From 4ad2c9d04b76d7c4a42d80a82c022cd60b43b8b2 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Fri, 24 May 2024 12:38:51 -0500
Subject: [PATCH 04/10] x86: Add seperate non-temporal tunable for memset

The tuning for non-temporal stores for memset vs memcpy is not always
the same. This includes both the exact value and whether non-temporal
stores are profitable at all for a given arch.

This patch add `x86_memset_non_temporal_threshold`. Currently we
disable non-temporal stores for non Intel vendors as the only
benchmarks showing its benefit have been on Intel hardware.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
---
 manual/tunables.texi                             | 16 +++++++++++++++-
 sysdeps/x86/cacheinfo.h                          |  8 +++++++-
 sysdeps/x86/dl-cacheinfo.h                       | 16 ++++++++++++++++
 sysdeps/x86/dl-diagnostics-cpu.c                 |  2 ++
 sysdeps/x86/dl-tunables.list                     |  3 +++
 sysdeps/x86/include/cpu-features.h               |  4 +++-
 .../x86_64/multiarch/memset-vec-unaligned-erms.S |  6 +++---
 7 files changed, 49 insertions(+), 6 deletions(-)

diff --git a/manual/tunables.texi b/manual/tunables.texi
index 6493904bae..2a2877884c 100644
--- a/manual/tunables.texi
+++ b/manual/tunables.texi
@@ -52,6 +52,7 @@ glibc.elision.skip_lock_busy: 3 (min: 0, max: 2147483647)
 glibc.malloc.top_pad: 0x20000 (min: 0x0, max: 0xffffffffffffffff)
 glibc.cpu.x86_rep_stosb_threshold: 0x800 (min: 0x1, max: 0xffffffffffffffff)
 glibc.cpu.x86_non_temporal_threshold: 0xc0000 (min: 0x4040, max: 0xfffffffffffffff)
+glibc.cpu.x86_memset_non_temporal_threshold: 0xc0000 (min: 0x4040, max: 0xfffffffffffffff)
 glibc.cpu.x86_shstk:
 glibc.pthread.stack_cache_size: 0x2800000 (min: 0x0, max: 0xffffffffffffffff)
 glibc.cpu.hwcap_mask: 0x6 (min: 0x0, max: 0xffffffffffffffff)
@@ -486,7 +487,8 @@ thread stack originally backup by Huge Pages to default pages.
 @cindex shared_cache_size tunables
 @cindex tunables, shared_cache_size
 @cindex non_temporal_threshold tunables
-@cindex tunables, non_temporal_threshold
+@cindex memset_non_temporal_threshold tunables
+@cindex tunables, non_temporal_threshold, memset_non_temporal_threshold
 
 @deftp {Tunable namespace} glibc.cpu
 Behavior of @theglibc{} can be tuned to assume specific hardware capabilities
@@ -562,6 +564,18 @@ like memmove and memcpy.
 This tunable is specific to i386 and x86-64.
 @end deftp
 
+@deftp Tunable glibc.cpu.x86_memset_non_temporal_threshold
+The @code{glibc.cpu.x86_memset_non_temporal_threshold} tunable allows
+the user to set threshold in bytes for non temporal store in
+memset. Non temporal stores give a hint to the hardware to move data
+directly to memory without displacing other data from the cache. This
+tunable is used by some platforms to determine when to use non
+temporal stores memset.
+
+This tunable is specific to i386 and x86-64.
+@end deftp
+
+
 @deftp Tunable glibc.cpu.x86_rep_movsb_threshold
 The @code{glibc.cpu.x86_rep_movsb_threshold} tunable allows the user to
 set threshold in bytes to start using "rep movsb".  The value must be
diff --git a/sysdeps/x86/cacheinfo.h b/sysdeps/x86/cacheinfo.h
index ec1bc142c4..fd2b2ae66b 100644
--- a/sysdeps/x86/cacheinfo.h
+++ b/sysdeps/x86/cacheinfo.h
@@ -35,9 +35,12 @@ long int __x86_data_cache_size attribute_hidden = 32 * 1024;
 long int __x86_shared_cache_size_half attribute_hidden = 1024 * 1024 / 2;
 long int __x86_shared_cache_size attribute_hidden = 1024 * 1024;
 
