<div dir="ltr">Hi,<br><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Mar 30, 2015 at 12:30 PM, Justin Holewinski <span dir="ltr"><<a href="mailto:jholewinski@nvidia.com" target="_blank">jholewinski@nvidia.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Author: jholewinski<br>
Date: Mon Mar 30 14:30:55 2015<br>
New Revision: 233583<br>
<br>
URL: <a href="http://llvm.org/viewvc/llvm-project?rev=233583&view=rev" target="_blank">http://llvm.org/viewvc/llvm-project?rev=233583&view=rev</a><br>
Log:<br>
[NVPTX] Associate a minimum PTX version for each SM architecture<br>
<br></blockquote><div>... </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">URL: <a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/NVPTX/NVPTX.td?rev=233583&r1=233582&r2=233583&view=diff" target="_blank">http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/NVPTX/NVPTX.td?rev=233583&r1=233582&r2=233583&view=diff</a><br>
==============================================================================<br>
--- llvm/trunk/lib/Target/NVPTX/NVPTX.td (original)<br>
+++ llvm/trunk/lib/Target/NVPTX/NVPTX.td Mon Mar 30 14:30:55 2015<br>
@@ -46,10 +46,6 @@ def SM53 : SubtargetFeature<"sm_53", "Sm<br>
                             "Target SM 5.3">;<br>
<br>
 // PTX Versions<br>
-def PTX30 : SubtargetFeature<"ptx30", "PTXVersion", "30",<br>
-                             "Use PTX version 3.0">;<br>
-def PTX31 : SubtargetFeature<"ptx31", "PTXVersion", "31",<br>
-                             "Use PTX version 3.1">;<br>
 def PTX32 : SubtargetFeature<"ptx32", "PTXVersion", "32",<br>
                              "Use PTX version 3.2">;<br>
 def PTX40 : SubtargetFeature<"ptx40", "PTXVersion", "40",<br>
@@ -69,12 +65,12 @@ class Proc<string Name, list<SubtargetFe<br>
 def : Proc<"sm_20", [SM20]>;<br>
 def : Proc<"sm_21", [SM21]>;<br>
 def : Proc<"sm_30", [SM30]>;<br>
-def : Proc<"sm_32", [SM32]>;<br>
+def : Proc<"sm_32", [SM32, PTX40]>;<br>
 def : Proc<"sm_35", [SM35]>;<br></blockquote><div><br></div><div>Does it mean that SM35/SM37 would still be using PTX3.2? </div><div><br></div><div>libdevice.compute_35.10.bc that ships with CUDA-7.0 uses "rsqrt.approx.ftz.f64". That instruction was added in PTX4.0 which suggests that SM35/37 may have to bumped to PTX 4.0, too.</div><div><br></div><div>--Artem </div></div>
</div></div>