Bar Charts

Complete reference for bar chart visualizations in RING-5.

Overview

Bar charts display categorical data with rectangular bars representing values. RING-5 supports four bar chart variants:

• Simple Bar Chart
• Grouped Bar Chart
• Stacked Bar Chart
• Grouped Stacked Bar Chart

Simple Bar Chart

Purpose

Compare values across categories with one bar per category.

Configuration

Required:

• X-axis: Categorical column (benchmark, config)
• Y-axis: Numeric column (ipc, throughput)

Optional:

• Color: Single color or column-based
• Title and labels
• Legend settings

Example Use Cases

Performance Comparison:

X: config
Y: ipc
# Compares IPC across configurations

Benchmark Analysis:

X: benchmark
Y: execution_time
# Shows execution time per benchmark

Best Practices

1. Limit categories to 15-20 for readability
2. Sort bars by value for easier comparison
3. Use horizontal bars for long category names
4. Add value labels for precise comparisons

Grouped Bar Chart

Purpose

Compare multiple groups side-by-side within each category.

Configuration

Required:

• X-axis: Primary category (benchmark)
• Y-axis: Numeric value (ipc)
• Group by: Secondary category (config)

Optional:

• Color scheme per group
• Bar width and spacing
• Legend position

Example Use Cases

Multi-Configuration Comparison:

X: benchmark
Y: ipc
Group by: config
# Shows all configs per benchmark side-by-side

A/B Testing:

X: workload
Y: throughput
Group by: version
# Compares versions across workloads

Best Practices

1. Limit groups to 5-7 per category
2. Use consistent color scheme
3. Position legend clearly
4. Consider stacked bars if showing composition

Stacked Bar Chart

Purpose

Show composition of totals with components stacked vertically.

Configuration

Required:

• X-axis: Category (benchmark)
• Y-axis: Numeric value (cycles, bytes)
• Stack by: Component (operation_type, cache_level)

Optional:

• Normalize to 100% (percentage stacking)
• Color per stack component
• Show totals on top

Example Use Cases

Cache Hierarchy:

X: benchmark
Y: misses
Stack by: cache_level
# Shows L1, L2, L3 miss breakdown

Operation Breakdown:

X: config
Y: cycles
Stack by: operation
# Stacks read, write, compute cycles

Best Practices

1. Limit stack components to 5-7
2. Order stacks logically (largest at bottom)
3. Use diverging colors for contrast
4. Consider percentage mode for proportion analysis

Grouped Stacked Bar Chart

Purpose

Combine grouping and stacking for complex multi-dimensional comparisons.

Configuration

Required:

• X-axis: Primary category
• Y-axis: Numeric value
• Group by: Groups within category
• Stack by: Components within groups

Example:

X: benchmark
Y: memory_bytes
Group by: config
Stack by: allocation_type
# Groups configs, stacks allocation types

Best Practices

1. Use sparingly - can be complex
2. Ensure clear legend
3. Consider alternatives (faceted plots)
4. Limit to 3-4 groups and 3-4 stacks

Styling Guidelines

Colors

For Simple Bars:

• Single color: Professional blue/gray
• Color by category: Use consistent palette

For Grouped Bars:

• Sequential colors per group
• High contrast between groups
• Colorblind-friendly palettes

For Stacked Bars:

• Diverging colors for components
• Consistent colors across charts
• Ordered by intensity

Labels

Axis Labels:

X-axis: "Configuration"
Y-axis: "Instructions Per Cycle (IPC)"
Title: "IPC Comparison Across Configurations"

Value Labels:

• Show for < 10 bars
• Hide for dense charts
• Format consistently (2 decimal places)

Fonts

Publication Quality:

• Title: 18-20pt bold
• Axis labels: 14-16pt
• Tick labels: 12-14pt
• Value labels: 10-12pt

Data Preparation

Column Selection

Use Column Selector shaper:

{
    "type": "columnSelector",
    "columns": ["benchmark", "config", "ipc"]
}

Sorting

Control bar order with Sort shaper:

{
    "type": "sort",
    "order_dict": {
        "benchmark": ["mcf", "omnetpp", "xalancbmk"]
    }
}

Normalization

Show relative performance:

{
    "type": "normalize",
    "normalizeVars": ["ipc"],
    "normalizerColumn": "config",
    "normalizerValue": "baseline"
}

Common Patterns

Speedup Chart

Pipeline:
  1. Normalize to baseline
  2. Sort by value
  3. Add geomean row
Styling:
  - Horizontal bars
  - Value labels
  - Baseline reference line

Performance Comparison

Pipeline:
  1. Filter top N benchmarks
  2. Sort alphabetically
  3. Add mean aggregation
Styling:
  - Grouped bars by config
  - Consistent colors
  - Clear legend

Troubleshooting

Bars Too Narrow

• Reduce number of categories
• Increase plot width
• Use horizontal orientation

Overlapping Labels

• Rotate tick labels 45°
• Use abbreviations
• Increase plot width

Too Many Groups

• Filter data to top N
• Create separate plots
• Use faceted visualization

Next Steps

• Line Plots: Line-Plots.md
• Scatter Plots: Scatter-Plots.md
• Histograms: ../histogram-plot.md