diff options
| author | Tom de Vries <tdevries@suse.de> | 2019-01-12 22:17:42 +0000 |
|---|---|---|
| committer | Tom de Vries <vries@gcc.gnu.org> | 2019-01-12 22:17:42 +0000 |
| commit | 2b9d9e393766d2fa6e2dd5f361d0db14872cf261 (patch) | |
| tree | f7686faf57ee3b6e79c8818d2179d328251d77d4 /gcc | |
| parent | 6f7814d08fd89dc9f41807568ef26c3ffe498dba (diff) | |
| download | gcc-2b9d9e393766d2fa6e2dd5f361d0db14872cf261.zip gcc-2b9d9e393766d2fa6e2dd5f361d0db14872cf261.tar.gz gcc-2b9d9e393766d2fa6e2dd5f361d0db14872cf261.tar.bz2 | |
[nvptx] Enable large vectors
Allow vector_length clauses to accept values larger than warp size. Note that
this does not enable setting vector_length to values larger than warp size using
-fopenacc-dim.
2019-01-12 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Take larger vector
lengths into account.
* testsuite/libgomp.oacc-c-c++-common/vector-length-128-1.c: Expect
vector length to be 128.
* testsuite/libgomp.oacc-c-c++-common/parallel-dims.c: Expect vector
length 2097152 to be reduced to 1024 instead of 32.
From-SVN: r267889
Diffstat (limited to 'gcc')
| -rw-r--r-- | gcc/ChangeLog | 5 | ||||
| -rw-r--r-- | gcc/config/nvptx/nvptx.c | 2 |
2 files changed, 6 insertions, 1 deletions
diff --git a/gcc/ChangeLog b/gcc/ChangeLog index a2735a5..6f18a34 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,8 @@ +2019-01-12 Tom de Vries <tdevries@suse.de> + + * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Take larger vector + lengths into account. + 2019-01-12 Svante Signell <svante.signell@gmail.com> * config/i386/gnu.h (TARGET_THREAD_SSP_OFFSET): Define. diff --git a/gcc/config/nvptx/nvptx.c b/gcc/config/nvptx/nvptx.c index 1d97045..8d2740c 100644 --- a/gcc/config/nvptx/nvptx.c +++ b/gcc/config/nvptx/nvptx.c @@ -96,7 +96,7 @@ #define PTX_NUM_PER_WORKER_BARRIERS (PTX_CTA_NUM_BARRIERS - PTX_NUM_PER_CTA_BARRIERS) #define PTX_DEFAULT_VECTOR_LENGTH PTX_WARP_SIZE -#define PTX_MAX_VECTOR_LENGTH PTX_WARP_SIZE +#define PTX_MAX_VECTOR_LENGTH PTX_CTA_SIZE #define PTX_WORKER_LENGTH 32 #define PTX_DEFAULT_RUNTIME_DIM 0 /* Defer to runtime. */ |