diff --git a/runtime/hsa-runtime/core/runtime/trap_handler/trap_handler.s b/runtime/hsa-runtime/core/runtime/trap_handler/trap_handler.s
index fdbbfabfe7..1803f4fa10 100644
--- a/runtime/hsa-runtime/core/runtime/trap_handler/trap_handler.s
+++ b/runtime/hsa-runtime/core/runtime/trap_handler/trap_handler.s
@@ -225,7 +225,7 @@ trap_entry:
   //
   // ttmp7:  pc_lo[31:0]
   // ttmp11: 1st_level_ttmp11[31:23] pc_hi[15:0] 1st_level_ttmp11[6:0]
-.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor < 4) || (.amdgcn.gfx_generation_number == 10 && .amdgcn.gfx_generation_minor < 3)
+.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor < 4) || (.amdgcn.gfx_generation_number == 10 && .amdgcn.gfx_generation_minor < 3) || (.amdgcn.gfx_generation_number == 11)
   // Save the PC
   s_mov_b32            ttmp7, ttmp0
   s_and_b32            ttmp1, ttmp1, SQ_WAVE_PC_HI_ADDRESS_MASK
@@ -283,3 +283,11 @@ trap_entry:
 .parked:
   s_trap               0x2
   s_branch             .parked
+
+// For gfx11, add padding instructions so we can ensure instruction cache
+// prefetch always has something to load.
+.if .amdgcn.gfx_generation_number == 11
+.rept (256 - ((. - trap_entry) % 64)) / 4
+  s_code_end
+.endr
+.endif
diff --git a/runtime/hsa-runtime/loader/executable.cpp b/runtime/hsa-runtime/loader/executable.cpp
index b659d757e5..2a3b96f243 100644
--- a/runtime/hsa-runtime/loader/executable.cpp
+++ b/runtime/hsa-runtime/loader/executable.cpp
@@ -80,8 +80,9 @@ __attribute__((noinline)) static void _loader_debug_state() {
 // 5: New trap handler ABI. Save the PC in ttmp11[22:7] ttmp6[31:0], and park the wave if stopped
 // 6: New trap handler ABI. ttmp6[25:0] contains dispatch index modulo queue size
 // 7: New trap handler ABI. Send interrupts as a bitmask, coalescing concurrent exceptions.
-// 8: New trap handler ABI for gfx940: Initialize ttmp[4:5] if ttmp11[31] == 0.
-HSA_API r_debug _amdgpu_r_debug = {8,
+// 8: New trap handler ABI. for gfx940: Initialize ttmp[4:5] if ttmp11[31] == 0.
+// 9: New trap handler API. For gfx11: Save PC in ttmp11[22:7] ttmp6[31:0], and park the wave if stopped.
+HSA_API r_debug _amdgpu_r_debug = {9,
                            nullptr,
                            reinterpret_cast<uintptr_t>(&_loader_debug_state),
                            r_debug::RT_CONSISTENT,