diff --git a/library/axi_dmac/address_generator.v b/library/axi_dmac/address_generator.v
index 4b9593b4d..6404fe921 100644
--- a/library/axi_dmac/address_generator.v
+++ b/library/axi_dmac/address_generator.v
@@ -42,7 +42,8 @@ module address_generator #(
   parameter DMA_ADDR_WIDTH = 32,
   parameter BEATS_PER_BURST_WIDTH = 4,
   parameter BYTES_PER_BEAT_WIDTH = $clog2(DMA_DATA_WIDTH/8),
-  parameter LENGTH_WIDTH = 8)(
+  parameter LENGTH_WIDTH = 8,
+  parameter CACHE_COHERENT = 0)(
 
   input                        clk,
   input                        resetn,
@@ -80,7 +81,9 @@ localparam MAX_LENGTH = {BEATS_PER_BURST_WIDTH{1'b1}};
 
 assign burst = 2'b01;
 assign prot = 3'b000;
-assign cache = 4'b0011;
+// If CACHE_COHERENT is set, signal downstream that this transaction must be
+// looked up in cache. Otherwise default to "normal non-cachable bufferable".
+assign cache = CACHE_COHERENT ? 4'b1110 : 4'b0011;
 assign size = DMA_DATA_WIDTH == 1024 ? 3'b111 :
               DMA_DATA_WIDTH ==  512 ? 3'b110 :
               DMA_DATA_WIDTH ==  256 ? 3'b101 :
diff --git a/library/axi_dmac/axi_dmac.v b/library/axi_dmac/axi_dmac.v
index 0ba515ef8..4a93001d9 100644
--- a/library/axi_dmac/axi_dmac.v
+++ b/library/axi_dmac/axi_dmac.v
@@ -62,7 +62,8 @@ module axi_dmac #(
   parameter DMA_AXIS_DEST_W = 4,
   parameter DISABLE_DEBUG_REGISTERS = 0,
   parameter ENABLE_DIAGNOSTICS_IF = 0,
-  parameter ALLOW_ASYM_MEM = 0
+  parameter ALLOW_ASYM_MEM = 0,
+  parameter CACHE_COHERENT_DEST = 0
 ) (
   // Slave AXI interface
   input s_axi_aclk,
@@ -407,7 +408,8 @@ axi_dmac_regmap #(
   .HAS_DEST_ADDR(HAS_DEST_ADDR),
   .HAS_SRC_ADDR(HAS_SRC_ADDR),
   .DMA_2D_TRANSFER(DMA_2D_TRANSFER),
-  .SYNC_TRANSFER_START(SYNC_TRANSFER_START)
+  .SYNC_TRANSFER_START(SYNC_TRANSFER_START),
+  .CACHE_COHERENT_DEST(CACHE_COHERENT_DEST)
 ) i_regmap (
   .s_axi_aclk(s_axi_aclk),
   .s_axi_aresetn(s_axi_aresetn),
@@ -489,7 +491,8 @@ axi_dmac_transfer #(
   .AXI_LENGTH_WIDTH_SRC(8-(4*DMA_AXI_PROTOCOL_SRC)),
   .AXI_LENGTH_WIDTH_DEST(8-(4*DMA_AXI_PROTOCOL_DEST)),
   .ENABLE_DIAGNOSTICS_IF(ENABLE_DIAGNOSTICS_IF),
-  .ALLOW_ASYM_MEM(ALLOW_ASYM_MEM)
+  .ALLOW_ASYM_MEM(ALLOW_ASYM_MEM),
+  .CACHE_COHERENT_DEST(CACHE_COHERENT_DEST)
 ) i_transfer (
   .ctrl_clk(s_axi_aclk),
   .ctrl_resetn(s_axi_aresetn),
diff --git a/library/axi_dmac/axi_dmac_ip.tcl b/library/axi_dmac/axi_dmac_ip.tcl
index 9b13d1fd6..a7c2322b0 100644
--- a/library/axi_dmac/axi_dmac_ip.tcl
+++ b/library/axi_dmac/axi_dmac_ip.tcl
@@ -239,6 +239,7 @@ foreach {k v} { \
 		"AXI_SLICE_DEST" "false" \
 		"DISABLE_DEBUG_REGISTERS" "false" \
     "ENABLE_DIAGNOSTICS_IF" "false" \
+    "CACHE_COHERENT_DEST" "false" \
 	} { \
 	set_property -dict [list \
 			"value_format" "bool" \
@@ -326,6 +327,18 @@ set_property -dict [list \
 	"display_name" "Transfer Start Synchronization Support" \
 ] $p
 
