From 0454ad786d10e7fc4fd72dcfe1a3cbcb44d6aac3 Mon Sep 17 00:00:00 2001 From: Sebastian Huber Date: Thu, 12 Nov 2020 10:36:32 +0100 Subject: eng: Simplify "Load" environment reporting Report all runtime measurement environments with a name only and encode the worker count of the "Load" environment in the name. Update #3199. --- eng/test-framework.rst | 21 +++++++-------------- 1 file changed, 7 insertions(+), 14 deletions(-) diff --git a/eng/test-framework.rst b/eng/test-framework.rst index 94d2f35..589caf7 100644 --- a/eng/test-framework.rst +++ b/eng/test-framework.rst @@ -1303,14 +1303,14 @@ DirtyCache You can disable this variant with the `T_MEASURE_RUNTIME_DISABLE_DIRTY_CACHE` request flag. -Load +Load/ This variant tries to get close to worst-case conditions. The cache is set up according to the `DirtyCache` variant. In addition, other processors try to fully load the memory system. The load is produced through writes to a memory area with twice the size of the outer-most data cache. The load variant is performed multiple times with a different set of active - load worker threads (`M:L`). The active workers range from one up to the - processor count. + load worker threads. The value is the count of active + workers which ranges from one to the processor count. You can disable these variants with the `T_MEASURE_RUNTIME_DISABLE_MINOR_LOAD` and @@ -1393,8 +1393,7 @@ reported. M:D:0.000033244 M:E:Empty:D:1.887834875 M:B:Empty - M:V:Load - M:L:1 + M:V:Load/1 M:N:1024 M:MI:0.000000000 M:Q1:0.000000002 @@ -1407,8 +1406,7 @@ reported. [... 22 more load variants ...] M:E:Empty:D:0.021252583 M:B:Empty - M:V:Load - M:L:24 + M:V:Load/24 M:N:1024 M:MI:0.000000001 M:Q1:0.000000002 @@ -1907,8 +1905,6 @@ M **M:V:** - **M:L:** - **M:N:** **M:S::** @@ -1937,11 +1933,8 @@ M The execution variant which is one of **FullCache**, **HotCache**, - **DirtyCache**, or **Load**. - - - The active load workers count which ranges from one to the processor - count. + **DirtyCache**, or **Load/**. The is the + count of active workers which ranges from one to the processor count. The sample count as defined by the runtime measurement configuration. -- cgit v1.2.3