[llvm-exegesis] Improve documentation.

Summary:
- Better flag names.
- Fix flag reference in doc.
- Add usage examples in doc.

Fixes PR37497.

Reviewers: gchatelet

Subscribers: llvm-commits, tschuett

Differential Revision: https://reviews.llvm.org/D47015

llvm-svn: 332708

docs/CommandGuide/llvm-exegesis.rst

@@ -22,7 +22,116 @@ The code snippet is jitted and executed on the host subtarget. The time taken
 result is printed out as YAML to the standard output.
 
 The main goal of this tool is to automatically (in)validate the LLVM's TableDef
-scheduling models.
+scheduling models. To that end, we also provide analysis of the results.
+
+EXAMPLES: benchmarking
+----------------------
+
+Assume you have an X86-64 machine. To measure the latency of a single
+instruction, run:
+
+.. code-block:: bash
+
+    $ llvm-exegesis -mode=latency -opcode-name=ADD64rr
+
+Measuring the uop decomposition of an instruction works similarly:
+
+.. code-block:: bash
+
+    $ llvm-exegesis -mode=uops -opcode-name=ADD64rr
+
+The output is a YAML document (the default is to write to stdout, but you can
+redirect the output to a file using `-benchmarks-file`):
+
+.. code-block:: none
+
+  ---
+  key:
+    opcode_name:     ADD64rr
+    mode:            latency
+    config:          ''
+  cpu_name:        haswell
+  llvm_triple:     x86_64-unknown-linux-gnu
+  num_repetitions: 10000
+  measurements:
+    - { key: latency, value: 1.0058, debug_string: '' }
+  error:           ''
+  info:            'explicit self cycles, selecting one aliasing configuration.
+  Snippet:
+  ADD64rr R8, R8, R10
+  '
+  ...
+
+To measure the latency of all instructions for the host architecture, run:
+
+.. code-block:: bash
+
+  #!/bin/bash
+  readonly INSTRUCTIONS=$(grep INSTRUCTION_LIST_END build/lib/Target/X86/X86GenInstrInfo.inc | cut -f2 -d=)
+  for INSTRUCTION in $(seq 1 ${INSTRUCTIONS});
+  do
+    ./build/bin/llvm-exegesis -mode=latency -opcode-index=${INSTRUCTION} | sed -n '/---/,$p'
+  done
+
+FIXME: Provide an :program:`llvm-exegesis` option to test all instructions.
+
+EXAMPLES: analysis
+----------------------
+
+Assuming you have a set of benchmarked instructions (either latency or uops) as
+YAML in file `/tmp/benchmarks.yaml`, you can analyze the results using the
+following command:
+
+.. code-block:: bash
+
+    $ llvm-exegesis -mode=analysis \
+  -benchmarks-file=/tmp/benchmarks.yaml \
+  -analysis-clusters-output-file=/tmp/clusters.csv \
+  -analysis-inconsistencies-output-file=/tmp/inconsistencies.txt
+
+This will group the instructions into clusters with the same performance
+characteristics. The clusters will be written out to `/tmp/clusters.csv` in the
+following format:
+
+.. code-block:: none
+
+  cluster_id,opcode_name,config,sched_class
+  ...
+  2,ADD32ri8_DB,,WriteALU,1.00
+  2,ADD32ri_DB,,WriteALU,1.01
+  2,ADD32rr,,WriteALU,1.01
+  2,ADD32rr_DB,,WriteALU,1.00
+  2,ADD32rr_REV,,WriteALU,1.00
+  2,ADD64i32,,WriteALU,1.01
+  2,ADD64ri32,,WriteALU,1.01
+  2,MOVSX64rr32,,BSWAP32r_BSWAP64r_MOVSX64rr32,1.00
+  2,VPADDQYrr,,VPADDBYrr_VPADDDYrr_VPADDQYrr_VPADDWYrr_VPSUBBYrr_VPSUBDYrr_VPSUBQYrr_VPSUBWYrr,1.02
+  2,VPSUBQYrr,,VPADDBYrr_VPADDDYrr_VPADDQYrr_VPADDWYrr_VPSUBBYrr_VPSUBDYrr_VPSUBQYrr_VPSUBWYrr,1.01
+  2,ADD64ri8,,WriteALU,1.00
+  2,SETBr,,WriteSETCC,1.01
+  ...
+
+:program:`llvm-exegesis` will also analyze the clusters to point out
+inconsistencies in the scheduling information. For example,
+`/tmp/inconsistencies.txt` will contain messages like:
+
+.. code-block:: none
+
+  Sched Class EXTRACTPSrr_VEXTRACTPSrr contains instructions with distinct performance characteristics, falling into 2 clusters:
+  4,EXTRACTPSrr,,3.00
+  3,VEXTRACTPSrr,,2.01
+
+  Sched Class WriteCRC32 contains instructions with distinct performance characteristics, falling into 2 clusters:
+  4,CRC32r32r16,,3.01
+  4,CRC32r32r32,,3.00
+  11,CRC32r32r8,,4.01
+  4,CRC32r64r64,,3.01
+  4,CRC32r64r8,,3.00
+
+Note that the scheduling class names will be resolved only when
+:program:`llvm-exegesis` is compiled in debug mode, else only the class id will
+be shown. This does not invalidate any of the analysis results though.
+
 
 OPTIONS
 -------
@@ -41,15 +150,40 @@ OPTIONS
  Specify the opcode to measure, by name.
  Either `opcode-index` or `opcode-name` must be set.
 
-.. option:: -benchmark-mode=[Latency|Uops]
+.. option:: -mode=[latency|uops|analysis]
 
- Specify which characteristic of the opcode to measure.
+ Specify the run mode.
 
 .. option:: -num-repetitions=<Number of repetition>
 
  Specify the number of repetitions of the asm snippet.
  Higher values lead to more accurate measurements but lengthen the benchmark.
 
+ .. option:: -benchmarks-file=</path/to/file>
+
+ File to read (`analysis` mode) or write (`latency`/`uops` modes) benchmark
+ results. "-" uses stdin/stdout.
+
+.. option:: -analysis-clusters-output-file=</path/to/file>
+
+ If provided, write the analysis clusters as CSV to this file. "-" prints to
+ stdout.
+
+.. option:: -analysis-inconsistencies-output-file=</path/to/file>
+
+ If non-empty, write inconsistencies found during analysis to this file. `-`
+ prints to stdout.
+
+.. option:: -analysis-numpoints=<dbscan numPoints parameter>
+
+ Specify the numPoints parameters to be used for DBSCAN clustering
+ (`analysis` mode).
+
+.. option:: -analysis-espilon=<dbscan epsilon parameter>
+
+ Specify the numPoints parameters to be used for DBSCAN clustering
+ (`analysis` mode).
+
 
 EXIT STATUS
 -----------

tools/llvm-exegesis/llvm-exegesis.cpp

@@ -49,7 +49,7 @@ static llvm::cl::opt<std::string>
 
 enum class BenchmarkModeE { Latency, Uops, Analysis };
 static llvm::cl::opt<BenchmarkModeE> BenchmarkMode(
-    "benchmark-mode", llvm::cl::desc("the benchmark mode to run"),
+    "mode", llvm::cl::desc("the mode to run"),
     llvm::cl::values(
         clEnumValN(BenchmarkModeE::Latency, "latency", "Instruction Latency"),
         clEnumValN(BenchmarkModeE::Uops, "uops", "Uop Decomposition"),