diff options
| author | Peter Bergner <bergner@linux.ibm.com> | 2020-07-22 11:44:35 -0500 |
|---|---|---|
| committer | Peter Bergner <bergner@linux.ibm.com> | 2020-07-22 13:36:28 -0500 |
| commit | ae575662833d70cb7d74b9538096c7becc79af14 (patch) | |
| tree | 54c823e58d7e47cc8d6752ac099a147bddabf64e /gcc | |
| parent | 6e1e0decc9e17a4283d1b5508e892be5215b8ab9 (diff) | |
| download | gcc-ae575662833d70cb7d74b9538096c7becc79af14.zip gcc-ae575662833d70cb7d74b9538096c7becc79af14.tar.gz gcc-ae575662833d70cb7d74b9538096c7becc79af14.tar.bz2 | |
rs6000: __builtin_mma_disassemble_acc() doesn't store elements correctly in LE mode
PR96236 shows a problem where we don't correctly store our 512-bit accumulators
correctly in little-endian mode. The patch below detects when we're doing a
little-endian memory access and stores to the correct memory locations.
2020-07-22 Peter Bergner <bergner@linux.ibm.com>
gcc/
PR target/96236
* config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Handle
little-endian memory ordering.
gcc/testsuite/
PR target/96236
* gcc.target/powerpc/mma-double-test.c: Update storing results for
correct little-endian ordering.
* gcc.target/powerpc/mma-single-test.c: Likewise.
Diffstat (limited to 'gcc')
| -rw-r--r-- | gcc/config/rs6000/rs6000-call.c | 3 | ||||
| -rwxr-xr-x | gcc/testsuite/gcc.target/powerpc/mma-double-test.c | 8 | ||||
| -rwxr-xr-x | gcc/testsuite/gcc.target/powerpc/mma-single-test.c | 16 |
3 files changed, 14 insertions, 13 deletions
diff --git a/gcc/config/rs6000/rs6000-call.c b/gcc/config/rs6000/rs6000-call.c index 5ec3f2c..bb0fdf2 100644 --- a/gcc/config/rs6000/rs6000-call.c +++ b/gcc/config/rs6000/rs6000-call.c @@ -11154,11 +11154,12 @@ rs6000_gimple_fold_mma_builtin (gimple_stmt_iterator *gsi) tree src_array = build1 (VIEW_CONVERT_EXPR, array_type, src); for (unsigned i = 0; i < 4; i++) { + unsigned index = WORDS_BIG_ENDIAN ? i : 3 - i; tree ref = build4 (ARRAY_REF, unsigned_V16QI_type_node, src_array, build_int_cst (size_type_node, i), NULL_TREE, NULL_TREE); tree dst = build2 (MEM_REF, unsigned_V16QI_type_node, dst_base, - build_int_cst (dst_type, i * 16)); + build_int_cst (dst_type, index * 16)); gimplify_assign (dst, ref, &new_seq); } pop_gimplify_context (NULL); diff --git a/gcc/testsuite/gcc.target/powerpc/mma-double-test.c b/gcc/testsuite/gcc.target/powerpc/mma-double-test.c index ac84ae3..044a288 100755 --- a/gcc/testsuite/gcc.target/powerpc/mma-double-test.c +++ b/gcc/testsuite/gcc.target/powerpc/mma-double-test.c @@ -12,13 +12,13 @@ typedef double v4sf_t __attribute__ ((vector_size (16))); #define SAVE_ACC(ACC, ldc, J) \ __builtin_mma_disassemble_acc (result, ACC); \ rowC = (v4sf_t *) &CO[0*ldc+J]; \ - rowC[0] += result[3] ; \ + rowC[0] += result[0]; \ rowC = (v4sf_t *) &CO[1*ldc+J]; \ - rowC[0] += result[2] ; \ + rowC[0] += result[1]; \ rowC = (v4sf_t *) &CO[2*ldc+J]; \ - rowC[0] += result[1] ; \ + rowC[0] += result[2]; \ rowC = (v4sf_t *) &CO[3*ldc+J]; \ - rowC[0] += result[0] ; + rowC[0] += result[3]; void MMA (int m, int n, int k, double *A, double *B, double *C) diff --git a/gcc/testsuite/gcc.target/powerpc/mma-single-test.c b/gcc/testsuite/gcc.target/powerpc/mma-single-test.c index 15369a6..7e628df 100755 --- a/gcc/testsuite/gcc.target/powerpc/mma-single-test.c +++ b/gcc/testsuite/gcc.target/powerpc/mma-single-test.c @@ -12,24 +12,24 @@ typedef float v4sf_t __attribute__ ((vector_size (16))); #define SAVE_ACC(ACC, ldc,J) \ __builtin_mma_disassemble_acc (result, ACC); \ rowC = (v4sf_t *) &CO[0*ldc+J]; \ - rowC[0] += result[3] ; \ + rowC[0] += result[0]; \ rowC = (v4sf_t *) &CO[1*ldc+J]; \ - rowC[0] += result[2] ; \ + rowC[0] += result[1]; \ rowC = (v4sf_t *) &CO[2*ldc+J]; \ - rowC[0] += result[1] ; \ + rowC[0] += result[2]; \ rowC = (v4sf_t *) &CO[3*ldc+J]; \ - rowC[0] += result[0] ; + rowC[0] += result[3]; #define SAVE_ACC1(ACC,ldc, J) \ __builtin_mma_disassemble_acc (result, ACC); \ rowC = (v4sf_t *) &CO[4* ldc+J]; \ - rowC[0] += result[3] ; \ + rowC[0] += result[0]; \ rowC = (v4sf_t *) &CO[5*ldc+J]; \ - rowC[0] += result[2] ; \ + rowC[0] += result[1]; \ rowC = (v4sf_t *) &CO[6*ldc+J]; \ - rowC[0] += result[1] ; \ + rowC[0] += result[2]; \ rowC = (v4sf_t *) &CO[7*ldc+J]; \ - rowC[0] += result[0] ; + rowC[0] += result[3]; void MMA (int m, int n, int k, float *A, float *B, float *C) { |