Simplify the AMD-specific codepath.

See TODO comment being removed.

Change-Id: I92ce7018f88c24b3e2e61441397fda36b977d3b8
Reviewed-on: https://boringssl-review.googlesource.com/5435
Reviewed-by: Adam Langley <agl@google.com>
This commit is contained in:
David Benjamin 2015-07-12 12:32:46 -04:00 committed by Adam Langley
parent 7315251d4e
commit f21fe46764

View File

@ -63,7 +63,6 @@
#if !defined(OPENSSL_NO_ASM) && (defined(OPENSSL_X86) || defined(OPENSSL_X86_64)) #if !defined(OPENSSL_NO_ASM) && (defined(OPENSSL_X86) || defined(OPENSSL_X86_64))
#include <assert.h>
#include <inttypes.h> #include <inttypes.h>
#include <stdlib.h> #include <stdlib.h>
#include <stdio.h> #include <stdio.h>
@ -154,13 +153,12 @@ void OPENSSL_cpuid_setup(void) {
ecx == 0x444d4163 /* cAMD */; ecx == 0x444d4163 /* cAMD */;
int has_amd_xop = 0; int has_amd_xop = 0;
uint32_t num_amd_extended_ids = 0;
if (is_amd) { if (is_amd) {
/* AMD-specific logic. /* AMD-specific logic.
* See http://developer.amd.com/wordpress/media/2012/10/254811.pdf */ * See http://developer.amd.com/wordpress/media/2012/10/254811.pdf */
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 0x80000000); OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 0x80000000);
num_amd_extended_ids = eax; uint32_t num_extended_ids = eax;
if (num_amd_extended_ids >= 0x80000001) { if (num_extended_ids >= 0x80000001) {
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 0x80000001); OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 0x80000001);
if (ecx & (1 << 11)) { if (ecx & (1 << 11)) {
has_amd_xop = 1; has_amd_xop = 1;
@ -174,47 +172,27 @@ void OPENSSL_cpuid_setup(void) {
extended_features = ebx; extended_features = ebx;
} }
/* Query the main feature flags and adjust the hyper-threading bit. /* Determine the number of cores sharing an L1 data cache to adjust the
* * hyper-threading bit. */
* TODO(davidben): Can the AMD half of logic be trimmed down? I haven't found uint32_t cores_per_cache = 0;
* evidence that any current AMD CPUs share an L1 data cache between threads, if (is_amd) {
* and the CPUID manual (see section 3) suggests this code will always clear /* AMD CPUs never share an L1 data cache between threads but do set the HTT
* HTT on AMD. Only aes-586.pl queries this, so hopefully any future CPUs will * bit on multi-core CPUs. */
* use better implementations anyway. */ cores_per_cache = 1;
if (num_amd_extended_ids >= 0x80000008) { } else if (num_ids >= 4) {
/* See AMD CPUID manual (2.34), page 27. */ /* TODO(davidben): The Intel manual says this CPUID leaf enumerates all
assert(is_amd); * caches using ECX and doesn't say which is first. Does this matter? */
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 0x80000008); OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 4);
uint32_t num_physical_cores = 1 + (ecx & 0xff); cores_per_cache = 1 + ((eax >> 14) & 0xfff);
}
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 1); OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 1);
/* Correct the hyper-threading bit. */ /* Adjust the hyper-threading bit. */
if (edx & (1 << 28)) { if (edx & (1 << 28)) {
/* See AMD CPUID manual (2.34), pages 11 and 13. */ uint32_t num_logical_cores = (ebx >> 16) & 0xff;
uint32_t num_logical_cores = (ebx >> 16) & 0xff; if (cores_per_cache == 1 || num_logical_cores <= 1) {
if (num_logical_cores <= num_physical_cores) { edx &= ~(1 << 28);
edx &= ~(1 << 28);
}
}
} else {
uint32_t cores_per_cache = 0;
if (num_ids >= 4) {
/* TODO(davidben): The Intel manual says this CPUID leaf enumerates all
* caches using ECX and doesn't say which is first. Does this matter? */
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 4);
cores_per_cache = 1 + ((eax >> 14) & 0xfff);
}
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 1);
/* Correct the hyper-threading bit if the data cache isn't shared between
* logical cores. */
if (edx & (1 << 28)) {
uint32_t num_logical_cores = (ebx >> 16) & 0xff;
if (cores_per_cache == 1 || num_logical_cores <= 1) {
edx &= ~(1 << 28);
}
} }
} }