[llvm-commits] [llvm] r169730 - in /llvm/trunk/lib/Target/X86: X86.td X86Subtarget.cpp

Chandler Carruth chandlerc at gmail.com
Mon Dec 10 01:18:44 PST 2012


Author: chandlerc
Date: Mon Dec 10 03:18:44 2012
New Revision: 169730

URL: http://llvm.org/viewvc/llvm-project?rev=169730&view=rev
Log:
Address a FIXME and update the fast unaligned memory feature for newer
Intel chips.

The model number rules were determined by inspecting Intel's
documentation for their newer chip model numbers. My understanding is
that all of the newer Intel chips have fast unaligned memory access, but
if anyone is concerned about a particular chip, just shout.

No tests updated; it's not clear we have dedicated tests for the chips'
various features, but if anyone would like tests (or can point me at
some existing ones), I'm happy to oblige.

Modified:
    llvm/trunk/lib/Target/X86/X86.td
    llvm/trunk/lib/Target/X86/X86Subtarget.cpp

Modified: llvm/trunk/lib/Target/X86/X86.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86.td?rev=169730&r1=169729&r2=169730&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86.td (original)
+++ llvm/trunk/lib/Target/X86/X86.td Mon Dec 10 03:18:44 2012
@@ -182,17 +182,17 @@
 // Sandy Bridge
 // SSE is not listed here since llvm treats AVX as a reimplementation of SSE,
 // rather than a superset.
-def : Proc<"corei7-avx",      [FeatureAVX, FeatureCMPXCHG16B, FeaturePOPCNT,
-                               FeatureAES, FeaturePCLMUL]>;
+def : Proc<"corei7-avx",      [FeatureAVX, FeatureCMPXCHG16B, FeatureFastUAMem,
+                               FeaturePOPCNT, FeatureAES, FeaturePCLMUL]>;
 // Ivy Bridge
-def : Proc<"core-avx-i",      [FeatureAVX, FeatureCMPXCHG16B, FeaturePOPCNT,
-                               FeatureAES, FeaturePCLMUL,
+def : Proc<"core-avx-i",      [FeatureAVX, FeatureCMPXCHG16B, FeatureFastUAMem,
+                               FeaturePOPCNT, FeatureAES, FeaturePCLMUL,
                                FeatureRDRAND, FeatureF16C, FeatureFSGSBase]>;
 
 // Haswell
-def : Proc<"core-avx2",       [FeatureAVX2, FeatureCMPXCHG16B, FeaturePOPCNT,
-                               FeatureAES, FeaturePCLMUL, FeatureRDRAND,
-                               FeatureF16C, FeatureFSGSBase,
+def : Proc<"core-avx2",       [FeatureAVX2, FeatureCMPXCHG16B, FeatureFastUAMem,
+                               FeaturePOPCNT, FeatureAES, FeaturePCLMUL,
+                               FeatureRDRAND, FeatureF16C, FeatureFSGSBase,
                                FeatureMOVBE, FeatureLZCNT, FeatureBMI,
                                FeatureBMI2, FeatureFMA,
                                FeatureRTM]>;

Modified: llvm/trunk/lib/Target/X86/X86Subtarget.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Subtarget.cpp?rev=169730&r1=169729&r2=169730&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86Subtarget.cpp Mon Dec 10 03:18:44 2012
@@ -234,12 +234,20 @@
       ToggleFeature(X86::FeatureSlowBTMem);
     }
 
-    // If it's Nehalem, unaligned memory access is fast.
-    // Include Westmere and Sandy Bridge as well.
-    // FIXME: add later processors.
-    if (IsIntel && ((Family == 6 && Model == 26) ||
-        (Family == 6 && Model == 44) ||
-        (Family == 6 && Model == 42))) {
+    // If it's an Intel chip since Nehalem and not an Atom chip, unaligned
+    // memory access is fast. We hard code model numbers here because they
+    // aren't strictly increasing for Intel chips it seems.
+    if (IsIntel &&
+        ((Family == 6 && Model == 0x1E) || // Nehalem: Clarksfield, Lynnfield,
+                                           //          Jasper Froest
+         (Family == 6 && Model == 0x2A) || // Nehalem: Bloomfield, Nehalem-EP
+         (Family == 6 && Model == 0x2E) || // Nehalem: Nehalem-EX
+         (Family == 6 && Model == 0x25) || // Westmere: Arrandale, Clarksdale
+         (Family == 6 && Model == 0x2C) || // Westmere: Gulftown, Westmere-EP
+         (Family == 6 && Model == 0x2F) || // Westmere: Westmere-EX
+         (Family == 6 && Model == 0x2A) || // SandyBridge
+         (Family == 6 && Model == 0x2D) || // SandyBridge: SandyBridge-E*
+         (Family == 6 && Model == 0x3A))) {// IvyBridge
       IsUAMemFast = true;
       ToggleFeature(X86::FeatureFastUAMem);
     }





More information about the llvm-commits mailing list