Import latest aicc hipcc fp8 pa snapshot.

Source: feature/aicc-hipcc-unified-attn-fp8-pa @ fc89765

Import latest aicc hipcc fp8 pa snapshot.
Source: feature/aicc-hipcc-unified-attn-fp8-pa @ fc89765
5248d7d2 · hly · c2a1b310 · 5248d7d2 · 5248d7d2 · 5248d7d2
Commit 5248d7d2 authored Jun 05, 2026 by hly
20 changed files
--- a/csrc/flash_attn_hg/src/flash_fwd_b4_pa.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_b4_pa.h
--- a/csrc/flash_attn_hg/src/flash_fwd_b8_fa.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_b8_fa.h
--- a/csrc/flash_attn_hg/src/flash_fwd_b8_mla.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_b8_mla.h
--- a/csrc/flash_attn_hg/src/flash_fwd_b8_pa.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_b8_pa.h
--- a/csrc/flash_attn_hg/src/flash_fwd_kernel.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_kernel.h
--- a/csrc/flash_attn_hg/src/flash_fwd_launch_template.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_launch_template.h
--- a/csrc/flash_attn_hg/src/flash_fwd_launch_template_mla.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_launch_template_mla.h
--- a/csrc/flash_attn_hg/src/flash_fwd_launch_template_pa.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_launch_template_pa.h
--- a/csrc/flash_attn_hg/src/flash_fwd_permute_hdim128.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_permute_hdim128.h
--- a/csrc/flash_attn_hg/src/flash_fwd_reduce.h
+++ b/csrc/flash_attn_hg/src/flash_fwd_reduce.h
--- a/csrc/flash_attn_hg/src/flash_permute_api.h
+++ b/csrc/flash_attn_hg/src/flash_permute_api.h
--- a/csrc/flash_attn_hg/src/flash_permute_hdim128.h
+++ b/csrc/flash_attn_hg/src/flash_permute_hdim128.h
--- a/csrc/flash_attn_hg/src/flash_sumout_api.h
+++ b/csrc/flash_attn_hg/src/flash_sumout_api.h
--- a/csrc/flash_attn_hg/src/generate_kernels.py
+++ b/csrc/flash_attn_hg/src/generate_kernels.py
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim128_bf16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim128_bf16.cpp
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim128_fp16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim128_fp16.cpp
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim160_bf16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim160_bf16.cpp
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim160_fp16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim160_fp16.cpp
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim192_bf16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim192_bf16.cpp
--- a/csrc/flash_attn_hg/src/target/flash_bwd_hdim192_fp16.cpp
+++ b/csrc/flash_attn_hg/src/target/flash_bwd_hdim192_fp16.cpp