-/* Threshold to use non temporal store.  */
+/* Threshold to use non temporal store in memmove.  */
 long int __x86_shared_non_temporal_threshold attribute_hidden;
 
+/* Threshold to use non temporal store in memset.  */
+long int __x86_memset_non_temporal_threshold attribute_hidden;
+
 /* Threshold to use Enhanced REP MOVSB.  */
 long int __x86_rep_movsb_threshold attribute_hidden = 2048;
 
@@ -77,6 +80,9 @@ init_cacheinfo (void)
   __x86_shared_non_temporal_threshold
     = cpu_features->non_temporal_threshold;
 
+  __x86_memset_non_temporal_threshold
+      = cpu_features->memset_non_temporal_threshold;
+
   __x86_rep_movsb_threshold = cpu_features->rep_movsb_threshold;
   __x86_rep_stosb_threshold = cpu_features->rep_stosb_threshold;
   __x86_rep_movsb_stop_threshold =  cpu_features->rep_movsb_stop_threshold;
diff --git a/sysdeps/x86/dl-cacheinfo.h b/sysdeps/x86/dl-cacheinfo.h
index ce2e6927e4..9f27da21ce 100644
--- a/sysdeps/x86/dl-cacheinfo.h
+++ b/sysdeps/x86/dl-cacheinfo.h
@@ -1048,6 +1048,13 @@ dl_init_cacheinfo (struct cpu_features *cpu_features)
   if (CPU_FEATURE_USABLE_P (cpu_features, FSRM))
     rep_movsb_threshold = 2112;
 
+  /* Non-temporal stores in memset have only been tested on Intel hardware.
+     Until we benchmark data on other x86 processor, disable non-temporal
+     stores in memset. */
+  unsigned long int memset_non_temporal_threshold = SIZE_MAX;
+  if (cpu_features->basic.kind == arch_kind_intel)
+      memset_non_temporal_threshold = non_temporal_threshold;
+
    /* For AMD CPUs that support ERMS (Zen3+), REP MOVSB is in a lot of
       cases slower than the vectorized path (and for some alignments,
       it is really slow, check BZ #30994).  */
@@ -1074,6 +1081,11 @@ dl_init_cacheinfo (struct cpu_features *cpu_features)
       && tunable_size <= maximum_non_temporal_threshold)
     non_temporal_threshold = tunable_size;
 
