1. This API returns memory allocation size of pointer
2. Added test to check its functionality
Change-Id: I87976d817b5a6ca5530336c09e7cb0420601cb2c
[ROCm/hip commit: 1546732604]
Squashed commit of the following:
commit c111b5bd10d7c2a5b0b1ad8b07f6e81185b47b39
Author: Wen-Heng (Jack) Chung <whchung@gmail.com>
Date: Sat Mar 4 17:06:46 2017 +0800
Use __device__ for all variables and functions to be used in kernel path
Abolish __device and adopt [[hc]] in HIP implementation, so __device__ can be
used on all HIP applications, no matter they are variables or functions.
Change-Id: I20ca25857ce3bc3e42a5ebf65cafea2c8492f4c7
commit 30c0e4e4701bbf6bd9a7182e0320a71ff73d3a83
Author: Wen-Heng (Jack) Chung <whchung@gmail.com>
Date: Thu Mar 2 12:14:11 2017 +0800
XXX FIXME get around LDS spills caused in Promote-free HCC
hipDynamicShared2 uses all 64KB of LDS for computation. But in Promote-free HCC
there are cases where LDS spills would occur, which would make the test case to
hang.
In this workaround commit we reduce the size of dynamic LDS used to get around
this known issue, and will revert this commit when LDS spills are resolved in
HCC.
Change-Id: If648b36200a4f9143951a8129192bcb7ed0bef5e
commit e803173be2d73e2f132a7ff7f61e7a20b4083d34
Author: Wen-Heng (Jack) Chung <whchung@gmail.com>
Date: Wed Mar 1 21:41:41 2017 +0800
Fix math functions which take pointer arguments
Change-Id: I332c997e640edbc44824691e2a9434c6b3dadefa
commit de590c469e213c42090ff83dbd060f25bb1d6047
Author: Wen-Heng (Jack) Chung <whchung@gmail.com>
Date: Wed Mar 1 18:38:54 2017 +0800
Changes to cope with Promote-free HCC
- abolish usage of address_space GNU attribute
- use __device in file-scope global variables which would be accessed by GPU kernels
- temporarily disable some math functions which take pointer arguments
Change-Id: I730311dee848e20e763e35cd3980317fce0dce0d
Change-Id: I1f6b970b53b9401eeaaab08f04a7b9fed0fb8cf0
[ROCm/hip commit: efb9b9e86c]
1. It is an integer containing gfx values 701, 801, 802, 803
2. On NV path, it is zero
Change-Id: I2b4c7f48981d0214d8c6b1905d2cc85b16203419
[ROCm/hip commit: f86f3b3b33]
-
-Contexts across threads are listed under device
-Device reset cleans up all contexts and re-initializes _primaryCtx
Change-Id: Ie1cfbb26d43a8dc6869be3e6ebaf7344ce374643
[ROCm/hip commit: bddd6b73c0]
1. Changed test to assert for same hipFunction values
2. Added better memory management for hipModule
Change-Id: I10d7aef13c215a2211e262f3c79017f26a17d9a7
[ROCm/hip commit: 378eb3fa55]
1. Split hip_ir.ll to hip_hc.ll and hip_hc_gfx803.ll
a. hip_hc.ll contains arch generic ir implementations
b. hip_hc_gfx803.ll contains gfx803 (fiji, polaris) specific ir
2. HIPCC can now parse --amdgpu-target=*.
a. Usage: hipcc --amdgpu-target=gfx803 --amdgpu-target=gfx701
b. TODO: Convert to --amdgpu-target=gfx803,gfx701
3. With LLC in HCC able to generate native f16 isa, removed inline half asm math ops
4. Fixed threadfence and threadfence_block to use functions in rocdl
Change-Id: Ic9a9e3e04139b0d75d2c2a263c030ca77adc1019
[ROCm/hip commit: 01b66dd998]
Fail sub-optimal-copies rather than perform them slowly.
