Skip to content

coll/accelerator initial enhancements #12986

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Dec 19, 2024
Merged

coll/accelerator initial enhancements #12986

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 9 additions & 6 deletions ompi/mca/coll/accelerator/coll_accelerator.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
* reserved.
* Copyright (c) 2014-2024 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -87,22 +88,24 @@ mca_coll_accelerator_reduce_scatter_block(const void *sbuf, void *rbuf, size_t r
* @retval >0 The buffer belongs to a managed buffer in
* device memory.
*/
static inline int mca_coll_accelerator_check_buf(void *addr)
static inline int mca_coll_accelerator_check_buf(void *addr, int *dev_id)
{
uint64_t flags;
int dev_id;

if (OPAL_LIKELY(NULL != addr)) {
return opal_accelerator.check_addr(addr, &dev_id, &flags);
return opal_accelerator.check_addr(addr, dev_id, &flags);
} else {
*dev_id = MCA_ACCELERATOR_NO_DEVICE_ID;
return 0;
}
}

static inline void *mca_coll_accelerator_memcpy(void *dest, const void *src, size_t size)
static inline void *mca_coll_accelerator_memcpy(void *dest, int dest_dev, const void *src, int src_dev, size_t size,
opal_accelerator_transfer_type_t type)
{
int res;
res = opal_accelerator.mem_copy(MCA_ACCELERATOR_NO_DEVICE_ID, MCA_ACCELERATOR_NO_DEVICE_ID,
dest, src, size, MCA_ACCELERATOR_TRANSFER_UNSPEC);

res = opal_accelerator.mem_copy(dest_dev, src_dev, dest, src, size, type);
if (res != 0) {
opal_output(0, "coll/accelerator: Error in mem_copy: res=%d, dest=%p, src=%p, size=%d", res, dest, src,
(int) size);
Expand Down
15 changes: 10 additions & 5 deletions ompi/mca/coll/accelerator/coll_accelerator_allreduce.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
* Copyright (c) 2014-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2022 Amazon.com, Inc. or its affiliates. All Rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -37,11 +38,12 @@ mca_coll_accelerator_allreduce(const void *sbuf, void *rbuf, size_t count,
mca_coll_accelerator_module_t *s = (mca_coll_accelerator_module_t*) module;
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int sbuf_dev, rbuf_dev;
size_t bufsize;
int rc;

bufsize = opal_datatype_span(&dtype->super, count, &gap);
rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -50,10 +52,11 @@ mca_coll_accelerator_allreduce(const void *sbuf, void *rbuf, size_t count,
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, bufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev,
bufsize, MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}
rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -63,7 +66,8 @@ mca_coll_accelerator_allreduce(const void *sbuf, void *rbuf, size_t count,
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, bufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev,
bufsize, MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -73,7 +77,8 @@ mca_coll_accelerator_allreduce(const void *sbuf, void *rbuf, size_t count,
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, bufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, bufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand Down
15 changes: 10 additions & 5 deletions ompi/mca/coll/accelerator/coll_accelerator_exscan.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
* Copyright (c) 2014-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2022 Amazon.com, Inc. or its affiliates. All Rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand All @@ -29,11 +30,12 @@ int mca_coll_accelerator_exscan(const void *sbuf, void *rbuf, size_t count,
mca_coll_accelerator_module_t *s = (mca_coll_accelerator_module_t*) module;
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int sbuf_dev, rbuf_dev;
size_t bufsize;
int rc;

bufsize = opal_datatype_span(&dtype->super, count, &gap);
rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -43,10 +45,11 @@ int mca_coll_accelerator_exscan(const void *sbuf, void *rbuf, size_t count,
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, bufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}
rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -56,7 +59,8 @@ int mca_coll_accelerator_exscan(const void *sbuf, void *rbuf, size_t count,
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, bufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -68,7 +72,8 @@ int mca_coll_accelerator_exscan(const void *sbuf, void *rbuf, size_t count,
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, bufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, bufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand Down
29 changes: 19 additions & 10 deletions ompi/mca/coll/accelerator/coll_accelerator_reduce.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@
* Copyright (c) 2014-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2022 Amazon.com, Inc. or its affiliates. All Rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -39,12 +40,13 @@ mca_coll_accelerator_reduce(const void *sbuf, void *rbuf, size_t count,
int rank = ompi_comm_rank(comm);
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int rbuf_dev, sbuf_dev;
size_t bufsize;
int rc;

bufsize = opal_datatype_span(&dtype->super, count, &gap);

rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -53,11 +55,12 @@ mca_coll_accelerator_reduce(const void *sbuf, void *rbuf, size_t count,
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, bufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}

rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -67,7 +70,8 @@ mca_coll_accelerator_reduce(const void *sbuf, void *rbuf, size_t count,
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, bufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -80,7 +84,8 @@ mca_coll_accelerator_reduce(const void *sbuf, void *rbuf, size_t count,
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, bufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, bufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand All @@ -94,12 +99,13 @@ mca_coll_accelerator_reduce_local(const void *sbuf, void *rbuf, size_t count,
{
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int sbuf_dev, rbuf_dev;
size_t bufsize;
int rc;

bufsize = opal_datatype_span(&dtype->super, count, &gap);

rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -109,11 +115,12 @@ mca_coll_accelerator_reduce_local(const void *sbuf, void *rbuf, size_t count,
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, bufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}

rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -124,7 +131,8 @@ mca_coll_accelerator_reduce_local(const void *sbuf, void *rbuf, size_t count,
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, bufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -137,7 +145,8 @@ mca_coll_accelerator_reduce_local(const void *sbuf, void *rbuf, size_t count,
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, bufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, bufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand Down
15 changes: 10 additions & 5 deletions ompi/mca/coll/accelerator/coll_accelerator_reduce_scatter_block.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
* Copyright (c) 2014-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2022 Amazon.com, Inc. or its affiliates. All Rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -41,13 +42,14 @@ mca_coll_accelerator_reduce_scatter_block(const void *sbuf, void *rbuf, size_t r
mca_coll_accelerator_module_t *s = (mca_coll_accelerator_module_t*) module;
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int sbuf_dev, rbuf_dev;
size_t sbufsize, rbufsize;
int rc;

rbufsize = opal_datatype_span(&dtype->super, rcount, &gap);

sbufsize = rbufsize * ompi_comm_size(comm);
rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -56,10 +58,11 @@ mca_coll_accelerator_reduce_scatter_block(const void *sbuf, void *rbuf, size_t r
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, sbufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev, sbufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}
rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -69,7 +72,8 @@ mca_coll_accelerator_reduce_scatter_block(const void *sbuf, void *rbuf, size_t r
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, rbufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev, rbufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -80,7 +84,8 @@ mca_coll_accelerator_reduce_scatter_block(const void *sbuf, void *rbuf, size_t r
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, rbufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand Down
15 changes: 10 additions & 5 deletions ompi/mca/coll/accelerator/coll_accelerator_scan.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
* Copyright (c) 2014-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2022 Amazon.com, Inc. or its affiliates. All Rights reserved.
* Copyright (c) 2024 Triad National Security, LLC. All rights reserved.
* Copyright (c) 2024 Advanced Micro Devices, Inc. All Rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -36,11 +37,12 @@ int mca_coll_accelerator_scan(const void *sbuf, void *rbuf, size_t count,
mca_coll_accelerator_module_t *s = (mca_coll_accelerator_module_t*) module;
ptrdiff_t gap;
char *rbuf1 = NULL, *sbuf1 = NULL, *rbuf2 = NULL;
int sbuf_dev, rbuf_dev;
size_t bufsize;
int rc;

bufsize = opal_datatype_span(&dtype->super, count, &gap);
rc = mca_coll_accelerator_check_buf((void *)sbuf);
rc = mca_coll_accelerator_check_buf((void *)sbuf, &sbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -49,10 +51,11 @@ int mca_coll_accelerator_scan(const void *sbuf, void *rbuf, size_t count,
if (NULL == sbuf1) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(sbuf1, sbuf, bufsize);
mca_coll_accelerator_memcpy(sbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, sbuf, sbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
sbuf = sbuf1 - gap;
}
rc = mca_coll_accelerator_check_buf(rbuf);
rc = mca_coll_accelerator_check_buf(rbuf, &rbuf_dev);
if (rc < 0) {
return rc;
}
Expand All @@ -62,7 +65,8 @@ int mca_coll_accelerator_scan(const void *sbuf, void *rbuf, size_t count,
if (NULL != sbuf1) free(sbuf1);
return OMPI_ERR_OUT_OF_RESOURCE;
}
mca_coll_accelerator_memcpy(rbuf1, rbuf, bufsize);
mca_coll_accelerator_memcpy(rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, rbuf, rbuf_dev, bufsize,
MCA_ACCELERATOR_TRANSFER_DTOH);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - gap;
}
Expand All @@ -73,7 +77,8 @@ int mca_coll_accelerator_scan(const void *sbuf, void *rbuf, size_t count,
}
if (NULL != rbuf1) {
rbuf = rbuf2;
mca_coll_accelerator_memcpy(rbuf, rbuf1, bufsize);
mca_coll_accelerator_memcpy(rbuf, rbuf_dev, rbuf1, MCA_ACCELERATOR_NO_DEVICE_ID, bufsize,
MCA_ACCELERATOR_TRANSFER_HTOD);
free(rbuf1);
}
return rc;
Expand Down
Loading