Changes between Version 10 and Version 11 of AppCoprocessor

Timestamp:

Mar 12, 2008, 4:40:35 PM (17 years ago)

Author:

davea

Comment:

--

AppCoprocessor

@@ -11 +11 @@
 
 The BOINC client probes for coprocessors and reports them in scheduler requests.
-The XML looks like:
+
+== Deploying a coprocessor app ==
+
+BOINC uses the [AppPlan application planning] mechanism to
+coordinate the scheduling of multi-threaded applications.
+
+Suppose you've developed a coprocessor program,
+that it uses a CUDA GPU and 1 GFLOPS of the CPU.
+To deploy it:
+
+ * Choose a "planning class" name for the program, say "cuda_1.1" (see below).
+ * Create an [UpdateVersions app version].  Include a file '''app_plan''' containing "cuda_1.1".
+ * Link the following function into your scheduler:
 {{{
-<coprocs>
-   <coproc_cuda>
-      <count>1</count>
-      <name>GeForce 8800 GT (1)</name>
-      <totalGlobalMem>...</totalGlobalMem>
-      ...
-   </coproc_cuda>
-</coprocs>
+bool app_plan(HOST& host, char* plan_class, HOST_USAGE& hu) {
+    if (!strcmp(plan_class, "cuda_1.1")) {
+        for (i=0; i<host.coprocs.size(); i++) {
+            COPROC cp = host.coprocs[i];
+            if (cp.type == COPROC_CUDA) {
+                COPROC_USAGE cu;
+                strcpy(cu.name, cp.name);
+                cu.count = 1;
+                hu.coprocs.push_back(cu);
+                double x = 1e9/host.p_fpops;
+                if (x > 1) x = 1;
+                hu.ncpus = x;
+                hu.fpops = 5e11;
+                return true;
+            }
+        }
+    }
+    return false;
+}
 }}}
-
-
 == Questions ==