Enhancement - Fix hipblaslt-gemm result parsing for hipBLASLt v1500+ output format#791
Open
polarG wants to merge 6 commits into
Open
Enhancement - Fix hipblaslt-gemm result parsing for hipBLASLt v1500+ output format#791polarG wants to merge 6 commits into
polarG wants to merge 6 commits into
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
Updates hipblaslt-gemm result parsing to support hipBLASLt v1500+ CSV output by using header-based column lookup instead of hardcoded field counts/positions.
Changes:
- Parse the CSV header to dynamically locate the
hipblaslt-Gflopscolumn and validate row width against the header. - Add a unit test covering the new v1500+ (34-column) output format.
- Clarify parsing intent via inline comments.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| tests/benchmarks/micro_benchmarks/test_hipblaslt_function.py | Adds a new test case for v1500+ output and annotates the existing old-format test. |
| superbench/benchmarks/micro_benchmarks/hipblaslt_function.py | Switches parsing to header-based hipblaslt-Gflops extraction and header/data column-count validation. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #791 +/- ##
==========================================
+ Coverage 85.69% 85.70% +0.01%
==========================================
Files 103 103
Lines 7890 7907 +17
==========================================
+ Hits 6761 6777 +16
- Misses 1129 1130 +1
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
added 2 commits
May 2, 2026 00:11
…sserts - Use exact match (col_name.strip() == 'hipblaslt-Gflops') for the FLOPS column lookup, avoiding accidental matches against future names like 'hipblaslt-Gflops-peak'. - Rewrite the inline comment to drop misleading positional index claims and use the correct 'hipblaslt-GB/s' name instead of 'GB_s'. - Replace brittle assertEqual(2, len(...)) and exact float assertEqual with assertIn(...) + assertAlmostEqual(..., places=3) for the new-format test.
- Harden first-header-field rank-marker stripping with an anchored regex (re.sub r'^\s*\[\d+\]:?'), so a column whose name happens to contain ']' is not truncated. - Add an explicit IndexError guard before reading lines[index + 1] so hipblaslt-bench output that ends on the header (e.g., crash after printing it) yields a clear ValueError instead of a generic IndexError.
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
WenqingLan1
approved these changes
May 12, 2026
guoshzhao
approved these changes
May 13, 2026
Resolves the remaining unresolved Copilot review comment on PR #791 (comment id 3177652631): the previous fix only resolved the hipblaslt-Gflops column by header name; the metric key still used fixed positions 3-6 for batch_count/m/n/k, defeating the forward-compatibility claim. Changes: - Build a column-name -> index map from the parsed header (first-occurrence-wins via dict.setdefault) and resolve every key column (batch_count, m, n, k, hipblaslt-Gflops) by name. - Raise ValueError listing all missing required columns so unknown output formats fail loudly instead of silently producing a wrong metric key. - Strip whitespace from each selected data value before building the metric key, so any CSV padding does not bleed into the key. - Replace the misleading 'indices 3/4-6 are stable' comment with a note explaining that all key fields are header-driven. Tests: - Add forward-compat positive test that inserts an extra column before batch_count and pads data values with whitespace, asserting the correct key is produced and that no key derived from the wrong positional field leaks through. - Add negative test asserting MICROBENCHMARK_RESULT_PARSING_FAILURE when a required column (batch_count) is missing from the header.
Resolves Copilot inline review comment id 3277063504 on PR #791. The v1500 and future-format tests already followed the assertIn + assertAlmostEqual pattern after the previous review round; this commit applies the same hardening to the v600 positive test that was missed. - Drop self.assertEqual(2, len(benchmark.result)) which coupled the test to the always-present return_code metric instead of the behavior under test. - Replace self.assertEqual(58.6245, ...) (exact float equality) with self.assertIn(...) + self.assertAlmostEqual(..., places=4) so the numeric check is not flaky on floating-point representation. places=4 matches the four decimal places in the expected fixture value.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
The hipblaslt-gemm benchmark result parser fails with MICROBENCHMARK_RESULT_PARSING_FAILURE (return code 33) when running against hipBLASLt v1500+. The benchmark kernels execute successfully and produce valid TFLOPS data, but SuperBench cannot parse the results into structured metrics.
Root cause: The parser hardcodes len(fields) != 23 to validate the output CSV, but hipBLASLt v1500 outputs 34 columns — it added a_type, b_type, c_type, d_type, scaleA, scaleB, scaleC, scaleD, amaxD, bias_type, aux_type, and hipblaslt-GB/s. This causes two bugs:
Fix
Replace the hardcoded field-count check and positional index with header-based column lookup:
This approach is: