diff --git a/tensorflow/core/grappler/optimizers/auto_mixed_precision.cc b/tensorflow/core/grappler/optimizers/auto_mixed_precision.cc index 4da6f454d2486e..f44540792d7b22 100644 --- a/tensorflow/core/grappler/optimizers/auto_mixed_precision.cc +++ b/tensorflow/core/grappler/optimizers/auto_mixed_precision.cc @@ -116,7 +116,7 @@ bool HasFastFP16Support(const DeviceProperties& props) { #elif TENSORFLOW_USE_ROCM absl::flat_hash_set FP16SupportedDevices = { {"gfx906"}, {"gfx908"}, {"gfx90a"}, {"gfx910"}, {"gfx940"}, {"gfx941"}, - {"gfx942"}, {"gfx1010"}, {"gfx1012"}, {"gfx1030"}, {"gfx1100"} + {"gfx942"}, {"gfx1010"}, {"gfx1012"}, {"gfx1030"}, {"gfx1100"}, {"gfx1103"} }; std::string gcnArchName = props.environment().at("architecture"); std::vector gpu_arch = absl::StrSplit(gcnArchName, ":"); diff --git a/tensorflow/tools/ci_build/Dockerfile.rocm b/tensorflow/tools/ci_build/Dockerfile.rocm index fd7a78a657543a..c96584f299bc0c 100644 --- a/tensorflow/tools/ci_build/Dockerfile.rocm +++ b/tensorflow/tools/ci_build/Dockerfile.rocm @@ -7,7 +7,7 @@ ARG ROCM_DEB_REPO=https://repo.radeon.com/rocm/apt/6.2/ ARG ROCM_BUILD_NAME=ubuntu ARG ROCM_BUILD_NUM=main ARG ROCM_PATH=/opt/rocm-6.2.0 -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ARG DEBIAN_FRONTEND=noninteractive ENV TF_NEED_ROCM 1 diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux2014 b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux2014 index 428d84e283fdd8..68030cf73e872e 100644 --- a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux2014 +++ b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux2014 @@ -8,7 +8,7 @@ COPY setup.packages.rocm.cs7.sh setup.packages.rocm.cs7.sh COPY builder.packages.rocm.cs7.txt builder.packages.rocm.cs7.txt RUN /setup.packages.rocm.cs7.sh /builder.packages.rocm.cs7.txt -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ENV GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS} # Install ROCM diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux_2_28 b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux_2_28 index 62fa83359940a9..31efa92f0b2774 100644 --- a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux_2_28 +++ b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.manylinux_2_28 @@ -17,7 +17,7 @@ COPY setup.packages.rocm.el8.sh setup.packages.rocm.el8.sh COPY builder.packages.rocm.el8.txt builder.packages.rocm.el8.txt RUN /setup.packages.rocm.el8.sh /builder.packages.rocm.el8.txt -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ENV GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS} # Install ROCM diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub20 b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub20 index b576b9ca7dc85f..841e94010e38b6 100644 --- a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub20 +++ b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub20 @@ -2,7 +2,7 @@ FROM ubuntu:20.04 ################################################################################ -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ENV GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS} # Install build dependencies diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub22 b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub22 index 1c325ec37f4b0c..8b167f594a1e76 100644 --- a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub22 +++ b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub22 @@ -2,7 +2,7 @@ FROM ubuntu:22.04 ################################################################################ -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ENV GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS} # Install build dependencies diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub24 b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub24 index 21875e44ef50e0..77caf99977ed7d 100644 --- a/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub24 +++ b/tensorflow/tools/tf_sig_build_dockerfiles/Dockerfile.rocm.ub24 @@ -2,7 +2,7 @@ FROM ubuntu:24.04 ################################################################################ -ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100" +ARG GPU_DEVICE_TARGETS="gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103" ENV GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS} # Install build dependencies diff --git a/tensorflow/tools/tf_sig_build_dockerfiles/setup.rocm.sh b/tensorflow/tools/tf_sig_build_dockerfiles/setup.rocm.sh index 7d3a78d5c50dad..06545bac680ef8 100755 --- a/tensorflow/tools/tf_sig_build_dockerfiles/setup.rocm.sh +++ b/tensorflow/tools/tf_sig_build_dockerfiles/setup.rocm.sh @@ -145,7 +145,7 @@ then echo "build:rocm_base --copt=-fclang-abi-compat=17" >> /etc/bazel.bazelrc fi -GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS:-"gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100"} +GPU_DEVICE_TARGETS=${GPU_DEVICE_TARGETS:-"gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx1100 gfx1103"} echo $ROCM_VERSION echo $ROCM_REPO diff --git a/third_party/xla/xla/stream_executor/device_description.h b/third_party/xla/xla/stream_executor/device_description.h index 99d7f1ce5d83c8..32193b37f3a2e1 100644 --- a/third_party/xla/xla/stream_executor/device_description.h +++ b/third_party/xla/xla/stream_executor/device_description.h @@ -190,13 +190,15 @@ class RocmComputeCapability { bool gfx11_rx7900() const { return gfx_version() == "gfx1100"; } + bool gfx11_r780m() const { return gfx_version() == "gfx1103"; } + bool has_nhwc_layout_support() const { return gfx9_mi100_or_later(); } bool has_bf16_dtype_support() const { return gfx9_mi100_or_later(); } bool has_fast_fp16_support() const { return gfx9_mi100_or_later() || gfx10_rx68xx() || gfx10_rx69xx() || - gfx11_rx7900(); + gfx11_rx7900() || gfx11_r780m(); } bool has_mfma_instr_support() const { return gfx9_mi100_or_later(); } @@ -239,7 +241,8 @@ class RocmComputeCapability { "gfx90a", // MI200 "gfx940", "gfx941", "gfx942", // MI300 "gfx1030", // RX68xx / RX69xx - "gfx1100" // RX7900 + "gfx1100", // RX7900 + "gfx1103" // Radeon 780M }; };