+  tunable_size = TUNABLE_GET (x86_memset_non_temporal_threshold, long int, NULL);
+  if (tunable_size > minimum_non_temporal_threshold
+      && tunable_size <= maximum_non_temporal_threshold)
+    memset_non_temporal_threshold = tunable_size;
+
   tunable_size = TUNABLE_GET (x86_rep_movsb_threshold, long int, NULL);
   if (tunable_size > minimum_rep_movsb_threshold)
     rep_movsb_threshold = tunable_size;
@@ -1089,6 +1101,9 @@ dl_init_cacheinfo (struct cpu_features *cpu_features)
   TUNABLE_SET_WITH_BOUNDS (x86_non_temporal_threshold, non_temporal_threshold,
 			   minimum_non_temporal_threshold,
 			   maximum_non_temporal_threshold);
+  TUNABLE_SET_WITH_BOUNDS (
+      x86_memset_non_temporal_threshold, memset_non_temporal_threshold,
+      minimum_non_temporal_threshold, maximum_non_temporal_threshold);
   TUNABLE_SET_WITH_BOUNDS (x86_rep_movsb_threshold, rep_movsb_threshold,
 			   minimum_rep_movsb_threshold, SIZE_MAX);
   TUNABLE_SET_WITH_BOUNDS (x86_rep_stosb_threshold, rep_stosb_threshold, 1,
@@ -1102,6 +1117,7 @@ dl_init_cacheinfo (struct cpu_features *cpu_features)
   cpu_features->data_cache_size = data;
   cpu_features->shared_cache_size = shared;
   cpu_features->non_temporal_threshold = non_temporal_threshold;
+  cpu_features->memset_non_temporal_threshold = memset_non_temporal_threshold;
   cpu_features->rep_movsb_threshold = rep_movsb_threshold;
   cpu_features->rep_stosb_threshold = rep_stosb_threshold;
   cpu_features->rep_movsb_stop_threshold = rep_movsb_stop_threshold;
diff --git a/sysdeps/x86/dl-diagnostics-cpu.c b/sysdeps/x86/dl-diagnostics-cpu.c
index 5aab63e532..05d54b5eba 100644
--- a/sysdeps/x86/dl-diagnostics-cpu.c
+++ b/sysdeps/x86/dl-diagnostics-cpu.c
@@ -83,6 +83,8 @@ _dl_diagnostics_cpu (void)
                             cpu_features->shared_cache_size);
   print_cpu_features_value ("non_temporal_threshold",
                             cpu_features->non_temporal_threshold);
+  print_cpu_features_value ("memset_non_temporal_threshold",
+                            cpu_features->memset_non_temporal_threshold);
   print_cpu_features_value ("rep_movsb_threshold",
                             cpu_features->rep_movsb_threshold);
   print_cpu_features_value ("rep_movsb_stop_threshold",
diff --git a/sysdeps/x86/dl-tunables.list b/sysdeps/x86/dl-tunables.list
index d1442d88ba..53852d6a07 100644
--- a/sysdeps/x86/dl-tunables.list
+++ b/sysdeps/x86/dl-tunables.list
@@ -30,6 +30,9 @@ glibc {
     x86_non_temporal_threshold {
       type: SIZE_T
     }
+    x86_memset_non_temporal_threshold {
+      type: SIZE_T
+    }
     x86_rep_movsb_threshold {
       type: SIZE_T
       # Since there is overhead to set up REP MOVSB operation, REP
diff --git a/sysdeps/x86/include/cpu-features.h b/sysdeps/x86/include/cpu-features.h
index 594feeb2f4..e2d641dcd0 100644
--- a/sysdeps/x86/include/cpu-features.h
+++ b/sysdeps/x86/include/cpu-features.h
@@ -918,8 +918,10 @@ struct cpu_features
   /* Shared cache size for use in memory and string routines, typically
      L2 or L3 size.  */
   unsigned long int shared_cache_size;
-  /* Threshold to use non temporal store.  */
+  /* Threshold to use non temporal store in memmove.  */
   unsigned long int non_temporal_threshold;
+  /* Threshold to use non temporal store in memset.  */
+  unsigned long int memset_non_temporal_threshold;
   /* Threshold to use "rep movsb".  */
   unsigned long int rep_movsb_threshold;
   /* Threshold to stop using "rep movsb".  */
diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
index aba45e3da0..d95750b516 100644
--- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
@@ -24,9 +24,9 @@
    5. If size is more to 4 * VEC_SIZE, align to 1 * VEC_SIZE with
       4 VEC stores and store 4 * VEC at a time until done.
    6. On machines ERMS feature, if size is range
-	  [__x86_rep_stosb_threshold, __x86_shared_non_temporal_threshold)
+	  [__x86_rep_stosb_threshold, __x86_memset_non_temporal_threshold)
 	  then REP STOSB will be used.
-   7. If size >= __x86_shared_non_temporal_threshold, use a
+   7. If size >= __x86_memset_non_temporal_threshold, use a
 	  non-temporal stores.  */
 
 #include <sysdep.h>
@@ -318,7 +318,7 @@ L(return_vzeroupper):
 	/* If no USE_LESS_VEC_MASK put L(stosb_local) here. Will be in
 	   range for 2-byte jump encoding.  */
 L(stosb_local):
-	cmp	__x86_shared_non_temporal_threshold(%rip), %RDX_LP
+	cmp	__x86_memset_non_temporal_threshold(%rip), %RDX_LP
 	jae	L(nt_memset)
 	movzbl	%sil, %eax
 	mov	%RDX_LP, %RCX_LP
-- 
2.17.1