BLAS compatibility library #7
Open
Conversation
This file contains 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
|
I've just added a unit test (currently failing) for SYRK although we probably don't want to run it with the regular unit tests because it requires a hardware kernel. Do you have any suggestions here? |
|
You can see what I did for GEMM, it's a test but not a unit test |
|
There's a nice routine to search for the kernel at |
definelicht
requested changes
Jan 4, 2022
Co-authored-by: definelicht <definelicht@inf.ethz.ch>
definelicht
requested changes
Jan 7, 2022
| const int k) { | ||
| #pragma HLS INLINE | ||
| ReadA_N: | ||
| for (int n1 = 0; n1 < kTileSizeN; ++n1) { | ||
| #pragma HLS PIPELINE II = 1 | ||
| #pragma HLS LOOP_FLATTEN | ||
| DramLine num[1]; | ||
| num[0] = mem[(n0 * kTileSizeN + n1) * size_k + k]; | ||
| num[0] = mem[((n0 * kTileSizeN + n1) + k * size_n) * kLinesPerNumber]; |
There was a problem hiding this comment.
Suggested change
| num[0] = mem[((n0 * kTileSizeN + n1) + k * size_n) * kLinesPerNumber]; | |
| num[0] = mem[k * size_n + n0 * kTileSizeN + n1]; |
| @@ -19,7 +24,7 @@ void ReadAInner(DramLine const *const mem, hlslib::Stream<PackedFloat> &a_to_fee | |||
| for (int i = 0; i < kLinesPerNumber; ++i) { | |||
| #pragma HLS PIPELINE II = 1 | |||
| #pragma HLS LOOP_FLATTEN | |||
| num[i] = mem[((n0 * kTileSizeN + n1) * size_k + k) * kLinesPerNumber + i]; | |||
| num[i] = mem[((n0 * kTileSizeN + n1) + k * size_n) * kLinesPerNumber + i]; | |||
There was a problem hiding this comment.
Suggested change
| num[i] = mem[((n0 * kTileSizeN + n1) + k * size_n) * kLinesPerNumber + i]; | |
| num[i] = mem[(k * size_n + n0 * kTileSizeN + n1) * kLinesPerNumber + i]; |
| @@ -100,7 +105,7 @@ void ReadBInner(DramLine const *const mem, hlslib::Stream<PackedFloat> &b_to_fee | |||
| for (int i = 0; i < kLinesPerNumber; ++i) { | |||
| #pragma HLS PIPELINE II = 1 | |||
| #pragma HLS LOOP_FLATTEN | |||
| num[i] = mem[(k * size_m + m0 * kTileSizeM + m1) * kLinesPerNumber + i]; | |||
| num[i] = mem[(k + (m0 * kTileSizeM + m1) * size_k) * kLinesPerNumber + i]; | |||
There was a problem hiding this comment.
Suggested change
| num[i] = mem[(k + (m0 * kTileSizeM + m1) * size_k) * kLinesPerNumber + i]; | |
| num[i] = mem[((m0 * kTileSizeM + m1) * size_k + k) * kLinesPerNumber + i]; |
| const int k) { | ||
| #pragma HLS INLINE | ||
| ReadB_M: | ||
| for (int m1 = 0; m1 < kTileSizeM; ++m1) { | ||
| #pragma HLS PIPELINE II = 1 | ||
| #pragma HLS LOOP_FLATTEN | ||
| DramLine num[1]; | ||
| num[0] = mem[k * size_m + m0 * kTileSizeM + m1]; | ||
| num[0] = mem[(k + (m0 * kTileSizeM + m1) * size_k) * kLinesPerNumber]; |
There was a problem hiding this comment.
Suggested change
| num[0] = mem[(k + (m0 * kTileSizeM + m1) * size_k) * kLinesPerNumber]; | |
| num[0] = mem[(m0 * kTileSizeM + m1) * size_k + k]; |
43d8a88
to
3860e6e
Compare
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.
We will want some integration tests for this, but it's probably more important to not have a huge diff piling up