SOC occur on async copy of unpinned memory, or P2P copy between GPUs
that are not peers.
[ROCm/hip commit: 0409bf639c]
1. Added math_functions.h to hip_runtime.h
2. Changed operator overloading classifier static to static inline
3. Added vector types test for gpu
4. Seperated __host__ and __device__ for math functions in headers
Change-Id: I499862fad5d7b10da686da9011d7ecefe523f8e2
[ROCm/hip commit: f537d96633]
1. Fixed compilation issues for tests
2. Added missing intrinsics + math functions
3. Disabled some device functions as they are causing linking error with HCC
Change-Id: I79d52c4c7a539cc8ef40580247ad97ffcb975f09
[ROCm/hip commit: ea382e15f8]
1. All fp32, fp64 math device/host functions should be in math_functions.h/.cpp
2. All fp32, fp64 fast math intrinsics for device/host functions should be in device_functions.h/.cpp
3. All the device code implementations should be in device_util.h/.cpp
4. Hence, made changes appropriately by moving code and creating new header files
5. Added math_functions.cpp/.h
6. Changed #ifndef signature to make sure no conflicts between headers with same names in hip/hip_runtime.h and hip/hcc_detail/hip_runtime.h
7. Changed tests to fit the code changes, making them to include appropriate headers
8. Added math_functions.cpp to CMakeLists.txt
9. Some of the tests are still broken, mostly host math functions will fix them in next commit
10. TODO: FIX compilation issues for host math functions
Change-Id: I7a17637d7e294a7d224ffba932c1a08668febd26
[ROCm/hip commit: b723169ee9]
1. As we use holder data structure, we move all the cmp, math, cvt apis to cpp file
2. All the tests passed
3. Add more extensive testing for half
Change-Id: I92c6399dace602a0a24432728e3f2a07124e6fb1
[ROCm/hip commit: 18631efbc0]
1. Added all type conversion intrinsics
2. NO TESTS have been added. (Will add in next commit)
3. Sanatized code in hip_runtime.h
4. Added passed() to hipTestHalf to make it pass on HIT
Change-Id: I0987963c802fc7ff4d7e07d7b88d86da35da53c9
[ROCm/hip commit: 6f2cfddc67]
1. They use SDWA + LLVM IR
2. Added these functions to test
3. Need to do exp, exp10, log, log10, rint
Change-Id: I06176acc6cb8bb054495310531777406a41b54e4
[ROCm/hip commit: 0e576295b4]
1. Added SDWA implementation inside IR file
2. Added device functions to header + used them in test
Change-Id: Ib4e059a58eee201cc82438689e3e9bc5f9d26653
[ROCm/hip commit: 5ef8ef3bd7]
1. Removed HIP_EXPERIMENTAL env variable so that device code will be accessed from LLVM IR
2. Removed soft support from headers and moved to hip_fp16.cpp
3. Added LLVM IR + inline asm to hip_ir.ll
4. Added test for fp16
5. Added barriers for hcc 3.5 and hcc 4.0 for half support
a. Which means, hcc 4.0 can parse __fp16 but hcc 3.5 cant
b. HCC 4.0 code is implemented now, hcc 3.5 will be added later
Change-Id: Ic37859b2688ebb02e168bab643d1882bf4727952
[ROCm/hip commit: d180fdaae0]
Includes some tricky manipulation of the locks for contexts and streams.
issue is that stealing a stream requires we lock the context to
walk the streams to find a victim. To avoid deadlock, we can't
have a stream locked when we lock the context. This implementation
releases the stream lock, then acquires the context and selects the
victim.
A more stable implemenation might be to copy the stream list
from a context so that a lock is not required to walk all streams.
Smart shared_ptr could be used to prevent the streams from being
deallocated during the walk.
[ROCm/hip commit: a3e0012567]
Wen-Heng (Jack) Chung
Maneesh Gupta
Ben Sander
Rahul Garg
Aditya Atluri