+set p [ipgui::get_guiparamspec -name "CACHE_COHERENT_DEST" -component $cc]
+ipgui::move_param -component $cc -order 4 $p -parent $dest_group
+set_property -dict [list \
+	"tooltip" "Assume destination port ensures cache coherency (e.g. Ultrascale HPC port)" \
+] $p
+set_property -dict [list \
+	"display_name" "Assume cache coherent" \
+	"enablement_tcl_expr" "\$DMA_TYPE_DEST == 0 && \$DMA_AXI_PROTOCOL_DEST == 0" \
+	"value_tcl_expr" "\$DMA_TYPE_DEST == 0 && \$DMA_AXI_PROTOCOL_DEST == 0" \
+	"enablement_value" "false" \
+] [ipx::get_user_parameters CACHE_COHERENT_DEST -of_objects $cc]
+
 set general_group [ipgui::add_group -name "General Configuration" -component $cc \
 		-parent $page0 -display_name "General Configuration"]
 
diff --git a/library/axi_dmac/axi_dmac_regmap.v b/library/axi_dmac/axi_dmac_regmap.v
index 1274949a9..2905973ae 100644
--- a/library/axi_dmac/axi_dmac_regmap.v
+++ b/library/axi_dmac/axi_dmac_regmap.v
@@ -50,7 +50,8 @@ module axi_dmac_regmap #(
   parameter HAS_DEST_ADDR = 1,
   parameter HAS_SRC_ADDR = 1,
   parameter DMA_2D_TRANSFER = 0,
-  parameter SYNC_TRANSFER_START = 0
+  parameter SYNC_TRANSFER_START = 0,
+  parameter CACHE_COHERENT_DEST = 0
 ) (
   // Slave AXI interface
   input s_axi_aclk,
@@ -114,7 +115,7 @@ module axi_dmac_regmap #(
   input [31:0] dbg_ids1
 );
 
-localparam PCORE_VERSION = 'h00040361;
+localparam PCORE_VERSION = 'h00040461;
 
