File amd-cacheinfo.patch of Package glibc
From 103a469dc7755fd9e8ccf362f3dd4c55dc761908 Mon Sep 17 00:00:00 2001
From: Sajan Karumanchi <sajan.karumanchi@amd.com>
Date: Wed, 18 Jan 2023 18:29:04 +0100
Subject: [PATCH] x86: Cache computation for AMD architecture.
All AMD architectures cache details will be computed based on
__cpuid__ `0x8000_001D` and the reference to __cpuid__ `0x8000_0006` will be
zeroed out for future architectures.
Reviewed-by: Premachandra Mallappa <premachandra.mallappa@amd.com>
---
sysdeps/x86/dl-cacheinfo.h | 204 ++++++++-----------------------------
1 file changed, 45 insertions(+), 159 deletions(-)
Index: glibc-2.31/sysdeps/x86/cacheinfo.c
===================================================================
--- glibc-2.31.orig/sysdeps/x86/cacheinfo.c
+++ glibc-2.31/sysdeps/x86/cacheinfo.c
@@ -326,111 +326,41 @@ handle_amd (int name)
unsigned int ebx;
unsigned int ecx;
unsigned int edx;
- __cpuid (0x80000000, eax, ebx, ecx, edx);
+ unsigned int count = 0x1;
/* No level 4 cache (yet). */
if (name > _SC_LEVEL3_CACHE_LINESIZE)
return 0;
- unsigned int fn = 0x80000005 + (name >= _SC_LEVEL2_CACHE_SIZE);
- if (eax < fn)
- return 0;
-
- __cpuid (fn, eax, ebx, ecx, edx);
+ if (name >= _SC_LEVEL3_CACHE_SIZE)
+ count = 0x3;
+ else if (name >= _SC_LEVEL2_CACHE_SIZE)
+ count = 0x2;
+ else if (name >= _SC_LEVEL1_DCACHE_SIZE)
+ count = 0x0;
- if (name < _SC_LEVEL1_DCACHE_SIZE)
- {
- name += _SC_LEVEL1_DCACHE_SIZE - _SC_LEVEL1_ICACHE_SIZE;
- ecx = edx;
- }
+ __cpuid_count (0x8000001D, count, eax, ebx, ecx, edx);
switch (name)
{
- case _SC_LEVEL1_DCACHE_SIZE:
- return (ecx >> 14) & 0x3fc00;
-
- case _SC_LEVEL1_DCACHE_ASSOC:
- ecx >>= 16;
- if ((ecx & 0xff) == 0xff)
- /* Fully associative. */
- return (ecx << 2) & 0x3fc00;
- return ecx & 0xff;
-
- case _SC_LEVEL1_DCACHE_LINESIZE:
- return ecx & 0xff;
-
- case _SC_LEVEL2_CACHE_SIZE:
- return (ecx & 0xf000) == 0 ? 0 : (ecx >> 6) & 0x3fffc00;
-
- case _SC_LEVEL2_CACHE_ASSOC:
- switch ((ecx >> 12) & 0xf)
- {
- case 0:
- case 1:
- case 2:
- case 4:
- return (ecx >> 12) & 0xf;
- case 6:
- return 8;
- case 8:
- return 16;
- case 10:
- return 32;
- case 11:
- return 48;
- case 12:
- return 64;
- case 13:
- return 96;
- case 14:
- return 128;
- case 15:
- return ((ecx >> 6) & 0x3fffc00) / (ecx & 0xff);
- default:
- return 0;
- }
- /* NOTREACHED */
-
- case _SC_LEVEL2_CACHE_LINESIZE:
- return (ecx & 0xf000) == 0 ? 0 : ecx & 0xff;
-
- case _SC_LEVEL3_CACHE_SIZE:
- return (edx & 0xf000) == 0 ? 0 : (edx & 0x3ffc0000) << 1;
-
- case _SC_LEVEL3_CACHE_ASSOC:
- switch ((edx >> 12) & 0xf)
- {
- case 0:
- case 1:
- case 2:
- case 4:
- return (edx >> 12) & 0xf;
- case 6:
- return 8;
- case 8:
- return 16;
- case 10:
- return 32;
- case 11:
- return 48;
- case 12:
- return 64;
- case 13:
- return 96;
- case 14:
- return 128;
- case 15:
- return ((edx & 0x3ffc0000) << 1) / (edx & 0xff);
- default:
- return 0;
- }
- /* NOTREACHED */
-
- case _SC_LEVEL3_CACHE_LINESIZE:
- return (edx & 0xf000) == 0 ? 0 : edx & 0xff;
-
- default:
- assert (! "cannot happen");
+ case _SC_LEVEL1_ICACHE_ASSOC:
+ case _SC_LEVEL1_DCACHE_ASSOC:
+ case _SC_LEVEL2_CACHE_ASSOC:
+ case _SC_LEVEL3_CACHE_ASSOC:
+ return ecx?((ebx >> 22) & 0x3ff) + 1 : 0;
+ case _SC_LEVEL1_ICACHE_LINESIZE:
+ case _SC_LEVEL1_DCACHE_LINESIZE:
+ case _SC_LEVEL2_CACHE_LINESIZE:
+ case _SC_LEVEL3_CACHE_LINESIZE:
+ return ecx?(ebx & 0xfff) + 1 : 0;
+ case _SC_LEVEL1_ICACHE_SIZE:
+ case _SC_LEVEL1_DCACHE_SIZE:
+ case _SC_LEVEL2_CACHE_SIZE:
+ case _SC_LEVEL3_CACHE_SIZE:
+ return ecx?(((ebx >> 22) & 0x3ff) + 1)*((ebx & 0xfff) + 1)\
+ *(ecx + 1):0;
+ default:
+ assert (! "cannot happen");
}
return -1;
}
@@ -710,36 +640,8 @@ intel_bug_no_cache_info:
__cpuid (0x80000000, max_cpuid_ex, ebx, ecx, edx);
if (shared <= 0)
- /* No shared L3 cache. All we have is the L2 cache. */
- shared = core;
- else
- {
- /* Figure out the number of logical threads that share L3. */
- if (max_cpuid_ex >= 0x80000008)
- {
- /* Get width of APIC ID. */
- __cpuid (0x80000008, max_cpuid_ex, ebx, ecx, edx);
- threads = 1 << ((ecx >> 12) & 0x0f);
- }
-
- if (threads == 0)
- {
- /* If APIC ID width is not available, use logical
- processor count. */
- __cpuid (0x00000001, max_cpuid_ex, ebx, ecx, edx);
-
- if ((edx & (1 << 28)) != 0)
- threads = (ebx >> 16) & 0xff;
- }
-
- /* Cap usage of highest cache level to the number of
- supported threads. */
- if (threads > 0)
- shared /= threads;
-
- /* Account for exclusive L2 and L3 caches. */
- shared += core;
- }
+ /* No shared L3 cache. All we have is the L2 cache. */
+ shared = core;
#ifndef DISABLE_PREFETCHW
if (max_cpuid_ex >= 0x80000001)