GL1

GL1C is the shared L1 cache layer on gfx1151, supplied by TCP (GL0). For TCP panels and Memory Chart rows through the TCP-GL1 boundary, see GL0. For downstream GL2C, GCEA, and DRAM panels after GL1C, see GL2.

GL1C panels

GL1C utilization

RDNA 3.5 (gfx1151)

Metric	Description	Unit
GL1C Busy	Percentage of cycles where the GL1C cache was busy processing requests. GL1C is the shared L1 cache within each Shader Engine. Per sample: 100 * GL1C_BUSY / GL1C_CYCLE (when GL1C_CYCLE != 0).	Percent
GL1C Starve	Percentage of cycles where GL1C had no pending requests from TCP. Per sample: 100 * GL1C_STARVE / GL1C_CYCLE (when GL1C_CYCLE != 0).	Percent

GL1C request statistics

RDNA 3.5 (gfx1151)

Metric	Description	Unit
Total Requests	Total number of requests to the GL1C cache from TCP.	Count per Normalization Unit
Read Requests	Number of read requests to the GL1C cache.	Count per Normalization Unit
Write Requests	Number of write requests to the GL1C cache.	Count per Normalization Unit
Miss Requests	Number of requests that missed in the GL1C cache.	Count per Normalization Unit

GL1C cache performance

RDNA 3.5 (gfx1151)

Metric	Description	Unit
Hit Rate	Percentage of GL1C requests that hit in the cache.	Percent

GL1C-GL2 interface

RDNA 3.5 (gfx1151)

Metric	Description	Unit
GL2 Read Requests	Number of read requests sent from GL1C to GL2C due to cache misses.	Count per Normalization Unit
GL2 Read 128B Requests	GL1C→GL2 read requests using 128-byte lines (GL1C_GL2_REQ_READ_128B_sum / $denom).	Count per Normalization Unit
GL2 Write Requests	GL1C→GL2 write requests (GL1C_GL2_REQ_WRITE_sum / $denom).	Count per Normalization Unit

GL1C stalls

RDNA 3.5 (gfx1151)

Metric	Description	Unit
GL2 Stall	Cycles GL1C was stalled due to backpressure from GL2C.	Cycles per Normalization Unit
LFIFO Full Stall	Cycles GL1C stalled because the return/load FIFO was full (GL1C_STALL_LFIFO_FULL_sum / $denom), often indicating downstream or consumer backpressure.	Cycles per Normalization Unit