/pg_cpu_x86.c
+++ b/src/port/pg_cpu_x86.c
@@ -31,31 +31,28 @@
 
 #include "port/pg_cpu.h"
 
+/* XSAVE state component bits that we need */
+#define XMM (1<<1)
+#define YMM  (1<<2)
+#define OPMASK  (1<<5)
+#define ZMM0_15  (1<<6)
+#define ZMM16_31  (1<<7)
+
 
 bool		X86Features[X86FeaturesSize] = {0};
 
-/*
- * Does XGETBV say the ZMM registers are enabled?
- *
- * NB: Caller is responsible for verifying that osxsave is available
- * before calling this.
- */
-#ifdef HAVE_XSAVE_INTRINSICS
-pg_attribute_target("xsave")
-#endif
 static bool
-zmm_regs_available(void)
+mask_available(uint32 value, uint32 mask)
 {
-#ifdef HAVE_XSAVE_INTRINSICS
-	return (_xgetbv(0) & 0xe6) == 0xe6;
-#else
-	return false;
-#endif
+	return (value & mask) == mask;
 }
 
 /*
  * Parse the CPU ID info for runtime checks.
  */
+#ifdef HAVE_XSAVE_INTRINSICS
+pg_attribute_target("xsave")
+#endif
 void
 set_x86_features(void)
 {
@@ -75,22 +72,29 @@ set_x86_features(void)
 	/* All these features depend on OSXSAVE */
 	if (exx[2] & (1 << 27))
 	{
-		/* second cpuid call on leaf 7 to check extended AVX-512 support */
+		uint32		xcr0_val = 0;
 
+		/* second cpuid call on leaf 7 to check extended AVX-512 support */
 		memset(exx, 0, 4 * sizeof(exx[0]));
-
 #if defined(HAVE__GET_CPUID_COUNT)
 		__get_cpuid_count(7, 0, &exx[0], &exx[1], &exx[2], &exx[3]);
 #elif defined(HAVE__CPUIDEX)
 		__cpuidex(exx, 7, 0);
 #endif
 
-		if (zmm_regs_available())
+#ifdef HAVE_XSAVE_INTRINSICS
+		/* get value of Extended Control Register */
+		xcr0_val = _xgetbv(0);
+#endif
+
+		/* Are ZMM registeres enabled? */
+		if (mask_available(xcr0_val, XMM | YMM |
+						   OPMASK | ZMM0_15 | ZMM16_31))
 		{
 			X86Features[PG_AVX512_BW] = exx[1] >> 30 & 1;
 			X86Features[PG_AVX512_VL] = exx[1] >> 31 & 1;
 
-			X86Features[PG_VPCLMULQDQ] = exx[2] >> 10 & 1;
+			X86Features[PG_AVX512_VPCLMULQDQ] = exx[2] >> 10 & 1;
 			X86Features[PG_AVX512_VPOPCNTDQ] = exx[2] >> 14 & 1;
 		}
 	}
diff --git a/src/port/pg_crc32c_sse42.c b/src/port/pg_crc32c_sse42.c
index 2e740e12a7a..d1d9d74e5ab 100644
--- a/src/port/pg_crc32c_sse42.c
+++ b/src/port/pg_crc32c_sse42.c
@@ -179,7 +179,7 @@ pg_comp_crc32c_choose(pg_crc32c crc, const void *data, size_t len)
 
 #ifdef USE_AVX512_CRC32C_WITH_RUNTIME_CHECK
 	if (x86_feature_available(PG_AVX512_VL) &&
-		x86_feature_available(PG_VPCLMULQDQ))
+		x86_feature_available(PG_AVX512_VPCLMULQDQ))
 		pg_comp_crc32c = pg_comp_crc32c_avx512;
 #endif
 
-- 
2.53.0