 // Register interface signals
 reg [31:0] up_rdata = 32'h00;
@@ -205,6 +206,7 @@ always @(posedge s_axi_aclk) begin
                          4'b0,BYTES_PER_BURST_WIDTH[3:0],
                          2'b0,DMA_TYPE_SRC[1:0],BYTES_PER_BEAT_WIDTH_SRC[3:0],
                          2'b0,DMA_TYPE_DEST[1:0],BYTES_PER_BEAT_WIDTH_DEST[3:0]};
+    9'h005: up_rdata <= {31'd0, CACHE_COHERENT_DEST};
     9'h020: up_rdata <= up_irq_mask;
     9'h021: up_rdata <= up_irq_pending;
     9'h022: up_rdata <= up_irq_source;
diff --git a/library/axi_dmac/axi_dmac_transfer.v b/library/axi_dmac/axi_dmac_transfer.v
index 45dc1456d..0597b6021 100644
--- a/library/axi_dmac/axi_dmac_transfer.v
+++ b/library/axi_dmac/axi_dmac_transfer.v
@@ -58,7 +58,8 @@ module axi_dmac_transfer #(
   parameter AXI_LENGTH_WIDTH_SRC = 8,
   parameter AXI_LENGTH_WIDTH_DEST = 8,
   parameter ENABLE_DIAGNOSTICS_IF = 0,
-  parameter ALLOW_ASYM_MEM = 0
+  parameter ALLOW_ASYM_MEM = 0,
+  parameter CACHE_COHERENT_DEST = 0
 ) (
   input ctrl_clk,
   input ctrl_resetn,
@@ -337,7 +338,8 @@ request_arb #(
   .AXI_LENGTH_WIDTH_DEST (AXI_LENGTH_WIDTH_DEST),
   .AXI_LENGTH_WIDTH_SRC (AXI_LENGTH_WIDTH_SRC),
   .ENABLE_DIAGNOSTICS_IF(ENABLE_DIAGNOSTICS_IF),
-  .ALLOW_ASYM_MEM (ALLOW_ASYM_MEM)
+  .ALLOW_ASYM_MEM (ALLOW_ASYM_MEM),
+  .CACHE_COHERENT_DEST(CACHE_COHERENT_DEST)
 ) i_request_arb (
   .req_clk (req_clk),
   .req_resetn (req_resetn),
diff --git a/library/axi_dmac/dest_axi_mm.v b/library/axi_dmac/dest_axi_mm.v
index 251cf42df..c52ad9852 100644
--- a/library/axi_dmac/dest_axi_mm.v
+++ b/library/axi_dmac/dest_axi_mm.v
@@ -44,7 +44,8 @@ module dest_axi_mm #(
   parameter BEATS_PER_BURST_WIDTH = 4,
   parameter MAX_BYTES_PER_BURST = 128,
   parameter BYTES_PER_BURST_WIDTH = $clog2(MAX_BYTES_PER_BURST),
-  parameter AXI_LENGTH_WIDTH = 8)(
+  parameter AXI_LENGTH_WIDTH = 8,
+  parameter CACHE_COHERENT = 0)(
 
   input                               m_axi_aclk,
   input                               m_axi_aresetn,
@@ -115,7 +116,8 @@ address_generator #(
   .BYTES_PER_BEAT_WIDTH(BYTES_PER_BEAT_WIDTH),
   .DMA_DATA_WIDTH(DMA_DATA_WIDTH),
   .LENGTH_WIDTH(AXI_LENGTH_WIDTH),
-  .DMA_ADDR_WIDTH(DMA_ADDR_WIDTH)
+  .DMA_ADDR_WIDTH(DMA_ADDR_WIDTH),
+  .CACHE_COHERENT(CACHE_COHERENT)
 ) i_addr_gen (
   .clk(m_axi_aclk),
   .resetn(m_axi_aresetn),
diff --git a/library/axi_dmac/request_arb.v b/library/axi_dmac/request_arb.v
index 879b586dd..1c26619f0 100644
--- a/library/axi_dmac/request_arb.v
+++ b/library/axi_dmac/request_arb.v
@@ -57,7 +57,8 @@ module request_arb #(
   parameter AXI_LENGTH_WIDTH_SRC = 8,
   parameter AXI_LENGTH_WIDTH_DEST = 8,
   parameter ENABLE_DIAGNOSTICS_IF = 0,
-  parameter ALLOW_ASYM_MEM = 0
+  parameter ALLOW_ASYM_MEM = 0,
+  parameter CACHE_COHERENT_DEST = 0
 )(
   input req_clk,
   input req_resetn,
@@ -361,7 +362,8 @@ dest_axi_mm #(
   .BYTES_PER_BEAT_WIDTH(BYTES_PER_BEAT_WIDTH_DEST),
   .MAX_BYTES_PER_BURST(MAX_BYTES_PER_BURST),
   .BYTES_PER_BURST_WIDTH(BYTES_PER_BURST_WIDTH),
-  .AXI_LENGTH_WIDTH(AXI_LENGTH_WIDTH_DEST)
+  .AXI_LENGTH_WIDTH(AXI_LENGTH_WIDTH_DEST),
+  .CACHE_COHERENT(CACHE_COHERENT_DEST)
 ) i_dest_dma_mm (
   .m_axi_aclk(m_dest_axi_aclk),
   .m_axi_aresetn(dest_resetn),